1
00:00:01,470 --> 00:00:04,140
- Hello, my name is Peter Molfese.

2
00:00:04,140 --> 00:00:05,620
I'm the science lead here

3
00:00:05,620 --> 00:00:07,610
at the Center for Multimodal Neuroimaging

4
00:00:07,610 --> 00:00:09,720
at the National Institutes
of Mental Health.

5
00:00:09,720 --> 00:00:11,983
Welcome to our workshop on PET-MRI.

6
00:00:13,370 --> 00:00:15,660
This is an entirely online workshop.

7
00:00:15,660 --> 00:00:18,240
You can access more
information on our website,

8
00:00:18,240 --> 00:00:23,240
cmn.nimh.nih.gov/cmnworkshop2022.

9
00:00:24,640 --> 00:00:28,240
You can also rewatch this workshop anytime

10
00:00:28,240 --> 00:00:31,293
at the videocast.nih.gov link below.

11
00:00:32,200 --> 00:00:35,370
We've collected seven
leading experts on PET-MRI

12
00:00:35,370 --> 00:00:36,530
to talk today,

13
00:00:36,530 --> 00:00:40,360
and their presentations range
from 30 minutes to 60 minutes.

14
00:00:40,360 --> 00:00:41,630
Following today's workshop,

15
00:00:41,630 --> 00:00:44,310
we will have a round-table
discussion this fall,

16
00:00:44,310 --> 00:00:47,200
yet to be scheduled, in
either August or September.

17
00:00:47,200 --> 00:00:49,510
It will also be online streaming only.

18
00:00:49,510 --> 00:00:51,070
And you can register for information

19
00:00:51,070 --> 00:00:54,200
about this on Eventbrite
or using this link,

20
00:00:54,200 --> 00:00:59,200
tinyurl.com/3tzfuuv5.

21
00:01:00,580 --> 00:01:02,170
You can also find out more information

22
00:01:02,170 --> 00:01:04,340
on how to submit
questions to this workshop

23
00:01:04,340 --> 00:01:08,663
using the link cmn-meetings.nimh.nih.gov.

24
00:01:10,940 --> 00:01:12,820
The agenda for the workshop today

25
00:01:12,820 --> 00:01:15,520
is this talk, "You Are Here,"

26
00:01:15,520 --> 00:01:18,550
followed by a kickoff by Dr. Bruce Rosen,

27
00:01:18,550 --> 00:01:22,360
covering basics of PET-MRI
and different classes

28
00:01:22,360 --> 00:01:25,800
of experiments that can
be conceptualized within.

29
00:01:25,800 --> 00:01:29,780
We then proceed alphabetically
by last name of our speakers.

30
00:01:29,780 --> 00:01:33,780
So next is Dr. Richard
Carson, then Dr. Audrey Fan,

31
00:01:33,780 --> 00:01:37,983
then Dr. Andreas Hahn,
then Dr. Sharna Jamadar,

32
00:01:39,080 --> 00:01:40,940
followed by Dr. Christine Sanders

33
00:01:40,940 --> 00:01:43,043
and, finally, Dr. Dardo Tomasi.

34
00:01:45,300 --> 00:01:46,700
If at any time you have questions,

35
00:01:46,700 --> 00:01:49,700
please refer to our websites or email me

36
00:01:49,700 --> 00:01:52,820
at Peter.Molfese@nih.gov.

37
00:01:52,820 --> 00:01:53,653
Thank you.

38
00:01:56,623 --> 00:01:58,393
- Hi. My name is Bruce Rosen.

39
00:01:58,393 --> 00:02:01,353
It's my pleasure to be
here today to talk to you

40
00:02:01,353 --> 00:02:03,093
about MR/PET.

41
00:02:03,093 --> 00:02:04,873
Give a bit of an introduction,

42
00:02:04,873 --> 00:02:08,776
both with a backward and
mostly a forward look.

43
00:02:09,763 --> 00:02:13,773
When we think about the
history of the field of MR/PET,

44
00:02:13,773 --> 00:02:16,723
it's worth remembering that
both of these modalities

45
00:02:16,723 --> 00:02:19,693
have been around for
more than 50 years now.

46
00:02:19,693 --> 00:02:22,956
So when I think about the
early history of MR/PET,

47
00:02:24,283 --> 00:02:26,283
this is what comes to mind.

48
00:02:26,283 --> 00:02:28,653
The notion of parallel play.

49
00:02:28,653 --> 00:02:30,223
Here are two toddlers.

50
00:02:30,223 --> 00:02:31,493
They're playing next to each other,

51
00:02:31,493 --> 00:02:34,103
but not really interacting directly.

52
00:02:34,103 --> 00:02:35,813
And this was largely the case

53
00:02:35,813 --> 00:02:40,543
for the first 30, even the
first 40 years of MR And PET.

54
00:02:40,543 --> 00:02:42,526
They were both around doing great work,

55
00:02:43,373 --> 00:02:46,413
but not really interacting very much.

56
00:02:46,413 --> 00:02:48,503
And in fact, this was equally true,

57
00:02:48,503 --> 00:02:50,403
not just of individuals within the field,

58
00:02:50,403 --> 00:02:52,056
but really of our societies.

59
00:02:53,023 --> 00:02:57,363
Large groups of investigators
working in each domain

60
00:02:57,363 --> 00:03:00,906
with not very much
interaction between them.

61
00:03:02,663 --> 00:03:04,493
Yet this is surprising in some ways,

62
00:03:04,493 --> 00:03:07,043
given that we're really not so different

63
00:03:07,043 --> 00:03:08,863
in terms of our technologies.

64
00:03:08,863 --> 00:03:11,703
First of course, we both
have nuclear in our name.

65
00:03:11,703 --> 00:03:13,113
And it's worth their memory, of course,

66
00:03:13,113 --> 00:03:14,953
that magnetic resonance imaging

67
00:03:14,953 --> 00:03:18,143
really started as nuclear
magnetic resonance.

68
00:03:18,143 --> 00:03:23,143
So, we both can image carbon-11
or 13, as in this case.

69
00:03:23,493 --> 00:03:27,193
Though, in case the PET
folks are complaining

70
00:03:27,193 --> 00:03:32,193
about the challenges of the
short half-lives of carbon-11,

71
00:03:32,293 --> 00:03:36,513
carbon-13's half-life
is less than 60 seconds.

72
00:03:36,513 --> 00:03:39,863
So, you better get your
procedures for working quickly

73
00:03:40,853 --> 00:03:44,413
in the pharmacy in order.

74
00:03:44,413 --> 00:03:47,313
In addition to carbon, oxygen.

75
00:03:47,313 --> 00:03:49,183
We can image oxygen-15 with PET.

76
00:03:49,183 --> 00:03:53,233
We can image oxygen-17 with MR

77
00:03:53,233 --> 00:03:55,823
And just like the expense

78
00:03:55,823 --> 00:03:59,143
of generating radioisotope tracers,

79
00:03:59,143 --> 00:04:03,013
oxygen 17-is about $500 in ml.

80
00:04:03,013 --> 00:04:07,303
And because it's not a tracer,
we need lots and lots of mls

81
00:04:07,303 --> 00:04:08,853
if we're gonna image a human.

82
00:04:08,853 --> 00:04:13,466
So, we both have the challenges
of paying for our agents.

83
00:04:14,923 --> 00:04:19,263
Fluorine, of course, is
another common tracer.

84
00:04:19,263 --> 00:04:21,493
Probably the most common in PET.

85
00:04:21,493 --> 00:04:23,033
And fluorine-19

86
00:04:23,033 --> 00:04:28,033
is a viable MR of visible nucleus.

87
00:04:29,103 --> 00:04:31,653
However, in this case,

88
00:04:31,653 --> 00:04:35,353
perhaps the vaunted image quality of MR

89
00:04:35,353 --> 00:04:37,133
really may be on PET side.

90
00:04:37,133 --> 00:04:40,133
So perhaps, we should be thinking
about using the PET image

91
00:04:40,133 --> 00:04:43,113
as the priors for our
reconstruction of the MR data,

92
00:04:43,113 --> 00:04:44,806
rather than the other way around.

93
00:04:46,503 --> 00:04:49,773
Finally, even imaging
perfusion with water.

94
00:04:49,773 --> 00:04:52,453
We can do it with PET.

95
00:04:52,453 --> 00:04:54,293
We can do it with MR.

96
00:04:54,293 --> 00:04:56,933
And it's rather difficult to choose

97
00:04:56,933 --> 00:04:59,143
which is your preferred image,

98
00:04:59,143 --> 00:05:01,106
as you can see in these pictures here.

99
00:05:02,913 --> 00:05:06,113
In addition, of course, we
both have our little issues.

100
00:05:06,113 --> 00:05:08,353
Spatial fidelity is an issue.

101
00:05:08,353 --> 00:05:10,773
We have distortions on the MR side.

102
00:05:10,773 --> 00:05:13,503
We have geometric
distortions on the PET side

103
00:05:13,503 --> 00:05:14,803
that we need to deal with.

104
00:05:16,263 --> 00:05:19,793
We also have lively discussions,

105
00:05:19,793 --> 00:05:21,663
sometimes heated discussions,

106
00:05:21,663 --> 00:05:24,183
over optimal modeling of our data.

107
00:05:24,183 --> 00:05:28,743
And I especially like the
title of the presentation

108
00:05:28,743 --> 00:05:31,893
that Rich Carson recently gave

109
00:05:31,893 --> 00:05:36,383
talking about the challenges
on the PET side of modeling.

110
00:05:36,383 --> 00:05:38,123
Plenty of madness in that field, perhaps,

111
00:05:38,123 --> 00:05:40,566
but the same is equally
true on the MR side.

112
00:05:42,263 --> 00:05:44,723
Finally, of course, safety is an issue

113
00:05:44,723 --> 00:05:46,793
for all of our imaging modalities.

114
00:05:46,793 --> 00:05:50,663
On the MR side, RF
heating, metal projectiles,

115
00:05:50,663 --> 00:05:52,443
contrast agent reactions.

116
00:05:52,443 --> 00:05:55,303
On the PET side, we have the QA associated

117
00:05:55,303 --> 00:05:57,193
with the radioisotope production,

118
00:05:57,193 --> 00:05:58,743
radiation safety.

119
00:05:58,743 --> 00:06:00,013
Combined device, of course,

120
00:06:00,013 --> 00:06:03,136
has all of these safety issues
that need to be dealt with.

121
00:06:04,263 --> 00:06:07,073
But if you're having trouble
sorting them out at this point,

122
00:06:07,073 --> 00:06:11,203
just remember we do MRI
mostly with stable nuclei

123
00:06:11,203 --> 00:06:13,623
in reference to our world leaders here.

124
00:06:13,623 --> 00:06:15,176
PET, unstable nuclei.

125
00:06:17,353 --> 00:06:21,003
Now given their similarities,

126
00:06:21,003 --> 00:06:23,943
why are we interested
in combining PET and MR?

127
00:06:23,943 --> 00:06:25,513
Of course, it's because

128
00:06:25,513 --> 00:06:29,043
that they largely have
complementary features

129
00:06:29,043 --> 00:06:31,283
as well as these areas of overlap.

130
00:06:31,283 --> 00:06:33,623
MR, of course, generally does

131
00:06:33,623 --> 00:06:36,193
have very high spatial resolution.

132
00:06:36,193 --> 00:06:40,363
And thus, is the generic
anatomical framework

133
00:06:40,363 --> 00:06:43,133
for most of our multimodal studies.

134
00:06:43,133 --> 00:06:46,123
It's quite a good tool for
measuring different facets

135
00:06:46,123 --> 00:06:48,173
of physiology and biophysics.

136
00:06:48,173 --> 00:06:49,493
It's not invasive.

137
00:06:49,493 --> 00:06:53,123
On the other hand, it generally
has quite poor sensitivity

138
00:06:53,123 --> 00:06:54,533
and it's quite challenging.

139
00:06:54,533 --> 00:06:57,233
They're not impossible to
quantify parameters from MR.

140
00:06:58,313 --> 00:07:00,663
PET has its limitations.

141
00:07:00,663 --> 00:07:03,323
Lesser spatial resolution perhaps.

142
00:07:03,323 --> 00:07:05,623
And with that, limited anatomical detail,

143
00:07:05,623 --> 00:07:10,103
especially with tracers
that have focal uptake.

144
00:07:10,103 --> 00:07:11,973
It also uses ionizing radiation,

145
00:07:11,973 --> 00:07:15,073
which is a limitation in
our ability in particular

146
00:07:15,073 --> 00:07:17,943
to study brain development in children.

147
00:07:17,943 --> 00:07:20,923
On the other hand, it is
exquisitely sensitive.

148
00:07:20,923 --> 00:07:22,563
Readily quantifiable,

149
00:07:22,563 --> 00:07:26,853
albeit with all the caveats
that Rich talked about.

150
00:07:26,853 --> 00:07:30,033
And most important, it allows us to assay

151
00:07:30,033 --> 00:07:33,916
a wide range of biological processes.

152
00:07:34,863 --> 00:07:37,773
The reason we're interested in
combining the two modalities

153
00:07:37,773 --> 00:07:40,093
of course, is to get
the strengths of both.

154
00:07:40,093 --> 00:07:43,280
High resolution physiology with MR

155
00:07:43,280 --> 00:07:44,953
and the exquisite sensitivity

156
00:07:44,953 --> 00:07:48,316
and wide range of biological
processes accessible with PET.

157
00:07:49,173 --> 00:07:52,213
And if we look at the list of features

158
00:07:52,213 --> 00:07:54,573
that we can get from PET and MR,

159
00:07:54,573 --> 00:07:57,513
this group, of course, is well familiar

160
00:07:57,513 --> 00:08:00,873
with this list and could
certainly add to it.

161
00:08:00,873 --> 00:08:02,046
On the PET side,

162
00:08:02,963 --> 00:08:07,073
great measures of metabolism.

163
00:08:07,073 --> 00:08:10,960
Hemodynamics perhaps will
increasingly fall to MR

164
00:08:10,960 --> 00:08:13,193
Although, we certainly
can measure it with PET.

165
00:08:13,193 --> 00:08:17,603
Protein synthesis, amino acid
transport, DNA synthesis,

166
00:08:17,603 --> 00:08:20,903
and especially in the
brain neurotransmitter

167
00:08:20,903 --> 00:08:25,903
and neuromodulator action,
all can be studied with PET

168
00:08:26,463 --> 00:08:27,953
really uniquely.

169
00:08:27,953 --> 00:08:28,983
MR, of course,

170
00:08:28,983 --> 00:08:33,113
is our anatomical modality of choice

171
00:08:33,113 --> 00:08:36,693
to for look for subtle changes
over time in brain anatomy.

172
00:08:36,693 --> 00:08:39,273
Hemodynamics with MR is good

173
00:08:39,273 --> 00:08:43,983
in getting better biophysical
properties like diffusion,

174
00:08:43,983 --> 00:08:47,023
functional properties like with the BOLD

175
00:08:47,023 --> 00:08:50,883
or hemodynamic measures with CBF and CBV

176
00:08:50,883 --> 00:08:53,263
to allow us to infer neural activity

177
00:08:53,263 --> 00:08:56,173
through its correlate to hemodynamics.

178
00:08:56,173 --> 00:08:58,463
MR spectroscopy is a way

179
00:08:58,463 --> 00:09:01,173
to look at endogenous neurochemistry.

180
00:09:01,173 --> 00:09:03,843
MR angiography, et cetera.

181
00:09:03,843 --> 00:09:07,183
Combine all of these
capabilities in one device

182
00:09:07,183 --> 00:09:09,353
and you have a pretty powerful tool

183
00:09:09,353 --> 00:09:13,046
to begin to study the brain.

184
00:09:14,223 --> 00:09:16,493
So, given that these technologies

185
00:09:16,493 --> 00:09:18,263
have been around for so long,

186
00:09:18,263 --> 00:09:19,963
why is it that we're having this

187
00:09:19,963 --> 00:09:24,443
a first of its kind conference
now almost 50 years on?

188
00:09:24,443 --> 00:09:27,263
Well, there were two
fundamental technical advances

189
00:09:27,263 --> 00:09:28,683
that were required

190
00:09:28,683 --> 00:09:31,953
in order to bring these
two modalities together.

191
00:09:31,953 --> 00:09:33,243
The first on the PET side

192
00:09:33,243 --> 00:09:35,893
was the development of
solid state detectors

193
00:09:35,893 --> 00:09:38,763
that were insensitive to magnetic fields.

194
00:09:38,763 --> 00:09:40,763
The conventional photo multiplier tube,

195
00:09:40,763 --> 00:09:44,353
unfortunately, was exquisitely
sensitive to magnetic fields

196
00:09:44,353 --> 00:09:48,213
and would not operate within
the high fields required

197
00:09:48,213 --> 00:09:49,593
for MRI.

198
00:09:49,593 --> 00:09:51,159
Of course, on the PET side,

199
00:09:51,159 --> 00:09:54,623
these new detectors, like
avalanche photodiodes

200
00:09:54,623 --> 00:09:57,353
and more recently
silicon photomultipliers,

201
00:09:57,353 --> 00:09:58,553
have made all the difference

202
00:09:58,553 --> 00:10:03,076
because of their insensitivity
to magnetic field.

203
00:10:03,076 --> 00:10:05,643
And this is just an example
from the wonderful work

204
00:10:05,643 --> 00:10:10,443
of Craig Levin looking at
silicon photomultipliers

205
00:10:10,443 --> 00:10:15,443
as a key element in moving PET technology

206
00:10:15,483 --> 00:10:17,336
into high magnetic fields.

207
00:10:18,173 --> 00:10:21,293
In addition to the
developments on the PET side,

208
00:10:21,293 --> 00:10:23,583
there were important
advances on the MR side.

209
00:10:23,583 --> 00:10:25,293
The most important of which

210
00:10:25,293 --> 00:10:28,003
was the design of new gradients

211
00:10:28,003 --> 00:10:30,743
which allowed a high quality performance

212
00:10:30,743 --> 00:10:33,323
with larger inner diameters.

213
00:10:33,323 --> 00:10:37,153
Now, Siemens was really the first

214
00:10:37,153 --> 00:10:41,843
to be able to move MR systems

215
00:10:41,843 --> 00:10:46,373
from the conventional 55 or
60 centimeter inner diameters

216
00:10:46,373 --> 00:10:47,743
out to 70 centimeters.

217
00:10:47,743 --> 00:10:49,843
And that gave the bore space.

218
00:10:49,843 --> 00:10:53,513
Of course, they developed
these wide bore systems

219
00:10:53,513 --> 00:10:56,063
mostly for wide bore Americans.

220
00:10:56,063 --> 00:11:00,413
When we say that Americans
are as fat as a horse,

221
00:11:00,413 --> 00:11:02,673
indeed, they sometimes can be.

222
00:11:02,673 --> 00:11:05,463
And that was the clinical driver

223
00:11:05,463 --> 00:11:07,063
for wide bore systems.

224
00:11:07,063 --> 00:11:08,723
But of course, that extra bore size

225
00:11:08,723 --> 00:11:12,173
could be used to incorporate
the technology required

226
00:11:12,173 --> 00:11:13,983
for PET detection.

227
00:11:13,983 --> 00:11:17,773
And indeed, the ability

228
00:11:17,773 --> 00:11:20,363
to create high performance
gradient systems

229
00:11:20,363 --> 00:11:23,243
with that extra bore real estate was key

230
00:11:23,243 --> 00:11:25,003
to being able to develop

231
00:11:25,003 --> 00:11:28,796
human scale simultaneous MR/PET devices.

232
00:11:29,873 --> 00:11:33,003
And since that time,

233
00:11:33,003 --> 00:11:35,923
many of the commercial MR manufacturers

234
00:11:35,923 --> 00:11:39,503
have subsequently gone to
these wide bore systems

235
00:11:39,503 --> 00:11:41,383
for their conventional MR

236
00:11:41,383 --> 00:11:42,663
And thus, have the advantage

237
00:11:42,663 --> 00:11:47,483
to be able to integrate PET
scanners into their MR devices.

238
00:11:47,483 --> 00:11:50,553
And at least three of the
large commercial vendors,

239
00:11:50,553 --> 00:11:54,023
General Electric, Siemens, United,

240
00:11:54,023 --> 00:11:57,573
all now have fully integrated PET scanners

241
00:11:57,573 --> 00:11:59,563
into their MR scanners.

242
00:11:59,563 --> 00:12:02,453
And of course, in addition
to the human scale scanners

243
00:12:02,453 --> 00:12:06,573
shown here, several pre-clinical companies

244
00:12:06,573 --> 00:12:08,163
now have pre-clinical scanners.

245
00:12:08,163 --> 00:12:12,646
So from everything from mice
all the way up to humans,

246
00:12:13,793 --> 00:12:17,496
we can now do fully integrated
PET and MR scanning.

247
00:12:19,703 --> 00:12:23,323
Well, how are we going to
use this combined tool?

248
00:12:23,323 --> 00:12:25,003
I'd like to kind of break it

249
00:12:25,003 --> 00:12:27,523
into several different
classes of experiments

250
00:12:27,523 --> 00:12:30,793
that we can consider now that
we have a combined device.

251
00:12:30,793 --> 00:12:33,823
And what I'll call Class 1 experiments,

252
00:12:33,823 --> 00:12:36,393
we're really thinking
about ways to use the data

253
00:12:36,393 --> 00:12:39,953
from one machine to infer
properties about the other

254
00:12:39,953 --> 00:12:42,813
or perhaps use the data
from each of these machines

255
00:12:42,813 --> 00:12:44,303
independently.

256
00:12:44,303 --> 00:12:47,183
Here, the focus is really
on unique information

257
00:12:47,183 --> 00:12:49,303
and properties of each modality,

258
00:12:49,303 --> 00:12:52,376
but simultaneous acquisition
is not essential.

259
00:12:53,473 --> 00:12:55,193
However, I'll just say parenthetically

260
00:12:55,193 --> 00:12:57,433
that we shouldn't discount the value

261
00:12:57,433 --> 00:12:59,303
of simultaneous acquisition,

262
00:12:59,303 --> 00:13:04,223
especially in studies in
sensitive patient populations,

263
00:13:04,223 --> 00:13:07,383
where moving from one
scanner to another scanner

264
00:13:07,383 --> 00:13:09,803
actually is a significant prohibition.

265
00:13:09,803 --> 00:13:14,803
So this in typically includes
this class of experiments,

266
00:13:15,003 --> 00:13:18,073
typically includes many
of the clinical studies

267
00:13:18,073 --> 00:13:20,953
that we routinely use MR And PET for,

268
00:13:20,953 --> 00:13:21,953
but as we'll see

269
00:13:21,953 --> 00:13:25,973
and there are a number of
other examples of this class.

270
00:13:25,973 --> 00:13:29,363
So, this single image

271
00:13:29,363 --> 00:13:31,883
from my colleague Onofrio Catalano

272
00:13:31,883 --> 00:13:36,013
just highlights the utility
of the MR and PET together

273
00:13:37,233 --> 00:13:39,273
in clinical use.

274
00:13:39,273 --> 00:13:42,603
Most clinical use is focused on oncology

275
00:13:42,603 --> 00:13:44,913
and much of it in the body.

276
00:13:44,913 --> 00:13:47,133
In this case, the combined MR

277
00:13:47,133 --> 00:13:51,033
relative to say the traditional PET CT

278
00:13:51,033 --> 00:13:56,033
was able to rule out the
presence of a metastasis

279
00:13:56,763 --> 00:14:00,053
in a patient that was suspicious for that.

280
00:14:00,053 --> 00:14:03,213
And thus, made this patient
potentially resectable.

281
00:14:03,213 --> 00:14:07,743
The combined MR/PET study
made an important difference

282
00:14:07,743 --> 00:14:09,856
in the treatment path of this patient.

283
00:14:11,913 --> 00:14:13,573
In the brain side,

284
00:14:13,573 --> 00:14:18,573
we are increasingly using
the anatomical frameworks

285
00:14:19,073 --> 00:14:22,433
that we can get from MR to
better inform our PET data.

286
00:14:22,433 --> 00:14:26,403
And this very nice paper
by Vincent Beliveau.

287
00:14:26,403 --> 00:14:29,153
He worked with Doug Greve,
who's one of the pioneers

288
00:14:29,153 --> 00:14:32,773
of the so-called cortical
flattening technique.

289
00:14:32,773 --> 00:14:36,293
The FreeSurfer tool to
basically take receptor maps

290
00:14:36,293 --> 00:14:38,583
across five different

291
00:14:38,583 --> 00:14:42,293
5-HT receptor classes.

292
00:14:42,293 --> 00:14:46,363
And not only image them
in the tomographic space,

293
00:14:46,363 --> 00:14:49,353
but bring them into the
natural framework of the cortex

294
00:14:49,353 --> 00:14:52,433
to begin to more readily identify

295
00:14:52,433 --> 00:14:56,183
the different cortical
distributions of these patients,

296
00:14:56,183 --> 00:14:58,243
of these receptors, I should say.

297
00:14:58,243 --> 00:15:02,763
So, this is FreeSurfer
meets PET or PETSurfer.

298
00:15:02,763 --> 00:15:04,763
And it's certainly worth highlighting

299
00:15:04,763 --> 00:15:07,923
that the NIH's own colleague Robert Innis

300
00:15:07,923 --> 00:15:11,453
is leading a large scale
effort from multiple groups

301
00:15:11,453 --> 00:15:15,973
to develop a new database of PET data

302
00:15:15,973 --> 00:15:18,993
that directly incorporates the MR data.

303
00:15:18,993 --> 00:15:20,023
Notice that Doug Greve

304
00:15:20,023 --> 00:15:23,283
is an important part of this consortium.

305
00:15:23,283 --> 00:15:25,953
So that the anatomical framework

306
00:15:25,953 --> 00:15:27,933
that the PET data is analyzed from

307
00:15:27,933 --> 00:15:32,063
can be visualized in all of the
different anatomical formats

308
00:15:32,063 --> 00:15:35,926
that we can use routinely with MR.

309
00:15:37,573 --> 00:15:41,083
Of course, there are other
studies that people are doing

310
00:15:41,083 --> 00:15:43,303
that are thinking about how we can begin

311
00:15:43,303 --> 00:15:45,713
to compare the individual measurements

312
00:15:45,713 --> 00:15:48,223
made from the different modalities.

313
00:15:48,223 --> 00:15:51,683
And this very nice study
of neuroinflammation

314
00:15:51,683 --> 00:15:53,043
using the NIH's PBR28

315
00:15:55,173 --> 00:15:57,523
and studying a patient group in this case,

316
00:15:57,523 --> 00:16:00,183
ALS and Nicole.

317
00:16:00,183 --> 00:16:03,153
Zurcher and her colleagues were
looking at the relationship

318
00:16:03,153 --> 00:16:07,483
between TSPO uptake and expression

319
00:16:07,483 --> 00:16:11,693
and the changes in the
associated white matter bundles

320
00:16:11,693 --> 00:16:16,203
within the ALS patients,

321
00:16:16,203 --> 00:16:17,953
within their cortico spinal tracts.

322
00:16:19,713 --> 00:16:23,043
And another nice study from Nicole here,

323
00:16:23,043 --> 00:16:27,463
she was looking at C-11 Martinostat

324
00:16:27,463 --> 00:16:30,103
and an HDAC targeted agent

325
00:16:30,103 --> 00:16:33,283
looking at correlates with age

326
00:16:33,283 --> 00:16:36,033
and the relationship between changes

327
00:16:36,033 --> 00:16:38,033
in HDAC uptake

328
00:16:38,033 --> 00:16:40,413
and changes in white
matter microstructure.

329
00:16:40,413 --> 00:16:42,283
Again, independent measurements

330
00:16:42,283 --> 00:16:45,213
but where we can begin to
think about the relationship

331
00:16:45,213 --> 00:16:47,116
between these two quantities.

332
00:16:48,573 --> 00:16:50,283
My colleague Greg Zaharchuk,

333
00:16:50,283 --> 00:16:52,443
when I asked him to share his thoughts

334
00:16:52,443 --> 00:16:56,423
on the combined modality,
mentioned to me his own interest

335
00:16:56,423 --> 00:16:59,813
and his thought that PET and
MR really the ideal modality

336
00:16:59,813 --> 00:17:02,483
for deep learning applications,

337
00:17:02,483 --> 00:17:06,723
where we can combine PET
and MR physiological imaging

338
00:17:06,723 --> 00:17:09,046
to co-inform each other.

339
00:17:10,013 --> 00:17:11,593
So for example,

340
00:17:11,593 --> 00:17:14,933
he's been using AI-based
image transformations,

341
00:17:14,933 --> 00:17:19,463
taking multiple different input data,

342
00:17:19,463 --> 00:17:23,663
both the PET and multimodal
MR data, and contrasts

343
00:17:23,663 --> 00:17:26,893
to try to infer novel features

344
00:17:26,893 --> 00:17:28,646
on the output side.

345
00:17:30,253 --> 00:17:34,863
So his post-doctoral fellow, Kevin Chen,

346
00:17:34,863 --> 00:17:37,813
a wonderful young scientist,

347
00:17:37,813 --> 00:17:40,733
has been developing AI algorithms

348
00:17:40,733 --> 00:17:43,903
focused, for example,
on radiation reduction,

349
00:17:43,903 --> 00:17:46,113
looking at simultaneous amyloid imaging

350
00:17:46,113 --> 00:17:47,430
with MR/PET.

351
00:17:47,430 --> 00:17:52,430
And hereby, including the
MR data in the analysis,

352
00:17:53,163 --> 00:17:57,363
he's was able to show
significant reductions

353
00:17:57,363 --> 00:18:02,363
in dose required when MR data

354
00:18:02,453 --> 00:18:06,966
is included in priors for the network.

355
00:18:08,556 --> 00:18:11,493
And here's just another
example of a more recent paper

356
00:18:11,493 --> 00:18:16,253
from Kevin showing the utility

357
00:18:16,253 --> 00:18:20,983
and potential for the
CNN enhanced PET images

358
00:18:20,983 --> 00:18:24,853
to significantly reduce radiation dose.

359
00:18:24,853 --> 00:18:29,233
And as noted here, this
may enable the opportunity

360
00:18:29,233 --> 00:18:32,226
to do low and high dose
dual tracer protocols.

361
00:18:34,663 --> 00:18:38,173
Taken to its extreme, Ouyang et al,

362
00:18:38,173 --> 00:18:41,413
again working on Gregs Zaharchuk lab,

363
00:18:41,413 --> 00:18:44,643
has actually begun using
the MR data essentially

364
00:18:44,643 --> 00:18:46,673
as a prediction for the PET data,

365
00:18:46,673 --> 00:18:50,353
so-called Zero-dose FDG Imaging.

366
00:18:50,353 --> 00:18:55,046
Maybe this is an example
of a Class 0 experiment.

367
00:18:56,073 --> 00:18:59,496
We'll have to see how this
particular field evolves.

368
00:19:01,183 --> 00:19:02,863
But of course, this is not the only way

369
00:19:02,863 --> 00:19:06,076
that we can think about
using MR and PET together.

370
00:19:07,073 --> 00:19:09,913
In what I'll call Class 2 experiments,

371
00:19:09,913 --> 00:19:12,713
here we'd like to use the ability

372
00:19:12,713 --> 00:19:16,743
to acquire these two
data sets simultaneously

373
00:19:16,743 --> 00:19:19,703
to make measures with each modality.

374
00:19:19,703 --> 00:19:20,783
And in this case,

375
00:19:20,783 --> 00:19:23,183
the real goal is to
understand the relationship

376
00:19:23,183 --> 00:19:25,873
between the measures we
make in each of the modality

377
00:19:25,873 --> 00:19:28,653
in the face of dynamic state changes.

378
00:19:28,653 --> 00:19:30,583
This then are at least strongly benefits

379
00:19:30,583 --> 00:19:32,773
from simultaneous acquisitions,

380
00:19:32,773 --> 00:19:35,213
because we can be assured that the brain

381
00:19:35,213 --> 00:19:40,153
is in the same state during
each of the MR and PET measures.

382
00:19:40,153 --> 00:19:42,023
And a great example of this, for example,

383
00:19:42,023 --> 00:19:45,613
is fPET studies combined with fMRI studies

384
00:19:45,613 --> 00:19:48,543
in a variety of different
dynamic brain states.

385
00:19:48,543 --> 00:19:52,023
I think the fPET is a really nice example

386
00:19:52,023 --> 00:19:56,173
of how we can begin to
take some of the concepts

387
00:19:56,173 --> 00:20:00,703
that we developed in doing
fMRI and apply them to PET

388
00:20:00,703 --> 00:20:02,023
to look at dynamic imaging.

389
00:20:02,023 --> 00:20:06,703
In this case, the dynamics
is a simple behavioral state

390
00:20:06,703 --> 00:20:10,903
of the brain responding
to a visual checkerboard.

391
00:20:10,903 --> 00:20:13,663
Marjorie Villien et al
published several years ago

392
00:20:13,663 --> 00:20:16,216
that using a constant infusion protocol,

393
00:20:18,153 --> 00:20:22,093
it is possible to be
able to look for changes

394
00:20:22,093 --> 00:20:23,220
in CMRglucose

395
00:20:24,413 --> 00:20:27,823
in dynamics of just a few minutes.

396
00:20:27,823 --> 00:20:31,963
And thus, use the same kind
of block paradigm acquisitions

397
00:20:31,963 --> 00:20:35,053
that we had routinely
used for fMRI acquisitions

398
00:20:35,053 --> 00:20:37,275
for the previous 20 years.

399
00:20:37,275 --> 00:20:40,123
And perhaps not surprisingly

400
00:20:40,123 --> 00:20:43,353
in such a simple stimulus,

401
00:20:43,353 --> 00:20:46,800
we can now directly
relate the FDG changes,

402
00:20:46,800 --> 00:20:49,283
the changes in CMRglucose associated

403
00:20:49,283 --> 00:20:51,203
with the flashing checkerboards

404
00:20:51,203 --> 00:20:54,543
with perfusion changes measured
with arterial spin labeling,

405
00:20:54,543 --> 00:20:58,696
that's the ASL here, as
well as with BOLD contrast.

406
00:21:00,463 --> 00:21:04,693
A number of investigators
have since replicated

407
00:21:04,693 --> 00:21:08,243
and extended these
interesting initial findings.

408
00:21:08,243 --> 00:21:12,783
Here's two papers from
Hahn and from Rischka,

409
00:21:12,783 --> 00:21:15,243
which are looking at
primary sensory tasks.

410
00:21:15,243 --> 00:21:17,993
And again, showing the ability

411
00:21:17,993 --> 00:21:21,003
to measure CMRglucose changes

412
00:21:21,003 --> 00:21:25,203
and functional changes with fMRI,

413
00:21:25,203 --> 00:21:27,533
in this case BOLD contrast,

414
00:21:27,533 --> 00:21:32,286
in the same acquisition during
the same behavioral paradigm.

415
00:21:33,953 --> 00:21:36,073
Now, of course,

416
00:21:36,073 --> 00:21:37,823
just seeing the same results

417
00:21:37,823 --> 00:21:40,513
perhaps is it's good to see,

418
00:21:40,513 --> 00:21:43,103
but not the most interesting finding.

419
00:21:43,103 --> 00:21:44,913
What I think is more interesting

420
00:21:44,913 --> 00:21:47,843
is that people are now
beginning to look at these tools

421
00:21:47,843 --> 00:21:50,953
and the relationship between
the changes in metabolism

422
00:21:50,953 --> 00:21:54,013
and the changes in hemodynamics.

423
00:21:54,013 --> 00:21:56,563
I begin to try to understand

424
00:21:56,563 --> 00:21:58,563
when we not only see similarities,

425
00:21:58,563 --> 00:22:00,513
but when we see differences.

426
00:22:00,513 --> 00:22:02,913
For example, in the
setting of neuromodulation,

427
00:22:02,913 --> 00:22:04,090
either by tDCS

428
00:22:06,983 --> 00:22:08,923
or with optogenetics,

429
00:22:08,923 --> 00:22:11,703
we're beginning to get an
understanding of the fact

430
00:22:11,703 --> 00:22:14,073
that there are sometimes
spatial mismatches

431
00:22:14,073 --> 00:22:16,203
between the fPET signals

432
00:22:16,203 --> 00:22:20,023
and the fMRI activations
and deactivations.

433
00:22:20,023 --> 00:22:23,083
And that's highlighted in
these two very nice papers

434
00:22:24,353 --> 00:22:27,096
or abstracts highlighted here.

435
00:22:29,173 --> 00:22:32,183
Another example, and I apologize

436
00:22:32,183 --> 00:22:37,183
that I'm now covering
up the covering up the,

437
00:22:37,183 --> 00:22:39,736
covering up the authors of this paper.

438
00:22:40,983 --> 00:22:41,816
Let's see.

439
00:22:45,733 --> 00:22:48,133
Here we go. This is how
it was supposed to look.

440
00:22:49,523 --> 00:22:53,493
Here in this paper, by the same group,

441
00:22:53,493 --> 00:22:56,273
they've now moved from
simple sensory tasks

442
00:22:56,273 --> 00:22:59,593
up to more sophisticated studies

443
00:22:59,593 --> 00:23:02,063
looking at brain network function

444
00:23:02,063 --> 00:23:07,003
and the metabolic cost
associated with task performance.

445
00:23:07,003 --> 00:23:08,653
And their conclusion

446
00:23:10,123 --> 00:23:12,743
was that in addition
to the regional changes

447
00:23:12,743 --> 00:23:14,773
that they had previously documented,

448
00:23:14,773 --> 00:23:19,063
that in this cognitively
challenging task in this paper,

449
00:23:19,063 --> 00:23:23,603
they showed the association
between glucose metabolism

450
00:23:23,603 --> 00:23:26,023
and functional network dynamics.

451
00:23:26,023 --> 00:23:28,863
And interestingly, that these changes

452
00:23:28,863 --> 00:23:31,583
were specific to feed forward connections,

453
00:23:31,583 --> 00:23:35,473
linking visual and
dorsal attention systems.

454
00:23:35,473 --> 00:23:38,323
And the conclusion was that these roles

455
00:23:38,323 --> 00:23:40,843
highlight the key role
of metabolic factors

456
00:23:40,843 --> 00:23:43,713
in supporting brain
network reconfigurations

457
00:23:43,713 --> 00:23:45,686
as a function of cognitive engagement.

458
00:23:47,325 --> 00:23:49,976
Very interesting set of studies.

459
00:23:51,033 --> 00:23:54,513
Similarly, Jingyuan Chen

460
00:23:54,513 --> 00:23:56,663
within our own Martino Center

461
00:23:56,663 --> 00:23:59,993
has been studying sleep states again

462
00:23:59,993 --> 00:24:02,563
and arousal-induced dynamics

463
00:24:02,563 --> 00:24:04,773
where we can measure electrophysiology,

464
00:24:04,773 --> 00:24:06,673
hemodynamics, and metabolic signals

465
00:24:06,673 --> 00:24:10,366
all within a single experimental paradigm.

466
00:24:11,603 --> 00:24:14,143
And again, another example

467
00:24:14,143 --> 00:24:16,293
of what I would call
this Class 2 experiment,

468
00:24:16,293 --> 00:24:20,323
where you really want to
be imaging simultaneously

469
00:24:20,323 --> 00:24:24,253
because of the idiosyncratic nature

470
00:24:24,253 --> 00:24:26,856
of the dynamics of the brain state.

471
00:24:28,923 --> 00:24:30,533
She was able to show that

472
00:24:30,533 --> 00:24:32,183
while there's a, of course, a general,

473
00:24:32,183 --> 00:24:35,843
a positive correlation between
the changes in hemodynamics

474
00:24:35,843 --> 00:24:38,453
measured in BOLD and the
changes in metabolism,

475
00:24:38,453 --> 00:24:40,923
there were, this was not uniformly true

476
00:24:40,923 --> 00:24:42,553
across all cortical patches.

477
00:24:42,553 --> 00:24:47,553
And now, we're left to try
to understand those areas

478
00:24:47,553 --> 00:24:50,753
where there're positively
correlated in other areas,

479
00:24:50,753 --> 00:24:54,243
where in fact, we actually
see negative correlations

480
00:24:54,243 --> 00:24:55,926
during sleep states.

481
00:24:58,003 --> 00:25:01,273
Finally, another study.

482
00:25:01,273 --> 00:25:04,893
This was shared with me
from the Copenhagen group,

483
00:25:04,893 --> 00:25:07,183
Joergenson and Gitte Knudsen.

484
00:25:07,183 --> 00:25:11,876
Here, they're looking
at a DBS and a model,

485
00:25:13,153 --> 00:25:17,723
and they observed BOLD
responses in the limbic system

486
00:25:17,723 --> 00:25:20,856
related to changes in C-11,

487
00:25:22,423 --> 00:25:24,986
5-HT1B tracer.

488
00:25:25,943 --> 00:25:30,133
They measured interstitial 5-HT levels.

489
00:25:30,133 --> 00:25:34,593
And again, using the DBS,
they were looking at,

490
00:25:34,593 --> 00:25:39,593
able to look at network
dynamics in these animals,

491
00:25:39,823 --> 00:25:41,353
with the goal, ultimately,

492
00:25:41,353 --> 00:25:44,753
towards understanding
psychiatric disorders,

493
00:25:44,753 --> 00:25:47,426
which may involve
dysfunctions of these systems.

494
00:25:48,713 --> 00:25:52,293
DBS is actually a wonderful testbed

495
00:25:52,293 --> 00:25:56,476
for activation-based and
displacement studies in general.

496
00:25:57,503 --> 00:25:59,113
It can be repeatable.

497
00:25:59,113 --> 00:26:00,843
It can be titratable.

498
00:26:00,843 --> 00:26:02,503
It can be lateralized.

499
00:26:02,503 --> 00:26:07,063
So, for example, in this study
performed by Joe Mandeville,

500
00:26:07,063 --> 00:26:10,783
Wim Vanduffel, Christin
Sanders, and others.

501
00:26:10,783 --> 00:26:15,023
There was a unilateral
stimulation of the VTA,

502
00:26:15,023 --> 00:26:17,383
which gave us the opportunity to look

503
00:26:17,383 --> 00:26:21,226
at the evoked dopamine release
in the nucleus accumbens.

504
00:26:22,723 --> 00:26:24,213
Now, the challenge in this of course,

505
00:26:24,213 --> 00:26:27,693
is that the changes that we
saw are really quite subtle.

506
00:26:27,693 --> 00:26:30,863
And I think highlight the
challenges we're going to have

507
00:26:30,863 --> 00:26:33,523
in doing dynamic displacement studies.

508
00:26:33,523 --> 00:26:36,773
Kind of analogous to fMRI
studies, but doing so

509
00:26:36,773 --> 00:26:41,032
looking at neurotransmitter
specific modulations.

510
00:26:41,032 --> 00:26:44,200
And this study, for example,
the conventional SRTM model

511
00:26:45,503 --> 00:26:48,656
showed a significant
biases in its patterns.

512
00:26:49,673 --> 00:26:52,053
However, Joe has developed

513
00:26:52,053 --> 00:26:54,983
a new forward model fRTM,

514
00:26:54,983 --> 00:26:57,923
which produces results

515
00:26:57,923 --> 00:27:01,353
that can minimize those biases

516
00:27:01,353 --> 00:27:05,393
and interestingly seem
to now show responses

517
00:27:05,393 --> 00:27:08,123
within the nucleus accumbens quite similar

518
00:27:08,123 --> 00:27:09,306
to that seen in fMRI.

519
00:27:12,153 --> 00:27:16,713
Now, perhaps the ultimate use

520
00:27:16,713 --> 00:27:18,803
of the combined modalities

521
00:27:18,803 --> 00:27:21,683
is to make simultaneous measurements

522
00:27:21,683 --> 00:27:24,023
or to use simultaneous data, I should say,

523
00:27:24,023 --> 00:27:26,263
to make completely novel measurements.

524
00:27:26,263 --> 00:27:28,086
And here the key is to use,

525
00:27:29,913 --> 00:27:32,223
to make new kinds of measurements.

526
00:27:32,223 --> 00:27:33,633
Measurements that can't be made

527
00:27:33,633 --> 00:27:35,803
with either modality by itself

528
00:27:35,803 --> 00:27:39,093
and which require input
from both modalities.

529
00:27:39,093 --> 00:27:43,363
And I think studies, for
example, of receptor trafficking,

530
00:27:43,363 --> 00:27:45,343
receptor network interactions, et cetera,

531
00:27:45,343 --> 00:27:47,283
are good examples of that.

532
00:27:47,283 --> 00:27:49,883
One of the nicest examples, I think,

533
00:27:49,883 --> 00:27:53,343
comes from Christin Sanders and her work

534
00:27:53,343 --> 00:27:57,713
on the dopamine system and her modeling,

535
00:27:57,713 --> 00:28:01,953
which now has allowed her
using both the data streams

536
00:28:01,953 --> 00:28:05,213
from the PET and the MR to measure things

537
00:28:05,213 --> 00:28:08,953
like baseline occupancy
or internalization rates.

538
00:28:08,953 --> 00:28:11,393
These are measurements
that couldn't be made

539
00:28:11,393 --> 00:28:13,803
with either modality on its own.

540
00:28:13,803 --> 00:28:16,983
But when the modalities are combined

541
00:28:16,983 --> 00:28:21,033
and the pharmacologic
response can be combined

542
00:28:21,033 --> 00:28:22,583
with the functional response,

543
00:28:22,583 --> 00:28:26,343
we now have the ability to
make these unique measurements.

544
00:28:26,343 --> 00:28:27,853
And I won't say more about this,

545
00:28:27,853 --> 00:28:30,103
but you should definitely
listen to her talk.

546
00:28:31,123 --> 00:28:33,213
She, of course, though is not the only one

547
00:28:33,213 --> 00:28:37,366
using this tool in this way.

548
00:28:38,492 --> 00:28:41,063
Hsiao-Ying Wey or Monica
Wey, you may know her as,

549
00:28:41,063 --> 00:28:43,846
has been studying the mu opioid system.

550
00:28:45,233 --> 00:28:47,583
Now this is, of course, a complex receptor

551
00:28:47,583 --> 00:28:52,583
that has both G protein, coupling,

552
00:28:52,633 --> 00:28:54,753
and a beta-arrestin pathways.

553
00:28:54,753 --> 00:28:57,333
And so in such systems,

554
00:28:57,333 --> 00:28:59,653
understanding receptor trafficking

555
00:28:59,653 --> 00:29:02,273
requires both understanding
desensitization

556
00:29:02,273 --> 00:29:04,573
as well as internalization.

557
00:29:04,573 --> 00:29:06,323
And so, her work is really targeted

558
00:29:06,323 --> 00:29:09,983
to try to teasing those apart
and teasing out the influences

559
00:29:09,983 --> 00:29:13,643
of the G protein and
beta-arrestin pathways.

560
00:29:13,643 --> 00:29:16,223
In this case, she's using TRV130,

561
00:29:16,223 --> 00:29:21,223
which is a G protein, a biased
mu opioid receptor agonist,

562
00:29:21,273 --> 00:29:25,883
which does not induce the same
kind of receptor trafficking.

563
00:29:25,883 --> 00:29:29,693
And has now begun to perform experiments,

564
00:29:29,693 --> 00:29:34,393
comparing morphine, which
involves both the G coupled

565
00:29:34,393 --> 00:29:38,593
and beta-arrestin pathways
and then TRV130 which doesn't,

566
00:29:38,593 --> 00:29:42,953
showing in fact that there
significant discrepancies

567
00:29:42,953 --> 00:29:45,953
between the PET responses,

568
00:29:45,953 --> 00:29:50,413
even though there are
similar functional responses.

569
00:29:50,413 --> 00:29:54,203
Here, again, using the
combined fMRI and PET data

570
00:29:54,203 --> 00:29:56,613
to try to give us greater insights

571
00:29:56,613 --> 00:29:58,853
in terms of the underlying
molecular processes

572
00:29:58,853 --> 00:30:00,876
in these complex receptor systems.

573
00:30:03,353 --> 00:30:06,543
Hanne Hansen from the Copenhagen group

574
00:30:06,543 --> 00:30:09,913
has been taking a similar approach studies

575
00:30:09,913 --> 00:30:13,623
in non-human primates looking
at serotonergic system

576
00:30:15,473 --> 00:30:19,323
and has drawn a number of
interesting conclusions

577
00:30:19,323 --> 00:30:21,833
based on non-penetrable agonists

578
00:30:21,833 --> 00:30:25,643
and their comparison against
agonists and antagonists.

579
00:30:25,643 --> 00:30:28,833
Again, using the dynamics
of MR to try to tease out

580
00:30:28,833 --> 00:30:31,253
some of the underlying
functional properties

581
00:30:31,253 --> 00:30:32,626
of the receptor system.

582
00:30:33,823 --> 00:30:38,173
And here for those of you interested,

583
00:30:38,173 --> 00:30:40,743
there was a very nice review paper

584
00:30:40,743 --> 00:30:43,153
written by these three
wonderful colleague,

585
00:30:43,153 --> 00:30:47,223
Christin Sanders, Hanne
Hansen, Hsiao-Ying Wey,

586
00:30:47,223 --> 00:30:49,763
talking about the use of MR/PET

587
00:30:49,763 --> 00:30:54,273
in looking at dopamine, opioid,
and serotonergic system.

588
00:30:54,273 --> 00:30:56,673
And I definitely would encourage you

589
00:30:56,673 --> 00:30:58,576
to read that review paper.

590
00:30:59,893 --> 00:31:01,343
So, what's next?

591
00:31:01,343 --> 00:31:04,253
And what might a Class 4 experiment be?

592
00:31:04,253 --> 00:31:06,783
So, what does the future
hold for the technology?

593
00:31:06,783 --> 00:31:08,876
And then, how might we be able to use it?

594
00:31:10,123 --> 00:31:12,573
Well, certainly the technology of MR/PET

595
00:31:13,803 --> 00:31:15,463
will continue to expand

596
00:31:15,463 --> 00:31:17,713
and that will give us
some new opportunities.

597
00:31:18,793 --> 00:31:23,773
One example that I'm, I
think, may be quite important

598
00:31:23,773 --> 00:31:26,183
is the work that Ciprian Catana is doing

599
00:31:26,183 --> 00:31:28,383
based on a brain initiative grant

600
00:31:28,383 --> 00:31:32,603
to develop a 70 compatible device,

601
00:31:32,603 --> 00:31:35,553
with at least an order
of magnitude improvement

602
00:31:35,553 --> 00:31:36,813
in sensitivity.

603
00:31:36,813 --> 00:31:40,613
As you can see, it uses this
novel spherical PET geometry

604
00:31:40,613 --> 00:31:44,233
with a greater than 70%
solid angle coverage.

605
00:31:44,233 --> 00:31:47,823
So, the equivalent of systems

606
00:31:47,823 --> 00:31:52,243
like these very long bore scanners

607
00:31:52,243 --> 00:31:55,253
that United and now as
Siemens are developing,

608
00:31:55,253 --> 00:31:57,693
but in a, excuse me, in a compact design.

609
00:31:57,693 --> 00:32:00,633
Design to image the human brain

610
00:32:00,633 --> 00:32:02,673
and developed in such a way

611
00:32:02,673 --> 00:32:04,063
as to be fully compatible

612
00:32:04,063 --> 00:32:08,053
with ultra high field MR Scanner.

613
00:32:08,053 --> 00:32:11,653
Of course, showing that the technology

614
00:32:11,653 --> 00:32:15,033
can be made compatible with MR

615
00:32:15,033 --> 00:32:16,803
even at very high magnetic fields,

616
00:32:16,803 --> 00:32:19,943
where very subtle changes diamagnetic

617
00:32:21,043 --> 00:32:22,903
and subtle power magnetic effects

618
00:32:22,903 --> 00:32:27,603
can significantly distort
the MR field of you

619
00:32:27,603 --> 00:32:28,853
was very important.

620
00:32:28,853 --> 00:32:32,773
Fortunately, work from
colleagues Bastien Guerin

621
00:32:32,773 --> 00:32:36,123
and Larry Wald has shown
that it's certainly possible

622
00:32:36,123 --> 00:32:41,123
to build such systems that
do not distort the field

623
00:32:41,383 --> 00:32:44,063
and don't distort the PET detectors.

624
00:32:44,063 --> 00:32:45,673
And the bottom line

625
00:32:45,673 --> 00:32:49,143
is that when we compare
what this scanner is,

626
00:32:49,143 --> 00:32:51,823
we hope likely to deliver to us relative

627
00:32:51,823 --> 00:32:56,313
to the current class of
human PET/MR scanners.

628
00:32:56,313 --> 00:32:57,693
Notice that the sensitivity

629
00:32:57,693 --> 00:33:01,033
is almost in order of magnitude more

630
00:33:01,033 --> 00:33:03,963
than what is currently measured

631
00:33:03,963 --> 00:33:08,963
from the existing human MR/PET scanners.

632
00:33:10,303 --> 00:33:12,643
Of course now, if we combine this say

633
00:33:12,643 --> 00:33:16,323
with the AI-based
reconstructions and denoising

634
00:33:16,323 --> 00:33:19,363
a la Zaharchuk's work,

635
00:33:19,363 --> 00:33:22,363
maybe there's yet another
order of magnitude

636
00:33:22,363 --> 00:33:23,943
improvement to be gained.

637
00:33:23,943 --> 00:33:26,553
So, that would be almost
a hundred fold improvement

638
00:33:26,553 --> 00:33:28,363
over today's machines.

639
00:33:28,363 --> 00:33:32,376
And that's not a
completely fanciful target.

640
00:33:33,563 --> 00:33:36,343
And so, the question then
is, well, what would we do

641
00:33:36,343 --> 00:33:39,066
if we had a hundred fold
sensitive improvement?

642
00:33:41,193 --> 00:33:42,943
Another point, of course,

643
00:33:42,943 --> 00:33:46,033
is that MR/PET technology
continues to move

644
00:33:46,033 --> 00:33:49,053
to higher and higher spatial resolution.

645
00:33:49,053 --> 00:33:52,873
This is another brain initiative funded

646
00:33:54,233 --> 00:33:58,403
technical improvement grant.

647
00:33:58,403 --> 00:34:01,133
In this case, led by Georgia Pakref

648
00:34:01,133 --> 00:34:04,583
from the Gordon Center at
the Mass General Hospital.

649
00:34:04,583 --> 00:34:08,033
Here pushing a spatial resolution

650
00:34:08,033 --> 00:34:13,033
to a millimeter or sub-millimeter
for the human brain.

651
00:34:13,313 --> 00:34:15,133
Again, this particular design

652
00:34:15,133 --> 00:34:17,003
not designed to work in an MR scanner,

653
00:34:17,003 --> 00:34:19,343
but it certainly points to the possibility

654
00:34:19,343 --> 00:34:22,453
to develop very high spatial resolution,

655
00:34:22,453 --> 00:34:25,273
in addition to the very
high sensitivity scanners

656
00:34:25,273 --> 00:34:28,043
that Ciprian Catana was working on.

657
00:34:28,043 --> 00:34:32,513
And of course, MR technology
is also advancing in ways.

658
00:34:32,513 --> 00:34:35,823
For example, this again,

659
00:34:35,823 --> 00:34:38,823
brain initiative supported project

660
00:34:38,823 --> 00:34:43,103
on an MR corticography
is how David Feinberg,

661
00:34:43,103 --> 00:34:45,083
the PI of this grant, has labeled it.

662
00:34:45,083 --> 00:34:49,653
Very high resolution imaging
of the brain's cortex,

663
00:34:49,653 --> 00:34:51,413
where it should be possible to get

664
00:34:51,413 --> 00:34:55,153
less than half millimeter
isotropic voxels.

665
00:34:55,153 --> 00:34:57,713
Certainly, well within the
size that should allow us

666
00:34:57,713 --> 00:34:59,283
to be able to measure function

667
00:34:59,283 --> 00:35:02,996
at the level of cortico
laminar and cortical columns.

668
00:35:04,643 --> 00:35:07,313
Another brain initiative
funded technical advance

669
00:35:07,313 --> 00:35:08,543
on the MR side.

670
00:35:08,543 --> 00:35:12,803
This is the "Connectome
2.0," led by Suzy Huang

671
00:35:12,803 --> 00:35:14,683
and her colleagues.

672
00:35:14,683 --> 00:35:19,133
Here, we're pushing that
initial generation of machines

673
00:35:19,133 --> 00:35:23,326
designed to measure
microstructure and connectivity

674
00:35:24,593 --> 00:35:29,033
using diffusion techniques
to more than quadruple

675
00:35:29,033 --> 00:35:32,803
the fundamental capabilities
of Gmax and slew rates.

676
00:35:32,803 --> 00:35:34,263
And then,

677
00:35:34,263 --> 00:35:37,733
looking at how we can validate
microstructure measures

678
00:35:37,733 --> 00:35:40,163
using these advanced tools.

679
00:35:40,163 --> 00:35:42,246
Again, both of these tools,

680
00:35:44,003 --> 00:35:45,603
the last two I talked about,

681
00:35:45,603 --> 00:35:48,403
require high performance gradient systems

682
00:35:48,403 --> 00:35:51,133
that frankly would be
challenging to put an MR in,

683
00:35:51,133 --> 00:35:52,873
but never say never.

684
00:35:52,873 --> 00:35:53,910
I think it's certainly...

685
00:35:53,910 --> 00:35:56,093
Excuse me, to put a PET in.

686
00:35:56,093 --> 00:35:58,823
But I think it's certainly
possible and in fact likely

687
00:35:58,823 --> 00:36:01,329
that we're gonna see more and more devices

688
00:36:01,329 --> 00:36:04,173
that take advantage of
the technical improvements

689
00:36:04,173 --> 00:36:05,803
in both domains.

690
00:36:05,803 --> 00:36:08,601
So, what are we gonna do
with these kind of machines

691
00:36:08,601 --> 00:36:10,883
as they come around?

692
00:36:10,883 --> 00:36:12,203
Just some things to think about.

693
00:36:12,203 --> 00:36:15,893
Certainly, we're in a
position I think now,

694
00:36:15,893 --> 00:36:17,103
but certainly in the future

695
00:36:17,103 --> 00:36:18,883
with a great improvement in sensitivity,

696
00:36:18,883 --> 00:36:23,043
to think about and study
the role of nerve modulators

697
00:36:23,043 --> 00:36:26,943
on hemodynamics, on network function,

698
00:36:26,943 --> 00:36:28,563
and on behavior.

699
00:36:28,563 --> 00:36:32,313
And to link these regional
neuromodulatory dynamics

700
00:36:32,313 --> 00:36:35,103
to the cytoarchitecture within the brain.

701
00:36:35,103 --> 00:36:38,503
The work in the brain
initiative on the cell census

702
00:36:38,503 --> 00:36:40,983
certainly is going to give
us exquisite information

703
00:36:40,983 --> 00:36:45,753
about the nature of brain
cells throughout the brain.

704
00:36:45,753 --> 00:36:49,973
Now, let's try to relate that to receptor

705
00:36:49,973 --> 00:36:52,426
and neuronal function.

706
00:36:53,493 --> 00:36:56,693
We can certainly think about
mapping neuroreceptor dynamics

707
00:36:56,693 --> 00:36:59,733
with network function in
diverse cognitive paradigms.

708
00:36:59,733 --> 00:37:02,983
I'm really hopeful that with
increase in sensitivity,

709
00:37:02,983 --> 00:37:06,253
we can begin to do more
and more experiments

710
00:37:06,253 --> 00:37:10,693
like the fPET, but now with neuroreceptor

711
00:37:10,693 --> 00:37:13,193
and neuromodulatory specificity.

712
00:37:13,193 --> 00:37:16,783
And with cognitive paradigms
that are much more analogous

713
00:37:16,783 --> 00:37:19,646
to those that we routinely use with fMRI.

714
00:37:21,283 --> 00:37:23,733
In order to do this, it's
possible that we may need

715
00:37:23,733 --> 00:37:25,983
to rethink the tracers that we use

716
00:37:25,983 --> 00:37:28,935
to take better advantage of
this increased sensitivity,

717
00:37:28,935 --> 00:37:31,933
the temporal resolution of our cameras

718
00:37:31,933 --> 00:37:33,083
and of course, the ability

719
00:37:33,083 --> 00:37:36,743
to measure physiology simultaneously.

720
00:37:36,743 --> 00:37:39,463
So for example, tracers,

721
00:37:39,463 --> 00:37:43,963
which may have a significant
flow-based modulation,

722
00:37:43,963 --> 00:37:47,373
which have been discounted
in the past because of that,

723
00:37:47,373 --> 00:37:49,923
may not be discountable in the future.

724
00:37:49,923 --> 00:37:53,453
Because, of course, we can
simultaneously quantify flow

725
00:37:53,453 --> 00:37:56,136
using MR and build that into our modeling.

726
00:37:58,013 --> 00:38:00,083
As Greg Zaharchuk mentioned

727
00:38:00,083 --> 00:38:03,553
in the slides that he
graciously shared with me,

728
00:38:03,553 --> 00:38:06,913
we can do multiple tracer injections

729
00:38:06,913 --> 00:38:09,043
with function and microstructure

730
00:38:09,043 --> 00:38:11,673
to better understand neurotransmitter

731
00:38:11,673 --> 00:38:14,196
and neuromodulator interactions.

732
00:38:15,073 --> 00:38:16,783
And if we really can gain

733
00:38:16,783 --> 00:38:19,163
a factor of a hundred in sensitivity,

734
00:38:19,163 --> 00:38:22,203
perhaps two is not the
limit to our ability

735
00:38:22,203 --> 00:38:24,776
to look for these kind of interactions.

736
00:38:25,643 --> 00:38:28,833
This will allow us to begin to
understand paradoxes in drugs

737
00:38:28,833 --> 00:38:31,503
that seem to be of the same class,

738
00:38:31,503 --> 00:38:33,553
but which we know to work differently.

739
00:38:33,553 --> 00:38:34,786
Or in some cases,

740
00:38:35,993 --> 00:38:38,363
different drugs seemingly
in the same class

741
00:38:39,223 --> 00:38:40,336
don't work at all.

742
00:38:41,663 --> 00:38:44,433
Finally, we can combine our efforts

743
00:38:44,433 --> 00:38:48,183
with the brain initiative, Armamentarium,

744
00:38:48,183 --> 00:38:51,333
an effort that I think
will be extremely relevant

745
00:38:51,333 --> 00:38:53,503
to the PET community especially,

746
00:38:53,503 --> 00:38:56,996
to study the effects of
molecular targeted treatments,

747
00:38:58,146 --> 00:39:01,943
chemogenetics, for example, like DREADDS.

748
00:39:01,943 --> 00:39:06,313
And MR/PET can be combined
to manipulate receptors

749
00:39:06,313 --> 00:39:08,583
and measure it and understand the changes

750
00:39:08,583 --> 00:39:11,213
in occupancy and how that
then relates to function

751
00:39:11,213 --> 00:39:12,876
and ultimately behavior.

752
00:39:14,363 --> 00:39:19,033
And finally and in a very, to
my mind, exciting development,

753
00:39:19,033 --> 00:39:21,073
will we be able to reduce dose enough

754
00:39:21,073 --> 00:39:23,403
to allow studies of normotypic

755
00:39:23,403 --> 00:39:25,283
and abnormal human brain development?

756
00:39:25,283 --> 00:39:27,846
I.e. can we ultimately study children?

757
00:39:29,023 --> 00:39:31,553
I think all of these give us hope

758
00:39:31,553 --> 00:39:33,783
that there are going to
be things that we can do

759
00:39:33,783 --> 00:39:36,463
with the combined MR/PET devices

760
00:39:36,463 --> 00:39:39,403
that really will extend the tool.

761
00:39:39,403 --> 00:39:41,773
Not only to do the kinds of experiments

762
00:39:41,773 --> 00:39:44,483
we've been doing in the parallel play,

763
00:39:44,483 --> 00:39:48,023
but to really do completely
novel experiments.

764
00:39:48,023 --> 00:39:49,653
You're gonna hear about many of them

765
00:39:49,653 --> 00:39:52,513
in the rest of the, of this conference.

766
00:39:52,513 --> 00:39:55,583
So again, I want to thank
all of my colleagues

767
00:39:55,583 --> 00:39:58,303
who shared slides with me.

768
00:39:58,303 --> 00:40:01,763
And I thank you for listening

769
00:40:01,763 --> 00:40:05,723
and I look forward to the
opportunity to discuss this

770
00:40:05,723 --> 00:40:09,193
in the other talks as
the conference goes on.

771
00:40:09,193 --> 00:40:10,206
Thanks very much.

772
00:40:12,370 --> 00:40:13,536
- I'd like to thank the organizers

773
00:40:13,536 --> 00:40:17,686
for inviting me to
present at this symposium.

774
00:40:17,686 --> 00:40:19,106
And the topic I'd like to share is

775
00:40:19,106 --> 00:40:22,416
our work in recent years
on imaging synaptic density

776
00:40:22,416 --> 00:40:24,916
in the human brain here using PET,

777
00:40:24,916 --> 00:40:27,426
and in the spirit of the event,

778
00:40:27,426 --> 00:40:29,016
to really look about
how we were using this

779
00:40:29,016 --> 00:40:32,616
in a single modality,
multi-modality, multi-tracer,

780
00:40:32,616 --> 00:40:34,566
how we integrate all of these together.

781
00:40:36,706 --> 00:40:37,696
So here's the outline.

782
00:40:37,696 --> 00:40:39,406
We'll talk about the
development of our tracer,

783
00:40:39,406 --> 00:40:41,296
initial human PET studies,

784
00:40:41,296 --> 00:40:43,246
how we're validating the use of the tracer

785
00:40:43,246 --> 00:40:45,056
as a measure of synaptic markers,

786
00:40:45,056 --> 00:40:46,926
spend some time on our
Alzheimer's results,

787
00:40:46,926 --> 00:40:48,336
which really gives a good understanding

788
00:40:48,336 --> 00:40:51,746
of both primary results
looking at SV2A differences

789
00:40:51,746 --> 00:40:55,299
as well as multi-modality
and multi-tracer.

790
00:40:56,246 --> 00:40:58,336
In understanding it, we'd like
to understand a little bit

791
00:40:58,336 --> 00:41:00,986
about its sensitivity
to neuronal activation.

792
00:41:00,986 --> 00:41:04,166
That is, is this a static
measure of synapsis?

793
00:41:04,166 --> 00:41:07,986
We'll look into two
relevant disorders for NIMH,

794
00:41:07,986 --> 00:41:09,796
looking in the schizophrenia

795
00:41:09,796 --> 00:41:11,456
and then major depressive disorder.

796
00:41:11,456 --> 00:41:14,256
And then I'll end with some
non-analytic strategies

797
00:41:14,256 --> 00:41:16,236
how we'd like to be able
to look at network-based.

798
00:41:16,236 --> 00:41:18,426
And a lot of these derive from fMRI,

799
00:41:18,426 --> 00:41:20,226
looking at independent component analysis

800
00:41:20,226 --> 00:41:23,566
or how we can relate resting
state fMRI measures to SV2A

801
00:41:23,566 --> 00:41:28,426
and how we can apply those
into various disease states.

802
00:41:28,426 --> 00:41:29,956
Well, in studying the brain,

803
00:41:29,956 --> 00:41:31,366
it's pretty obvious
that it would be useful

804
00:41:31,366 --> 00:41:33,616
to have an in vivo marker of synapses,

805
00:41:33,616 --> 00:41:35,926
a common factor in so many disorders,

806
00:41:35,926 --> 00:41:37,766
whether we're looking
at neurodegeneration,

807
00:41:37,766 --> 00:41:41,226
the loss of synapses, in
focal areas like epilepsy,

808
00:41:41,226 --> 00:41:43,936
or in interesting areas in
the neuropsychiatric realm,

809
00:41:43,936 --> 00:41:47,506
of looking at perhaps diseases
of pruning of synapses

810
00:41:47,506 --> 00:41:49,526
in autism and schizophrenia

811
00:41:49,526 --> 00:41:52,466
or other areas like depression
where we see dynamic changes

812
00:41:52,466 --> 00:41:54,366
in the amount of spines that we see

813
00:41:54,366 --> 00:41:55,696
under chronic uncontrolled stress

814
00:41:55,696 --> 00:41:58,576
or how that can be modulated
with drugs like ketamine.

815
00:41:58,576 --> 00:41:59,513
The target we're gonna use is

816
00:41:59,513 --> 00:42:02,316
the synaptic vesicle glycoprotein 2A.

817
00:42:02,316 --> 00:42:06,556
It is a protein that sits
in the synaptic vesicles,

818
00:42:06,556 --> 00:42:08,576
just like synaptophysin and other proteins

819
00:42:08,576 --> 00:42:10,496
that have been used for a long time.

820
00:42:10,496 --> 00:42:14,906
And this molecule, the target
was found based on the use

821
00:42:14,906 --> 00:42:18,646
of study of the epileptic
anticonvulsant drug Keppra,

822
00:42:18,646 --> 00:42:19,899
which binds to SV2A.

823
00:42:21,366 --> 00:42:23,516
Now here's an example of that,

824
00:42:23,516 --> 00:42:26,106
that this has been used
as a surrogate marker

825
00:42:26,106 --> 00:42:28,096
for synaptic density for years.

826
00:42:28,096 --> 00:42:29,736
Certainly synaptophysin was doing that.

827
00:42:29,736 --> 00:42:31,276
That goes back a long time,

828
00:42:31,276 --> 00:42:33,336
both in development or
Alzheimer's disease.

829
00:42:33,336 --> 00:42:35,136
And more recently, this is being used

830
00:42:35,136 --> 00:42:37,476
in terms of looking for
example, in animal models

831
00:42:37,476 --> 00:42:40,146
where we can show reductions
in a transgenic animal,

832
00:42:40,146 --> 00:42:43,456
which actually can be
recovered after therapy.

833
00:42:43,456 --> 00:42:44,576
So here's how we started.

834
00:42:44,576 --> 00:42:45,456
This is the molecule.

835
00:42:45,456 --> 00:42:46,436
It's called UCB-J.

836
00:42:46,436 --> 00:42:48,116
It's labeled with carbon 11,

837
00:42:48,116 --> 00:42:51,066
and we developed that
first in animal models.

838
00:42:51,066 --> 00:42:52,636
And then we took that into human beings.

839
00:42:52,636 --> 00:42:54,266
So here's our initial human data.

840
00:42:54,266 --> 00:42:55,906
We did this with healthy controls,

841
00:42:55,906 --> 00:42:57,706
doing test-retest studies.

842
00:42:57,706 --> 00:43:00,126
And we're doing that
the hard way with time

843
00:43:00,126 --> 00:43:02,276
with the bolus injections of our tracer,

844
00:43:02,276 --> 00:43:04,886
measurements of the input
function in arterial blood

845
00:43:04,886 --> 00:43:07,956
and studies on our high resolution HRT.

846
00:43:07,956 --> 00:43:09,936
And we measure the activity in the blood.

847
00:43:09,936 --> 00:43:11,946
We show how that drops over time,

848
00:43:11,946 --> 00:43:14,306
that we produce radiolabeled metabolites

849
00:43:14,306 --> 00:43:16,726
that are more polar than
our parent compound.

850
00:43:16,726 --> 00:43:18,576
We also found that we
have a good free fraction,

851
00:43:18,576 --> 00:43:21,649
which means it's gonna be
readily entering the brain.

852
00:43:23,476 --> 00:43:25,016
There's an example of
what the images look like.

853
00:43:25,016 --> 00:43:25,849
They're not surprising.

854
00:43:25,849 --> 00:43:28,970
We have high activity of our tracer

855
00:43:28,970 --> 00:43:29,956
throughout the gray matter

856
00:43:29,956 --> 00:43:30,959
and lower in the white matter.

857
00:43:30,959 --> 00:43:34,316
There's a very high uptake
in the brain, partially due

858
00:43:34,316 --> 00:43:37,256
to that nice lipophilic
characteristic of the tracer,

859
00:43:37,256 --> 00:43:39,796
but also due to the high
amount of binding sites

860
00:43:39,796 --> 00:43:41,726
compared to the white matter here

861
00:43:41,726 --> 00:43:43,976
in the centrum semiovale.

862
00:43:43,976 --> 00:43:46,746
We did quantification of that
with compartment modeling

863
00:43:46,746 --> 00:43:48,406
and fortunately found
that the simplest model,

864
00:43:48,406 --> 00:43:50,846
the one tissue model gives
us very good agreement.

865
00:43:50,846 --> 00:43:52,746
And we can therefore
measure two parameters,

866
00:43:52,746 --> 00:43:53,926
the volume of distribution,

867
00:43:53,926 --> 00:43:56,916
that's the equilibrium
ratio of tissue to plasma,

868
00:43:56,916 --> 00:43:59,056
as well as the net uptake rate, K1.

869
00:43:59,056 --> 00:44:00,876
And with values of 0.3 and 0.4,

870
00:44:00,876 --> 00:44:03,676
that says this is gonna be
very sensitive to blood flow.

871
00:44:04,676 --> 00:44:06,666
We also found that our
test-retest reliability

872
00:44:06,666 --> 00:44:08,466
was very good in these initial data.

873
00:44:08,466 --> 00:44:10,726
So we have a highly reliable measure

874
00:44:10,726 --> 00:44:12,356
and therefore we can calculate this

875
00:44:12,356 --> 00:44:13,916
on a voxel-by-voxel level.

876
00:44:13,916 --> 00:44:16,146
These are parametric
images that we can see

877
00:44:16,146 --> 00:44:19,226
to be able to look at a very high quality.

878
00:44:19,226 --> 00:44:22,006
And so finally, since we
know that the binding site

879
00:44:22,006 --> 00:44:24,076
is displayable by levetiracetam,

880
00:44:24,076 --> 00:44:26,836
we could do dynamic
displacement from the scan

881
00:44:26,836 --> 00:44:29,846
showing that we are specifically
binding to this target

882
00:44:29,846 --> 00:44:31,439
on the presynaptic terminals.

883
00:44:32,396 --> 00:44:33,936
Now we went on to do some validation.

884
00:44:33,936 --> 00:44:35,206
This was done in the baboon

885
00:44:35,206 --> 00:44:38,086
where we did an in vivo
scan and the uptake of that,

886
00:44:38,086 --> 00:44:39,776
and then sacrificed the animal,

887
00:44:39,776 --> 00:44:42,606
did the Western blots
and ligand binding assays

888
00:44:42,606 --> 00:44:44,996
and a little bit of confocal microscopy.

889
00:44:44,996 --> 00:44:46,486
Here's when we look at the Westerns,

890
00:44:46,486 --> 00:44:48,746
we can see, for example,
that we have high binding

891
00:44:48,746 --> 00:44:50,426
in all cortical regions

892
00:44:50,426 --> 00:44:52,776
and very little in the centrum
semiovale, the white matter,

893
00:44:52,776 --> 00:44:54,796
which can be used as a reference region.

894
00:44:54,796 --> 00:44:56,286
We found that we also have good agreement

895
00:44:56,286 --> 00:44:59,786
between SV2A binding and
synaptophysin binding there.

896
00:44:59,786 --> 00:45:01,196
We did the binding assays there

897
00:45:01,196 --> 00:45:03,526
to be able to look at what
the affinity would be.

898
00:45:03,526 --> 00:45:06,236
We basically were able to
find that we have an affinity

899
00:45:06,236 --> 00:45:07,726
of about four nanomolar.

900
00:45:07,726 --> 00:45:09,176
For PET traces, that's not that great,

901
00:45:09,176 --> 00:45:10,406
except that we have a huge Vmax.

902
00:45:10,406 --> 00:45:13,556
The amount of this target
of SV2A is very large

903
00:45:13,556 --> 00:45:17,376
and our centrum semiovale
region was extremely low.

904
00:45:17,376 --> 00:45:19,636
And we could show that
the in vivo measures

905
00:45:19,636 --> 00:45:22,396
could be nicely correlated
with the in vitro measures,

906
00:45:22,396 --> 00:45:25,206
that basically all the measures
from the binding assays

907
00:45:25,206 --> 00:45:28,006
and the Westerns and the in
vivo were looking the same.

908
00:45:28,006 --> 00:45:29,316
So this gave us some confidence

909
00:45:29,316 --> 00:45:30,896
that we were on the right track

910
00:45:30,896 --> 00:45:32,726
of being able to measure synapses.

911
00:45:32,726 --> 00:45:34,846
We are also collecting other data sets.

912
00:45:34,846 --> 00:45:37,516
And here's an example
with resected human tissue

913
00:45:37,516 --> 00:45:40,126
from epilepsy patients following surgery.

914
00:45:40,126 --> 00:45:42,446
We were able to show a
relationship to the SV2A

915
00:45:42,446 --> 00:45:44,676
and its optical density
and the binding site.

916
00:45:44,676 --> 00:45:45,586
And we have a correlation.

917
00:45:45,586 --> 00:45:47,336
It's a sloppy one with a small N,

918
00:45:47,336 --> 00:45:49,076
but it's certainly showing
we're still on the right track,

919
00:45:49,076 --> 00:45:51,206
even in resected human tissue.

920
00:45:51,206 --> 00:45:54,046
And we are doing more work there
looking at how our measure,

921
00:45:54,046 --> 00:45:56,126
in this case, a binding potential measure,

922
00:45:56,126 --> 00:45:58,256
correlates well with SV2A.

923
00:45:58,256 --> 00:45:59,836
When we look at other markers

924
00:45:59,836 --> 00:46:02,236
for excitatory or inhibitory markers,

925
00:46:02,236 --> 00:46:04,666
basically a little, the
correlation breaks down.

926
00:46:04,666 --> 00:46:06,716
We need more data to be able to understand

927
00:46:06,716 --> 00:46:08,556
to what extent is our SV2A reflecting

928
00:46:08,556 --> 00:46:12,309
both GABAergic and glutamatergic synapses.

929
00:46:13,316 --> 00:46:16,486
So let me take it on
to our first population

930
00:46:16,486 --> 00:46:19,336
going on in the AD world.

931
00:46:19,336 --> 00:46:20,456
This is an obvious one to do,

932
00:46:20,456 --> 00:46:22,626
and we've done quite a
number of studies in there.

933
00:46:22,626 --> 00:46:25,166
So in looking at a population
of both cognitively normal

934
00:46:25,166 --> 00:46:28,726
and relatively mild AD or MCI,

935
00:46:28,726 --> 00:46:33,266
all of these subjects have
have PIP scans to validate

936
00:46:33,266 --> 00:46:34,906
that they're am appropriately amyloid,

937
00:46:34,906 --> 00:46:35,926
amyloid positive for the AD,

938
00:46:35,926 --> 00:46:38,666
amyloid negative in the controls.

939
00:46:38,666 --> 00:46:41,136
Neuropsych tests are being
measured in all of these things.

940
00:46:41,136 --> 00:46:43,336
And here our measure is
normalized to the cerebellum.

941
00:46:43,336 --> 00:46:45,806
We found that provides
a more reliable measure.

942
00:46:45,806 --> 00:46:47,226
We're also correcting for atrophy

943
00:46:47,226 --> 00:46:49,206
with partial volume corrections.

944
00:46:49,206 --> 00:46:52,096
So here's the cohort that
we published recently there,

945
00:46:52,096 --> 00:46:53,466
19 cognitive with controls,

946
00:46:53,466 --> 00:46:57,466
34, both MCI and AD broken down,

947
00:46:57,466 --> 00:47:00,816
and pretty much no differences
there in education and age,

948
00:47:00,816 --> 00:47:04,196
and of course, great differences
in the cognitive measures,

949
00:47:04,196 --> 00:47:06,236
including the clinical dementia ratios

950
00:47:06,236 --> 00:47:08,016
and the other logical memory

951
00:47:08,016 --> 00:47:10,886
and auditory, verbal learning test scores.

952
00:47:10,886 --> 00:47:13,256
This is example what the
average images look like.

953
00:47:13,256 --> 00:47:16,426
And we are seeing reductions
across the cortex.

954
00:47:16,426 --> 00:47:17,259
In our initial study,

955
00:47:17,259 --> 00:47:20,316
we had our primary results
were in the hippocampus,

956
00:47:20,316 --> 00:47:21,286
but now with a bigger N,

957
00:47:21,286 --> 00:47:23,526
we're seeing that more across the brain.

958
00:47:23,526 --> 00:47:25,283
So if we look at that
in a regional fashion,

959
00:47:25,283 --> 00:47:27,986
and we're looking at the
cognitively normals in blue

960
00:47:27,986 --> 00:47:30,876
and the AD subjects in red,

961
00:47:30,876 --> 00:47:33,286
we're seeing significant
reductions in many regions

962
00:47:33,286 --> 00:47:35,356
with larger effect sizes

963
00:47:35,356 --> 00:47:37,376
in entorhinal cortex and hippocampus,

964
00:47:37,376 --> 00:47:39,786
smaller ones in other cortical regions.

965
00:47:39,786 --> 00:47:40,896
If we correct for atrophy,

966
00:47:40,896 --> 00:47:42,506
we lose some of these in the cortex.

967
00:47:42,506 --> 00:47:44,706
We still maintain our primary effects,

968
00:47:44,706 --> 00:47:46,649
especially in the hippocampus.

969
00:47:47,506 --> 00:47:51,296
And so here's if we map
the magnitude of changes.

970
00:47:51,296 --> 00:47:52,566
And we're looking at here

971
00:47:52,566 --> 00:47:56,106
at either with or without
partial volume correction

972
00:47:56,106 --> 00:47:59,013
or with partial volume, the
magnitude of the effect sizes.

973
00:47:59,013 --> 00:48:01,396
And we compare that to just
the gray matter volumes

974
00:48:01,396 --> 00:48:03,836
that we see, and we
have larger effect sizes

975
00:48:03,836 --> 00:48:05,826
that we're seeing with
the synaptic markers

976
00:48:05,826 --> 00:48:08,867
than we have with just the volume markers.

977
00:48:08,867 --> 00:48:11,016
And the most interesting one
that we've published recently

978
00:48:11,016 --> 00:48:12,466
is the correlations that we're getting

979
00:48:12,466 --> 00:48:16,266
of our synaptic markers here
using a global measure of that

980
00:48:16,266 --> 00:48:18,746
to a variety of the cognitive scores.

981
00:48:18,746 --> 00:48:21,496
And we're seeing that across
a myriad of different ranges

982
00:48:21,496 --> 00:48:23,173
of different cognitive measures there.

983
00:48:23,173 --> 00:48:25,136
And these are very strong correlations.

984
00:48:25,136 --> 00:48:26,266
Again, in this case,

985
00:48:26,266 --> 00:48:29,426
we're only looking here at
the MCI and AD subjects.

986
00:48:29,426 --> 00:48:31,416
We're not including the healthy controls,

987
00:48:31,416 --> 00:48:32,823
which would give us, you know,

988
00:48:32,823 --> 00:48:33,809
help along inappropriately

989
00:48:33,809 --> 00:48:36,826
to be able to look for those correlations.

990
00:48:36,826 --> 00:48:39,526
So we had in our original 28 pilot data,

991
00:48:39,526 --> 00:48:40,936
we saw those reductions.

992
00:48:40,936 --> 00:48:42,786
We've now improved this with cerebellum,

993
00:48:42,786 --> 00:48:44,576
and now we're seeing with a larger cohort

994
00:48:44,576 --> 00:48:46,606
these reductions are very widespread,

995
00:48:46,606 --> 00:48:49,256
and with a very importantly
significant correlation,

996
00:48:49,256 --> 00:48:50,756
so that the SV2A is giving us

997
00:48:50,756 --> 00:48:53,046
some measure of cognition here,

998
00:48:53,046 --> 00:48:56,426
at least cross-sectionally
in this population.

999
00:48:56,426 --> 00:48:57,516
Let's take it a little further

1000
00:48:57,516 --> 00:48:59,426
and start to begin looking at one aspect

1001
00:48:59,426 --> 00:49:01,556
of network effects in the brain.

1002
00:49:01,556 --> 00:49:03,116
And a natural thing to do is

1003
00:49:03,116 --> 00:49:05,776
to compare the synaptic
binding that we see

1004
00:49:05,776 --> 00:49:07,256
to the deposition of tau.

1005
00:49:07,256 --> 00:49:10,476
And here we're gonna be
using AV-1451 Flortaucipir

1006
00:49:10,476 --> 00:49:11,803
as a measure of tau.

1007
00:49:11,803 --> 00:49:14,086
And we're particularly
looking at the perforant path

1008
00:49:14,086 --> 00:49:16,696
of projections from entorhinal cortex

1009
00:49:16,696 --> 00:49:18,556
up to hippocampus, able to see.

1010
00:49:18,556 --> 00:49:19,626
So this was a small study

1011
00:49:19,626 --> 00:49:23,336
with 10 subject cognitively
normal, 10 Alzheimer's disease,

1012
00:49:23,336 --> 00:49:26,636
basically related group to
what you've seen before.

1013
00:49:26,636 --> 00:49:28,716
Here, we allowed a couple
of our cognitively normals

1014
00:49:28,716 --> 00:49:31,756
who are amyloid positive
to remain in the group.

1015
00:49:31,756 --> 00:49:32,956
If we look at the tau measures,

1016
00:49:32,956 --> 00:49:35,556
obviously we have nice strong
effects throughout that

1017
00:49:35,556 --> 00:49:36,766
with our 10 and 10.

1018
00:49:36,766 --> 00:49:38,196
With the 10 and 10 SV2A,

1019
00:49:38,196 --> 00:49:41,366
this is on the margin of
being able to be significant.

1020
00:49:41,366 --> 00:49:42,576
We have changes in hippocampus,

1021
00:49:42,576 --> 00:49:45,146
but the Ns are small
to be able to do that.

1022
00:49:45,146 --> 00:49:47,426
If we look at it visually, the tau load,

1023
00:49:47,426 --> 00:49:48,716
again, this is those averages,

1024
00:49:48,716 --> 00:49:51,306
the SV2A that you've
basically seen already,

1025
00:49:51,306 --> 00:49:53,636
large increases in tau, no surprises here.

1026
00:49:53,636 --> 00:49:56,676
But now is there a relationship
between these two measures?

1027
00:49:56,676 --> 00:49:57,846
And this is where it gets interesting.

1028
00:49:57,846 --> 00:49:59,456
So on the x-axis here,

1029
00:49:59,456 --> 00:50:03,456
We're plotting the amount
of entorhinal cortex tau,

1030
00:50:03,456 --> 00:50:05,576
here it's measured by an SUV ratio,

1031
00:50:05,576 --> 00:50:07,126
to the distribution volume measured

1032
00:50:07,126 --> 00:50:09,076
upstream in the hippocampus.

1033
00:50:09,076 --> 00:50:11,656
We get a very nice relationship,
nice separation, of course,

1034
00:50:11,656 --> 00:50:13,756
between the cognitively
normals and the ADs.

1035
00:50:13,756 --> 00:50:17,406
Interestingly, the two PIP
positive cognitively normals

1036
00:50:17,406 --> 00:50:19,146
tend to fall in between.

1037
00:50:19,146 --> 00:50:20,816
If you look at that
against hippocampal volume,

1038
00:50:20,816 --> 00:50:23,306
we still have that relationship as well,

1039
00:50:23,306 --> 00:50:26,626
but it is not as
statistically significant.

1040
00:50:26,626 --> 00:50:28,046
So it's interesting that we're finding

1041
00:50:28,046 --> 00:50:30,316
this relationship to entorhinal tau,

1042
00:50:30,316 --> 00:50:32,116
how it projects and affects

1043
00:50:32,116 --> 00:50:34,026
synaptic density in the hippocampus,

1044
00:50:34,026 --> 00:50:36,796
which we see that in AD
compared to controls.

1045
00:50:36,796 --> 00:50:38,496
And so this is a interesting relationship,

1046
00:50:38,496 --> 00:50:41,206
just beginning to talk about
a very specific network

1047
00:50:41,206 --> 00:50:43,006
that we can look at within the disease.

1048
00:50:43,006 --> 00:50:45,826
And this is an interesting
one to follow longitudinally

1049
00:50:45,826 --> 00:50:47,539
as our ongoing studies will do.

1050
00:50:49,006 --> 00:50:51,916
But in looking at our measures
and thinking about this,

1051
00:50:51,916 --> 00:50:53,176
what is it that we're measuring?

1052
00:50:53,176 --> 00:50:55,886
You know, we're measuring
these synaptic vesicles.

1053
00:50:55,886 --> 00:50:58,016
We're looking at these proteins
on the synaptic vesicles

1054
00:50:58,016 --> 00:50:59,406
but of course the purpose of the vesicles

1055
00:50:59,406 --> 00:51:02,556
is to release neurotransmitters
into the synaptic cleft.

1056
00:51:02,556 --> 00:51:03,766
And how those are released,

1057
00:51:03,766 --> 00:51:05,573
they recycle and they come around.

1058
00:51:05,573 --> 00:51:06,956
And that became the question

1059
00:51:06,956 --> 00:51:08,856
of, well, are we sensitive to this?

1060
00:51:08,856 --> 00:51:11,066
If we have an activated state,

1061
00:51:11,066 --> 00:51:12,556
is this going to be something
where we're gonna have

1062
00:51:12,556 --> 00:51:14,886
a change in our signal or not?

1063
00:51:14,886 --> 00:51:16,116
And so how do we go about that?

1064
00:51:16,116 --> 00:51:16,949
Well, we kind of went

1065
00:51:16,949 --> 00:51:19,286
back to old-fashioned looking
at activation studies.

1066
00:51:19,286 --> 00:51:21,886
So we took some healthy
subjects and they did two scans,

1067
00:51:21,886 --> 00:51:23,416
one that was done at baseline, and one was

1068
00:51:23,416 --> 00:51:27,446
with classic eight hertz radial
checkerboard activations,

1069
00:51:27,446 --> 00:51:30,446
being able to look and see
what's going on in there.

1070
00:51:30,446 --> 00:51:31,636
And at the same time,

1071
00:51:31,636 --> 00:51:34,916
these people were also
independently did fMRI studies.

1072
00:51:34,916 --> 00:51:37,626
Because of the way PET is
taking longer period of time,

1073
00:51:37,626 --> 00:51:39,766
we used the PET-optimized
where we were on and off

1074
00:51:39,766 --> 00:51:42,243
for longer periods of
time compared to fMRI.

1075
00:51:42,243 --> 00:51:45,396
And we did it both ways
to be able to compare.

1076
00:51:45,396 --> 00:51:46,806
And so here are our summary images.

1077
00:51:46,806 --> 00:51:49,413
What we're gonna get is we're
gonna get our binding image

1078
00:51:49,413 --> 00:51:51,476
of the volume of distribution.

1079
00:51:51,476 --> 00:51:53,036
We're also gonna get that delivery image,

1080
00:51:53,036 --> 00:51:57,316
that K1 image there, which we
know is related to blood flow.

1081
00:51:57,316 --> 00:51:59,756
And here's an example of what
the average of these images

1082
00:51:59,756 --> 00:52:01,596
would look like at baseline or activation.

1083
00:52:01,596 --> 00:52:02,956
If we look at the flow image,

1084
00:52:02,956 --> 00:52:05,726
we get a lovely activation
in the primary visual cortex,

1085
00:52:05,726 --> 00:52:07,106
just as we'd expect.

1086
00:52:07,106 --> 00:52:09,626
And in that V1 area, we're
getting a very large change,

1087
00:52:09,626 --> 00:52:12,796
about 35% increase in K1.

1088
00:52:12,796 --> 00:52:14,306
If we look at other control regions,

1089
00:52:14,306 --> 00:52:16,239
temporal cortex, the
white matters, no changes.

1090
00:52:16,239 --> 00:52:18,216
If we look at our binding measure,

1091
00:52:18,216 --> 00:52:19,776
our reflection of the synaptic density,

1092
00:52:19,776 --> 00:52:21,166
we've got no change at all.

1093
00:52:21,166 --> 00:52:24,186
So the fact that we're having
this dynamic change going on

1094
00:52:24,186 --> 00:52:25,906
with huge flow increases there,

1095
00:52:25,906 --> 00:52:28,466
it's not changing the signal that we see

1096
00:52:28,466 --> 00:52:30,506
in the primary visual cortex.

1097
00:52:30,506 --> 00:52:32,846
Active binding potentials again,

1098
00:52:32,846 --> 00:52:34,736
that measure again,
normalized to white matter,

1099
00:52:34,736 --> 00:52:36,226
no changes at all.

1100
00:52:36,226 --> 00:52:39,236
So that suggests that
based on this activation,

1101
00:52:39,236 --> 00:52:42,206
that something that produces
a very large flow change,

1102
00:52:42,206 --> 00:52:44,266
that our measure is gonna be stable.

1103
00:52:44,266 --> 00:52:46,896
There's a representation
of what's going on

1104
00:52:46,896 --> 00:52:50,076
in terms of the number of
this binding site of SV2A

1105
00:52:50,076 --> 00:52:51,931
in the synaptic vesicles.

1106
00:52:51,931 --> 00:52:54,697
And of course we also, as I
mentioned, did it with fMRI

1107
00:52:54,697 --> 00:52:57,076
and of course we get the very
clear obvious activation here

1108
00:52:57,076 --> 00:52:59,266
back in the primary visual cortex.

1109
00:52:59,266 --> 00:53:01,146
And one thing that was very satisfying is

1110
00:53:01,146 --> 00:53:03,706
that we looked at the percent
change in the BOLD signal

1111
00:53:03,706 --> 00:53:05,196
correlated at the percent change

1112
00:53:05,196 --> 00:53:06,676
in K1 that I just showed you.

1113
00:53:06,676 --> 00:53:07,776
And across our seven subjects,

1114
00:53:07,776 --> 00:53:09,086
we got a very good correlation.

1115
00:53:09,086 --> 00:53:11,636
Of course, the PET signals are
producing very large changes,

1116
00:53:11,636 --> 00:53:14,436
but the signal to noises are
gonna be pretty comparable.

1117
00:53:14,436 --> 00:53:16,646
So this was saying that one
thing we wanna think about is

1118
00:53:16,646 --> 00:53:18,616
that our measures, and
I'll go back one slide,

1119
00:53:18,616 --> 00:53:20,806
since we can get a measure of flow

1120
00:53:20,806 --> 00:53:23,006
and a measure of synaptic binding,

1121
00:53:23,006 --> 00:53:24,996
we have two pieces of information

1122
00:53:24,996 --> 00:53:26,266
that we can get at the same time.

1123
00:53:26,266 --> 00:53:27,896
And that's gonna affect our interpretation

1124
00:53:27,896 --> 00:53:30,426
in terms of how we use that to understand

1125
00:53:30,426 --> 00:53:32,983
both the structure and
the function of the brain.

1126
00:53:32,983 --> 00:53:34,826
And we had seen that before.

1127
00:53:34,826 --> 00:53:36,856
This was the result from our early cohort

1128
00:53:36,856 --> 00:53:40,026
in Alzheimer's disease, where
we saw in the VT measures,

1129
00:53:40,026 --> 00:53:42,696
our primary effects were in hippocampus

1130
00:53:42,696 --> 00:53:44,566
and up in medial temporal lobe.

1131
00:53:44,566 --> 00:53:46,646
And again, this was the
small N study there,

1132
00:53:46,646 --> 00:53:49,886
but we had noticed in our K1
changes, our flow changes,

1133
00:53:49,886 --> 00:53:51,246
that along with synaptic changes,

1134
00:53:51,246 --> 00:53:54,166
we were seeing changes in
temporal parietal regions.

1135
00:53:54,166 --> 00:53:56,166
We were seeing changes in precuneus.

1136
00:53:56,166 --> 00:53:59,376
So again, with the one image,
with the one injection,

1137
00:53:59,376 --> 00:54:00,976
we can get a binding measure

1138
00:54:00,976 --> 00:54:03,306
reflective of the synaptic vesicles

1139
00:54:03,306 --> 00:54:05,496
and a flow measure, two
for the price of one.

1140
00:54:05,496 --> 00:54:07,466
So we'll come back to
that as we go forward.

1141
00:54:07,466 --> 00:54:10,046
So we wanted to look at
that in Alzheimer's disease.

1142
00:54:10,046 --> 00:54:11,886
And so in Alzheimer's
what's been used before

1143
00:54:11,886 --> 00:54:14,096
as a functional measure has been FDG,

1144
00:54:14,096 --> 00:54:15,046
and being able to do that.

1145
00:54:15,046 --> 00:54:16,966
That's been classically used for years.

1146
00:54:16,966 --> 00:54:18,216
Well, now we have three measures.

1147
00:54:18,216 --> 00:54:19,406
We have a synaptic measure.

1148
00:54:19,406 --> 00:54:21,586
That'd be that distribution volume ratio

1149
00:54:21,586 --> 00:54:23,556
we get with our UCB-J PET,

1150
00:54:23,556 --> 00:54:26,396
a flow measure, and this is
with our delivery constant.

1151
00:54:26,396 --> 00:54:28,236
Here we're using a relative one R1

1152
00:54:28,236 --> 00:54:30,226
that would be normalized to cerebellum,

1153
00:54:30,226 --> 00:54:32,946
as well as FDG, where we
can get a glucose analog.

1154
00:54:32,946 --> 00:54:34,856
So the question is, how do these relate?

1155
00:54:34,856 --> 00:54:38,206
When we have multiple
measures in the same patients,

1156
00:54:38,206 --> 00:54:41,916
how can we merge and interact
those and put those together?

1157
00:54:41,916 --> 00:54:42,936
So one approach we use

1158
00:54:42,936 --> 00:54:45,206
is called canonical correlation analysis.

1159
00:54:45,206 --> 00:54:47,116
And basically it looks
at pairs of datasets,

1160
00:54:47,116 --> 00:54:48,666
two datasets at a time.

1161
00:54:48,666 --> 00:54:50,486
So we have a value of a regional value

1162
00:54:50,486 --> 00:54:51,906
across different subjects.

1163
00:54:51,906 --> 00:54:54,193
So for example, this
might be our SV2A data,

1164
00:54:54,193 --> 00:54:55,516
our synaptic data.

1165
00:54:55,516 --> 00:54:58,856
This might be our FDG data
or our functional data.

1166
00:54:58,856 --> 00:55:00,236
And what we will then do is say,

1167
00:55:00,236 --> 00:55:02,626
what kind of weighted sum
of the regions can we do

1168
00:55:02,626 --> 00:55:04,986
to produce a canonical variant for data,

1169
00:55:04,986 --> 00:55:06,386
for example, datas U and V.

1170
00:55:06,386 --> 00:55:08,606
And then we find those that
maximize the correlation.

1171
00:55:08,606 --> 00:55:10,776
We can find, where are
the patterns similar?

1172
00:55:10,776 --> 00:55:12,836
How do they support each other

1173
00:55:12,836 --> 00:55:14,896
in terms of their regional pattern?

1174
00:55:14,896 --> 00:55:16,686
So we could do, in this
case, we did three pairs.

1175
00:55:16,686 --> 00:55:21,636
We could look at the DVR of
UCB-J against glucose, the FDG,

1176
00:55:21,636 --> 00:55:23,806
or we could look at that
against the flow measure.

1177
00:55:23,806 --> 00:55:25,326
This would be a within scan effect.

1178
00:55:25,326 --> 00:55:28,326
Or just look at how blood flow
measures correlate with FDG.

1179
00:55:28,326 --> 00:55:30,756
But let's concentrate first on comparing

1180
00:55:30,756 --> 00:55:33,076
these two different scans
in the same subjects,

1181
00:55:33,076 --> 00:55:35,276
FDG looking at glucose metabolism

1182
00:55:35,276 --> 00:55:37,776
and UCB-J looking at the synaptic density.

1183
00:55:37,776 --> 00:55:39,606
So firstly, we look at
the regional loading.

1184
00:55:39,606 --> 00:55:42,396
So this is the values that
are applied for each region.

1185
00:55:42,396 --> 00:55:43,876
This is what we got for the pattern

1186
00:55:43,876 --> 00:55:45,996
for the synaptic measure,
for the glucose measure.

1187
00:55:45,996 --> 00:55:48,356
There's a pretty good
correlation between them, okay?

1188
00:55:48,356 --> 00:55:50,596
And that's saying that both measures

1189
00:55:50,596 --> 00:55:52,566
are containing some related information.

1190
00:55:52,566 --> 00:55:54,136
And remember here, when
we've done this analysis,

1191
00:55:54,136 --> 00:55:56,626
we've thrown in the healthy
controls and the Alzheimer's.

1192
00:55:56,626 --> 00:55:58,016
In that part, we're not telling them

1193
00:55:58,016 --> 00:55:59,899
about which patients are which.

1194
00:56:00,756 --> 00:56:03,086
And here's where those weights
look like for synapses.

1195
00:56:03,086 --> 00:56:03,919
So what are we seeing?

1196
00:56:03,919 --> 00:56:06,546
Not surprisingly, we're getting
weights on the hippocampus.

1197
00:56:06,546 --> 00:56:08,056
We're getting weights on precuneus,

1198
00:56:08,056 --> 00:56:09,416
we're getting weights
on temporal parietal,

1199
00:56:09,416 --> 00:56:11,166
so regions that are not surprising

1200
00:56:11,166 --> 00:56:14,496
and come on that do correlate
with the FDG measures.

1201
00:56:14,496 --> 00:56:15,336
So now what do we get

1202
00:56:15,336 --> 00:56:17,866
if I plot the values in this relationship?

1203
00:56:17,866 --> 00:56:20,916
So each of these is the loading
for the synaptic measure

1204
00:56:20,916 --> 00:56:24,306
and the glucose, coming
from UCB-J, coming from FDG.

1205
00:56:24,306 --> 00:56:27,336
We see a nice distribution
of our dementia subjects,

1206
00:56:27,336 --> 00:56:30,596
our MCI subjects, and our
cognitively normals control.

1207
00:56:30,596 --> 00:56:32,966
So we can do a principle component.

1208
00:56:32,966 --> 00:56:35,726
Let's turn that into one
variable along this dimension.

1209
00:56:35,726 --> 00:56:37,886
And now we get really elegant separation

1210
00:56:37,886 --> 00:56:39,606
between all three groups,

1211
00:56:39,606 --> 00:56:42,136
so the cognitively normals
which separate from the MCI,

1212
00:56:42,136 --> 00:56:45,976
separate from the ones with
classic Alzheimer's dementia.

1213
00:56:45,976 --> 00:56:48,556
And here by combining two
pieces of information,

1214
00:56:48,556 --> 00:56:51,626
here, both the FDG, the more
functional, the synaptic one,

1215
00:56:51,626 --> 00:56:53,156
probably more stationary

1216
00:56:53,156 --> 00:56:55,839
in terms of synapses that are there.

1217
00:56:57,006 --> 00:56:58,516
Now, how did that compare
with other measures?

1218
00:56:58,516 --> 00:57:00,726
I mean, before, if we just
looked at the hippocampus,

1219
00:57:00,726 --> 00:57:02,486
we have no problem separating

1220
00:57:02,486 --> 00:57:04,546
our cognitively normal
from our MCIs and ADs,

1221
00:57:04,546 --> 00:57:06,476
but pretty much we can't separate those.

1222
00:57:06,476 --> 00:57:08,606
FDG pretty much the same story.

1223
00:57:08,606 --> 00:57:10,006
So we're doing a little bit better

1224
00:57:10,006 --> 00:57:12,306
in this ability to separate MCI and AD

1225
00:57:12,306 --> 00:57:14,636
by combining the two measures together,

1226
00:57:14,636 --> 00:57:16,296
and to see how that would work.

1227
00:57:16,296 --> 00:57:18,616
And how does that work for our
other measures there again?

1228
00:57:18,616 --> 00:57:20,766
So here are the regional
patterns that we see again.

1229
00:57:20,766 --> 00:57:24,346
We saw some relationship, but
some variability around that.

1230
00:57:24,346 --> 00:57:26,286
If I look at the blood flow measure,

1231
00:57:26,286 --> 00:57:27,456
again, that's the early part

1232
00:57:27,456 --> 00:57:29,269
of our SV2A scan versus glucose,

1233
00:57:29,269 --> 00:57:31,566
they're very strongly correlated.

1234
00:57:31,566 --> 00:57:34,276
They're saying pretty much
it's the same measure.

1235
00:57:34,276 --> 00:57:35,736
So that opens up the question

1236
00:57:35,736 --> 00:57:37,056
about, could I use these two measures?

1237
00:57:37,056 --> 00:57:38,976
Now these are two values
that come from one scan,

1238
00:57:38,976 --> 00:57:42,516
a synaptic marker, the equilibrium
measure, the flow marker.

1239
00:57:42,516 --> 00:57:45,676
Can we use that with a single
UCB-J scan to do as well?

1240
00:57:45,676 --> 00:57:46,946
And the answer turns out to be yes,

1241
00:57:46,946 --> 00:57:49,366
that just with data where
I'm combining there,

1242
00:57:49,366 --> 00:57:50,936
the synaptic measure and the flow measure

1243
00:57:50,936 --> 00:57:52,576
both coming from one scan,

1244
00:57:52,576 --> 00:57:56,336
we could separate out MCI and
AD and get nice correlations

1245
00:57:56,336 --> 00:57:57,506
here with the Mini-Mental score.

1246
00:57:57,506 --> 00:57:59,426
I didn't do the other
cognitive measures here,

1247
00:57:59,426 --> 00:58:02,876
just within the AD and the MCI groups.

1248
00:58:02,876 --> 00:58:04,826
So that suggested that with
these three measures here

1249
00:58:04,826 --> 00:58:07,636
doing pairwise comparison
or synaptic measure,

1250
00:58:07,636 --> 00:58:10,096
our flow measure, the
delivery and glucose measure,

1251
00:58:10,096 --> 00:58:11,576
they all contain related information,

1252
00:58:11,576 --> 00:58:13,856
but the flow and glucose
are very, very similar.

1253
00:58:13,856 --> 00:58:15,356
And so we think that when we just use

1254
00:58:15,356 --> 00:58:17,686
both pieces of information from one scan,

1255
00:58:17,686 --> 00:58:19,456
we can provide additional information

1256
00:58:19,456 --> 00:58:21,566
in terms of either diagnostic accuracy

1257
00:58:21,566 --> 00:58:23,989
and/or propagation of measurements there.

1258
00:58:25,096 --> 00:58:28,306
So let's move on into a
different disorder here.

1259
00:58:28,306 --> 00:58:31,446
So let's look into depression.

1260
00:58:31,446 --> 00:58:34,146
And I mean, synaptic loss
has been demonstrated

1261
00:58:34,146 --> 00:58:36,416
to be a long hold in postmortem data,

1262
00:58:36,416 --> 00:58:39,216
about those loss of synapses
there going back quite a while

1263
00:58:39,216 --> 00:58:41,366
and be being very
significant measures there

1264
00:58:41,366 --> 00:58:42,199
if you're measuring

1265
00:58:42,199 --> 00:58:46,406
other synaptic vesicle
protein measurements there.

1266
00:58:46,406 --> 00:58:49,206
And as well, as we
mentioned that we know this,

1267
00:58:49,206 --> 00:58:50,736
but in the animal models,

1268
00:58:50,736 --> 00:58:52,796
that the chronic stresses
I mentioned before,

1269
00:58:52,796 --> 00:58:55,579
that can produce the
alterations in synaptic spines

1270
00:58:55,579 --> 00:58:57,536
and the synaptic density would be there,

1271
00:58:57,536 --> 00:59:00,026
coming out of Ron Duman's seminal work.

1272
00:59:00,026 --> 00:59:01,446
So naturally it made sense

1273
00:59:01,446 --> 00:59:04,596
to be able to do a study in depression.

1274
00:59:04,596 --> 00:59:07,266
And so this was a study
led by Irina Esterlis

1275
00:59:07,266 --> 00:59:09,236
looking at 21 healthy controls,

1276
00:59:09,236 --> 00:59:10,686
and then 26 with depression here

1277
00:59:10,686 --> 00:59:13,356
grouped into a low and
high severity group.

1278
00:59:13,356 --> 00:59:17,876
And so they're well matched
for age and other markers.

1279
00:59:17,876 --> 00:59:19,246
And we certainly have

1280
00:59:20,876 --> 00:59:22,710
no substantive changes

1281
00:59:22,710 --> 00:59:26,136
in terms of the imaging
markers that we used here.

1282
00:59:26,136 --> 00:59:27,756
So here's what those values look like.

1283
00:59:27,756 --> 00:59:31,261
And here we've focused on
a couple key regions here,

1284
00:59:31,261 --> 00:59:32,716
dorsolateral prefrontal cortex,

1285
00:59:32,716 --> 00:59:34,673
interior cingulate and the hippocampus.

1286
00:59:34,673 --> 00:59:36,716
And we're not seeing a change

1287
00:59:36,716 --> 00:59:40,036
if we compare the low severity,
that HAM-D scores are higher

1288
00:59:40,036 --> 00:59:42,326
compared to that of the healthy controls.

1289
00:59:42,326 --> 00:59:43,446
But in the high severity group,

1290
00:59:43,446 --> 00:59:46,826
we are seeing reductions across
the board in these measures,

1291
00:59:46,826 --> 00:59:49,476
here using a volume of
distribution measure.

1292
00:59:49,476 --> 00:59:51,676
And when we correlate that
volume of distribution measure,

1293
00:59:51,676 --> 00:59:53,226
again, in all three of these regions

1294
00:59:53,226 --> 00:59:55,486
against a HAM-D score as one example,

1295
00:59:55,486 --> 00:59:56,856
we get a negative correlation.

1296
00:59:56,856 --> 00:59:59,826
That is, higher depressive
symptoms are gonna be correlated

1297
00:59:59,826 --> 01:00:03,766
with lower synaptic density
in all of these three regions.

1298
01:00:03,766 --> 01:00:05,136
And we'd like to take
that a little further

1299
01:00:05,136 --> 01:00:07,756
and see what we can relate that into fMRI,

1300
01:00:07,756 --> 01:00:09,786
and so the measure that
we use a lot at Yale,

1301
01:00:09,786 --> 01:00:12,516
something called the intrinsic
connectivity distribution.

1302
01:00:12,516 --> 01:00:15,346
And so doing a resting state
fMRI, you get the time courses,

1303
01:00:15,346 --> 01:00:17,556
you generate your correlation map,

1304
01:00:17,556 --> 01:00:19,316
and then looking at a given voxel

1305
01:00:19,316 --> 01:00:20,676
or given region in the brain,

1306
01:00:20,676 --> 01:00:23,276
we can look at the distribution
of those correlations,

1307
01:00:23,276 --> 01:00:24,376
and be able to have a measure

1308
01:00:24,376 --> 01:00:27,056
of where we're showing
excessive correlations,

1309
01:00:27,056 --> 01:00:30,196
so this measure of this,
in the connectivity,

1310
01:00:30,196 --> 01:00:31,646
the distribution of the connectivities.

1311
01:00:31,646 --> 01:00:33,696
And by that measure, we
were able to show certainly

1312
01:00:33,696 --> 01:00:36,046
a significant difference
between a healthy control group

1313
01:00:36,046 --> 01:00:37,376
and a clinical depression group

1314
01:00:37,376 --> 01:00:39,148
that was shown some time ago.

1315
01:00:39,148 --> 01:00:41,026
But how can we tie that together

1316
01:00:41,026 --> 01:00:42,476
when we think about networks?

1317
01:00:42,476 --> 01:00:45,206
So we've long established
the kind of anti-correlations

1318
01:00:45,206 --> 01:00:47,376
one sees between a default-mode network

1319
01:00:47,376 --> 01:00:49,006
and central executive-mode networks.

1320
01:00:49,006 --> 01:00:51,126
And we can do that
within this data as well.

1321
01:00:51,126 --> 01:00:53,046
For example, if we look in DLPFC

1322
01:00:53,046 --> 01:00:54,736
in the central executive network,

1323
01:00:54,736 --> 01:00:57,126
and to see how does that
correlate with posterior cingulate

1324
01:00:57,126 --> 01:00:58,926
as part of the default-mode network,

1325
01:00:58,926 --> 01:01:00,986
and just the connectivity
that across subjects

1326
01:01:00,986 --> 01:01:02,986
is also a measure that correlates

1327
01:01:02,986 --> 01:01:04,469
with the HAM-D that we see,

1328
01:01:05,366 --> 01:01:07,586
that the anti-correlation is reduced

1329
01:01:07,586 --> 01:01:09,656
as you get to higher depression scores.

1330
01:01:09,656 --> 01:01:11,503
So does that all tie
together with our synapses?

1331
01:01:11,503 --> 01:01:14,860
And it does; I can use
this connectivity measure.

1332
01:01:14,860 --> 01:01:17,996
That means that the more
positive that connectivity is,

1333
01:01:17,996 --> 01:01:19,876
the less negative that that would be.

1334
01:01:19,876 --> 01:01:22,206
We are reflecting that
with a higher symptom score

1335
01:01:22,206 --> 01:01:24,916
and as well as lower
synaptic density in there.

1336
01:01:24,916 --> 01:01:26,256
So again, looking and trying to build

1337
01:01:26,256 --> 01:01:29,496
multiple measures together
to see how we can inform

1338
01:01:30,336 --> 01:01:32,786
our understanding of the brain networks.

1339
01:01:32,786 --> 01:01:33,836
Now let's take this one further,

1340
01:01:33,836 --> 01:01:36,106
look at a second tracer, a second target.

1341
01:01:36,106 --> 01:01:39,773
And that would be
looking at mGluR5 in MDD.

1342
01:01:39,773 --> 01:01:41,426
And we've done extensive studies there

1343
01:01:41,426 --> 01:01:42,363
using the tracer FPEB,

1344
01:01:42,363 --> 01:01:45,366
and we weren't seeing
baseline differences there.

1345
01:01:45,366 --> 01:01:47,926
Intrinsically the FPEB is a
little bit more variable there,

1346
01:01:47,926 --> 01:01:48,759
but we don't see that.

1347
01:01:48,759 --> 01:01:50,826
Even with higher symptom
severity we didn't see that.

1348
01:01:50,826 --> 01:01:52,926
But we're really
interested in this target,

1349
01:01:52,926 --> 01:01:55,036
obviously because of
the effects of ketamine.

1350
01:01:55,036 --> 01:01:57,266
Now, ketamine either in
healthy controls or depressed

1351
01:01:57,266 --> 01:02:01,056
can show dynamic changes
in the measures of mGluR5,

1352
01:02:01,056 --> 01:02:03,166
being able to show that
in short-term acting.

1353
01:02:03,166 --> 01:02:04,396
So it's an exciting target.

1354
01:02:04,396 --> 01:02:05,506
It's relevant to think

1355
01:02:05,506 --> 01:02:08,306
about how the excitatory marker for mGluR5

1356
01:02:08,306 --> 01:02:10,866
is gonna relate to our
total synaptic marker.

1357
01:02:10,866 --> 01:02:12,486
So here's the population that we have.

1358
01:02:12,486 --> 01:02:14,506
We have 12 controls and 20 MDD.

1359
01:02:14,506 --> 01:02:17,426
We've had both tracers,
and we've been careful

1360
01:02:17,426 --> 01:02:18,996
to be able to look, to try to match that.

1361
01:02:18,996 --> 01:02:20,706
Even though these were
done on different days,

1362
01:02:20,706 --> 01:02:22,616
the HAM-D scores don't change too much

1363
01:02:22,616 --> 01:02:24,526
from one day to the next.

1364
01:02:24,526 --> 01:02:25,816
So how does that relation work?

1365
01:02:25,816 --> 01:02:27,996
So this is just some
pilot data I wanna share

1366
01:02:27,996 --> 01:02:28,829
'cause it's just interesting

1367
01:02:28,829 --> 01:02:30,906
about directions that we may wanna go.

1368
01:02:30,906 --> 01:02:32,736
So what we're looking at
here is we have values,

1369
01:02:32,736 --> 01:02:35,816
and this is of looking at their mGluR5,

1370
01:02:35,816 --> 01:02:37,840
a measure with the tracer FPEB,

1371
01:02:37,840 --> 01:02:41,516
the SV2A marker, the
VT coming out of UCB-J.

1372
01:02:41,516 --> 01:02:43,776
But just look at our
healthy control populations.

1373
01:02:43,776 --> 01:02:46,677
We have very nice positive correlations,

1374
01:02:46,677 --> 01:02:50,636
0.5, 0.4, et cetera, between
the different tracers

1375
01:02:50,636 --> 01:02:53,016
within the cross-subjects
in these regions.

1376
01:02:53,016 --> 01:02:54,976
But when we go to the MDDs,

1377
01:02:54,976 --> 01:02:57,036
all those correlations fall apart.

1378
01:02:57,036 --> 01:02:58,886
We could see that in one region.

1379
01:02:58,886 --> 01:03:02,796
We're just looking at the DLPFC
with itself as one measure.

1380
01:03:02,796 --> 01:03:05,046
And in the healthy controls in green,

1381
01:03:05,046 --> 01:03:06,816
we have a nice positive correlation.

1382
01:03:06,816 --> 01:03:09,526
In the MDD, the correlation
is breaking down.

1383
01:03:09,526 --> 01:03:11,816
So we're suggesting some
interesting pattern here

1384
01:03:11,816 --> 01:03:13,516
where we're getting differential effects

1385
01:03:13,516 --> 01:03:16,516
on our synaptic marker and
our glutamate markers here

1386
01:03:16,516 --> 01:03:17,916
that are relevant in depression,

1387
01:03:17,916 --> 01:03:20,206
but not present in the healthy controls.

1388
01:03:20,206 --> 01:03:21,706
Much more work need to be done,

1389
01:03:21,706 --> 01:03:24,100
but an example of bringing
two tracers together

1390
01:03:24,100 --> 01:03:27,496
in the same patients in a different area.

1391
01:03:27,496 --> 01:03:29,076
Let's move to schizophrenia,

1392
01:03:29,076 --> 01:03:31,956
obviously another area where
there's been long field data

1393
01:03:31,956 --> 01:03:34,596
to suggest that there's
gonna be losses of synapses.

1394
01:03:34,596 --> 01:03:36,206
Measurements of spine density there

1395
01:03:36,206 --> 01:03:38,206
has been shown in postmortem data.

1396
01:03:38,206 --> 01:03:41,806
And also the interesting part
about how C4 overexpression

1397
01:03:41,806 --> 01:03:44,046
affects dendritic spine density

1398
01:03:44,046 --> 01:03:46,066
and how that is an effect that we can see

1399
01:03:46,066 --> 01:03:48,326
in schizophrenic patients,

1400
01:03:48,326 --> 01:03:50,606
increased expression of
C4 again in postmortem.

1401
01:03:50,606 --> 01:03:53,266
So naturally that led us to
doing a schizophrenia study.

1402
01:03:53,266 --> 01:03:55,466
Right now, this is a study
with 13 schizophrenia

1403
01:03:55,466 --> 01:03:57,286
and 15 healthy controls,

1404
01:03:57,286 --> 01:04:01,166
and some mixture about what
they're on, antipsychotics,

1405
01:04:01,166 --> 01:04:05,156
various cognitive measures,
structural clinical scores,

1406
01:04:05,156 --> 01:04:06,406
cog state, et cetera,

1407
01:04:06,406 --> 01:04:08,276
and following our
standard approaches here,

1408
01:04:08,276 --> 01:04:10,146
and also applying that with
partial volume correction

1409
01:04:10,146 --> 01:04:14,246
because of possible loss
of tissue measurements.

1410
01:04:14,246 --> 01:04:16,526
Here's the balance of
our patient populations,

1411
01:04:16,526 --> 01:04:19,846
and no differences there
in any of the measures

1412
01:04:19,846 --> 01:04:21,186
or the structural measures there.

1413
01:04:21,186 --> 01:04:22,706
And we'll also notice we're gonna use

1414
01:04:22,706 --> 01:04:25,186
a binding potential measure
normalized to the white matter

1415
01:04:25,186 --> 01:04:26,456
and no significant differences

1416
01:04:26,456 --> 01:04:28,836
in the white matter binding values.

1417
01:04:28,836 --> 01:04:30,826
Clinical characteristics of the patients

1418
01:04:30,826 --> 01:04:32,366
in terms of the typical scores,

1419
01:04:32,366 --> 01:04:34,846
typical exposure of antipsychotics,

1420
01:04:34,846 --> 01:04:37,596
and we can be using these as
covariates, as you'll see.

1421
01:04:38,566 --> 01:04:41,756
So if we look in the regions
that we were targeting there,

1422
01:04:41,756 --> 01:04:45,286
we are seeing mean reductions
and across multiple regions,

1423
01:04:45,286 --> 01:04:46,646
seeing that in the anterior cingulate,

1424
01:04:46,646 --> 01:04:48,396
in frontal cortex, hippocampus,

1425
01:04:48,396 --> 01:04:50,556
in fact, in multiple cortical
regions, so really feeling

1426
01:04:50,556 --> 01:04:53,676
like a widespread difference
throughout the brain.

1427
01:04:53,676 --> 01:04:55,876
And then that when we
look at those numbers,

1428
01:04:57,303 --> 01:04:59,576
we're seeing that these

1429
01:04:59,576 --> 01:05:02,056
were those values without corrections.

1430
01:05:02,056 --> 01:05:03,696
If we also look at volume,

1431
01:05:03,696 --> 01:05:05,956
we're also seeing volume
losses in some of these,

1432
01:05:05,956 --> 01:05:07,366
not all of them significant.

1433
01:05:07,366 --> 01:05:08,426
So that was why it was important

1434
01:05:08,426 --> 01:05:10,056
to do the partial volume corrections.

1435
01:05:10,056 --> 01:05:12,906
But we still have significance
in those primary regions

1436
01:05:12,906 --> 01:05:14,606
after correcting for the volume losses

1437
01:05:14,606 --> 01:05:15,766
that would be the same.

1438
01:05:15,766 --> 01:05:17,046
But the interesting part often becomes

1439
01:05:17,046 --> 01:05:18,986
when we begin to correlate our measures,

1440
01:05:18,986 --> 01:05:21,736
the in vivo binding
measures here using a BPND

1441
01:05:21,736 --> 01:05:24,276
against in this case, the
positive symptom score.

1442
01:05:24,276 --> 01:05:25,896
And that basically that the higher

1443
01:05:25,896 --> 01:05:29,306
of these positive symptom
scores that we get,

1444
01:05:29,306 --> 01:05:31,956
we're seeing lower in this
case, in the frontal cortex,

1445
01:05:31,956 --> 01:05:34,106
and the reverse effect when
we look at a different measure

1446
01:05:34,106 --> 01:05:36,416
and we look at the
social cognition measure.

1447
01:05:36,416 --> 01:05:39,326
So overall the study,
this was not the first,

1448
01:05:39,326 --> 01:05:42,506
the first one came from
the UK from Onwordi et al,

1449
01:05:42,506 --> 01:05:45,566
we replicated those
reductions in schizophrenia,

1450
01:05:45,566 --> 01:05:47,496
and we're seeing them in multiple regions,

1451
01:05:47,496 --> 01:05:49,019
so fairly widespread.

1452
01:05:50,046 --> 01:05:52,096
We also saw that we had
correlations in frontal

1453
01:05:52,096 --> 01:05:54,366
with varied symptom
severity cognitive scores,

1454
01:05:54,366 --> 01:05:57,116
but we didn't see associations
with duration of illness

1455
01:05:57,116 --> 01:06:00,036
or with the total amount
of antipsychotic exposure.

1456
01:06:00,036 --> 01:06:01,216
So more needs to be done here,

1457
01:06:01,216 --> 01:06:03,357
just beginning that relationship.

1458
01:06:03,357 --> 01:06:05,726
All right, so in the
last section of the talk,

1459
01:06:05,726 --> 01:06:09,356
I'd like to kind of go
through some newer methodology

1460
01:06:09,356 --> 01:06:11,676
that we're working with in
terms of trying to be able

1461
01:06:11,676 --> 01:06:14,766
to look at these in multiple
variates and multiple ways,

1462
01:06:14,766 --> 01:06:17,486
and especially with something
as interesting as synapses,

1463
01:06:17,486 --> 01:06:18,816
how that might correlate

1464
01:06:18,816 --> 01:06:21,476
with other measures that
we can get with fMRI.

1465
01:06:21,476 --> 01:06:23,126
So I don't need to tell
this audience probably

1466
01:06:23,126 --> 01:06:24,586
about resting state networks

1467
01:06:24,586 --> 01:06:26,286
that have been around for a while.

1468
01:06:26,286 --> 01:06:27,756
That has been this really interesting

1469
01:06:27,756 --> 01:06:30,196
and really reproducible
in general measures

1470
01:06:30,196 --> 01:06:32,996
that we can be getting
from resting state fMRI,

1471
01:06:32,996 --> 01:06:33,829
but not just from fMRI.

1472
01:06:33,829 --> 01:06:36,276
I mean, this has been
around from FDG as well.

1473
01:06:36,276 --> 01:06:37,856
And if we look at the FDG patterns,

1474
01:06:37,856 --> 01:06:39,506
we find that those patterns,

1475
01:06:39,506 --> 01:06:40,946
the spatial patterns that we get

1476
01:06:40,946 --> 01:06:42,836
have a lot of visual similarities

1477
01:06:42,836 --> 01:06:45,576
that we might get to the
resting state values,

1478
01:06:45,576 --> 01:06:48,026
although there's some
components that are unique.

1479
01:06:48,026 --> 01:06:49,386
And that makes sense if we're talking

1480
01:06:49,386 --> 01:06:53,256
about what's going on in the
kind of the functional states

1481
01:06:53,256 --> 01:06:55,126
that are being there that
are also being driven

1482
01:06:55,126 --> 01:06:56,596
by the metabolic signals.

1483
01:06:56,596 --> 01:06:59,006
So those things might
be very, very similar.

1484
01:06:59,006 --> 01:07:00,346
So that's the goal is let's go

1485
01:07:00,346 --> 01:07:02,756
and look in our synaptic
density data and say,

1486
01:07:02,756 --> 01:07:04,986
well, what can we do?

1487
01:07:04,986 --> 01:07:07,006
How can we use the patterns
and look in resting state

1488
01:07:07,006 --> 01:07:09,206
using independent component analysis

1489
01:07:09,206 --> 01:07:10,426
to be able to extract them

1490
01:07:10,426 --> 01:07:12,736
from our synaptic density markers?

1491
01:07:12,736 --> 01:07:14,316
All right, so independent
component analysis,

1492
01:07:14,316 --> 01:07:17,116
and many of you probably
know this better than me,

1493
01:07:17,116 --> 01:07:18,926
but the idea is to separate the signal.

1494
01:07:18,926 --> 01:07:22,056
So for example, if I looked
at this colorful square,

1495
01:07:22,056 --> 01:07:23,966
it really could be recognized

1496
01:07:23,966 --> 01:07:26,366
that it's broken down
into different intensities

1497
01:07:26,366 --> 01:07:28,866
of red, green, and blue,
and how they add up.

1498
01:07:28,866 --> 01:07:32,186
So we could break this
down into its sources there

1499
01:07:32,186 --> 01:07:34,624
with a different loading for each voxel.

1500
01:07:34,624 --> 01:07:36,116
And of course, when we do imaging data,

1501
01:07:36,116 --> 01:07:36,949
the same logic can be here.

1502
01:07:36,949 --> 01:07:39,056
There can be different patterns of there

1503
01:07:39,056 --> 01:07:41,196
that are merged in different ways.

1504
01:07:41,196 --> 01:07:43,996
Now in fMRI, you've got
that nice dimension of time

1505
01:07:43,996 --> 01:07:45,866
in the spatial loadings
with the resting state.

1506
01:07:45,866 --> 01:07:48,076
In PET we don't have
it organized that way.

1507
01:07:48,076 --> 01:07:49,016
So here, what we're gonna do

1508
01:07:49,016 --> 01:07:50,976
is doing that across different subjects.

1509
01:07:50,976 --> 01:07:53,106
So each subject we're
gonna have its PET data,

1510
01:07:53,106 --> 01:07:55,616
and for each voxel or
each region going across.

1511
01:07:55,616 --> 01:07:56,936
So this is our input matrix

1512
01:07:56,936 --> 01:07:59,456
to be able to do independent
component analysis.

1513
01:07:59,456 --> 01:08:00,646
And we'll then be able to use that

1514
01:08:00,646 --> 01:08:03,276
to break that down into
images that'll be source maps.

1515
01:08:03,276 --> 01:08:05,846
And for each subject, we're
gonna have a loading score

1516
01:08:05,846 --> 01:08:08,086
to be able to map towards
those source maps.

1517
01:08:08,086 --> 01:08:09,766
You may have to decide
how many of these to do

1518
01:08:09,766 --> 01:08:12,346
and how reproducible that they might be.

1519
01:08:12,346 --> 01:08:13,636
And that's kind of the interesting thing

1520
01:08:13,636 --> 01:08:15,186
is really understanding
those loading weights.

1521
01:08:15,186 --> 01:08:17,346
What can they tell us
about different patients

1522
01:08:17,346 --> 01:08:21,166
or how they correlate with
other individual measures?

1523
01:08:21,166 --> 01:08:23,786
Now we started doing
this back a while ago.

1524
01:08:23,786 --> 01:08:26,316
Pat Worhunsky pioneered this in our hands,

1525
01:08:26,316 --> 01:08:28,009
looking at in PHNO.

1526
01:08:28,009 --> 01:08:30,216
PHNO binds to dopamine D2 and D3s.

1527
01:08:30,216 --> 01:08:32,106
It is a sum of two patterns.

1528
01:08:32,106 --> 01:08:33,626
And it worked out really, really well

1529
01:08:33,626 --> 01:08:35,856
that we could look at this
both in a healthy control

1530
01:08:35,856 --> 01:08:37,946
and cocaine use disorder population,

1531
01:08:37,946 --> 01:08:39,756
where we found multiple measures

1532
01:08:39,756 --> 01:08:41,776
automatically pulling
that out of the data.

1533
01:08:41,776 --> 01:08:43,396
A striatopallidal portion,

1534
01:08:43,396 --> 01:08:46,866
which is really lined up
with the dopamine D2 values,

1535
01:08:46,866 --> 01:08:48,836
which showed differences,
which showed lower,

1536
01:08:48,836 --> 01:08:51,236
consistent with separate D2 studies

1537
01:08:51,236 --> 01:08:52,536
in relations to cocaine use.

1538
01:08:52,536 --> 01:08:54,186
But then when we focused on a separate one

1539
01:08:54,186 --> 01:08:56,396
that was leaning towards the D3 areas,

1540
01:08:56,396 --> 01:08:58,186
being able to pull out the nigra

1541
01:08:58,186 --> 01:09:00,086
and other areas in the globus pallidus

1542
01:09:00,086 --> 01:09:02,446
showing increases in CUD, which we saw

1543
01:09:02,446 --> 01:09:05,306
if we just looked
individually at those regions.

1544
01:09:05,306 --> 01:09:07,586
It was a nice thing
because this is a tracer

1545
01:09:07,586 --> 01:09:10,106
that is a combination of multiple sources.

1546
01:09:10,106 --> 01:09:12,436
ICA was able to find them.

1547
01:09:12,436 --> 01:09:14,086
So let's see how we're
doing this with SV2A.

1548
01:09:14,086 --> 01:09:17,236
We have a large cohort of
healthy controls, 80 subjects,

1549
01:09:17,236 --> 01:09:18,996
all of them that are rest-scanned,

1550
01:09:18,996 --> 01:09:20,766
60 minutes with arterial blood samples.

1551
01:09:20,766 --> 01:09:23,436
We're using that total binding measure VT,

1552
01:09:23,436 --> 01:09:25,436
trying to start with our
most conservative measure,

1553
01:09:25,436 --> 01:09:27,286
registering everybody in the MNI template

1554
01:09:27,286 --> 01:09:30,876
and smoothing the images there
by by about 12 millimeters.

1555
01:09:30,876 --> 01:09:34,556
And we used the source-based
morphom SBM toolbox,

1556
01:09:34,556 --> 01:09:35,389
the ICA toolbox there.

1557
01:09:36,916 --> 01:09:38,216
And part of what we wanted to know

1558
01:09:38,216 --> 01:09:40,476
is just, well, how
reliable might these be?

1559
01:09:40,476 --> 01:09:41,756
I mean, we looked at that FDG study

1560
01:09:41,756 --> 01:09:44,836
that had 18 sources there,
and we weren't really sure.

1561
01:09:44,836 --> 01:09:46,556
So we looked at multiple source levels

1562
01:09:46,556 --> 01:09:49,106
to be able to see what
would be most reliable.

1563
01:09:49,106 --> 01:09:51,596
We checked how stable
they were using ICASSO

1564
01:09:51,596 --> 01:09:53,606
to be able to iterate,
to be able to find that.

1565
01:09:53,606 --> 01:09:56,386
And then we also split the
sample to be able to see

1566
01:09:56,386 --> 01:09:58,816
which of those were reliably restored.

1567
01:09:58,816 --> 01:10:00,806
We wanted to know
basically how far to take

1568
01:10:00,806 --> 01:10:02,816
our PET data in this way.

1569
01:10:02,816 --> 01:10:04,726
So here are the images, and
I'm sorry that they're small.

1570
01:10:04,726 --> 01:10:06,816
But basically be taking the same data set

1571
01:10:06,816 --> 01:10:11,586
where we extracted 8 or 12
or 18 or 24 components there.

1572
01:10:11,586 --> 01:10:12,676
And what I'm showing you are those

1573
01:10:12,676 --> 01:10:14,676
that were reliably characterized

1574
01:10:14,676 --> 01:10:17,366
as we moved into multiple
components there.

1575
01:10:17,366 --> 01:10:18,576
So for example, out of the eight,

1576
01:10:18,576 --> 01:10:20,666
four of them were replicated very nicely.

1577
01:10:20,666 --> 01:10:22,426
As we went to 12, we added new ones,

1578
01:10:22,426 --> 01:10:24,006
which were replicated nicely,

1579
01:10:24,006 --> 01:10:27,356
when we went to 18 here at
that quality coefficient.

1580
01:10:27,356 --> 01:10:28,466
And when we got up to 24,

1581
01:10:28,466 --> 01:10:30,536
we were starting to see
many of the components

1582
01:10:30,536 --> 01:10:33,436
that just were not there to be
able to be measured reliably.

1583
01:10:33,436 --> 01:10:35,446
So we focus that down back on 18.

1584
01:10:35,446 --> 01:10:39,066
Of those 18, looking a
little bit more carefully,

1585
01:10:39,066 --> 01:10:41,146
we've really focused that basically 13

1586
01:10:41,146 --> 01:10:43,756
we would consider to
be reasonable networks,

1587
01:10:43,756 --> 01:10:45,086
able to say in terms of that.

1588
01:10:45,086 --> 01:10:45,919
And then we looked again.

1589
01:10:45,919 --> 01:10:46,916
This is just a healthy control.

1590
01:10:46,916 --> 01:10:49,276
We just said, are there
any relationships here

1591
01:10:49,276 --> 01:10:50,206
with sex or age?

1592
01:10:50,206 --> 01:10:51,206
We saw nothing with sex,

1593
01:10:51,206 --> 01:10:52,686
but we saw these relationship

1594
01:10:52,686 --> 01:10:54,626
in these four networks with age,

1595
01:10:54,626 --> 01:10:56,166
suggesting that we might have a pattern

1596
01:10:56,166 --> 01:10:57,786
that would be valuable there.

1597
01:10:57,786 --> 01:10:59,806
And here we're bothering
to reconvert the data

1598
01:10:59,806 --> 01:11:03,276
back into VT units so we know
the magnitude of the effects

1599
01:11:03,276 --> 01:11:06,086
in terms of what the
synapses that might be there.

1600
01:11:06,086 --> 01:11:08,236
So overall, we identified a total of 13

1601
01:11:08,236 --> 01:11:11,586
that were reliably obtained
from our UCB-J data.

1602
01:11:11,586 --> 01:11:14,896
There was some spatial
consistency, more with FDG perhaps

1603
01:11:14,896 --> 01:11:16,846
than what was seen with
the resting states.

1604
01:11:16,846 --> 01:11:19,866
And again, we noticed this effect on age.

1605
01:11:19,866 --> 01:11:21,766
So now we wanted to take
this to the next step.

1606
01:11:21,766 --> 01:11:24,136
And kind of that next step is to combine,

1607
01:11:24,136 --> 01:11:25,616
to look in the same subject where we have

1608
01:11:25,616 --> 01:11:28,516
SV2A PET and resting state fMRI.

1609
01:11:28,516 --> 01:11:31,066
When we combine those
together, will that work?

1610
01:11:31,066 --> 01:11:33,046
As I mentioned, there was
some similarity in sources,

1611
01:11:33,046 --> 01:11:35,186
but how do these tie together in some way?

1612
01:11:35,186 --> 01:11:38,496
Well, we realized we
didn't wanna open this

1613
01:11:38,496 --> 01:11:39,946
up to a full fishing expedition.

1614
01:11:39,946 --> 01:11:41,546
So we were really focusing in one area,

1615
01:11:41,546 --> 01:11:42,676
looking at the relationships

1616
01:11:42,676 --> 01:11:44,596
between the striatum and the cortex

1617
01:11:44,596 --> 01:11:47,253
based on years of
preclinical studies there

1618
01:11:47,253 --> 01:11:48,776
and tractography studies there,

1619
01:11:48,776 --> 01:11:51,536
being able to show that those
relationships should be there.

1620
01:11:51,536 --> 01:11:52,736
So that's what we wanted to focus on.

1621
01:11:52,736 --> 01:11:55,106
We wanted to look at
resting state networks

1622
01:11:55,106 --> 01:11:56,146
in the same subject

1623
01:11:56,146 --> 01:11:59,186
compared to the synaptic density
measures that we could get.

1624
01:11:59,186 --> 01:12:01,506
How do those relate to each other

1625
01:12:01,506 --> 01:12:04,556
and focusing on corticostriatal regions?

1626
01:12:04,556 --> 01:12:05,996
So we have 35 data sets,

1627
01:12:05,996 --> 01:12:08,096
all in cognitively
normal, healthy controls,

1628
01:12:08,096 --> 01:12:10,196
again, with the VT images there.

1629
01:12:10,196 --> 01:12:12,716
We had a short fMRI
sequence, 10 minutes there.

1630
01:12:12,716 --> 01:12:13,946
Everything was motion-corrected

1631
01:12:13,946 --> 01:12:16,686
and brought into standard template space.

1632
01:12:16,686 --> 01:12:19,856
So for the fMRI, we were able
to estimate 30 components.

1633
01:12:19,856 --> 01:12:21,736
And the measure that
we're taking it out is

1634
01:12:21,736 --> 01:12:22,826
the fractional amplitude

1635
01:12:22,826 --> 01:12:25,276
of low frequency fluctuations or fALFF.

1636
01:12:25,276 --> 01:12:26,306
Again, that's looking at,

1637
01:12:26,306 --> 01:12:29,416
when we look at the Fourier
transform of the data,

1638
01:12:29,416 --> 01:12:31,236
what is the low frequency components

1639
01:12:31,236 --> 01:12:32,866
as normalized to the high frequency,

1640
01:12:32,866 --> 01:12:34,146
trying to look at just a reflection

1641
01:12:34,146 --> 01:12:38,166
that's been used commonly as a
measure of neuronal activity.

1642
01:12:38,166 --> 01:12:41,376
In UCB-J we, as again,
went with our 18 components

1643
01:12:41,376 --> 01:12:42,806
as a starting point to be able to do that,

1644
01:12:42,806 --> 01:12:45,626
but we're gonna focus
down on a few networks.

1645
01:12:45,626 --> 01:12:47,266
And remember, keeping something in mind

1646
01:12:47,266 --> 01:12:50,366
that the SV2A measures,
we said that the fMRI

1647
01:12:50,366 --> 01:12:53,556
is gonna be more functional,
more flow related going on.

1648
01:12:53,556 --> 01:12:55,276
UCB-J with the VT is gonna be

1649
01:12:55,276 --> 01:12:57,916
more of that static synaptic measure.

1650
01:12:57,916 --> 01:13:00,984
So of the fMRI, we really felt like 17

1651
01:13:00,984 --> 01:13:04,276
were visually consistent
with resting state networks.

1652
01:13:04,276 --> 01:13:05,216
We're gonna focus on these.

1653
01:13:05,216 --> 01:13:06,986
We have two of the dorsal,

1654
01:13:06,986 --> 01:13:09,436
the default-mode networks and
the anterior and posterior,

1655
01:13:09,436 --> 01:13:11,536
a sensory motor network, and two others,

1656
01:13:11,536 --> 01:13:14,586
the executive control
and a salience network

1657
01:13:14,586 --> 01:13:16,026
that were identified that came out.

1658
01:13:16,026 --> 01:13:18,946
In the UCB-J of those, we're
gonna focus on the ones

1659
01:13:18,946 --> 01:13:20,736
that are relevant for corticostriatal,

1660
01:13:20,736 --> 01:13:23,106
looking at one that is primarily striatal.

1661
01:13:23,106 --> 01:13:24,066
So we display that.

1662
01:13:24,066 --> 01:13:25,786
And these two are medial prefrontal

1663
01:13:25,786 --> 01:13:29,076
and medial parietal region
there, these two networks,

1664
01:13:29,076 --> 01:13:32,176
and much more needing to be
done with all of this stuff.

1665
01:13:32,176 --> 01:13:34,236
So now what I wanna show
you is when we relate

1666
01:13:34,236 --> 01:13:36,496
the loading of a subject

1667
01:13:36,496 --> 01:13:40,436
from their PET SV2A to, in this case,

1668
01:13:40,436 --> 01:13:42,116
to their measure, to their fALFF measure

1669
01:13:42,116 --> 01:13:45,855
coming out associated with that region.

1670
01:13:45,855 --> 01:13:48,846
So if I look here first,
we have a striatal network.

1671
01:13:48,846 --> 01:13:50,816
We're seeing some correlations there

1672
01:13:50,816 --> 01:13:54,266
when we look at the default-mode
networks across that.

1673
01:13:54,266 --> 01:13:57,816
The more interesting one is in
our medial prefrontal region.

1674
01:13:57,816 --> 01:13:58,649
And that we see in ProTec

1675
01:13:58,649 --> 01:14:00,036
we're seeing large correlations

1676
01:14:00,036 --> 01:14:02,886
across the loadings of multiple networks.

1677
01:14:02,886 --> 01:14:07,176
So here's an example of
the fALFF measurement

1678
01:14:07,176 --> 01:14:09,016
coming out from these subjects in that,

1679
01:14:09,016 --> 01:14:11,496
according to that network correlating that

1680
01:14:11,496 --> 01:14:13,216
within the medial frontal
loading that we see,

1681
01:14:13,216 --> 01:14:15,936
and we get this very
significant correlation here.

1682
01:14:15,936 --> 01:14:18,336
Interestingly, when we looked
at the medial parietal source,

1683
01:14:18,336 --> 01:14:20,786
we really didn't show any
of that relationship at all,

1684
01:14:20,786 --> 01:14:22,306
which was a little bit surprising

1685
01:14:22,306 --> 01:14:24,096
given the relationship we have

1686
01:14:24,096 --> 01:14:27,566
with the posterior dorsal
default-mode network.

1687
01:14:27,566 --> 01:14:31,466
So a lot of really early
kind of discussions

1688
01:14:31,466 --> 01:14:33,309
or thoughts going on.

1689
01:14:36,846 --> 01:14:39,056
We certainly saw that spatially,

1690
01:14:39,056 --> 01:14:41,606
we saw a couple of the
sources that looked similar.

1691
01:14:41,606 --> 01:14:43,876
And in fact, we saw that
the medial prefrontal

1692
01:14:43,876 --> 01:14:46,126
and the UCB-J source correlated nicely

1693
01:14:46,126 --> 01:14:48,236
with the resting state
default-mode network,

1694
01:14:48,236 --> 01:14:49,069
the anterior one,

1695
01:14:49,069 --> 01:14:51,066
but we did not see that
with the posterior.

1696
01:14:51,066 --> 01:14:51,976
In the cortical world,

1697
01:14:51,976 --> 01:14:53,636
we really did not see associations

1698
01:14:53,636 --> 01:14:56,066
with the striatal and the
sensory default network,

1699
01:14:56,066 --> 01:14:57,696
although we did see that
with the default-mode

1700
01:14:57,696 --> 01:14:59,206
and the salience network.

1701
01:14:59,206 --> 01:15:00,966
And really this is interesting

1702
01:15:00,966 --> 01:15:02,886
that this medial prefrontal
is the strongest one,

1703
01:15:02,886 --> 01:15:05,866
correlating really across
all the functional domains.

1704
01:15:05,866 --> 01:15:07,806
So we're just getting started with this.

1705
01:15:07,806 --> 01:15:11,336
We recognize that maybe there
should be differences here.

1706
01:15:11,336 --> 01:15:15,056
If our UCB-J measure is a more
static measure of synapses,

1707
01:15:15,056 --> 01:15:16,116
is that gonna be reflecting

1708
01:15:16,116 --> 01:15:17,783
the current state of those patients?

1709
01:15:17,783 --> 01:15:19,489
And so we're looking
more at that flow map,

1710
01:15:19,489 --> 01:15:22,956
that K1 map or VT map
as well that's going on

1711
01:15:22,956 --> 01:15:23,946
and really thinking about this

1712
01:15:23,946 --> 01:15:25,996
in terms of some network activities.

1713
01:15:25,996 --> 01:15:27,636
So finally, we really wanted
to be able to apply this.

1714
01:15:27,636 --> 01:15:30,166
So if we can take this
into the Alzheimer's world

1715
01:15:30,166 --> 01:15:31,396
or into some patient world,

1716
01:15:31,396 --> 01:15:33,676
if we can extract another better measure,

1717
01:15:33,676 --> 01:15:35,846
especially in a data-driven fashion

1718
01:15:35,846 --> 01:15:37,146
that we've got right here.

1719
01:15:39,176 --> 01:15:41,106
We've now done in a small cohort,

1720
01:15:41,106 --> 01:15:42,986
we take a subset of our AD subjects

1721
01:15:42,986 --> 01:15:46,106
where we're trying to now
go back to their SV2A data.

1722
01:15:46,106 --> 01:15:47,816
We haven't done this with the fMRI yet.

1723
01:15:47,816 --> 01:15:49,466
And say, okay, let's apply ICA.

1724
01:15:49,466 --> 01:15:50,496
Let's see what it extracts.

1725
01:15:50,496 --> 01:15:52,436
And again, remember,
this is a data-driven.

1726
01:15:52,436 --> 01:15:55,436
We're throwing in our cognitively
normals, our MCI and AD,

1727
01:15:55,436 --> 01:15:58,376
and let them find the patterns
and those loadings. Okay.

1728
01:15:58,376 --> 01:16:01,436
Again, we did that with SV2A
coming out with 18 measures.

1729
01:16:01,436 --> 01:16:02,396
And now the question was,

1730
01:16:02,396 --> 01:16:04,886
do the loadings tell us
something about the disease?

1731
01:16:04,886 --> 01:16:06,896
'Cause of course the ICA is not told

1732
01:16:06,896 --> 01:16:09,116
about what population is which.

1733
01:16:09,116 --> 01:16:10,666
And so we found that five patterns

1734
01:16:10,666 --> 01:16:11,856
if we looked at the loading,

1735
01:16:11,856 --> 01:16:15,266
so here we're plotting for
five of these components,

1736
01:16:15,266 --> 01:16:17,566
we looked at the loading
in the cognitively normal,

1737
01:16:17,566 --> 01:16:19,966
in the MCI and in the AD.

1738
01:16:19,966 --> 01:16:21,716
And these were the five components

1739
01:16:21,716 --> 01:16:23,776
that showed statistically
significant group differences,

1740
01:16:23,776 --> 01:16:26,656
at least between two of the groups.

1741
01:16:26,656 --> 01:16:28,886
And we have the regions that we're seeing

1742
01:16:28,886 --> 01:16:29,966
are perhaps not surprising.

1743
01:16:29,966 --> 01:16:31,356
We're seeing a lot of temporal lobe,

1744
01:16:31,356 --> 01:16:33,116
a lot of frontal lobe regions

1745
01:16:33,116 --> 01:16:34,766
and a little bit in occipital as well,

1746
01:16:34,766 --> 01:16:36,206
which we think actually may be reflecting

1747
01:16:36,206 --> 01:16:38,916
more of a normalization pattern there.

1748
01:16:38,916 --> 01:16:40,436
So how does that relate

1749
01:16:42,750 --> 01:16:43,666
when we look at that compared

1750
01:16:43,666 --> 01:16:45,336
to some of the cognitive scores?

1751
01:16:45,336 --> 01:16:46,504
And now of course, we're just looking

1752
01:16:46,504 --> 01:16:49,736
within the MCI and AD populations.

1753
01:16:49,736 --> 01:16:51,586
And now we're seeing some
really interesting results.

1754
01:16:51,586 --> 01:16:54,056
The first pattern in
particular is very strong,

1755
01:16:54,056 --> 01:16:56,996
and being able to separate
out and get a nice correlation

1756
01:16:56,996 --> 01:16:59,806
with a number, with the
various sums of boxes,

1757
01:16:59,806 --> 01:17:01,176
Mini-Mental scores, et cetera.

1758
01:17:01,176 --> 01:17:02,936
So we're seeing that we can extract that

1759
01:17:02,936 --> 01:17:04,916
in cognitively normal ways across that.

1760
01:17:04,916 --> 01:17:05,749
So this one is there.

1761
01:17:05,749 --> 01:17:06,826
And again, we're not using

1762
01:17:06,826 --> 01:17:08,266
including the cognitively normals.

1763
01:17:08,266 --> 01:17:10,616
We're just doing this within
the patient populations,

1764
01:17:10,616 --> 01:17:12,266
similar to what we saw

1765
01:17:12,266 --> 01:17:15,986
with when we used the full
large components there,

1766
01:17:15,986 --> 01:17:17,026
the large average,

1767
01:17:17,026 --> 01:17:19,526
but here doing this in
a data-independent way.

1768
01:17:19,526 --> 01:17:21,366
If we look at component two,

1769
01:17:21,366 --> 01:17:23,206
a little less convincing here,

1770
01:17:23,206 --> 01:17:26,106
but on the border of being
able to see correlations

1771
01:17:26,106 --> 01:17:27,536
of some of these measures there,

1772
01:17:27,536 --> 01:17:29,766
and just trying to begin to
understand to what extent

1773
01:17:29,766 --> 01:17:31,786
the regional pattern were related

1774
01:17:31,786 --> 01:17:35,596
to the specific cognitive
domain that is in deficit.

1775
01:17:35,596 --> 01:17:38,376
So we've been excited with this ICA.

1776
01:17:38,376 --> 01:17:40,176
It's interesting how it will separate

1777
01:17:40,176 --> 01:17:42,636
out these source networks,
again, with no a priori knowledge

1778
01:17:42,636 --> 01:17:44,366
of the subject characteristics.

1779
01:17:44,366 --> 01:17:46,316
And we certainly found
again without telling it,

1780
01:17:46,316 --> 01:17:47,496
that we had certain components

1781
01:17:47,496 --> 01:17:49,886
that significantly differentiate the AD,

1782
01:17:49,886 --> 01:17:53,076
and that we found particular
sums that separate MCI and AD

1783
01:17:53,076 --> 01:17:54,826
with their cognitive measures.

1784
01:17:54,826 --> 01:17:56,806
And so we're still left with
a lot of questions here.

1785
01:17:56,806 --> 01:18:00,146
And just one of them is
just, will this method

1786
01:18:00,146 --> 01:18:01,236
give us more sensitivity

1787
01:18:01,236 --> 01:18:03,966
to be able to both follow
patients over time,

1788
01:18:03,966 --> 01:18:07,356
differentiate patients
or look at drug effects?

1789
01:18:07,356 --> 01:18:08,916
So let me give an overall summary

1790
01:18:08,916 --> 01:18:11,136
of the issues I've talked about here today

1791
01:18:11,136 --> 01:18:13,499
with our SV2A synaptic marker.

1792
01:18:14,386 --> 01:18:16,556
We believe and although
there's obviously more

1793
01:18:16,556 --> 01:18:18,116
that needs to be done, that this provides

1794
01:18:18,116 --> 01:18:20,546
a really good general imaging biomarker

1795
01:18:20,546 --> 01:18:22,056
for synaptic density.

1796
01:18:22,056 --> 01:18:23,936
We are looking at a vesicular protein,

1797
01:18:23,936 --> 01:18:25,146
and that's been used traditionally

1798
01:18:25,146 --> 01:18:27,386
to be able to assess
that in postmortem data.

1799
01:18:27,386 --> 01:18:30,756
And so far the data are
consistent and supportive of that.

1800
01:18:30,756 --> 01:18:33,076
The tracer that I've shown
you here is an excellent one.

1801
01:18:33,076 --> 01:18:34,546
It has very high test-retest,

1802
01:18:34,546 --> 01:18:36,316
very high-quality measurements.

1803
01:18:36,316 --> 01:18:38,596
It really lets us look
in very small regions.

1804
01:18:38,596 --> 01:18:40,506
Of course, carbon 11 has
a 20 minute half-life.

1805
01:18:40,506 --> 01:18:41,686
That's not ideal.

1806
01:18:41,686 --> 01:18:44,976
We now have versions
of that, F18-SynVesT-1

1807
01:18:44,976 --> 01:18:47,826
that is basically identical in
terms of its characteristics,

1808
01:18:47,826 --> 01:18:49,166
very similar structure,

1809
01:18:49,166 --> 01:18:50,986
and it's something that
is gonna be very useful

1810
01:18:50,986 --> 01:18:52,746
in multicenter trials.

1811
01:18:52,746 --> 01:18:54,826
When we've applied that in
our patient populations,

1812
01:18:54,826 --> 01:18:56,446
I've shown you Alzheimer's,

1813
01:18:56,446 --> 01:18:58,446
depression and schizophrenia today,

1814
01:18:58,446 --> 01:19:00,836
we are seeing significant
results clinically

1815
01:19:00,836 --> 01:19:02,756
and basically biologically
relevant results.

1816
01:19:02,756 --> 01:19:05,136
I have not shown you our other populations

1817
01:19:05,136 --> 01:19:07,856
in Parkinson's, HIV, epilepsy
that we've already published,

1818
01:19:07,856 --> 01:19:11,696
ongoing work, even work in
development using a monkey model,

1819
01:19:11,696 --> 01:19:13,556
and then all sorts of small animal models

1820
01:19:13,556 --> 01:19:16,496
of these different
disorders, which are ongoing.

1821
01:19:16,496 --> 01:19:19,216
Overall, in the neurological disorders

1822
01:19:19,216 --> 01:19:21,516
we see the more focal losses.

1823
01:19:21,516 --> 01:19:24,906
We see it in the hippocampus in AD.

1824
01:19:24,906 --> 01:19:27,136
In the Parkinson's, we see it most clearly

1825
01:19:27,136 --> 01:19:28,176
in the substantia nigra,

1826
01:19:28,176 --> 01:19:30,626
and epilepsy in the site of the focus.

1827
01:19:30,626 --> 01:19:33,186
In the psychiatric diseases, in depression

1828
01:19:33,186 --> 01:19:34,496
in MDD and other areas like that,

1829
01:19:34,496 --> 01:19:37,956
we're seeing that as more
widespread of disorders.

1830
01:19:37,956 --> 01:19:39,556
And then I think what's
really interesting is

1831
01:19:39,556 --> 01:19:41,673
if we combine these measures together,

1832
01:19:41,673 --> 01:19:44,836
what's that gonna open in
terms of unique analytic tools,

1833
01:19:44,836 --> 01:19:46,386
combining what we can do in MR.

1834
01:19:47,456 --> 01:19:49,376
Looking just within the SV2A networks,

1835
01:19:49,376 --> 01:19:50,506
we saw the age effects.

1836
01:19:50,506 --> 01:19:53,546
We saw the ability to separate
that in Alzheimer's disease,

1837
01:19:53,546 --> 01:19:56,066
or even really multi-parametric
where we can be saying,

1838
01:19:56,066 --> 01:19:58,076
okay, I wanna look within the two measures

1839
01:19:58,076 --> 01:19:59,566
of volume of distribution,

1840
01:19:59,566 --> 01:20:02,026
a measure of the synaptic concentration

1841
01:20:02,026 --> 01:20:02,859
compared to the blood flow measure,

1842
01:20:02,859 --> 01:20:04,746
that that is very useful.

1843
01:20:04,746 --> 01:20:06,696
It's also exciting to
do these multi-tracers.

1844
01:20:06,696 --> 01:20:08,756
We looked at tau and FDG in AD.

1845
01:20:08,756 --> 01:20:11,996
We looked at our mGluR5
measure in depression,

1846
01:20:11,996 --> 01:20:14,516
and being able to learn more
about how we can combine that.

1847
01:20:14,516 --> 01:20:17,526
Obviously, all the more and
more combinations with fMRI,

1848
01:20:17,526 --> 01:20:18,846
how we do it's important

1849
01:20:18,846 --> 01:20:21,496
because, with this much data,
we're gonna find things.

1850
01:20:21,496 --> 01:20:24,156
How to do that in a reproducible
way will be fascinating.

1851
01:20:24,156 --> 01:20:25,616
And for me, a lot of anything

1852
01:20:25,616 --> 01:20:27,986
where we're showing correlations
with cognitive measures

1853
01:20:27,986 --> 01:20:30,436
suggests that we're on the right track,

1854
01:20:30,436 --> 01:20:32,296
being able to do that.

1855
01:20:32,296 --> 01:20:35,726
So finally, a lot of people to thank.

1856
01:20:35,726 --> 01:20:38,986
In the picture in the back is
the people in our PET center

1857
01:20:38,986 --> 01:20:41,396
back in pre-COVID times.

1858
01:20:41,396 --> 01:20:44,086
So many people contributing
to the work here.

1859
01:20:44,086 --> 01:20:46,076
I've highlighted in yellow

1860
01:20:46,076 --> 01:20:47,596
the people who have contributed slides

1861
01:20:47,596 --> 01:20:48,546
that I've shown today.

1862
01:20:48,546 --> 01:20:52,246
Many, many others involved in
different aspects of our SV2A.

1863
01:20:52,246 --> 01:20:54,296
Lots of wonderful support from the NIH.

1864
01:20:54,296 --> 01:20:56,316
Thank you, and from other foundations.

1865
01:20:56,316 --> 01:20:58,799
And I thank you for your attention.

1866
01:21:03,109 --> 01:21:04,609
- Hi, I'm Audrey Fan from

1867
01:21:04,609 --> 01:21:06,649
the University of California Davis,

1868
01:21:06,649 --> 01:21:09,229
and I'm in the Departments
of Biomedical Engineering

1869
01:21:09,229 --> 01:21:11,149
and Neurology here.

1870
01:21:11,149 --> 01:21:14,369
I'm delighted to participate
in the NIMH workshop

1871
01:21:14,369 --> 01:21:17,349
on PET fMRI and I'm hoping this talk

1872
01:21:17,349 --> 01:21:19,829
will provide a complimentary perspective

1873
01:21:19,829 --> 01:21:22,299
to some of the other talks you've seen

1874
01:21:22,299 --> 01:21:23,909
with the focus here specifically

1875
01:21:23,909 --> 01:21:28,909
on using PET MRI to
study brain physiology,

1876
01:21:29,259 --> 01:21:32,179
biomarkers and brain vascular health.

1877
01:21:32,179 --> 01:21:35,039
I'm really interested
in this topic because

1878
01:21:35,039 --> 01:21:37,969
the brain is so uniquely
metabolically demanding.

1879
01:21:37,969 --> 01:21:40,719
It's only two to 5% of our body mass,

1880
01:21:40,719 --> 01:21:45,199
but the brain takes up about
20% of our oxygen supply.

1881
01:21:45,199 --> 01:21:46,289
And because our brain tissues

1882
01:21:46,289 --> 01:21:49,469
cannot store oxygen as a reserve,

1883
01:21:49,469 --> 01:21:54,469
it relies on constant
circulation delivery of oxygen

1884
01:21:54,569 --> 01:21:56,929
and essential nutrients through

1885
01:21:56,929 --> 01:21:59,259
the cerebral blood flow, CBF,

1886
01:21:59,259 --> 01:22:02,669
on the flip side of the coin,
oxygen extraction fraction.

1887
01:22:02,669 --> 01:22:06,459
So the ability of
tissues to extract oxygen

1888
01:22:06,459 --> 01:22:09,129
to meet its metabolic needs,

1889
01:22:09,129 --> 01:22:12,529
and these vascular or
physiological parameters

1890
01:22:12,529 --> 01:22:17,529
may be familiar to you as
parts of the FMRI BOLD signal,

1891
01:22:17,739 --> 01:22:21,539
but are critical for
normal brain function.

1892
01:22:21,539 --> 01:22:22,469
And unfortunately,

1893
01:22:22,469 --> 01:22:25,249
when this circulation
and supply is disrupted,

1894
01:22:25,249 --> 01:22:28,159
it can lead to devastating consequences.

1895
01:22:28,159 --> 01:22:29,429
Most dramatically,

1896
01:22:29,429 --> 01:22:32,852
as many of us are aware
in acute ischemic stroke,

1897
01:22:35,688 --> 01:22:37,679
in acute stroke for many decades,

1898
01:22:37,679 --> 01:22:41,929
the standard of care has
been relatively crude.

1899
01:22:41,929 --> 01:22:45,079
And what I mean by that
is it's been primarily

1900
01:22:45,079 --> 01:22:46,899
based on the time window.

1901
01:22:46,899 --> 01:22:51,069
So if you have arrived to the hospital

1902
01:22:51,069 --> 01:22:54,819
with symptoms of acute
stroke, within 4.5 hours,

1903
01:22:54,819 --> 01:22:57,399
then you are eligible
for standard treatment

1904
01:22:57,399 --> 01:23:00,539
called TPA or clot buster drugs.

1905
01:23:00,539 --> 01:23:04,849
And beyond that, the risk
value benefit changes

1906
01:23:04,849 --> 01:23:06,519
so that you might have,

1907
01:23:06,519 --> 01:23:10,189
overall more risk of hemorrhage

1908
01:23:10,189 --> 01:23:11,669
and you're ineligible, right?

1909
01:23:11,669 --> 01:23:15,089
But what this means is you
actually miss a lot of patients

1910
01:23:15,089 --> 01:23:18,759
who perhaps don't know when
their symptoms started say,

1911
01:23:18,759 --> 01:23:21,339
you woke up with a stroke.

1912
01:23:21,339 --> 01:23:25,699
This would miss the opportunity
to help many patients.

1913
01:23:25,699 --> 01:23:28,819
Instead there's a complimentary framework,

1914
01:23:28,819 --> 01:23:33,449
which actually focuses more
on exactly the physiological

1915
01:23:33,449 --> 01:23:36,779
mechanisms that are happening
during this pathophysiology.

1916
01:23:36,779 --> 01:23:38,359
So in stroke,

1917
01:23:38,359 --> 01:23:40,239
it's well known for a long time

1918
01:23:40,239 --> 01:23:43,249
that there is an area of infarct,

1919
01:23:43,249 --> 01:23:46,039
this ischemic core of tissue in the brain

1920
01:23:46,039 --> 01:23:49,849
that's destined to die
because a lack of blood flow,

1921
01:23:49,849 --> 01:23:53,359
but that there is also this area of tissue

1922
01:23:53,359 --> 01:23:57,119
in some patients around
the core called penumbra.

1923
01:23:57,119 --> 01:24:00,179
And this penumbra is vulnerable.

1924
01:24:00,179 --> 01:24:03,482
at risk, but potentially saveable tissue.

1925
01:24:04,569 --> 01:24:09,569
And so one of the keys is
could we identify patients

1926
01:24:09,739 --> 01:24:12,439
who have penumbra and could benefit

1927
01:24:12,439 --> 01:24:15,499
from stroke treatments like TBA,

1928
01:24:15,499 --> 01:24:18,239
even beyond the 4.5 hours.

1929
01:24:18,239 --> 01:24:20,609
And I think from basic science histology

1930
01:24:20,609 --> 01:24:22,379
and preclinical studies,

1931
01:24:22,379 --> 01:24:25,614
there is a physiological basis for this.

1932
01:24:25,614 --> 01:24:28,349
So what I'm showing
here is a representation

1933
01:24:28,349 --> 01:24:32,249
of physiological changes as
we reduce the blood flow,

1934
01:24:32,249 --> 01:24:34,569
emulating what happens in the stroke.

1935
01:24:34,569 --> 01:24:37,499
And certainly there are neuronal features

1936
01:24:37,499 --> 01:24:38,759
that could be detected,

1937
01:24:38,759 --> 01:24:43,249
but physiologically some of the
hallmarks of this vulnerable

1938
01:24:43,249 --> 01:24:47,389
or at risk tissue in the
penumbra is characterized

1939
01:24:47,389 --> 01:24:50,729
by reduced cerebro blood
flow, not surprisingly,

1940
01:24:50,729 --> 01:24:55,179
but a compensatory OEF oxygen
extraction that's elevated.

1941
01:24:55,179 --> 01:24:58,109
The tissue is trying
to extract more oxygen

1942
01:24:58,109 --> 01:25:01,809
to compensate for the
lack that it's receiving.

1943
01:25:01,809 --> 01:25:06,039
And so I think this is
really a shift in how we're

1944
01:25:06,039 --> 01:25:09,399
identifying patients that could benefit

1945
01:25:09,399 --> 01:25:11,189
from treatment and stroke.

1946
01:25:11,189 --> 01:25:13,779
It was the basis of multiple studies in

1947
01:25:13,779 --> 01:25:14,839
the last five years.

1948
01:25:14,839 --> 01:25:18,909
One of them led by Stanford
called the DEFUSE-3 trial.

1949
01:25:18,909 --> 01:25:22,209
And what you're seeing here
are example MRI images,

1950
01:25:22,209 --> 01:25:24,589
where the purple represents diffusion

1951
01:25:24,589 --> 01:25:27,879
MR segmentations of this ischemic core.

1952
01:25:27,879 --> 01:25:31,839
Again, tissue destined to
die because of the stroke,

1953
01:25:31,839 --> 01:25:35,529
but also a complimentary hypoperfusion.

1954
01:25:35,529 --> 01:25:39,729
So a contrast based dynamic
susceptibility MR scan,

1955
01:25:39,729 --> 01:25:43,229
where the green indicates
area that's been affected

1956
01:25:43,229 --> 01:25:46,249
with reduced perfusion,
relatively speaking,

1957
01:25:46,249 --> 01:25:49,759
and in this trial imaging
played a huge role.

1958
01:25:49,759 --> 01:25:54,119
It was this mismatch
involved metric delineation

1959
01:25:54,119 --> 01:25:57,249
between these two regions that suspected

1960
01:25:57,249 --> 01:26:00,819
there likely is an area of tissue adjacent

1961
01:26:00,819 --> 01:26:03,599
to the core potentially saveable.

1962
01:26:03,599 --> 01:26:08,599
And in these trials that used
imaging to identify patients,

1963
01:26:08,789 --> 01:26:10,029
the patients who actually

1964
01:26:10,029 --> 01:26:13,939
had potential penumbra benefited from TPA,

1965
01:26:13,939 --> 01:26:16,559
or even newer thrombectomy treatments,

1966
01:26:16,559 --> 01:26:20,829
way beyond the 4.5 hours
up to 16 to 24 hours

1967
01:26:20,829 --> 01:26:24,089
in other trials where
we're reaching patients

1968
01:26:24,089 --> 01:26:27,243
seen based on their physiological state

1969
01:26:27,243 --> 01:26:31,549
and not just a standard time
window that we can help them.

1970
01:26:31,549 --> 01:26:33,729
And so I think this is
a very exciting time.

1971
01:26:33,729 --> 01:26:36,059
where imaging specifically

1972
01:26:36,059 --> 01:26:38,829
physiological biomarkers plays a big role,

1973
01:26:38,829 --> 01:26:40,839
and we can do even better than this

1974
01:26:40,839 --> 01:26:42,899
because when you look at this,

1975
01:26:42,899 --> 01:26:47,629
actually the hypoperfusion is
based on contrast injection

1976
01:26:47,629 --> 01:26:52,149
and also requires difficulty
in quantification.

1977
01:26:52,149 --> 01:26:55,269
And so usually this is a relative marker.

1978
01:26:55,269 --> 01:26:59,999
I think where PET and
MR both could actually

1979
01:26:59,999 --> 01:27:03,179
be pushed is to provide quantitative

1980
01:27:03,179 --> 01:27:07,199
imaging of these biomarkers
across the whole span

1981
01:27:07,199 --> 01:27:11,352
of oxygenation, blood flow and metabolism.

1982
01:27:12,191 --> 01:27:15,429
This has traditionally
been in the purview of PET

1983
01:27:15,429 --> 01:27:20,429
because it has established
kinetic modeling models.

1984
01:27:20,809 --> 01:27:24,279
It has been around for a long
time, but as we all know,

1985
01:27:24,279 --> 01:27:28,329
it has some complications
with radioactive tracers,

1986
01:27:28,329 --> 01:27:32,669
the need for arterial blood
sampling for quantification,

1987
01:27:32,669 --> 01:27:35,069
and specifically for some
of the biomarkers that

1988
01:27:36,449 --> 01:27:41,032
I'm laying out here
requires O-15 label tracers,

1989
01:27:42,078 --> 01:27:44,239
which only have a half-life of two minutes

1990
01:27:44,239 --> 01:27:48,452
and is quite prohibitive in
terms of requiring a cyclotron.

1991
01:27:49,499 --> 01:27:53,919
And so I think there's
complementary benefits

1992
01:27:53,919 --> 01:27:58,049
of combining PET quantification with MR,

1993
01:27:58,049 --> 01:27:59,879
especially some new sequences

1994
01:27:59,879 --> 01:28:02,319
that can directly assess cerebro

1995
01:28:02,319 --> 01:28:06,349
blood flow oxygenation
without ionizing radiation

1996
01:28:06,349 --> 01:28:10,449
and reduce some of
these barriers of set up

1997
01:28:10,449 --> 01:28:12,309
and cyclotron needs,

1998
01:28:12,309 --> 01:28:15,549
at the same time providing
accurate quantification.

1999
01:28:15,549 --> 01:28:19,491
And so this is the
synergy that we're hoping

2000
01:28:19,491 --> 01:28:21,679
to leverage with PET and MR going forward

2001
01:28:22,554 --> 01:28:23,699
for not only stroke,

2002
01:28:23,699 --> 01:28:27,719
but a broad set of
neurological and psychiatric

2003
01:28:27,719 --> 01:28:30,984
conditions that potentially have systemic

2004
01:28:30,984 --> 01:28:35,199
and brain focused
physiological disruptions.

2005
01:28:35,199 --> 01:28:36,519
So with that overview,

2006
01:28:36,519 --> 01:28:40,492
I hope these biomarkers
can be quite broadly used.

2007
01:28:41,468 --> 01:28:43,049
And I wanna with the rest of the talk,

2008
01:28:43,049 --> 01:28:45,959
indicate how with these examples

2009
01:28:47,741 --> 01:28:49,891
fully simultaneous PET
and MR have actually

2010
01:28:50,789 --> 01:28:52,149
added to our knowledge

2011
01:28:52,149 --> 01:28:54,862
of brain physiology and how to measure it.

2012
01:28:55,729 --> 01:28:58,999
So specifically I wanna
cover how we've used

2013
01:29:00,166 --> 01:29:02,245
(indistinct) been labeling.

2014
01:29:02,245 --> 01:29:05,709
So an MRI-based measure of brain perfusion

2015
01:29:05,709 --> 01:29:07,749
and used PET MR to validate it,

2016
01:29:07,749 --> 01:29:11,289
including during a stress test.

2017
01:29:11,289 --> 01:29:15,489
I wanna then highlight
how further quantification

2018
01:29:15,489 --> 01:29:18,659
can be achieved in a way
that's less invasive,

2019
01:29:18,659 --> 01:29:21,032
but still accurate using PET MR.

2020
01:29:22,569 --> 01:29:27,269
And beyond that, what
are other applications

2021
01:29:27,269 --> 01:29:29,819
that are not specifically cerebrovascular,

2022
01:29:29,819 --> 01:29:33,119
but target some of the
physiological contributions

2023
01:29:33,119 --> 01:29:37,756
to neurological conditions
and other biomarkers

2024
01:29:37,756 --> 01:29:39,699
that would benefit in the future

2025
01:29:39,699 --> 01:29:41,622
from this PET MR technology.

2026
01:29:46,259 --> 01:29:49,578
Let's move to this first section

2027
01:29:49,578 --> 01:29:52,609
where we will leverage
PET MR simultaneously

2028
01:29:52,609 --> 01:29:55,839
to validate brain profusion biomarkers.

2029
01:29:55,839 --> 01:29:57,059
And the case study I wanted

2030
01:29:57,059 --> 01:30:01,069
to show here is of intracranial stenosis.

2031
01:30:01,069 --> 01:30:05,289
And specifically of patients
with Moyamoya disease,

2032
01:30:05,289 --> 01:30:07,339
Moyamoya disease is a relatively rare,

2033
01:30:07,339 --> 01:30:09,219
but progressive disorder.

2034
01:30:09,219 --> 01:30:11,119
There are some genetic components to it,

2035
01:30:11,119 --> 01:30:14,409
but we don't really
understand why it happens.

2036
01:30:14,409 --> 01:30:16,639
Arteries of the base of the brain begin

2037
01:30:17,565 --> 01:30:21,149
to over time become stenosed or occluded.

2038
01:30:21,149 --> 01:30:22,649
And these patients tend

2039
01:30:22,649 --> 01:30:26,399
to have sevenfold
increased risk of stroke.

2040
01:30:26,399 --> 01:30:30,369
And it's really important
to capture these brain

2041
01:30:30,369 --> 01:30:33,399
physiological markers
in patients to decide

2042
01:30:33,399 --> 01:30:36,979
who might benefit from
surgical treatments.

2043
01:30:36,979 --> 01:30:41,979
An example of these moyamoya
cases as shown here is

2044
01:30:43,959 --> 01:30:47,849
highlighted in these
catheter based angiograms,

2045
01:30:47,849 --> 01:30:51,089
where compared to the healthy volunteers,

2046
01:30:51,089 --> 01:30:56,029
the moyamoya circulation
has blocked occlusions,

2047
01:30:56,029 --> 01:31:01,029
but also these wispy
collateral patterns of flow

2048
01:31:02,299 --> 01:31:05,479
that potentially is trying to compensate.

2049
01:31:05,479 --> 01:31:07,379
But it's unclear whether all

2050
01:31:07,379 --> 01:31:09,769
of these collaterals are
well formed in patients.

2051
01:31:09,769 --> 01:31:12,759
And so we need to actually
measure what's happening

2052
01:31:12,759 --> 01:31:15,699
in the tissue physiology to decide

2053
01:31:15,699 --> 01:31:19,399
which patients actually need surgery.

2054
01:31:19,399 --> 01:31:21,219
And one of the things about moyamoya

2055
01:31:21,219 --> 01:31:25,179
that is a great test
case for our situation,

2056
01:31:25,179 --> 01:31:27,429
it actually poses challenges.

2057
01:31:27,429 --> 01:31:29,789
It takes a very long time,

2058
01:31:29,789 --> 01:31:32,491
the longest transit delays among

2059
01:31:32,491 --> 01:31:34,439
many cerebrovascular disorders

2060
01:31:34,439 --> 01:31:37,409
to get through some of
these collateral patterns.

2061
01:31:37,409 --> 01:31:41,059
So we have this severe
pathology in relatively young

2062
01:31:41,059 --> 01:31:44,989
patients that we can
test in the vasculature

2063
01:31:44,989 --> 01:31:47,299
and really make sure that our measurements

2064
01:31:47,299 --> 01:31:52,299
are comparable to some of our
PET established standards.

2065
01:31:53,719 --> 01:31:55,849
One of the methods I'm going to focus

2066
01:31:55,849 --> 01:31:58,529
on here is arterial spin labeling,

2067
01:31:58,529 --> 01:32:03,529
and specifically a type of ASL
called pseudo continuous ASL.

2068
01:32:03,929 --> 01:32:06,609
This method with MRI does not require

2069
01:32:06,609 --> 01:32:09,702
any injection of a contrast agent.

2070
01:32:11,197 --> 01:32:13,449
Instead it uses radiofrequency pulses

2071
01:32:13,449 --> 01:32:18,069
in the MR scanner at the
approximate neck level.

2072
01:32:18,069 --> 01:32:22,469
And it labels spins into arterial
blood that are eventually

2073
01:32:22,469 --> 01:32:24,559
going to flow into the brain

2074
01:32:24,559 --> 01:32:27,249
where this blue imaging slot is.

2075
01:32:27,249 --> 01:32:29,269
And one of the controls that the user

2076
01:32:29,269 --> 01:32:32,486
as a imager has is this post label delay.

2077
01:32:32,486 --> 01:32:35,449
So how long between these RF pulses

2078
01:32:35,449 --> 01:32:39,179
do we wait before we capture
the image in the brain?

2079
01:32:39,179 --> 01:32:44,179
If we are able to take
this labeled acquisition

2080
01:32:44,489 --> 01:32:47,639
and repeat it with a control image,

2081
01:32:47,639 --> 01:32:48,999
the difference between

2082
01:32:48,999 --> 01:32:53,149
the two will take out all static spins

2083
01:32:53,149 --> 01:32:56,379
in the brain and in the end,

2084
01:32:56,379 --> 01:33:00,789
what you're left with is
this exactly the labeled flow

2085
01:33:00,789 --> 01:33:02,829
that has arrived to the imaging slab.

2086
01:33:02,829 --> 01:33:06,269
And what you see here are
example profusion maps,

2087
01:33:06,269 --> 01:33:10,039
where the scale, the brightness of each

2088
01:33:11,122 --> 01:33:14,286
voxel is indicative of the microvascular

2089
01:33:14,286 --> 01:33:17,279
perfusion that has arrived to the tissue.

2090
01:33:17,279 --> 01:33:20,119
So this is a general
framework for how ASL works,

2091
01:33:20,119 --> 01:33:23,369
but you can imagine, especially
in the moyamoya cases,

2092
01:33:23,369 --> 01:33:26,479
I was pointing out on the
previous slide that this

2093
01:33:26,479 --> 01:33:30,909
is challenging to do in patients
with abnormal vasculature.

2094
01:33:30,909 --> 01:33:34,329
How do we know what the
post label delay is?

2095
01:33:34,329 --> 01:33:39,329
And how can we validate
some of these ASL MRI maps

2096
01:33:40,229 --> 01:33:43,209
in such challenging patient cases?

2097
01:33:43,209 --> 01:33:46,099
So embarking on this study,

2098
01:33:46,099 --> 01:33:48,819
one of the first things I did
was I looked in the literature

2099
01:33:48,819 --> 01:33:51,531
and I compared previous studies

2100
01:33:51,531 --> 01:33:56,531
that performed both PET and
ASL, MRI profusion assessment.

2101
01:33:57,959 --> 01:34:00,639
And one thing that's very clear is

2102
01:34:02,050 --> 01:34:04,239
that the self-reported
R squared correlations

2103
01:34:04,239 --> 01:34:05,599
from the study itself.

2104
01:34:05,599 --> 01:34:09,039
So between PET and MR is higher

2105
01:34:09,039 --> 01:34:12,449
if the two scans are
spaced closer together,

2106
01:34:12,449 --> 01:34:14,432
that's not surprising.

2107
01:34:16,659 --> 01:34:21,659
With our daily diurnal
variations, with your diet,

2108
01:34:21,939 --> 01:34:25,119
if you've taken a couple
coffee now to stay awake during

2109
01:34:25,119 --> 01:34:28,779
this talk, your blood flow
could have reduced by 30%.

2110
01:34:28,779 --> 01:34:30,719
And so with these natural fluctuations,

2111
01:34:30,719 --> 01:34:33,039
it's really important to capture the PET

2112
01:34:33,039 --> 01:34:38,039
and MR together to
quantify brain physiology,

2113
01:34:39,129 --> 01:34:42,899
which varies so much across day to day.

2114
01:34:42,899 --> 01:34:45,389
So in order to do a more careful

2115
01:34:45,389 --> 01:34:49,109
physiologic controlled comparison,

2116
01:34:49,109 --> 01:34:52,239
we leveraged simultaneous
PET MR, specifically,

2117
01:34:52,239 --> 01:34:55,739
experiments were done
on the GE Sigma system

2118
01:34:55,739 --> 01:34:57,449
where we're truly capturing

2119
01:34:57,449 --> 01:34:58,992
with this whole body system.

2120
01:34:59,859 --> 01:35:04,859
Perfusion in the same
state by PET and by MRI.

2121
01:35:04,929 --> 01:35:07,432
So let me walk you through one example

2122
01:35:07,432 --> 01:35:11,639
in a moyamoya patient of what
these experiments look like.

2123
01:35:11,639 --> 01:35:15,729
Here's a standard ASL profusion map

2124
01:35:15,729 --> 01:35:20,065
in a 26 year old female with
bilateral moyamoya disease.

2125
01:35:20,065 --> 01:35:24,922
Moyamoya disease tends to
affect both hemispheres.

2126
01:35:25,869 --> 01:35:28,739
Sometimes it's unilateral,
many times it's bilateral.

2127
01:35:28,739 --> 01:35:32,249
And it typically is in
the anterior circulation.

2128
01:35:32,249 --> 01:35:36,299
So we see some heterogeneous signal here.

2129
01:35:36,299 --> 01:35:39,259
The standard ASL was acquired
with a post label delay

2130
01:35:39,259 --> 01:35:41,199
wait time of two seconds,

2131
01:35:41,199 --> 01:35:44,919
but it's really hard to tell
from this one perfusion map,

2132
01:35:44,919 --> 01:35:49,312
what exactly the interpretation should be.

2133
01:35:50,699 --> 01:35:53,339
In the simultaneous acquisition? However,

2134
01:35:53,339 --> 01:35:57,769
we see a reference from O-15 water PET,

2135
01:35:57,769 --> 01:35:59,429
which has been established

2136
01:36:00,293 --> 01:36:03,299
to be a gold standard for human imaging.

2137
01:36:03,299 --> 01:36:05,849
So this is with radio labeled water.

2138
01:36:05,849 --> 01:36:09,002
What we see is actually much more uniform,

2139
01:36:10,159 --> 01:36:15,159
less heterogeneous
signal in the PET image,

2140
01:36:15,399 --> 01:36:20,399
although with less perfusion
anteriorly, like we expect,

2141
01:36:20,859 --> 01:36:23,919
but certainly these two images do not look

2142
01:36:24,849 --> 01:36:28,539
in terms of the
heterogeneity quite similar.

2143
01:36:28,539 --> 01:36:31,069
And it actually gives us
some insights as to what's

2144
01:36:31,069 --> 01:36:34,749
happening underlying in
the brain, so for instance,

2145
01:36:34,749 --> 01:36:39,749
a hypothesis that we had
is on the left hemisphere,

2146
01:36:40,019 --> 01:36:42,776
there is these hotspots of higher signal,

2147
01:36:42,776 --> 01:36:45,589
and this may reflect labeled blood

2148
01:36:45,589 --> 01:36:49,499
that has arrived through the ASL labeling.

2149
01:36:49,499 --> 01:36:51,099
It has arrived into the brain,

2150
01:36:51,099 --> 01:36:54,569
but it's still stuck in low,

2151
01:36:54,569 --> 01:36:57,282
slow flowing, larger macro vessels.

2152
01:36:58,269 --> 01:37:00,469
It hasn't yet perfused
into the capillaries.

2153
01:37:02,579 --> 01:37:05,329
Whereas on the other hemisphere,

2154
01:37:05,329 --> 01:37:09,399
in the right hemisphere of the patient,

2155
01:37:09,399 --> 01:37:12,819
there's low signal compared to the PET.

2156
01:37:12,819 --> 01:37:17,079
And this may reflect an
even longer transit time.

2157
01:37:17,079 --> 01:37:19,909
These patients have even
more severe pathology.

2158
01:37:19,909 --> 01:37:24,549
It takes longer on this hemisphere
for the blood to arrive.

2159
01:37:24,549 --> 01:37:28,679
It's possible that the labeled
blood from the RF ASL tagging

2160
01:37:28,679 --> 01:37:31,222
has not even yet arrived
in the imaging voxel.

2161
01:37:32,249 --> 01:37:34,329
And so these are two hypotheses

2162
01:37:34,329 --> 01:37:39,329
that we can test and a
couple of ways that we can

2163
01:37:39,849 --> 01:37:44,109
add confirmatory information to this is

2164
01:37:44,109 --> 01:37:47,129
taking a contrast based MR Measurement

2165
01:37:47,129 --> 01:37:49,159
that does not give us a quantitative

2166
01:37:49,159 --> 01:37:52,476
MR Profusion map like ASL,

2167
01:37:52,476 --> 01:37:54,809
but it gives us a sense
of how long it takes

2168
01:37:54,809 --> 01:37:57,829
to arrive for the blood.

2169
01:37:57,829 --> 01:37:59,729
And this time to maximum, as you can see,

2170
01:37:59,729 --> 01:38:02,159
is elevated anteriorly,

2171
01:38:02,159 --> 01:38:05,779
especially on the right hemisphere here.

2172
01:38:05,779 --> 01:38:08,706
And another way to address this and test

2173
01:38:08,706 --> 01:38:13,706
and also improve the maps
is to use multi delay ASL.

2174
01:38:14,689 --> 01:38:18,789
So instead of only one post label delay,

2175
01:38:18,789 --> 01:38:21,329
their strategy's now to do sequential,

2176
01:38:21,329 --> 01:38:25,679
but also tying interleaved,
multi delay acquisitions.

2177
01:38:25,679 --> 01:38:27,689
Here I'm showing a simple example

2178
01:38:27,689 --> 01:38:30,799
with five labeling delay times.

2179
01:38:30,799 --> 01:38:34,349
And what that means is you
can actually capture different

2180
01:38:34,349 --> 01:38:38,209
times of arrival inflow
into the capillary bed,

2181
01:38:38,209 --> 01:38:43,209
as well as washout of this
RF labeled ASL signal.

2182
01:38:44,219 --> 01:38:47,609
And ultimately this gives us
a sense of how long it will

2183
01:38:47,609 --> 01:38:52,139
take for the blood to arrive
and also correct for it.

2184
01:38:52,139 --> 01:38:56,249
There are some limitations
in terms of time efficiency,

2185
01:38:56,249 --> 01:38:58,849
as well as signal to noise ratio.

2186
01:38:58,849 --> 01:39:03,849
If you're trying to acquire
many of these post label delays,

2187
01:39:03,889 --> 01:39:08,889
however, in the end, when you
compare to the PET signal,

2188
01:39:09,349 --> 01:39:11,709
this multi delay map looks

2189
01:39:11,709 --> 01:39:14,812
much more homogenous over the cortex,

2190
01:39:16,499 --> 01:39:17,799
and actually could also give you

2191
01:39:17,799 --> 01:39:20,499
at the same scan in arterial transit map,

2192
01:39:20,499 --> 01:39:24,089
which mimics the same
hypothesis that we have.

2193
01:39:24,089 --> 01:39:28,779
So we have longer delays due
to moyamoya vascular pathology

2194
01:39:28,779 --> 01:39:32,542
on both hemispheres, especially
in the right hemisphere.

2195
01:39:33,439 --> 01:39:35,219
If we were to wait even longer

2196
01:39:35,219 --> 01:39:39,519
and just capture one
long delay ASL signal,

2197
01:39:39,519 --> 01:39:42,839
you would actually see
the blood flow arrive

2198
01:39:42,839 --> 01:39:45,879
to the cortex with the ASL,
if you waited long enough,

2199
01:39:45,879 --> 01:39:48,449
but some of the signal
in the basal ganglia,

2200
01:39:48,449 --> 01:39:51,462
which has different relaxation
properties has decayed away.

2201
01:39:52,434 --> 01:39:55,369
So this is an overview in a
patient case of how it actually

2202
01:39:55,369 --> 01:39:59,099
teaches us something about
the transit time and also how

2203
01:39:59,099 --> 01:40:02,712
to best measure it in
comparison to a reference PET.

2204
01:40:04,719 --> 01:40:05,552
Really quickly,

2205
01:40:05,552 --> 01:40:09,579
here's a second example in
a moyamoya case where this

2206
01:40:09,579 --> 01:40:14,409
participant actually
had a previous stroke.

2207
01:40:14,409 --> 01:40:15,809
And in this case,

2208
01:40:15,809 --> 01:40:20,599
actually both with standard ASL
and multi delay acquisitions

2209
01:40:20,599 --> 01:40:25,159
up to three seconds waiting
time at post label delay,

2210
01:40:25,159 --> 01:40:29,622
both of these are not able
to capture the true remaining

2211
01:40:30,519 --> 01:40:33,339
unilateral signal of perfusion

2212
01:40:33,339 --> 01:40:36,179
on the right hemisphere
that we see on the 15 PET.

2213
01:40:36,179 --> 01:40:38,429
And it's really the longer delay

2214
01:40:38,429 --> 01:40:40,249
that allows us to capture this.

2215
01:40:40,249 --> 01:40:44,449
So I think it's a combination
of both capturing multi delay,

2216
01:40:44,449 --> 01:40:47,625
getting a sense of the transit times,

2217
01:40:47,625 --> 01:40:50,239
but potentially for these
cases where you suspect due

2218
01:40:50,239 --> 01:40:53,889
to pathology, or in other cases, age,

2219
01:40:53,889 --> 01:40:57,259
other conditions related
to vascular pathology,

2220
01:40:57,259 --> 01:40:59,999
you may have to wait longer
with the ASL compared

2221
01:40:59,999 --> 01:41:03,909
to what's standard for healthy volunteers.

2222
01:41:03,909 --> 01:41:08,629
If we look across regions
with very severe delays,

2223
01:41:08,629 --> 01:41:12,129
so time to maximum,
greater than five seconds,

2224
01:41:12,129 --> 01:41:16,689
we do see a high correlation for any

2225
01:41:16,689 --> 01:41:19,359
of these ASL signals with PET,

2226
01:41:19,359 --> 01:41:23,509
even for regions with
severe vascular delays

2227
01:41:24,829 --> 01:41:27,505
and still with the Bland Altman plot.

2228
01:41:27,505 --> 01:41:30,639
So I'm showing now the same data except

2229
01:41:30,639 --> 01:41:34,992
on the y-axis there's
MRI and PET difference.

2230
01:41:35,893 --> 01:41:39,449
And the x-axis is the mean
between the two modalities.

2231
01:41:39,449 --> 01:41:41,099
You'll see that actually

2232
01:41:41,099 --> 01:41:43,449
there's high correlation,

2233
01:41:43,449 --> 01:41:47,129
but there's bias until
we have longer delays.

2234
01:41:47,129 --> 01:41:50,969
So it's really the long
delay as well as multi delay

2235
01:41:50,969 --> 01:41:55,969
that kind of gets us
consistency between MRI and PET.

2236
01:41:56,639 --> 01:42:00,359
And what that allows us
to do on a group level

2237
01:42:00,359 --> 01:42:03,539
with these recommendations
is actually compare okay.

2238
01:42:03,539 --> 01:42:07,239
If we take PET and compare patients

2239
01:42:07,239 --> 01:42:11,359
to controls and each modality separately,

2240
01:42:11,359 --> 01:42:14,689
it's not surprising that the standard ASL

2241
01:42:14,689 --> 01:42:17,809
will overestimate the hypoperfusion.

2242
01:42:17,809 --> 01:42:20,139
So it's saying a lot of these patients

2243
01:42:20,139 --> 01:42:24,059
have under perfused regions,

2244
01:42:24,059 --> 01:42:26,019
but we know actually that some

2245
01:42:26,019 --> 01:42:28,789
of the signal has just not arrived.

2246
01:42:28,789 --> 01:42:32,249
And so we think that this
is actually artifactual.

2247
01:42:32,249 --> 01:42:36,409
And when we switch to a
recommended multi delay strategy,

2248
01:42:36,409 --> 01:42:39,349
especially with longer post label delays,

2249
01:42:39,349 --> 01:42:42,079
this allows us to better
discriminate patients

2250
01:42:42,079 --> 01:42:43,319
who are actually suffering

2251
01:42:43,319 --> 01:42:46,269
from low perfusion in a way that's more

2252
01:42:46,269 --> 01:42:49,659
consistent with PET
and more focused a lot,

2253
01:42:49,659 --> 01:42:52,002
especially in the anterior circulation.

2254
01:42:53,099 --> 01:42:54,389
So in summary,

2255
01:42:54,389 --> 01:42:58,399
these types of careful
validations with different ASL

2256
01:42:58,399 --> 01:43:03,399
parameters perform at the
same time with O-15 water PET,

2257
01:43:03,919 --> 01:43:08,919
gives us a better sense of
a recommended ASL strategy,

2258
01:43:09,109 --> 01:43:11,252
getting at quantitative profusion

2259
01:43:11,252 --> 01:43:13,999
in these challenging patients,

2260
01:43:13,999 --> 01:43:17,339
and also able to discriminate patients

2261
01:43:17,339 --> 01:43:18,439
from healthy controls.

2262
01:43:19,759 --> 01:43:21,399
Now you may ask,

2263
01:43:21,399 --> 01:43:24,569
is it really necessary to
do this at the same time?

2264
01:43:24,569 --> 01:43:26,399
And some of these pathologies,

2265
01:43:26,399 --> 01:43:31,399
these changes are different
enough across from a patient,

2266
01:43:31,919 --> 01:43:34,829
especially if it's like a unilateral case

2267
01:43:34,829 --> 01:43:36,179
and there's asymmetry,

2268
01:43:36,179 --> 01:43:38,679
it would be different enough that I think

2269
01:43:38,679 --> 01:43:41,069
the same profusion abnormalities

2270
01:43:41,069 --> 01:43:43,222
would show up on PET and ASL.

2271
01:43:44,169 --> 01:43:47,479
However, I think it's even more important

2272
01:43:47,479 --> 01:43:50,919
to capture the same physiological states

2273
01:43:50,919 --> 01:43:54,712
if you're looking at more
than one physiological state.

2274
01:43:55,970 --> 01:43:58,939
And this leads me to describe
some of the experiments

2275
01:43:58,939 --> 01:44:01,672
we've done to assess
cerebrovascular reactivity.

2276
01:44:02,759 --> 01:44:06,639
This is the ability of the
brain to augment perfusion

2277
01:44:06,639 --> 01:44:09,919
in response to an external challenge.

2278
01:44:09,919 --> 01:44:12,799
And it could test the ability

2279
01:44:12,799 --> 01:44:17,799
of your brain to auto-regulate
if there is a stenosis,

2280
01:44:18,119 --> 01:44:21,109
the downstream vessels or capillaries

2281
01:44:22,239 --> 01:44:25,439
may already be maximally dilated,

2282
01:44:25,439 --> 01:44:30,439
and unable to respond to
an exogenous challenge

2283
01:44:31,209 --> 01:44:35,469
such as hypercalcemia or a drug

2284
01:44:35,469 --> 01:44:39,752
that actually performs fatal
dilation like acetazolamide.

2285
01:44:40,839 --> 01:44:43,139
So in giving a stress test to the brain,

2286
01:44:43,139 --> 01:44:46,879
we're doing something
quite like a cardiologist

2287
01:44:46,879 --> 01:44:49,099
would have you run on a treadmill.

2288
01:44:49,099 --> 01:44:51,779
And we're trying to see
if there's pathology

2289
01:44:51,779 --> 01:44:55,859
that shows up not only at baseline,

2290
01:44:55,859 --> 01:44:59,472
but after stressing these
vessels in the brain.

2291
01:45:00,499 --> 01:45:05,142
This is important because you
not only is some pathology

2292
01:45:05,142 --> 01:45:09,459
only identified after the stress test,

2293
01:45:09,459 --> 01:45:13,029
but it could identify
a very poor prognosis.

2294
01:45:13,029 --> 01:45:14,299
In this case,

2295
01:45:14,299 --> 01:45:18,439
here's a PET image at
baseline and a PET image

2296
01:45:18,439 --> 01:45:21,842
after vasodilatation with a
drug called acetazolamide.

2297
01:45:23,934 --> 01:45:26,039
And even though we're expecting increase

2298
01:45:26,039 --> 01:45:28,748
in profusion with vasodilation,

2299
01:45:28,748 --> 01:45:30,959
there's actually a paradoxical decrease

2300
01:45:30,959 --> 01:45:32,899
in this right hemisphere.

2301
01:45:32,899 --> 01:45:36,742
And that's indicative of
cerebral vascular steal,

2302
01:45:37,894 --> 01:45:41,529
and it's a poor prognosis
for this particular patient.

2303
01:45:41,529 --> 01:45:44,986
We're starting to see that with ASL

2304
01:45:44,986 --> 01:45:49,986
we're able to do some
of the same assessment

2305
01:45:50,589 --> 01:45:53,519
in multiple physiological states.

2306
01:45:53,519 --> 01:45:55,759
And this is challenging because

2307
01:45:55,759 --> 01:45:58,829
it's hard to deliver the same exact gas

2308
01:45:58,829 --> 01:46:01,479
or ensure that the same dose

2309
01:46:01,479 --> 01:46:03,429
on separate sessions to create

2310
01:46:03,429 --> 01:46:06,036
vasodilation is exactly consistent.

2311
01:46:06,036 --> 01:46:08,199
And so I would argue that

2312
01:46:08,199 --> 01:46:11,469
specifically in this
case of a stress test,

2313
01:46:11,469 --> 01:46:15,639
we're looking at more than
one physiological condition.

2314
01:46:15,639 --> 01:46:18,629
This is even more reason
that the validation

2315
01:46:18,629 --> 01:46:20,369
for profusion has to happen

2316
01:46:22,219 --> 01:46:24,532
simultaneously between PET and MR.

2317
01:46:26,242 --> 01:46:30,212
There's been several studies
where we've looked at this,

2318
01:46:31,119 --> 01:46:34,942
not only in patients, but
in healthy controls as well.

2319
01:46:35,961 --> 01:46:38,150
And it's actually quite interesting

2320
01:46:38,150 --> 01:46:42,033
because areas that have
short transit times.

2321
01:46:42,033 --> 01:46:43,969
So we talked about the deep gray matter,

2322
01:46:43,969 --> 01:46:48,969
the basal ganglia previously
and intrinsically these deep

2323
01:46:49,009 --> 01:46:53,069
gray matter regions tend
to have short transit time.

2324
01:46:53,069 --> 01:46:56,369
And if the transit time is too short,

2325
01:46:56,369 --> 01:47:01,219
the ASL signal might
actually decay quicker.

2326
01:47:01,219 --> 01:47:03,549
And for the same post they will delay.

2327
01:47:03,549 --> 01:47:06,962
We have also underestimation.

2328
01:47:07,809 --> 01:47:11,469
So in blue, the ASL underestimating,

2329
01:47:11,469 --> 01:47:14,259
just in the same way
that long delay times led

2330
01:47:14,259 --> 01:47:16,379
to underestimation that ASL.

2331
01:47:16,379 --> 01:47:20,279
We're also seeing that
with short transit times,

2332
01:47:20,279 --> 01:47:22,609
if you give a vasal dilation challenge,

2333
01:47:22,609 --> 01:47:25,129
some of the blood arrives even faster,

2334
01:47:25,129 --> 01:47:28,539
you get these larger
areas after vasal dilation

2335
01:47:28,539 --> 01:47:33,209
of green indicating short transit times,

2336
01:47:33,209 --> 01:47:37,309
and thus much more
underestimation coincident

2337
01:47:37,309 --> 01:47:39,549
with these areas of short transit time.

2338
01:47:39,549 --> 01:47:42,529
So that was kind of a complicated logic,

2339
01:47:42,529 --> 01:47:44,519
but the main takeaway that we have

2340
01:47:44,519 --> 01:47:48,309
to be cautious when
there's fast blood flow,

2341
01:47:48,309 --> 01:47:51,119
as well after vasodilation
and the stress test

2342
01:47:52,914 --> 01:47:56,589
and in areas of short
ATT, short transit times,

2343
01:47:56,589 --> 01:47:59,609
as well as long transit
times that we discussed,

2344
01:47:59,609 --> 01:48:00,599
this could be a problem

2345
01:48:00,599 --> 01:48:04,112
when we compare to PET as in this case.

2346
01:48:05,179 --> 01:48:08,669
And my colleague Mo Dr.
Moss Sal at Stanford

2347
01:48:08,669 --> 01:48:11,759
is continuing to evaluate this with CBR

2348
01:48:11,759 --> 01:48:13,719
and has done a really beautiful job

2349
01:48:13,719 --> 01:48:17,559
of comparing PET super
vascular reactivity.

2350
01:48:17,559 --> 01:48:21,589
So this is the percent
response and profusion due

2351
01:48:21,589 --> 01:48:25,598
to acetazolamide and
showed nice correspondence

2352
01:48:25,598 --> 01:48:30,559
with multi post able delay ASL MRI

2353
01:48:30,559 --> 01:48:34,529
in comparing across these modalities.

2354
01:48:34,529 --> 01:48:35,469
So in summary,

2355
01:48:35,469 --> 01:48:40,069
I think it's really
critical to look at the same

2356
01:48:40,069 --> 01:48:44,499
physiological state for
accurate quantification.

2357
01:48:44,499 --> 01:48:47,559
And this is particularly so when you want

2358
01:48:47,559 --> 01:48:50,199
to track multiple physiological states

2359
01:48:51,495 --> 01:48:55,369
that you've induced in order
to see in a stressed condition

2360
01:48:55,369 --> 01:48:58,159
is the brain showing abnormalities

2361
01:48:58,159 --> 01:49:00,932
that we didn't detect before at baseline.

2362
01:49:03,529 --> 01:49:06,669
Now that we've motivated
the need for physiological

2363
01:49:06,669 --> 01:49:11,249
simultaneity in PET and
MRI validation studies,

2364
01:49:11,249 --> 01:49:14,380
I wanna share some
technical considerations

2365
01:49:14,380 --> 01:49:19,380
as to how PET MR can
reduce the invasiveness

2366
01:49:19,479 --> 01:49:22,289
while keeping the quantitative accuracy

2367
01:49:22,289 --> 01:49:25,439
of these physiological biomarkers.

2368
01:49:25,439 --> 01:49:30,439
So one established need for
quantification in standard PET

2369
01:49:30,889 --> 01:49:35,619
acquisitions is the need
to sample arterial blood

2370
01:49:35,619 --> 01:49:39,039
as an input function and understanding

2371
01:49:39,039 --> 01:49:42,859
how much tracer is
delivered at which times,

2372
01:49:42,859 --> 01:49:46,439
in order to perform a
quantitative kinetic model.

2373
01:49:46,439 --> 01:49:49,799
We are able to see some of the signal

2374
01:49:49,799 --> 01:49:52,639
in the head PET field of view,

2375
01:49:52,639 --> 01:49:56,679
but the arteries are
quite small to compare

2376
01:49:56,679 --> 01:50:00,349
with this typical PET spatial resolution.

2377
01:50:00,349 --> 01:50:02,189
And so with PET MR studies,

2378
01:50:02,189 --> 01:50:05,709
what we have is a really well aligned MR

2379
01:50:06,629 --> 01:50:09,639
high spatial angiogram, for instance,

2380
01:50:09,639 --> 01:50:12,489
shown here with the time of flight image,

2381
01:50:12,489 --> 01:50:16,749
we could compare this
to the early timeframes

2382
01:50:16,749 --> 01:50:18,639
where the tracer's being developed.

2383
01:50:18,639 --> 01:50:22,039
So if you look at the total PET counts

2384
01:50:22,039 --> 01:50:23,339
in an initial injection,

2385
01:50:23,339 --> 01:50:28,339
there's a window of time
where the counts are likely

2386
01:50:28,919 --> 01:50:31,909
in the vasculature before
it's actually stabilized

2387
01:50:31,909 --> 01:50:35,379
and arrived into the brain tissue.

2388
01:50:35,379 --> 01:50:37,889
So in taking a short timeframe

2389
01:50:37,889 --> 01:50:41,649
with high enough sensitivity,
you're able to see,

2390
01:50:41,649 --> 01:50:44,669
a similar angiogram from the PET,

2391
01:50:44,669 --> 01:50:47,699
albeit with lower spatial resolution.

2392
01:50:47,699 --> 01:50:51,339
And there are ways to model these spill in

2393
01:50:51,339 --> 01:50:55,259
and spill out effects that are due

2394
01:50:55,259 --> 01:50:57,909
to this limitation in spatial resolution,

2395
01:50:57,909 --> 01:51:00,889
and kind of look at these
different arterial mass from MR

2396
01:51:01,834 --> 01:51:05,409
and PET aligned from the
simultaneous acquisition

2397
01:51:05,409 --> 01:51:08,849
to then extract and remove spill

2398
01:51:08,849 --> 01:51:13,849
and swell effects in the
image derived in function.

2399
01:51:13,859 --> 01:51:16,519
So the blue curve here is a representation

2400
01:51:16,519 --> 01:51:19,602
of an injection of O-15 water.

2401
01:51:19,602 --> 01:51:23,499
This is the initial bolus
and input function delivery,

2402
01:51:23,499 --> 01:51:26,339
and using PET MR information,

2403
01:51:26,339 --> 01:51:27,539
we're able to correct for some

2404
01:51:27,539 --> 01:51:30,159
of the spill and spill out artifacts.

2405
01:51:30,159 --> 01:51:33,279
And this has a lot of
advantages first and foremost,

2406
01:51:33,279 --> 01:51:35,142
being noninvasive.

2407
01:51:36,171 --> 01:51:39,242
So not needing to sample arterial blood,

2408
01:51:40,449 --> 01:51:44,489
but it's also closer to the
brain area of interest compared

2409
01:51:44,489 --> 01:51:48,449
to sampling from the
peripheral arteries in the arm.

2410
01:51:48,449 --> 01:51:51,069
And it's also automatically calibrated

2411
01:51:51,069 --> 01:51:53,759
to the rest of the brain time activity

2412
01:51:53,759 --> 01:51:56,149
curves or brain tissue signals

2413
01:51:56,149 --> 01:51:59,172
for the later kinetic modeling part.

2414
01:52:00,249 --> 01:52:02,589
Other studies have taken this image

2415
01:52:02,589 --> 01:52:05,239
direct input function approach,

2416
01:52:05,239 --> 01:52:08,749
not only to take into account
point spread functions

2417
01:52:08,749 --> 01:52:10,719
and spill in and spill out,

2418
01:52:10,719 --> 01:52:13,469
but also to use in the early timeframes

2419
01:52:13,469 --> 01:52:15,512
where the bolus is being injected,

2420
01:52:16,546 --> 01:52:17,579
MR and navigator.

2421
01:52:17,579 --> 01:52:21,029
So very quick navigator acquisitions

2422
01:52:21,029 --> 01:52:24,839
that show you the position

2423
01:52:24,839 --> 01:52:29,209
with a MR acquisition and allow you

2424
01:52:29,209 --> 01:52:33,849
to then adjust the mask over the time

2425
01:52:33,849 --> 01:52:36,409
of the dynamic acquisition, say,

2426
01:52:36,409 --> 01:52:41,019
in this case up to 60
minutes and adjust the mask

2427
01:52:41,019 --> 01:52:43,189
of the vasculature so that you're

2428
01:52:43,189 --> 01:52:45,982
innately correcting for motion.

2429
01:52:47,763 --> 01:52:52,109
So not only are we able to
address spacial resolution

2430
01:52:52,109 --> 01:52:55,859
limitations but we're
able to use MR navigators

2431
01:52:55,859 --> 01:53:00,859
to address patient motion
and get again in this study,

2432
01:53:02,389 --> 01:53:06,162
very high fidelity image
derived input functions,

2433
01:53:07,542 --> 01:53:10,729
low difference from the
actual arterial blood

2434
01:53:10,729 --> 01:53:13,899
that was sampled as the ground truth,

2435
01:53:13,899 --> 01:53:18,899
and ultimately quantitative profusion,

2436
01:53:19,019 --> 01:53:22,189
or in this case, glucose maps,

2437
01:53:22,189 --> 01:53:25,622
after you use an image
derived input function.

2438
01:53:26,721 --> 01:53:30,909
So these strategies which
do involve extraction

2439
01:53:30,909 --> 01:53:34,579
of the vessels are broadly applicable.

2440
01:53:34,579 --> 01:53:39,109
And I think play a role both
on in physiological biomarkers,

2441
01:53:39,109 --> 01:53:43,629
but other tracers that you
may want to use in order

2442
01:53:43,629 --> 01:53:47,789
to account for these motion artifacts

2443
01:53:47,789 --> 01:53:52,429
or spatial resolution artifacts
on the input function,

2444
01:53:52,429 --> 01:53:56,379
beyond that, for physiological biomarkers,

2445
01:53:56,379 --> 01:54:01,179
there's actually an
opportunity that even overcomes

2446
01:54:01,179 --> 01:54:05,756
the need for input
functions to begin with.

2447
01:54:05,756 --> 01:54:10,649
And this concept kind
of is based on the fact

2448
01:54:10,649 --> 01:54:14,639
that there are acquisitions that globally

2449
01:54:14,639 --> 01:54:18,149
with MRI can be sensitive to some

2450
01:54:18,149 --> 01:54:20,069
of these parameters that we care about,

2451
01:54:20,069 --> 01:54:22,359
like brain perfusion.

2452
01:54:22,359 --> 01:54:25,839
So here's a example paradigm

2453
01:54:25,839 --> 01:54:30,079
that acquires both MR and O-15 water PET.

2454
01:54:30,079 --> 01:54:35,079
And the main MR acquisition
is called phase contrast,

2455
01:54:37,339 --> 01:54:40,879
phase contrast uses the phase MRI signal

2456
01:54:40,879 --> 01:54:43,409
to look at major arteries and give you

2457
01:54:43,409 --> 01:54:48,329
a global profusion
representation in the brain.

2458
01:54:48,329 --> 01:54:50,619
And what previous studies have shown

2459
01:54:50,619 --> 01:54:55,619
is that phase contrast and O-15 water PET,

2460
01:54:55,659 --> 01:55:00,029
the reference standard are
highly correlated across

2461
01:55:00,029 --> 01:55:03,399
different conditions on the global scale.

2462
01:55:03,399 --> 01:55:08,329
So this is again a whole
brain profusion measurement.

2463
01:55:08,329 --> 01:55:13,209
This has been tested in different
gas breathing conditions,

2464
01:55:13,209 --> 01:55:15,419
which actually changed the profusion

2465
01:55:15,419 --> 01:55:16,999
so much like the stress test

2466
01:55:16,999 --> 01:55:19,209
we were talking about before.

2467
01:55:19,209 --> 01:55:22,279
But phase contrast is actually
going to be a nice global

2468
01:55:23,241 --> 01:55:25,639
representation that has been validated

2469
01:55:25,639 --> 01:55:28,779
with O-15 water PET,

2470
01:55:28,779 --> 01:55:32,299
and what that allowed other investigators,

2471
01:55:32,299 --> 01:55:35,609
including our group to do
is to think about, well,

2472
01:55:35,609 --> 01:55:38,319
maybe we don't even need to capture

2473
01:55:38,319 --> 01:55:43,319
the input function itself if
we trust this global scaling.

2474
01:55:44,829 --> 01:55:48,939
So if a static O-15 water PET scan,

2475
01:55:48,939 --> 01:55:53,701
such as this, say over two
minutes is a good representation

2476
01:55:53,701 --> 01:55:58,701
of the distribution of
the profusion tracer.

2477
01:56:00,059 --> 01:56:05,059
If our goal is to get to a
quantitative map of profusion,

2478
01:56:06,179 --> 01:56:11,179
we could calculate a scaling
from this PET relative static

2479
01:56:12,569 --> 01:56:17,569
scan to a quantitative scan
using a main whole brain CPF.

2480
01:56:19,249 --> 01:56:21,309
And that main whole brain CPF

2481
01:56:21,309 --> 01:56:24,239
could exactly be from phase contrast,

2482
01:56:24,239 --> 01:56:28,699
'cause it's been validated as
based on my previous slide.

2483
01:56:28,699 --> 01:56:30,979
And this paradigm, as you can see,

2484
01:56:30,979 --> 01:56:34,199
would require the phase contrast MR

2485
01:56:34,199 --> 01:56:37,369
to be acquired at the
same time as the PET,

2486
01:56:37,369 --> 01:56:39,599
but it would simplify the need for

2487
01:56:40,779 --> 01:56:44,079
extraction vessel segmentation

2488
01:56:44,079 --> 01:56:46,559
of image derived in per function,

2489
01:56:46,559 --> 01:56:49,599
and still avoid our
arterial blood sampling.

2490
01:56:49,599 --> 01:56:53,749
So specifically for perfusion
and using phase contrast

2491
01:56:53,749 --> 01:56:58,599
these large vessels to get
a global scaling factor,

2492
01:56:58,599 --> 01:57:03,409
this is another clear way in
which physiology from both

2493
01:57:03,409 --> 01:57:07,859
PET and MR can be leveraged
together and get at something

2494
01:57:07,859 --> 01:57:11,292
quantitative like this
absolute map of CPF.

2495
01:57:13,259 --> 01:57:18,109
This PC PET approach again
requires phase contrast

2496
01:57:18,109 --> 01:57:22,059
to be acquired with the PET
in different conditions,

2497
01:57:22,059 --> 01:57:24,179
but it is sensitive.

2498
01:57:24,179 --> 01:57:26,654
We've shown that it shows

2499
01:57:26,654 --> 01:57:31,089
clear reactivity in healthy controls.

2500
01:57:31,089 --> 01:57:35,469
If you look at baseline
and postal dilation CPF,

2501
01:57:35,469 --> 01:57:38,849
as well as limitations and reactivity,

2502
01:57:38,849 --> 01:57:43,006
both reductions and
paradoxical negative reactivity

2503
01:57:44,939 --> 01:57:47,609
in the presence of a vasal dilator

2504
01:57:47,609 --> 01:57:50,092
for patients with moyamoya disease.

2505
01:57:51,055 --> 01:57:54,049
So again, these images are profusion maps

2506
01:57:54,049 --> 01:57:57,729
that have been scaled
with phase contrast MR

2507
01:57:57,729 --> 01:58:02,169
and it's really leveraging
both physiological measurements

2508
01:58:02,169 --> 01:58:03,719
in different conditions in order

2509
01:58:03,719 --> 01:58:07,609
to show this and actually
in this graph here,

2510
01:58:07,609 --> 01:58:12,039
show that we can stratify
cerebrovascular reactivity

2511
01:58:12,039 --> 01:58:15,289
with this hybrid approach and really

2512
01:58:15,289 --> 01:58:19,359
understand differences in CBR
that stratify across normal,

2513
01:58:19,359 --> 01:58:23,422
mild to moderate and
severe occluded stenosis.

2514
01:58:24,859 --> 01:58:29,859
Finally, beyond this hybrid approach

2515
01:58:30,079 --> 01:58:33,179
there's another way to
synergize PET and MR

2516
01:58:33,179 --> 01:58:36,689
which there's a lot of
excitement in the field,

2517
01:58:36,689 --> 01:58:39,609
of course around AI and the ability

2518
01:58:39,609 --> 01:58:44,609
to leverage more complicated models

2519
01:58:44,999 --> 01:58:47,649
with deep learning networks.

2520
01:58:47,649 --> 01:58:49,059
In this case,

2521
01:58:49,059 --> 01:58:53,189
we go a step further
and actually ask, okay,

2522
01:58:53,189 --> 01:58:58,189
we have initial ASL and
anatomical MRI as inputs.

2523
01:58:58,319 --> 01:59:03,319
Can we actually synthesize
a quantitative CBF map based

2524
01:59:03,619 --> 01:59:08,619
on these sort of ASL and anatomical scans.

2525
01:59:09,009 --> 01:59:10,699
And this is work for my colleague,

2526
01:59:10,699 --> 01:59:14,989
Professor Jia Guo at UC Riverside.

2527
01:59:14,989 --> 01:59:19,989
The way that this architecture
was developed is to try

2528
01:59:20,429 --> 01:59:24,559
to avoid actually injecting PET tracers

2529
01:59:25,639 --> 01:59:30,139
and asking whether the
ASL maps are sufficient

2530
01:59:30,139 --> 01:59:31,629
to make a prediction.

2531
01:59:31,629 --> 01:59:33,969
And the training itself in this algorithm

2532
01:59:34,809 --> 01:59:39,079
was based on simultaneous O-15 PET.

2533
01:59:39,079 --> 01:59:42,249
So if you have these combined

2534
01:59:42,249 --> 01:59:44,279
simultaneous PET MR acquisitions,

2535
01:59:44,279 --> 01:59:47,819
you have the ASLs at
multiple post label delays,

2536
01:59:47,819 --> 01:59:51,079
you can train the network to synthesize

2537
01:59:51,079 --> 01:59:55,609
a PET like gold standard
image of perfusion.

2538
01:59:55,609 --> 01:59:57,579
And this was achieved with an architecture

2539
01:59:57,579 --> 02:00:00,569
that is familiar to many in
the medical imaging field.

2540
02:00:00,569 --> 02:00:04,819
It's a unit with both an
encoder and decoder arm

2541
02:00:05,719 --> 02:00:08,529
in order to learn hidden

2542
02:00:08,529 --> 02:00:10,919
or deep features from the images

2543
02:00:10,919 --> 02:00:13,049
and then actually recreate

2544
02:00:13,049 --> 02:00:16,522
a synthesized quantitative profusion map.

2545
02:00:17,469 --> 02:00:22,069
What's really interesting in
this case is that, of course,

2546
02:00:22,069 --> 02:00:25,359
this comparison between standard ASL

2547
02:00:26,309 --> 02:00:30,309
and O-15 water PET as the gold standard,

2548
02:00:30,309 --> 02:00:33,029
we know from our previous discussion

2549
02:00:33,029 --> 02:00:34,329
in the first part of this talk,

2550
02:00:34,329 --> 02:00:37,499
that there are limitations in artifacts

2551
02:00:37,499 --> 02:00:39,066
with standard delay ASL,

2552
02:00:40,229 --> 02:00:45,159
these errors are slightly
reduced with multi delay ASL,

2553
02:00:45,159 --> 02:00:50,159
but in particular, the neural
network output outperforms

2554
02:00:50,399 --> 02:00:53,509
both of the ASL acquisitions,

2555
02:00:53,509 --> 02:00:58,239
the neural network is only
derived from MRI inputs,

2556
02:00:58,239 --> 02:01:00,999
but it's able to reproduce

2557
02:01:00,999 --> 02:01:05,999
after training a CBF map
that is even more consistent,

2558
02:01:07,789 --> 02:01:09,529
even more quantitatively accurate

2559
02:01:09,529 --> 02:01:11,042
with the PET reference standard

2560
02:01:11,042 --> 02:01:12,872
than the other ASL scans.

2561
02:01:13,869 --> 02:01:18,379
And when we're training across
these different cohorts,

2562
02:01:18,379 --> 02:01:22,896
it's not surprising that it
does depend on the training set.

2563
02:01:25,437 --> 02:01:28,259
So one of the metrics you could consider

2564
02:01:28,259 --> 02:01:32,059
in evaluating the deep learning network

2565
02:01:32,059 --> 02:01:34,052
is structural similarity.

2566
02:01:35,654 --> 02:01:39,699
And it's not surprising that
if you only train on controls

2567
02:01:40,949 --> 02:01:44,649
in the initial establishment of weights

2568
02:01:44,649 --> 02:01:47,559
in the architecture, your
structure similarity's

2569
02:01:47,559 --> 02:01:49,949
not going to be as good as if you trained

2570
02:01:49,949 --> 02:01:54,649
on patients or on a mixed
group with more data sets.

2571
02:01:54,649 --> 02:01:57,129
And that's because the
controls wouldn't have some

2572
02:01:57,129 --> 02:02:01,349
of the problems of long transit times

2573
02:02:01,349 --> 02:02:04,009
that the network could potentially learn

2574
02:02:04,009 --> 02:02:07,589
when it sees patient cases
and learn to correct for.

2575
02:02:07,589 --> 02:02:11,819
So it does matter in
terms of generalizability,

2576
02:02:11,819 --> 02:02:13,349
what patient cohorts and what's

2577
02:02:13,349 --> 02:02:16,339
the balance that you
include in the training.

2578
02:02:16,339 --> 02:02:18,369
We see a similar trend where

2579
02:02:19,209 --> 02:02:23,269
you get a benefit in root
mean squared error reduction

2580
02:02:23,269 --> 02:02:28,269
if you train on a mixed cohort
of healthy patient cases,

2581
02:02:29,139 --> 02:02:33,786
but overall, this is a very
promising approach that

2582
02:02:34,869 --> 02:02:38,369
could avoid the need for tracer injection.

2583
02:02:38,369 --> 02:02:40,209
And in our recent work,

2584
02:02:40,209 --> 02:02:43,934
actually even avoid the
need for acetazolamide

2585
02:02:43,934 --> 02:02:46,709
or vasodilation injection
to predict reactivity.

2586
02:02:48,461 --> 02:02:52,149
And it's a approach that is sophisticated

2587
02:02:52,149 --> 02:02:55,372
in terms of the deep learning network,

2588
02:02:56,299 --> 02:02:58,039
addressing some of the artifacts

2589
02:02:58,039 --> 02:03:00,019
that we talked about before,

2590
02:03:00,019 --> 02:03:02,039
but the deep learning network needs

2591
02:03:02,039 --> 02:03:04,969
very good quality training pairs,

2592
02:03:04,969 --> 02:03:09,199
and we believe the best
quality training pairs

2593
02:03:10,108 --> 02:03:14,269
are simultaneous ASL or
other physiological maps

2594
02:03:14,269 --> 02:03:17,209
at the same time in the
same physiological condition

2595
02:03:17,209 --> 02:03:19,262
as the PET scans.

2596
02:03:21,596 --> 02:03:23,459
And the last part of this talk,

2597
02:03:23,459 --> 02:03:25,789
I just wanna project forward a little bit

2598
02:03:25,789 --> 02:03:27,709
and broaden some of the topics

2599
02:03:27,709 --> 02:03:29,889
that we've talked about beyond just

2600
02:03:29,889 --> 02:03:33,029
specifically to cerebrovascular disease

2601
02:03:33,029 --> 02:03:36,899
to other neurological
conditions and other biomarkers

2602
02:03:37,768 --> 02:03:41,729
that you know you may not
have been familiar with.

2603
02:03:41,729 --> 02:03:46,279
So one example that has
gained a lot of traction

2604
02:03:47,158 --> 02:03:51,379
is not only the shift from
understanding Alzheimer's disease

2605
02:03:51,379 --> 02:03:54,199
as a set of clinical symptoms,

2606
02:03:54,199 --> 02:03:56,959
but actually looking at the underlying

2607
02:03:56,959 --> 02:04:01,529
biological construct,
and this A/T/N framework,

2608
02:04:01,529 --> 02:04:06,329
amyloid, tau and neuro
degeneration framework

2609
02:04:06,329 --> 02:04:11,329
is highly critical in terms
of both CSF and flood,

2610
02:04:12,369 --> 02:04:17,369
but also imaging biomarkers
to detect amyloid

2611
02:04:17,519 --> 02:04:21,729
and tau as Alzheimer's
specific pathologies,

2612
02:04:21,729 --> 02:04:25,409
but also non-Alzheimer's
pathological changes.

2613
02:04:25,409 --> 02:04:27,009
And I think one thing that belongs

2614
02:04:27,009 --> 02:04:32,009
within neurodegenerative like
cortical or tissue atrophy,

2615
02:04:32,659 --> 02:04:36,209
it could be part of the
neurodegenerative process that there

2616
02:04:36,209 --> 02:04:40,509
are also vascular risk factors
and vascular conditions

2617
02:04:40,509 --> 02:04:44,059
such as hypertension and diabetes,

2618
02:04:44,059 --> 02:04:48,739
which lead to cerebrovascular
impairment that is more subtle

2619
02:04:48,739 --> 02:04:51,899
than a stroke or a steno occlusion

2620
02:04:51,899 --> 02:04:53,799
like we've been talking about,

2621
02:04:53,799 --> 02:04:56,519
but still either
independently or interacts

2622
02:04:56,519 --> 02:05:00,009
with amyloid and tau to
create cognitive impairment.

2623
02:05:00,009 --> 02:05:04,849
So I think vascular conditions
and imaging approaches

2624
02:05:04,849 --> 02:05:08,369
to understand
cerebrovascular changes falls

2625
02:05:08,369 --> 02:05:12,159
within this category of neurodegeneration

2626
02:05:12,159 --> 02:05:14,249
within this biological construct

2627
02:05:14,249 --> 02:05:16,672
of cognitive impairment in ED.

2628
02:05:18,994 --> 02:05:19,827
And what this means is,

2629
02:05:19,827 --> 02:05:23,749
so we can't just understand
one facet of the disease,

2630
02:05:23,749 --> 02:05:25,679
but because there are multi

2631
02:05:27,119 --> 02:05:30,679
pronged sources for this pathology,

2632
02:05:30,679 --> 02:05:32,449
we need to be able to measure all of them.

2633
02:05:32,449 --> 02:05:36,329
And I think a clever way in
which we can actually leverage

2634
02:05:36,329 --> 02:05:40,359
PET MR is to get more out
of some of these biomarkers.

2635
02:05:40,359 --> 02:05:44,029
So here's an example that
I find very inspiring,

2636
02:05:44,029 --> 02:05:48,009
which is to use the amyloid PET signal

2637
02:05:49,169 --> 02:05:53,279
and actually in a PET MR framework,

2638
02:05:53,279 --> 02:05:58,019
reduce the time and potentially
get multiple biomarkers.

2639
02:05:58,019 --> 02:06:03,019
So a typical PET acquisition
for amyloid is to acquire,

2640
02:06:04,349 --> 02:06:08,519
maybe a static scan
only at the later part.

2641
02:06:08,519 --> 02:06:10,839
It depends on the exact amyloid tracer,

2642
02:06:10,839 --> 02:06:12,762
but let's say Florbetaben,

2643
02:06:13,769 --> 02:06:16,602
sometimes even ninety to a 110 minutes,

2644
02:06:18,159 --> 02:06:23,039
what you could do to get
quantitative information

2645
02:06:23,039 --> 02:06:27,379
is perhaps not have the subject have

2646
02:06:27,379 --> 02:06:31,299
to stay in this entire dynamic period

2647
02:06:31,299 --> 02:06:33,869
where you would see the initial bolus

2648
02:06:33,869 --> 02:06:36,189
and first pass of the tracer,

2649
02:06:36,189 --> 02:06:40,999
but have this data acquisition
only in the later parts,

2650
02:06:40,999 --> 02:06:45,299
but really use simultaneous
MR measurements,

2651
02:06:45,299 --> 02:06:49,439
say of perfusion to estimate actually

2652
02:06:49,439 --> 02:06:51,649
what happened in this earlier part.

2653
02:06:51,649 --> 02:06:54,879
So you're cutting down the scan time

2654
02:06:54,879 --> 02:06:58,609
that the patient has to tolerate in order

2655
02:06:58,609 --> 02:07:02,172
to get quantitative amyloid binding maps,

2656
02:07:04,882 --> 02:07:08,229
and actually with
simultaneous MR using the ASL,

2657
02:07:10,209 --> 02:07:13,559
not only as a separate
marker of perfusion,

2658
02:07:13,559 --> 02:07:17,839
which is important in and of
itself, but also to capture,

2659
02:07:17,839 --> 02:07:22,299
recreate the early part of
this amyloid time signal

2660
02:07:22,299 --> 02:07:24,619
and what these initial studies

2661
02:07:24,619 --> 02:07:27,279
have shown is that the binding potential

2662
02:07:28,819 --> 02:07:32,199
not only can capture regional
differences quantitatively,

2663
02:07:32,199 --> 02:07:36,429
it's quite consistent
with the gold standard,

2664
02:07:36,429 --> 02:07:39,099
gold standard meaning you
have actually acquired

2665
02:07:39,099 --> 02:07:42,043
the whole time duration in the patient,

2666
02:07:42,043 --> 02:07:45,179
and it's more consistent than something

2667
02:07:45,179 --> 02:07:49,779
that's just static like an
SUVR at the late time points.

2668
02:07:49,779 --> 02:07:51,259
So I think this is very promising

2669
02:07:51,259 --> 02:07:53,529
because you could get multiple information

2670
02:07:53,529 --> 02:07:55,909
about quantitative amyloid

2671
02:07:55,909 --> 02:07:59,729
and profusion from a shorter scan.

2672
02:07:59,729 --> 02:08:02,239
And I would encourage you to think

2673
02:08:02,239 --> 02:08:07,239
for your own tracers, whether
that's a neuroreceptor system,

2674
02:08:07,459 --> 02:08:11,489
could you also leverage
one scan with PET MR

2675
02:08:11,489 --> 02:08:14,569
to get multiple biomarkers
or cut down on the time

2676
02:08:14,569 --> 02:08:16,832
that the patient has to tolerate.

2677
02:08:18,839 --> 02:08:20,509
In other settings,

2678
02:08:20,509 --> 02:08:22,899
there are also clinical studies

2679
02:08:22,899 --> 02:08:27,899
where FDG PET as a marker
of glucose metabolism

2680
02:08:28,619 --> 02:08:32,469
is critical for patients
with cognitive impairment.

2681
02:08:32,469 --> 02:08:35,649
For instance, in delineating
Alzheimer's disease

2682
02:08:35,649 --> 02:08:39,959
from frontal parietal dementia.

2683
02:08:39,959 --> 02:08:44,079
And one question is instead of sending

2684
02:08:44,079 --> 02:08:48,369
a patient to get a separate
FDG scan in these situations,

2685
02:08:48,369 --> 02:08:52,559
would an MR, which typically
is required for structural

2686
02:08:53,509 --> 02:08:57,229
atrophy assessments or
longitudinal assessments,

2687
02:08:57,229 --> 02:09:00,159
would an MRI suffice.

2688
02:09:00,159 --> 02:09:04,009
In this case, we're showing ASL images

2689
02:09:04,009 --> 02:09:06,969
in healthy controls and AD patients.

2690
02:09:06,969 --> 02:09:10,359
And you can see not
surprisingly that the ASL

2691
02:09:10,359 --> 02:09:15,359
has a lot of spatial features
that are similar to FDG PET.

2692
02:09:15,729 --> 02:09:19,459
This is not surprising
because glucose metabolism

2693
02:09:19,459 --> 02:09:23,959
is so tightly coupled to profusion.

2694
02:09:23,959 --> 02:09:27,278
And I also thought it was
interesting in this study

2695
02:09:27,278 --> 02:09:29,149
where these arrows have pointed out

2696
02:09:29,149 --> 02:09:33,219
the deep gray matter as problematic areas

2697
02:09:33,219 --> 02:09:36,489
with ASL that would take
advanced strategies,

2698
02:09:36,489 --> 02:09:38,469
including deep learning to address,

2699
02:09:39,724 --> 02:09:44,209
but this special feature of
deep gray matter decaying

2700
02:09:44,209 --> 02:09:48,319
quicker was something that we
saw earlier in the talk too.

2701
02:09:48,319 --> 02:09:49,302
But on general,

2702
02:09:51,609 --> 02:09:52,919
if you look at this holistically

2703
02:09:53,807 --> 02:09:55,812
in both healthy controls and AD patients,

2704
02:09:56,880 --> 02:10:00,249
you see ASL could potentially
recapitulate the FDG signal

2705
02:10:00,249 --> 02:10:03,739
and help distinguish between
different patient types.

2706
02:10:03,739 --> 02:10:06,089
In this case, the statistical parametric

2707
02:10:07,239 --> 02:10:12,239
mapping results in this
study actually showed quite

2708
02:10:12,461 --> 02:10:16,606
a good overlap between FTG

2709
02:10:16,606 --> 02:10:19,269
and ASL based discrimination.

2710
02:10:19,269 --> 02:10:22,999
So what the clusters, the
colored clusters represent

2711
02:10:22,999 --> 02:10:27,009
here are areas of the brain where there's

2712
02:10:27,009 --> 02:10:30,909
less perfusion in AD patients
compared to controls.

2713
02:10:30,909 --> 02:10:33,419
And again, on visual average,

2714
02:10:33,419 --> 02:10:37,109
you can see quite a bit
of consistency between FDG

2715
02:10:37,109 --> 02:10:40,769
only based parametric statistics

2716
02:10:40,769 --> 02:10:42,074
and ASL only based,

2717
02:10:42,074 --> 02:10:43,749
there are some discrepancies

2718
02:10:43,749 --> 02:10:46,399
as shown in this right column here,

2719
02:10:46,399 --> 02:10:50,609
but I think on the balance,
it's also very promising.

2720
02:10:50,609 --> 02:10:51,979
And in this case,

2721
02:10:51,979 --> 02:10:56,339
could consider using ASL as a surrogate

2722
02:10:56,339 --> 02:11:00,419
to assess certain parts of the
vascular or metabolic health,

2723
02:11:00,419 --> 02:11:03,699
if FDT is typically used.

2724
02:11:03,699 --> 02:11:07,139
And that way you would avoid
using an additional tracer,

2725
02:11:07,139 --> 02:11:09,272
especially in neurodegenerative cases.

2726
02:11:10,359 --> 02:11:13,329
And finally, I wanna give an example

2727
02:11:14,229 --> 02:11:18,349
of beyond profusion studies now are taking

2728
02:11:18,349 --> 02:11:21,109
the next step and more challenging step

2729
02:11:21,109 --> 02:11:23,829
of validating other biomarkers,

2730
02:11:23,829 --> 02:11:28,004
and it's more challenging on both fronts.

2731
02:11:28,004 --> 02:11:32,282
So with PET, it actually
requires three separate tracers.

2732
02:11:34,952 --> 02:11:37,499
The established or reference
way to measure oxygen

2733
02:11:37,499 --> 02:11:41,819
metabolism is requiring
3 O-15 label tracers

2734
02:11:41,819 --> 02:11:46,279
to create CMR O2 maps
or oxygen extraction,

2735
02:11:46,279 --> 02:11:49,199
fraction maps that are shown here.

2736
02:11:49,199 --> 02:11:51,679
And with MR there's also more

2737
02:11:51,679 --> 02:11:55,646
and more new methods
to be able to map OEF,

2738
02:11:56,519 --> 02:11:59,659
but all of them are quite
complicated in terms

2739
02:11:59,659 --> 02:12:03,089
of modeling and require testing,

2740
02:12:03,089 --> 02:12:05,459
which is what this group from Cornell

2741
02:12:05,459 --> 02:12:07,522
is pursuing at this point.

2742
02:12:08,963 --> 02:12:11,509
So the MR maps here are
showing quantitative

2743
02:12:11,509 --> 02:12:14,639
BOLD plus a combination of susceptibility

2744
02:12:15,485 --> 02:12:18,609
mapping reconstructions
based on base signal,

2745
02:12:18,609 --> 02:12:21,369
again, a complicated MR metric.

2746
02:12:21,369 --> 02:12:25,799
But I think the same ideas
that we talked about earlier

2747
02:12:25,799 --> 02:12:29,879
in terms of using both
PET and MR simultaneously

2748
02:12:29,879 --> 02:12:32,259
will be a good framework in terms

2749
02:12:32,259 --> 02:12:36,209
of moving forward for even
more complicated biomarkers.

2750
02:12:36,209 --> 02:12:38,329
And I think this is very promising.

2751
02:12:38,329 --> 02:12:42,609
The study was nicely designed
because of repeat scans.

2752
02:12:42,609 --> 02:12:44,639
There is room for improvement.

2753
02:12:44,639 --> 02:12:46,989
So in this Bland Altman plot,

2754
02:12:46,989 --> 02:12:49,889
the OEF difference between PET and MR

2755
02:12:51,039 --> 02:12:55,179
had a limited range of difference
in terms of percentage.

2756
02:12:55,179 --> 02:12:58,409
So it was promising,
but you definitely see

2757
02:12:58,409 --> 02:13:03,208
a linear correlation where higher

2758
02:13:03,208 --> 02:13:07,679
OEF values tend to have
more positive difference.

2759
02:13:07,679 --> 02:13:10,052
And so there's this bias that,

2760
02:13:10,969 --> 02:13:15,492
I think with further PET MR
studies, with challenge studies,

2761
02:13:16,373 --> 02:13:19,049
like the cerebrovascular
reactivity we discuss

2762
02:13:19,049 --> 02:13:22,169
can actually change the oxygen
extraction fraction as well,

2763
02:13:22,169 --> 02:13:24,319
and use PET MR to continue

2764
02:13:24,319 --> 02:13:29,086
to improve the parameters on the MR side,

2765
02:13:29,086 --> 02:13:33,482
get better quantitative comparisons
with PET and validation.

2766
02:13:35,199 --> 02:13:38,689
And if you want more framework for this,

2767
02:13:38,689 --> 02:13:43,059
there's a review paper
that I wrote with some

2768
02:13:43,059 --> 02:13:46,099
of my Stanford colleagues
that kind of has a perspective

2769
02:13:46,099 --> 02:13:49,629
on the next step of O-15 gas PET

2770
02:13:49,629 --> 02:13:53,019
and how we could use PET
MR to actually validate

2771
02:13:53,019 --> 02:13:57,319
even more difficult
biomarkers like blood volume,

2772
02:13:57,319 --> 02:14:01,169
oxygen extraction fraction
beyond the perfusion

2773
02:14:01,169 --> 02:14:03,349
that we focused on in this talk.

2774
02:14:03,349 --> 02:14:05,119
So in summary,

2775
02:14:05,119 --> 02:14:07,939
I hope I've given you a
nice perspective today,

2776
02:14:07,939 --> 02:14:12,839
with example cases in stroke
and cerebrovascular disorders

2777
02:14:12,839 --> 02:14:14,839
of how we could dial and tailor

2778
02:14:14,839 --> 02:14:19,839
ASL MRI to match a simultaneous

2779
02:14:19,979 --> 02:14:24,199
physiologically matched
PET reference scan,

2780
02:14:24,199 --> 02:14:26,859
both at baseline and during a stress test,

2781
02:14:26,859 --> 02:14:27,909
that could be helpful

2782
02:14:28,856 --> 02:14:33,276
in teasing out additional
auto regulatory pathology.

2783
02:14:34,259 --> 02:14:37,419
I hope there's some new insight also

2784
02:14:37,419 --> 02:14:41,629
in terms of input functions scaling,

2785
02:14:41,629 --> 02:14:43,859
and deep learning methods

2786
02:14:43,859 --> 02:14:46,579
that really use simultaneous
PET MR information

2787
02:14:46,579 --> 02:14:50,669
to get better perfusion
maps quantitatively.

2788
02:14:50,669 --> 02:14:55,519
And that this framework
with PET MRI focused

2789
02:14:56,441 --> 02:14:59,869
on physiology is not just
applicable to stroke,

2790
02:14:59,869 --> 02:15:03,319
but I encourage you to think beyond

2791
02:15:03,319 --> 02:15:04,939
the tracers that we've talked about,

2792
02:15:04,939 --> 02:15:08,599
but also to the patients that
you're most interested in

2793
02:15:08,599 --> 02:15:11,539
and see whether some of these concepts

2794
02:15:11,539 --> 02:15:15,042
can also help design studies
that are less invasive,

2795
02:15:15,042 --> 02:15:19,779
that are more accurate for your scientific

2796
02:15:19,779 --> 02:15:21,069
and clinical needs.

2797
02:15:21,069 --> 02:15:21,902
With that,

2798
02:15:21,902 --> 02:15:24,829
I wanna thank you for your
attention and acknowledge

2799
02:15:24,829 --> 02:15:28,559
many excellent collaborators,

2800
02:15:28,559 --> 02:15:30,429
I think, and my group,

2801
02:15:30,429 --> 02:15:31,929
as well as my funding sources,

2802
02:15:31,929 --> 02:15:35,089
none of this would've been
possible without your support.

2803
02:15:35,089 --> 02:15:36,939
And I look forward to chatting

2804
02:15:36,939 --> 02:15:39,369
with you more at the discussion session.

2805
02:15:39,369 --> 02:15:40,512
Thank you very much.

2807
02:15:43,672 --> 02:15:46,772
- Hello everybody and
welcome to my presentation

2807
02:15:46,772 --> 02:15:50,392
where I would like to highlight
some of the advantages

2808
02:15:50,392 --> 02:15:53,572
to combine functional
PET and functional MRI

2809
02:15:53,572 --> 02:15:57,532
in the assessment of task
specific brain responses.

2810
02:15:57,532 --> 02:15:59,082
My name is Andreas Hahn.

2811
02:15:59,082 --> 02:16:02,992
I'm located at the Medical
University of Vienna in Austria,

2812
02:16:02,992 --> 02:16:06,512
at the Department of
Psychiatry and Psychotherapy.

2813
02:16:06,512 --> 02:16:10,152
And there, I am part of
the neuroimaging labs,

2814
02:16:10,152 --> 02:16:12,115
whose head is Rupert Lanzenberger.

2815
02:16:13,462 --> 02:16:17,442
I would like to thank the NIMH
for organizing this workshop

2816
02:16:17,442 --> 02:16:20,292
and for inviting me and giving me the

2817
02:16:20,292 --> 02:16:21,982
opportunity to talk here.

2818
02:16:21,982 --> 02:16:23,395
It's a particular pleasure.

2819
02:16:24,802 --> 02:16:26,752
I would like to start
off with some background

2820
02:16:26,752 --> 02:16:28,825
about FDG imaging.

2821
02:16:29,972 --> 02:16:32,632
It's an irreversible radioligand

2822
02:16:32,632 --> 02:16:35,082
for imaging glucose metabolism,

2823
02:16:35,082 --> 02:16:37,872
which means that after
a bolus application,

2824
02:16:37,872 --> 02:16:41,542
you have this irreversible
uptake into the cell,

2825
02:16:41,542 --> 02:16:44,122
which results in the steady increase

2826
02:16:44,122 --> 02:16:45,535
of the time activity curve.

2827
02:16:46,792 --> 02:16:50,932
One can also image task specific
effects with this approach.

2828
02:16:50,932 --> 02:16:53,502
However, only if the task

2829
02:16:53,502 --> 02:16:56,372
is carried out in the
beginning of the scan.

2830
02:16:56,372 --> 02:16:58,632
And this applies that
repeated measurements

2831
02:16:58,632 --> 02:17:01,732
would be necessary,
which in turn increases

2832
02:17:01,732 --> 02:17:04,102
the test-retest reliability.

2833
02:17:04,102 --> 02:17:07,622
Another option would be to use

2834
02:17:07,622 --> 02:17:11,132
repeated applications within
the same scanning session

2835
02:17:11,132 --> 02:17:12,242
as shown here.

2836
02:17:12,242 --> 02:17:15,412
But this still would
require about 30-70 minutes

2837
02:17:15,412 --> 02:17:19,362
per state for a reliable quantification.

2838
02:17:19,362 --> 02:17:21,532
And also, it's kind of a complex setting

2839
02:17:21,532 --> 02:17:26,295
and complex modeling so it's
not a really popular approach.

2840
02:17:27,782 --> 02:17:31,442
A neat solution to this problem
is the constant infusion

2841
02:17:31,442 --> 02:17:35,962
of FDG because the metabolism
is proportional to the slope

2842
02:17:35,962 --> 02:17:38,832
of the time activity curve,
and this can be easily

2843
02:17:38,832 --> 02:17:41,612
quantified with graphical
solutions such as

2844
02:17:41,612 --> 02:17:42,505
the pattern plot.

2845
02:17:44,192 --> 02:17:47,382
So when using constant
infusion, there will be free

2846
02:17:47,382 --> 02:17:50,899
radioligand available throughout the scan

2847
02:17:50,899 --> 02:17:54,362
and also for the actual metabolism
and the metabolic demands

2848
02:17:54,362 --> 02:17:56,672
of a certain task.

2849
02:17:56,672 --> 02:17:59,305
And this means that any
change in the metabolism

2850
02:17:59,305 --> 02:18:02,512
will also result in a proportional change

2851
02:18:02,512 --> 02:18:04,165
of the slope of the TAC.

2852
02:18:05,342 --> 02:18:08,722
And task specific assessment
has been successfully

2853
02:18:08,722 --> 02:18:12,802
demonstrated by pioneering
work of (indistinct) et al,

2854
02:18:12,802 --> 02:18:14,502
in 2014.

2855
02:18:14,502 --> 02:18:17,782
However, this was only done
in three subjects so far.

2856
02:18:17,782 --> 02:18:22,782
So the first aim we had
was to replicate this and

2857
02:18:23,082 --> 02:18:24,532
validate this.

2858
02:18:24,532 --> 02:18:27,672
So we had a task, very simple,
eyes open, eyes closed,

2859
02:18:27,672 --> 02:18:31,415
and right finger tapping
two times 10 minutes each.

2860
02:18:33,822 --> 02:18:37,692
And with this example I will
show you the general workflow

2861
02:18:37,692 --> 02:18:41,732
of the fPET analysis with
general linear model.

2862
02:18:41,732 --> 02:18:44,802
So basically we extract
time activity curve here,

2863
02:18:44,802 --> 02:18:47,592
for instance, for the precentral gyrus

2864
02:18:47,592 --> 02:18:49,752
and we have our regressors.

2865
02:18:49,752 --> 02:18:52,982
One which describes the
baseline metabolism,

2866
02:18:52,982 --> 02:18:57,032
and one, the task for eyes open here,

2867
02:18:57,032 --> 02:18:59,872
and the other one for finger tapping.

2868
02:18:59,872 --> 02:19:03,872
And we feed all of this
into a general linear model

2869
02:19:03,872 --> 02:19:05,922
and that's, I think also
where the name comes from

2870
02:19:05,922 --> 02:19:07,642
because it's so similar to fMRI,

2871
02:19:08,732 --> 02:19:11,062
naming this technique functional PET.

2872
02:19:11,062 --> 02:19:15,972
And with the TLM, we're able
to distinguish or resolve

2873
02:19:15,972 --> 02:19:19,702
baseline activity, separate this from

2874
02:19:19,702 --> 02:19:22,432
task specific activity.

2875
02:19:22,432 --> 02:19:26,142
So basically, the curve here and here

2876
02:19:26,142 --> 02:19:29,902
summing together will again,
result in this curve here.

2877
02:19:29,902 --> 02:19:32,692
And after separation,
we can do quantification

2878
02:19:32,692 --> 02:19:35,426
for each of these conditions
with, for instance,

2879
02:19:35,426 --> 02:19:37,292
a pattern flow.

2880
02:19:37,292 --> 02:19:40,345
So let's have a more detailed
look at the quantification.

2881
02:19:41,382 --> 02:19:44,765
Here's the equation for
a bonus application.

2882
02:19:45,982 --> 02:19:47,742
And it's known as a graphical solution.

2883
02:19:47,742 --> 02:19:51,872
So basically plotting
this term on the Y-axis

2884
02:19:51,872 --> 02:19:53,712
and this term on the X-axis

2885
02:19:53,712 --> 02:19:56,452
yields a curve that will look like this,

2886
02:19:56,452 --> 02:20:00,032
and importantly, it gets
linear after a certain time.

2887
02:20:00,032 --> 02:20:01,872
And the slope of this

2888
02:20:03,212 --> 02:20:07,622
linear curve describes
our glucose metabolism,

2889
02:20:07,622 --> 02:20:10,625
or here in this case, the
net influx constant, Ki.

2890
02:20:12,832 --> 02:20:14,492
For the constant infusion, as I said,

2891
02:20:14,492 --> 02:20:16,342
we use the general linear model.

2892
02:20:16,342 --> 02:20:18,092
Here's the basic equation.

2893
02:20:18,092 --> 02:20:22,492
And our data, Y, is given
by the time activity curve,

2894
02:20:22,492 --> 02:20:25,082
then we have our beta
estimates, which we want to

2895
02:20:25,082 --> 02:20:28,702
get out of this, and the design matrix, X,

2896
02:20:28,702 --> 02:20:32,102
which is given by the
regressors as I showed before.

2897
02:20:32,102 --> 02:20:34,912
One our baseline, one for finger tapping,

2898
02:20:34,912 --> 02:20:38,632
for eyes open and an
additional one to correct for

2899
02:20:38,632 --> 02:20:39,465
head motion.

2900
02:20:41,442 --> 02:20:44,182
For the pattern plot, we could see

2901
02:20:44,182 --> 02:20:48,402
that the glucose metabolism
is directly proportional

2902
02:20:48,402 --> 02:20:51,402
to the concentration in tissue.

2903
02:20:51,402 --> 02:20:55,402
Which again, is the baseline regressor

2904
02:20:55,402 --> 02:20:57,102
times the beta estimate.

2905
02:20:58,122 --> 02:21:00,922
So if we then have a task, then

2906
02:21:00,922 --> 02:21:04,452
the total glucose metabolism will

2907
02:21:04,452 --> 02:21:08,552
be the sum of the baseline
metabolism and as well,

2908
02:21:08,552 --> 02:21:11,312
instead of the task given by
the regressor of the task,

2909
02:21:11,312 --> 02:21:13,262
again, multiplied by the beta estimate.

2910
02:21:14,152 --> 02:21:18,712
So this means that the task
specific metabolism is given

2911
02:21:18,712 --> 02:21:21,642
by our task regressor
times the beta estimate.

2912
02:21:21,642 --> 02:21:25,222
Because, as we said, any
change in the metabolism

2913
02:21:25,222 --> 02:21:28,012
will result in a proportional
change of the slope

2914
02:21:28,012 --> 02:21:29,505
of the time activity curve.

2915
02:21:32,662 --> 02:21:34,792
The baseline regressors
as we defined them,

2916
02:21:34,792 --> 02:21:36,735
where the baseline,

2917
02:21:38,592 --> 02:21:41,742
was given as the average
of the whole gray matter,

2918
02:21:41,742 --> 02:21:43,222
all gray matter voxels,

2919
02:21:43,222 --> 02:21:45,872
modeled as a third order polynomial,

2920
02:21:45,872 --> 02:21:48,335
and also correcting for task effects.

2921
02:21:49,782 --> 02:21:52,682
Then we had the task regressors.

2922
02:21:52,682 --> 02:21:55,072
They were defined as a ramp
function with the slope

2923
02:21:55,072 --> 02:21:56,412
of 1 kBq/frame.

2924
02:21:56,412 --> 02:22:00,772
So basically always when
this ramps up, this indicates

2925
02:22:00,772 --> 02:22:02,232
active task block.

2926
02:22:02,232 --> 02:22:05,032
And then as a movement regressor
we used the first principle

2927
02:22:05,032 --> 02:22:08,235
component of six realignment parameters.

2928
02:22:11,254 --> 02:22:16,002
For the first results, in the
first study, we compared the

2929
02:22:16,002 --> 02:22:18,622
baseline metabolism that
was obtained from this

2930
02:22:18,622 --> 02:22:22,612
infusion protocol to a
common baseline protocol,

2931
02:22:22,612 --> 02:22:26,492
just to validate if this
is actually very similar,

2932
02:22:26,492 --> 02:22:30,082
and indeed it is, we
saw very good agreement

2933
02:22:30,082 --> 02:22:32,172
despite with some variants, but this is

2934
02:22:32,172 --> 02:22:36,222
expected because the scans
took place on a separate day.

2935
02:22:36,222 --> 02:22:39,212
And as I said before, this
will be subject to some

2936
02:22:39,212 --> 02:22:40,605
test-retest variability.

2937
02:22:42,492 --> 02:22:45,652
More importantly, for the
task specific metabolism,

2938
02:22:45,652 --> 02:22:48,572
we could absolutely quantify these changes

2939
02:22:48,572 --> 02:22:52,612
in micro mole per hundred gram per minute.

2940
02:22:52,612 --> 02:22:56,472
And it was about 20% change from baseline.

2941
02:22:56,472 --> 02:22:59,422
So the approach was sensitive to the task.

2942
02:22:59,422 --> 02:23:01,632
It was originally specific
because for instance,

2943
02:23:01,632 --> 02:23:03,662
for the eyes open condition,
we didn't see anything

2944
02:23:03,662 --> 02:23:06,562
in the motor cortex
and for finger tapping,

2945
02:23:06,562 --> 02:23:09,712
we didn't see anything
in the visual cortex.

2946
02:23:09,712 --> 02:23:11,662
And the most important
point is that we could image

2947
02:23:11,662 --> 02:23:14,822
all of these effects in a single PET scan.

2948
02:23:14,822 --> 02:23:19,822
So this is the perfect basis
for combining this with fMRI

2949
02:23:19,992 --> 02:23:21,715
with hybrid PET-MR scanners.

2950
02:23:23,722 --> 02:23:27,182
Of course the approach is
subject to several assumptions

2951
02:23:27,182 --> 02:23:28,512
and limitations.

2952
02:23:28,512 --> 02:23:31,032
For instance, the Patlak plot assumes that

2953
02:23:31,032 --> 02:23:33,992
the rate consent, k4, is zero,

2954
02:23:33,992 --> 02:23:36,822
which is not exactly the case.

2955
02:23:36,822 --> 02:23:39,852
So we have an underestimation
of our glucose metabolism,

2956
02:23:39,852 --> 02:23:42,342
which is basically known.

2957
02:23:42,342 --> 02:23:44,902
Another assumption is that
the coupling between the

2958
02:23:44,902 --> 02:23:47,762
neuronal response and
the glucose metabolism

2959
02:23:47,762 --> 02:23:51,582
is linear and time-invariant.

2960
02:23:51,582 --> 02:23:55,382
Linear means that any change
in the neural response

2961
02:23:55,382 --> 02:23:58,992
will translate into a proportional change

2962
02:23:58,992 --> 02:24:03,122
of the glucose metabolism,
and time-invariant means that

2963
02:24:03,122 --> 02:24:05,232
these affects

2964
02:24:06,331 --> 02:24:08,532
are also temporarily coupled.

2965
02:24:08,532 --> 02:24:10,792
These assumptions basically
come from fMRI imaging,

2966
02:24:10,792 --> 02:24:13,202
but I think they're still very useful here

2967
02:24:13,202 --> 02:24:14,395
and also apply here.

2968
02:24:16,212 --> 02:24:18,572
We also assume that the glucose metabolism

2969
02:24:18,572 --> 02:24:21,642
is at steady state also during the task.

2970
02:24:21,642 --> 02:24:24,912
That's not really the case
for all of the functional

2971
02:24:24,912 --> 02:24:28,432
imaging approaches, but still
it's a required approximation.

2972
02:24:28,432 --> 02:24:32,342
And furthermore, we assume
that the glucose metabolism's

2973
02:24:32,342 --> 02:24:34,205
constant throughout the task.

2974
02:24:36,722 --> 02:24:39,452
It has been proposed
that different regressors

2975
02:24:40,542 --> 02:24:43,242
that are a bit different
from a simple ramp function

2976
02:24:43,242 --> 02:24:45,662
might be more accurate,

2977
02:24:45,662 --> 02:24:48,025
but this is subject to further research.

2978
02:24:50,822 --> 02:24:55,652
We also want to improve the
fPET methodology a little bit.

2979
02:24:55,652 --> 02:24:58,682
So there are several approaches
to obtain a task-free

2980
02:24:58,682 --> 02:25:00,092
baseline regressor.

2981
02:25:00,092 --> 02:25:03,452
I would just like to stress
that it's very important

2982
02:25:03,452 --> 02:25:06,812
that your baseline regressor
is free of any task effects,

2983
02:25:06,812 --> 02:25:10,972
because otherwise it will
bias your task estimate.

2984
02:25:10,972 --> 02:25:14,352
So you could choose, for
instance, a third order polynomial

2985
02:25:14,352 --> 02:25:15,582
as we have done before.

2986
02:25:15,582 --> 02:25:16,902
It worked pretty good.

2987
02:25:16,902 --> 02:25:18,582
I would just like to mention that

2988
02:25:18,582 --> 02:25:21,442
it looks very much as
like a linear function,

2989
02:25:21,442 --> 02:25:24,622
the time activity occurs,
but clearly it's not.

2990
02:25:24,622 --> 02:25:28,482
And again, an ill defined baseline will

2991
02:25:28,482 --> 02:25:30,155
mess up your task effects.

2992
02:25:31,432 --> 02:25:34,872
One could also choose
an atlas-based approach

2993
02:25:34,872 --> 02:25:38,432
basically by excluding all the

2994
02:25:38,432 --> 02:25:41,452
brain regions that are assumed
to have some task activation

2995
02:25:41,452 --> 02:25:44,752
and take just the time activity
curve from the remaining

2996
02:25:44,752 --> 02:25:48,802
gray matter voxels or since
we are talking about PET-MRI,

2997
02:25:48,802 --> 02:25:51,642
one could use the
individual fMRI estimates,

2998
02:25:51,642 --> 02:25:53,642
the individual estimates

2999
02:25:55,949 --> 02:25:58,675
which voxels are active in a certain task.

3000
02:26:00,183 --> 02:26:02,482
We have basically shown
that the last approach

3001
02:26:02,482 --> 02:26:04,942
is the most robust one.

3002
02:26:04,942 --> 02:26:09,642
But if one doesn't want
to rely on the fMRI,

3003
02:26:09,642 --> 02:26:12,815
the other two approaches
are still quite valid.

3004
02:26:14,972 --> 02:26:17,572
Another thing is that we

3005
02:26:17,572 --> 02:26:22,332
also included a bolus
plus an initial bolus,

3006
02:26:22,332 --> 02:26:27,332
and then started with the
infusion in a ratio of 20% to 80%.

3007
02:26:27,586 --> 02:26:29,182
This is basically done to increase the

3008
02:26:29,182 --> 02:26:30,722
signal to noise ratio.

3009
02:26:30,722 --> 02:26:34,972
And I think the images
here after 10 minutes

3010
02:26:34,972 --> 02:26:36,212
speak for themselves.

3011
02:26:36,212 --> 02:26:39,822
You can clearly see much
better signal in the

3012
02:26:39,822 --> 02:26:41,922
bolus plus infusion protocol on the right.

3013
02:26:43,182 --> 02:26:46,382
So with this increased SNR,

3014
02:26:46,382 --> 02:26:50,382
this might enable us to
decrease the task duration,

3015
02:26:50,382 --> 02:26:52,432
because in the beginning,
we had a task duration

3016
02:26:52,432 --> 02:26:56,062
of 10 minutes and show
very stable task effects.

3017
02:26:56,062 --> 02:26:59,162
Was still very nice for five minutes here.

3018
02:26:59,162 --> 02:27:04,162
It's also possible to go
lower with the timing.

3019
02:27:04,512 --> 02:27:07,712
But of course, then the
frames will also get shorter

3020
02:27:07,712 --> 02:27:09,402
and

3021
02:27:09,402 --> 02:27:10,762
this means less counts

3022
02:27:10,762 --> 02:27:13,195
and this then means more noise.

3023
02:27:14,712 --> 02:27:19,012
But still doing so and increasing SNR,

3024
02:27:19,012 --> 02:27:21,055
decreasing task duration, we could show

3025
02:27:21,055 --> 02:27:25,612
a very nice spatial overlap
between glucose metabolism

3026
02:27:25,612 --> 02:27:28,442
shown here on the right
and the bolus signal,

3027
02:27:28,442 --> 02:27:31,132
sorry, glucose metabolism on
the left and the bolus signal

3028
02:27:31,132 --> 02:27:34,442
on the right, and also
some nice specificity

3029
02:27:34,442 --> 02:27:37,552
in this example during
finger tapping especially

3030
02:27:37,552 --> 02:27:38,695
in the motor cortex.

3031
02:27:42,782 --> 02:27:46,062
Next step, we wanted to go
a little bit more complex

3032
02:27:46,062 --> 02:27:49,542
and getting away from these
simple visual and motor

3033
02:27:49,542 --> 02:27:50,952
stimulation tasks.

3034
02:27:50,952 --> 02:27:52,262
So we had a task

3035
02:27:53,613 --> 02:27:55,272
with a task design

3036
02:27:55,272 --> 02:27:58,872
that included PET imaging for 52 minutes

3037
02:28:00,032 --> 02:28:01,822
with simultaneous MRI imaging.

3038
02:28:01,822 --> 02:28:05,382
For instance, old imaging for connectivity

3039
02:28:05,382 --> 02:28:07,272
as well as arterial spin labeling

3040
02:28:09,252 --> 02:28:11,982
for estimation of cerebral blood flow.

3041
02:28:11,982 --> 02:28:14,512
And in the end, a
conventional block design

3042
02:28:15,412 --> 02:28:19,942
for fMRI BOLD based neural activation.

3043
02:28:19,942 --> 02:28:23,022
And that as a task, we use
the video game, "Tetris,"

3044
02:28:23,022 --> 02:28:26,882
which is very well known,
that requires rapid visual

3045
02:28:26,882 --> 02:28:30,702
spatial motor coordination
and we had two task levels,

3046
02:28:30,702 --> 02:28:34,592
and the hard one basically the
bricks were falling down with

3047
02:28:36,042 --> 02:28:40,302
more speed and there were
more incomplete lines

3048
02:28:40,302 --> 02:28:42,225
filled at the bottom.

3049
02:28:43,522 --> 02:28:46,322
So I would just like
to say that this was a

3050
02:28:46,322 --> 02:28:48,862
impolitic continuous task design here,

3051
02:28:48,862 --> 02:28:53,392
so people really played
six minutes continuously,

3052
02:28:53,392 --> 02:28:56,312
and also the acquisition
of bolus was continuously.

3053
02:28:56,312 --> 02:28:59,462
So there was no on/off design here.

3054
02:28:59,462 --> 02:29:03,852
In contrast, one could also
employ a hierarchical design

3055
02:29:03,852 --> 02:29:08,342
where basically the BOLD
blocks, fast BOLD block design

3056
02:29:08,342 --> 02:29:11,292
is embedded in shorter blocks,

3057
02:29:11,292 --> 02:29:14,965
in longer blocks of the PET acquisition.

3058
02:29:17,442 --> 02:29:20,752
It really just depends basically
on which kind of question

3059
02:29:20,752 --> 02:29:21,702
you want to answer.

3060
02:29:23,672 --> 02:29:27,122
As a result, we really
saw a nice overlap between

3061
02:29:27,122 --> 02:29:30,212
the different parameters
of neuronal activation

3062
02:29:30,212 --> 02:29:31,922
as shown for instance on the right,

3063
02:29:31,922 --> 02:29:34,952
this is the spatial overlap of all three

3064
02:29:34,952 --> 02:29:38,315
different parameters that
show significant activation.

3065
02:29:39,692 --> 02:29:44,132
And most of these were also
sensitive to subtle changes

3066
02:29:44,132 --> 02:29:47,372
in cognitive demands.

3067
02:29:47,372 --> 02:29:50,492
Which means that we found
changes in glucose metabolism

3068
02:29:50,492 --> 02:29:53,962
between the easy and hard task level for

3069
02:29:55,262 --> 02:29:57,692
functional PETs, for BOLD imaging

3070
02:29:57,692 --> 02:30:01,455
and some effects also for
CBF, but not as strong.

3071
02:30:03,252 --> 02:30:04,962
Still, I would like to
highlight that these are

3072
02:30:04,962 --> 02:30:07,912
complementary aspects
of neural activation.

3073
02:30:07,912 --> 02:30:11,082
Because the BOLD signal
basically emerged from

3074
02:30:12,136 --> 02:30:14,642
a very high increase
in cerebral blood flow,

3075
02:30:14,642 --> 02:30:19,212
but less pronounced increase in the

3076
02:30:19,212 --> 02:30:22,385
metabolism of oxygen.

3077
02:30:24,372 --> 02:30:28,282
On the other hand, glutamate
release is actually very

3078
02:30:28,282 --> 02:30:31,772
highly involved in neuronal
activation and this

3079
02:30:31,772 --> 02:30:34,382
causes an increase in cerebral blood flow,

3080
02:30:34,382 --> 02:30:39,382
but this also lead to an increased

3081
02:30:39,962 --> 02:30:43,305
use of glucose in the neuronal cells.

3082
02:30:44,142 --> 02:30:47,482
So basically, we are
imaging the same phenomenon,

3083
02:30:47,482 --> 02:30:50,465
namely neural activation,
from different points of view.

3084
02:30:52,692 --> 02:30:55,392
We also assessed the
test-retest reliability

3085
02:30:55,392 --> 02:30:56,662
of this approach,

3086
02:30:56,662 --> 02:30:59,345
all of these approaches
and compared this directly.

3087
02:31:00,562 --> 02:31:02,572
The strongest reliability giving us the

3088
02:31:02,572 --> 02:31:04,725
interclass correlation
coefficient was given for

3089
02:31:04,725 --> 02:31:07,325
glucose metabolism at resting state.

3090
02:31:08,482 --> 02:31:11,312
Was also quite good for
cerebral blood flow.

3091
02:31:11,312 --> 02:31:13,352
It was not available for
both because we didn't use

3092
02:31:13,352 --> 02:31:15,112
the calibrated bolt.

3093
02:31:15,112 --> 02:31:17,502
On the other hand, for task effects,

3094
02:31:17,502 --> 02:31:20,032
we saw a good test-retest variability.

3095
02:31:20,032 --> 02:31:24,602
That was actually done with a
temporal delay of four weeks.

3096
02:31:24,602 --> 02:31:27,182
And was still good for glucose metabolism

3097
02:31:27,182 --> 02:31:29,642
and cerebral blood flow, very similar.

3098
02:31:29,642 --> 02:31:31,132
But it was not that good

3099
02:31:33,222 --> 02:31:35,872
for BOLD imaging.

3100
02:31:35,872 --> 02:31:38,892
And actually, this low
number is very much in line

3101
02:31:38,892 --> 02:31:41,202
with the recent meta analysis,

3102
02:31:41,202 --> 02:31:45,292
which has been done on
numerous BOLD tasks.

3103
02:31:45,292 --> 02:31:47,802
So it could be related
to the fact that BOLD

3104
02:31:47,802 --> 02:31:51,242
is more a relative signal,
whereas the other two parameters

3105
02:31:51,242 --> 02:31:53,295
can be quantified absolutely.

3106
02:31:57,592 --> 02:32:00,132
Let's come back to our study design.

3107
02:32:00,132 --> 02:32:04,832
We did an identification of
brain regions and the networks

3108
02:32:04,832 --> 02:32:07,092
that are active during these tasks.

3109
02:32:07,092 --> 02:32:08,742
And in the next step,

3110
02:32:08,742 --> 02:32:13,202
we wanted to assess the
interaction between these networks.

3111
02:32:13,202 --> 02:32:16,732
And we did this in a collaboration
with Luke (indistinct)

3112
02:32:16,732 --> 02:32:20,465
and Michael (indistinct) from Australia.

3113
02:32:22,232 --> 02:32:25,232
So in the first step we used
the method that's called

3114
02:32:25,232 --> 02:32:26,922
metabolic connectivity mapping,

3115
02:32:26,922 --> 02:32:30,092
and this combines functional connectivity

3116
02:32:30,092 --> 02:32:33,042
and metabolism.

3117
02:32:33,042 --> 02:32:35,695
I will explain this in a second.

3118
02:32:36,572 --> 02:32:38,872
So basically this gives an estimate

3119
02:32:38,872 --> 02:32:41,142
of directional connectivity.

3120
02:32:41,142 --> 02:32:44,052
But the gold standard
really for estimating this

3121
02:32:44,052 --> 02:32:45,652
is dynamic causal modeling.

3122
02:32:45,652 --> 02:32:49,692
So we use this method as well

3123
02:32:49,692 --> 02:32:52,335
to validate the MCM results.

3124
02:32:54,712 --> 02:32:58,052
As I said, metabolic connectivity mapping

3125
02:32:58,052 --> 02:33:01,322
relates the spatial patterns
of glucose metabolism

3126
02:33:01,322 --> 02:33:02,505
and connectivity.

3127
02:33:04,157 --> 02:33:07,152
In the first step, one computes
seed-to- voxel connectivity.

3128
02:33:07,152 --> 02:33:10,752
This is all known, basically
correlating fMRI time series,

3129
02:33:10,752 --> 02:33:14,242
and then one gets a connectivity
value between two regions,

3130
02:33:14,242 --> 02:33:16,175
but this is undirected connectivity.

3131
02:33:17,982 --> 02:33:22,012
So we can now include the
information of glucose metabolism

3132
02:33:22,012 --> 02:33:23,995
to identify the target region.

3133
02:33:27,372 --> 02:33:30,052
Because it is known
that the energy demands

3134
02:33:30,052 --> 02:33:32,642
mostly emerge post-synaptically.

3135
02:33:32,642 --> 02:33:35,592
And this can be used to
imply some directionality

3136
02:33:35,592 --> 02:33:37,482
to these regions.

3137
02:33:37,482 --> 02:33:41,092
So basically, one uses the pattern,

3138
02:33:41,092 --> 02:33:42,852
the spatial pattern of
(indistinct) pattern,

3139
02:33:42,852 --> 02:33:46,042
of functional connectivity
and correlates this

3140
02:33:46,042 --> 02:33:48,505
with the spatial pattern of the underlying

3141
02:33:48,505 --> 02:33:50,312
glucose metabolism.

3142
02:33:50,312 --> 02:33:54,802
So if connectivity between region Y

3143
02:33:54,802 --> 02:33:57,412
to region X is really causal,

3144
02:33:57,412 --> 02:34:00,052
then the pattern of
functional connectivity

3145
02:34:00,052 --> 02:34:03,752
should match that of glucose metabolism

3146
02:34:03,752 --> 02:34:05,732
because of the strong coupling

3147
02:34:05,732 --> 02:34:07,835
between the different signals.

3148
02:34:11,212 --> 02:34:13,842
We assessed this for the
three major brain regions

3149
02:34:13,842 --> 02:34:15,602
that were active during the task,

3150
02:34:15,602 --> 02:34:18,862
namely, the occipital
cortex, intraparietal sulcus

3151
02:34:18,862 --> 02:34:21,162
and frontal eye field.

3152
02:34:21,162 --> 02:34:25,052
And what we saw that the
connection from IPS to FEF was

3153
02:34:26,742 --> 02:34:30,432
very significant and this was
stable for rest (indistinct),

3154
02:34:30,432 --> 02:34:33,782
so it didn't really
change across conditions.

3155
02:34:33,782 --> 02:34:35,682
However, for the other connections,

3156
02:34:35,682 --> 02:34:40,112
we saw a strong increase
in the association of

3157
02:34:40,112 --> 02:34:42,542
glucose metabolism and connectivity.

3158
02:34:42,542 --> 02:34:45,022
And this was mostly for the
feet forward connections

3159
02:34:45,022 --> 02:34:47,815
from the occipital
cortex to other regions.

3160
02:34:49,272 --> 02:34:52,162
We did not see this for
the feedback connection,

3161
02:34:52,162 --> 02:34:54,115
so it was a specific finding.

3162
02:34:56,432 --> 02:34:59,882
And interestingly, once the task started,

3163
02:34:59,882 --> 02:35:01,672
this basically reached the plateau.

3164
02:35:01,672 --> 02:35:06,062
So there was no change from the
easy to the hard task level.

3165
02:35:06,062 --> 02:35:08,062
So when increasing cognitive demands,

3166
02:35:08,062 --> 02:35:10,615
it didn't change any more.

3167
02:35:11,662 --> 02:35:15,732
And also we confirmed this with
the dynamic causal modeling.

3168
02:35:15,732 --> 02:35:19,692
I just want to highlight
that the DCM model used

3169
02:35:19,692 --> 02:35:22,442
completely independent data basically

3170
02:35:22,442 --> 02:35:25,752
than used in the MCM and still we observed

3171
02:35:25,752 --> 02:35:28,255
a high correspondence
between these models.

3172
02:35:31,022 --> 02:35:35,492
So this means that the
non-reconfiguration of functional

3173
02:35:35,492 --> 02:35:39,252
networks during task
performance are indeed dependent

3174
02:35:39,252 --> 02:35:41,085
or related to glucose metabolism.

3175
02:35:43,102 --> 02:35:46,512
And importantly, the transition from rest

3176
02:35:48,235 --> 02:35:51,822
to the task carries actually
the major metabolic cost.

3177
02:35:53,252 --> 02:35:56,032
It would be interesting to
assess these task specific

3178
02:35:56,032 --> 02:35:59,772
changes also in disorders
such as Alzheimer's disease,

3179
02:35:59,772 --> 02:36:01,772
because it has already been shown that

3180
02:36:06,022 --> 02:36:08,082
baseline effects are altered and

3181
02:36:09,222 --> 02:36:12,522
show alterations in the
association of glucose metabolism

3182
02:36:12,522 --> 02:36:13,705
and connectivity.

3183
02:36:16,452 --> 02:36:20,042
So this result I showed before
was for the first PET scan.

3184
02:36:20,042 --> 02:36:24,392
However, we wanted to know
what happens now if the task

3185
02:36:24,392 --> 02:36:27,562
is trained for several weeks;
in this case four weeks.

3186
02:36:27,562 --> 02:36:30,723
So we did a second PET scan
after this training period.

3187
02:36:30,723 --> 02:36:34,512
We split up the groups
into training and control.

3188
02:36:34,512 --> 02:36:37,052
Control group did no
learning, so it was a passive

3189
02:36:37,052 --> 02:36:38,305
comparison basically.

3190
02:36:41,629 --> 02:36:45,882
The first step was,
again, to investigate the

3191
02:36:45,882 --> 02:36:48,392
energy demands regionally.

3192
02:36:48,392 --> 02:36:50,662
This was basically the same as done before

3193
02:36:50,662 --> 02:36:53,515
and confirm the previous
results in a larger sample.

3194
02:36:54,522 --> 02:36:57,222
And then we extended
metabolic connectivity mapping

3195
02:36:57,222 --> 02:36:58,462
to the voxel level.

3196
02:36:58,462 --> 02:36:59,515
So basically,

3197
02:37:01,162 --> 02:37:05,082
we are not only able to
assess this on a region

3198
02:37:05,082 --> 02:37:09,352
by region basis, but now we
get entire whole brain maps

3199
02:37:09,352 --> 02:37:12,702
of metabolic connectivity
mapping showing the influence

3200
02:37:12,702 --> 02:37:15,842
of one voxel to a certain brain region.

3201
02:37:15,842 --> 02:37:19,222
And finally, we wanted to
disentangle the contribution

3202
02:37:19,222 --> 02:37:23,025
of connectivity and
metabolism in simulations.

3203
02:37:25,072 --> 02:37:26,625
Showing the behavioral data,

3204
02:37:27,832 --> 02:37:30,062
we saw some changes.

3205
02:37:30,062 --> 02:37:32,722
We saw most of the changes
for the hard task level

3206
02:37:32,722 --> 02:37:36,402
and some changes also
in the control group.

3207
02:37:36,402 --> 02:37:38,022
This is expected because

3208
02:37:39,662 --> 02:37:44,002
the control buttons were
more difficult to use.

3209
02:37:44,002 --> 02:37:46,982
So there is some intrinsic training

3210
02:37:46,982 --> 02:37:49,642
even for the control group.

3211
02:37:49,642 --> 02:37:54,002
However, the behavioral effect
changes were much larger

3212
02:37:54,002 --> 02:37:57,962
for the training group
and also the interaction

3213
02:37:57,962 --> 02:37:59,822
was highly significant.

3214
02:37:59,822 --> 02:38:04,272
So we observed about a threefold
increase in performance

3215
02:38:04,272 --> 02:38:06,962
for the training group.

3216
02:38:06,962 --> 02:38:10,242
And we saw even a further
increase after the training

3217
02:38:10,242 --> 02:38:11,815
was discontinued.

3218
02:38:14,292 --> 02:38:17,652
And we did some testing
of the cognitive domains

3219
02:38:17,652 --> 02:38:20,302
and saw similar effects
for mental rotation

3220
02:38:20,302 --> 02:38:23,615
and visual search, but
not for spatial planning.

3221
02:38:25,892 --> 02:38:27,842
Coming to the imaging effects,

3222
02:38:27,842 --> 02:38:31,222
we used the occipital
cortex as a target region

3223
02:38:31,222 --> 02:38:34,902
and we saw that the input
of the salience network

3224
02:38:34,902 --> 02:38:37,822
to the occipital cortex really changed

3225
02:38:38,942 --> 02:38:40,475
after training.

3226
02:38:41,382 --> 02:38:45,612
So at resting state, we
saw that the influence

3227
02:38:45,612 --> 02:38:49,632
of the salience network to
occipital cortex increased

3228
02:38:49,632 --> 02:38:53,232
for the insula and dorsal
anterior cingulate cortex.

3229
02:38:53,232 --> 02:38:55,992
And interestingly, during task execution,

3230
02:38:55,992 --> 02:39:00,315
this decreased for the
very same connections.

3231
02:39:02,092 --> 02:39:05,782
And also this difference between

3232
02:39:05,782 --> 02:39:08,702
resting state and task specific MCM values

3233
02:39:08,702 --> 02:39:11,312
correlated very well with performance

3234
02:39:11,312 --> 02:39:14,242
at second PET scan, PET-MRI scan,

3235
02:39:14,242 --> 02:39:17,662
and also correlated with
the overall training success

3236
02:39:17,662 --> 02:39:21,642
and also with the performance of the

3237
02:39:21,642 --> 02:39:22,892
mental rotation paradigm.

3238
02:39:25,922 --> 02:39:29,372
So now we wanted to
disentangle the effects

3239
02:39:29,372 --> 02:39:32,252
or contributions from
metabolism and connectivity

3240
02:39:32,252 --> 02:39:33,542
to these results.

3241
02:39:33,542 --> 02:39:38,542
So basically, since MCM is
based on a spatial correlation,

3242
02:39:38,752 --> 02:39:42,932
we removed voxels based
on the amplitude of either

3243
02:39:42,932 --> 02:39:47,662
connectivity or metabolism
and then recalculated

3244
02:39:47,662 --> 02:39:48,882
the training effects.

3245
02:39:48,882 --> 02:39:52,605
More specifically, the
interaction of group by time.

3246
02:39:53,712 --> 02:39:56,102
And we saw that at rest,

3247
02:39:56,102 --> 02:39:58,782
and this was mostly dependent
on glucose metabolism.

3248
02:39:58,782 --> 02:40:03,112
So basically, if we
remove voxels up to 90%,

3249
02:40:03,112 --> 02:40:06,202
then we see that the training
effects basically diminish

3250
02:40:06,202 --> 02:40:08,275
for both regions.

3251
02:40:09,522 --> 02:40:13,102
On the other hand, task specific
changes were more driven

3252
02:40:13,102 --> 02:40:15,082
by factional connectivity,

3253
02:40:15,082 --> 02:40:17,752
because if we remove the voxels here,

3254
02:40:17,752 --> 02:40:20,712
showing in the dash lines,
the training effects

3255
02:40:20,712 --> 02:40:22,772
are rapidly decreased.

3256
02:40:22,772 --> 02:40:27,612
Interestingly, we could remove
up to 90% of voxels randomly

3257
02:40:27,612 --> 02:40:30,842
and there was basically no
change in the effects at all,

3258
02:40:30,842 --> 02:40:35,132
which basically supports
the specificity of

3259
02:40:36,112 --> 02:40:38,585
the resting and task specific effects.

3260
02:40:40,672 --> 02:40:42,352
So what does this all mean?

3261
02:40:42,352 --> 02:40:46,872
The salience network is highly
involved in error monitoring

3262
02:40:46,872 --> 02:40:47,802
and also

3263
02:40:49,202 --> 02:40:52,862
representing prediction
error between different task

3264
02:40:52,862 --> 02:40:55,595
representations across
different brain regions.

3265
02:40:58,912 --> 02:41:01,812
At resting state, there will be,

3266
02:41:01,812 --> 02:41:04,022
basically the salience
network was not involved

3267
02:41:04,022 --> 02:41:07,372
without the training and therefore we see

3268
02:41:07,372 --> 02:41:09,952
very little and randomly

3269
02:41:09,952 --> 02:41:12,645
influence or communication
between these regions.

3270
02:41:14,135 --> 02:41:16,122
During task performance,
this will result in a

3271
02:41:16,122 --> 02:41:19,472
high prediction era because
the neural representations

3272
02:41:19,472 --> 02:41:22,552
of the occipital cortex
and the salience networks

3273
02:41:22,552 --> 02:41:25,325
do not match because
subjects hasn't trained yet.

3274
02:41:28,583 --> 02:41:31,382
So there is a high level
of cognitive effort

3275
02:41:31,382 --> 02:41:33,595
required to manage the task.

3276
02:41:35,442 --> 02:41:36,795
During the training,

3277
02:41:37,772 --> 02:41:39,562
it has been shown that

3278
02:41:40,542 --> 02:41:43,422
physiological processes
such as synaptic tagging

3279
02:41:43,422 --> 02:41:44,832
and capture occur.

3280
02:41:44,832 --> 02:41:48,552
Synaptic tagging is
basically trial and error

3281
02:41:48,552 --> 02:41:52,352
of transiently establishing
synaptic connections

3282
02:41:53,262 --> 02:41:58,262
and see which connections
would improve performance.

3283
02:41:58,462 --> 02:42:01,025
And then if these connections are,

3284
02:42:02,692 --> 02:42:04,792
I'm sorry, the synaptic connections are

3285
02:42:06,912 --> 02:42:09,562
known to improve, then they are captured.

3286
02:42:09,562 --> 02:42:13,312
And especially this capturing is very

3287
02:42:13,312 --> 02:42:15,772
energy demanding, and it's
been shown that this doubles

3288
02:42:15,772 --> 02:42:17,372
the energy consumption,

3289
02:42:17,372 --> 02:42:20,302
and it happens basically by anchoring

3290
02:42:20,302 --> 02:42:23,472
glutamatergic AMPA receptors for instance.

3291
02:42:23,472 --> 02:42:26,642
So this will then explain the increase

3292
02:42:26,642 --> 02:42:29,772
in the association
between glucose metabolism

3293
02:42:29,772 --> 02:42:33,432
and BOLD connectivity

3294
02:42:33,432 --> 02:42:35,482
after the training at resting state.

3295
02:42:35,482 --> 02:42:37,562
We would refer to this as a skill Ingram,

3296
02:42:37,562 --> 02:42:41,842
which is a formation
of synaptic connections

3297
02:42:41,842 --> 02:42:44,572
which support task performance.

3298
02:42:44,572 --> 02:42:47,372
Because during task
performance, this skill Ingram

3299
02:42:47,372 --> 02:42:49,272
can be retrieved

3300
02:42:49,272 --> 02:42:50,452
and then

3301
02:42:51,719 --> 02:42:54,692
the prediction error
decreases because the neural

3302
02:42:54,692 --> 02:42:57,852
representations between
the different brain regions

3303
02:42:59,182 --> 02:43:03,042
has better match so there's
less cognitive effort required

3304
02:43:03,042 --> 02:43:05,045
to solve the task efficiently.

3305
02:43:07,662 --> 02:43:10,225
I would now like to switch
to something different.

3306
02:43:11,912 --> 02:43:14,322
A more basic investigation.

3307
02:43:14,322 --> 02:43:16,842
Namely the investigation of the

3308
02:43:16,842 --> 02:43:20,602
underlying metabolic demands
of BOLD deactivations.

3309
02:43:20,602 --> 02:43:23,552
We did is in collaboration with
Swedish group (indistinct).

3310
02:43:26,895 --> 02:43:29,202
And they employed a working memory task

3311
02:43:29,202 --> 02:43:31,542
in a hierarchical design.

3312
02:43:31,542 --> 02:43:35,952
And as shown previously, for
basically several other tasks,

3313
02:43:35,952 --> 02:43:39,265
increases in BOLD signal
during task performance

3314
02:43:39,265 --> 02:43:43,742
were also accompanied by
increases in glucose metabolism.

3315
02:43:43,742 --> 02:43:47,005
This is really a stable finding
across most of the tasks.

3316
02:43:47,982 --> 02:43:49,602
However,

3317
02:43:49,602 --> 02:43:53,362
in regions that showed
decreases in the BOLD signal,

3318
02:43:53,362 --> 02:43:56,292
especially in the
posterior cingulate cortex,

3319
02:43:56,292 --> 02:43:58,922
they showed that there
was either no change

3320
02:43:58,922 --> 02:44:02,852
or even an increase in glucose metabolism.

3321
02:44:02,852 --> 02:44:05,192
So that was a mismatch basically.

3322
02:44:05,192 --> 02:44:07,712
And this could mean

3323
02:44:07,712 --> 02:44:11,512
that the decrease in the
BOLD signal really is an

3324
02:44:11,512 --> 02:44:13,632
energy demanding process

3325
02:44:13,632 --> 02:44:17,022
and it could imply an
active suppression of task

3326
02:44:17,022 --> 02:44:19,235
irrelevant activity.

3327
02:44:20,692 --> 02:44:22,992
So we wanted to further investigate this

3328
02:44:24,002 --> 02:44:25,232
in more detail.

3329
02:44:25,232 --> 02:44:28,062
So here is the same data shown
for the working memory task

3330
02:44:28,062 --> 02:44:30,322
at the same mismatch basically

3331
02:44:30,322 --> 02:44:34,512
just presented in voxel space,
not in the surface anymore.

3332
02:44:34,512 --> 02:44:39,232
And we compare this to our "Tetris" task.

3333
02:44:39,232 --> 02:44:43,152
And interestingly, we saw that
decreases in the BOLD signal,

3334
02:44:43,152 --> 02:44:45,322
shown in green, are actually accompanied

3335
02:44:46,642 --> 02:44:50,112
also by decreases in glucose
metabolism, shown in blue,

3336
02:44:50,112 --> 02:44:52,272
and the overlap is shown in white,

3337
02:44:52,272 --> 02:44:55,162
which means that we
have most of the overlap

3338
02:44:55,162 --> 02:44:59,602
in midline core regions of
the default mode network.

3339
02:44:59,602 --> 02:45:03,282
And interestingly, we also
saw in our previous work

3340
02:45:03,282 --> 02:45:07,842
decreases in glucose
metabolism in these regions

3341
02:45:07,842 --> 02:45:10,932
during eyes open and
finger tapping conditions.

3342
02:45:10,932 --> 02:45:14,745
So we were asking is this really
just dependent on the task?

3343
02:45:16,422 --> 02:45:18,152
And we had a closer look at this

3344
02:45:18,152 --> 02:45:21,292
and we think that this is
dependent more on the task

3345
02:45:21,292 --> 02:45:24,089
specific networks because

3346
02:45:24,089 --> 02:45:27,952
"Tetris," shown in these
bar graphs here in blue,

3347
02:45:27,952 --> 02:45:31,664
showing the percentage of voxels activated

3348
02:45:31,664 --> 02:45:33,389
within each network.

3349
02:45:33,389 --> 02:45:36,312
"Tetris" mostly activates
visual and dorsal

3350
02:45:36,312 --> 02:45:37,692
attention networks.

3351
02:45:37,692 --> 02:45:39,692
Whereas working memory, shown in green,

3352
02:45:39,692 --> 02:45:42,485
highly activates frontoparietal network.

3353
02:45:43,362 --> 02:45:46,292
So the difference between
these qualitative difference

3354
02:45:46,292 --> 02:45:49,535
is shown here with the open bars.

3355
02:45:50,982 --> 02:45:53,772
Conversely, the deactivation
between this task

3356
02:45:53,772 --> 02:45:57,232
was also different with "Tetris" showing,

3357
02:45:57,232 --> 02:46:00,952
as I showed before,
highest deactivations in

3358
02:46:00,952 --> 02:46:04,342
core regions of the default mode network,

3359
02:46:04,342 --> 02:46:07,152
whereas working memory
mostly deactivated in

3360
02:46:08,372 --> 02:46:10,075
lateral temporal regions.

3361
02:46:11,172 --> 02:46:13,752
So this is a first qualitative hint

3362
02:46:13,752 --> 02:46:17,082
explaining these
differences between tasks.

3363
02:46:17,082 --> 02:46:20,472
However, we went further on
and wanted to explain this

3364
02:46:20,472 --> 02:46:21,872
in a quantitative manner.

3365
02:46:21,872 --> 02:46:25,192
So we used the regression
analysis and entered

3366
02:46:25,192 --> 02:46:27,035
values of glucose metabolism,

3367
02:46:28,782 --> 02:46:32,392
into regression model to
explain glucose metabolism

3368
02:46:32,392 --> 02:46:36,352
of the PCC by that of the
frontoparietal network

3369
02:46:36,352 --> 02:46:38,572
and those attention and visual networks,

3370
02:46:38,572 --> 02:46:42,462
because these three networks
showed the biggest difference

3371
02:46:42,462 --> 02:46:44,829
between working memory and "Tetris."

3372
02:46:45,692 --> 02:46:47,652
And indeed, what we found is that

3373
02:46:48,612 --> 02:46:51,142
these two networks,
frontoparietal and dorsal tension

3374
02:46:51,142 --> 02:46:54,852
networks explained the glucose metabolism

3375
02:46:54,852 --> 02:46:58,192
in a posterior cingulate
in an opposite manner,

3376
02:46:58,192 --> 02:47:01,212
which means that the higher the metabolism

3377
02:47:01,212 --> 02:47:03,722
in the dorsal attention network,

3378
02:47:03,722 --> 02:47:06,872
and the lower the metabolism
in the frontoparietal network

3379
02:47:06,872 --> 02:47:11,872
is the more decrease you
have in glucose metabolism

3380
02:47:13,072 --> 02:47:15,005
in the posterior cingulate cortex.

3381
02:47:18,472 --> 02:47:21,412
This could be explained,
for instance, because

3382
02:47:21,412 --> 02:47:25,112
all deactivations may emerge from

3383
02:47:25,112 --> 02:47:27,482
a decreased level of glutamate signaling,

3384
02:47:27,482 --> 02:47:31,505
which also would be followed
by decreased energy demands.

3385
02:47:32,602 --> 02:47:35,022
However, it could also be that

3386
02:47:35,022 --> 02:47:39,752
all deactivations are
related to an increase in

3387
02:47:39,752 --> 02:47:41,332
GABA activity.

3388
02:47:41,332 --> 02:47:44,862
And this is a energy demanding
process so this would imply

3389
02:47:44,862 --> 02:47:47,912
an increase in glucose metabolism as well.

3390
02:47:47,912 --> 02:47:50,999
And this is exactly these two
effects that we see during

3391
02:47:50,999 --> 02:47:53,722
"Tetris" task and working memory.

3392
02:47:53,722 --> 02:47:56,762
So we speculate that

3393
02:47:56,762 --> 02:48:01,682
maybe the connections between
the dorsal attention networks

3394
02:48:02,972 --> 02:48:05,472
that is activated during "Tetris"

3395
02:48:05,472 --> 02:48:08,152
to the posterior cingulate
cortex might be more

3396
02:48:08,152 --> 02:48:09,472
glutamatergic.

3397
02:48:09,472 --> 02:48:13,152
Various projections from
the frontoparietal network,

3398
02:48:13,152 --> 02:48:17,032
which is activated during
working memory to the PCC

3399
02:48:17,032 --> 02:48:19,545
might be more of GABA nature.

3400
02:48:20,482 --> 02:48:23,922
And on a behavioral level,
as I mentioned before,

3401
02:48:23,922 --> 02:48:25,122
this could be explained

3402
02:48:26,512 --> 02:48:29,552
by the high external
tension that is acquired

3403
02:48:29,552 --> 02:48:32,902
to solve the "Tetris" task,
whereas working memory

3404
02:48:32,902 --> 02:48:36,705
also has some high internal demands.

3405
02:48:40,002 --> 02:48:42,472
Another thing that you could do with

3406
02:48:43,432 --> 02:48:46,712
combined fPET-MRI
imaging is to investigate

3407
02:48:46,712 --> 02:48:50,012
immediate drug effects
in pharmaco imaging.

3408
02:48:50,012 --> 02:48:52,992
So that's been previously
shown that the BOLD signal

3409
02:48:52,992 --> 02:48:57,722
changes during acute
administration of SSRIs

3410
02:48:57,722 --> 02:48:59,662
with intravenous injection.

3411
02:48:59,662 --> 02:49:02,912
However, we could not
really replicate this

3412
02:49:02,912 --> 02:49:07,592
in a larger sample using
lots of different models

3413
02:49:07,592 --> 02:49:11,042
such as that using plasma
concentration of (indistinct)

3414
02:49:11,042 --> 02:49:14,612
or even serotonin
transport binding effects.

3415
02:49:14,612 --> 02:49:17,462
So we thought that maybe

3416
02:49:17,462 --> 02:49:20,062
more stable signals such
as glucose metabolism,

3417
02:49:20,062 --> 02:49:23,332
cerebral blood flow
obtained with PET and ASL

3418
02:49:23,332 --> 02:49:26,805
might be better suited
to assess these effects.

3419
02:49:28,882 --> 02:49:32,582
Since the acute SSR
effects follow an unknown

3420
02:49:32,582 --> 02:49:34,375
spatial temporal pattern,

3421
02:49:35,502 --> 02:49:39,622
we used intersubject correlations
to solve this problem.

3422
02:49:39,622 --> 02:49:43,482
So basically, this means that
the time course of the PET

3423
02:49:43,482 --> 02:49:48,482
and the ASL were correlated
between subjects for each voxel.

3424
02:49:50,092 --> 02:49:54,032
And we assume that if this
correlation across subjects

3425
02:49:54,032 --> 02:49:57,242
is non random, then we would
see a specific pattern.

3426
02:49:57,242 --> 02:49:58,962
And this was indeed the case.

3427
02:49:58,962 --> 02:50:02,362
So we saw higher correlations.
Higher than average.

3428
02:50:02,362 --> 02:50:05,572
For instance, in the striatum
in the occipital cortex

3429
02:50:05,572 --> 02:50:07,095
and also temporal regions.

3430
02:50:07,942 --> 02:50:11,582
And interestingly, comparing
Verum with placebo effects,

3431
02:50:11,582 --> 02:50:14,152
we saw that these effects
in the occipital cortex

3432
02:50:14,152 --> 02:50:18,362
were significant in the striatum;
at least reach the trend.

3433
02:50:18,362 --> 02:50:22,892
So this might be an example
of a promising investigation

3434
02:50:22,892 --> 02:50:24,875
of pharmacological effects.

3435
02:50:27,002 --> 02:50:29,582
And the final thing I would
like to present is that

3436
02:50:29,582 --> 02:50:33,092
functional PET can also be
used for imaging dopamine

3437
02:50:33,092 --> 02:50:34,502
neurotransmission.

3438
02:50:34,502 --> 02:50:36,522
And we call this the Synthesis Model

3439
02:50:36,522 --> 02:50:40,272
because the radioligand
FDOPA's used which is surrogate

3440
02:50:41,722 --> 02:50:42,742
for imaging

3441
02:50:44,032 --> 02:50:45,475
dopamine synthesis.

3442
02:50:47,392 --> 02:50:51,552
Because, during stimulation,
dopamine is of course released,

3443
02:50:51,552 --> 02:50:55,572
and this will increase also
dopamine synthesis to refill

3444
02:50:55,572 --> 02:50:57,422
the synaptic vesicles.

3445
02:50:57,422 --> 02:51:01,302
And we argue that this will also

3446
02:51:01,302 --> 02:51:04,412
result in an increased radioligand uptake

3447
02:51:04,412 --> 02:51:06,432
because there is really lots of

3448
02:51:07,472 --> 02:51:09,702
previous work showing

3449
02:51:09,702 --> 02:51:11,712
such an association because for instance,

3450
02:51:11,712 --> 02:51:15,722
neuronal firing increases
the activity of the enzymes

3451
02:51:15,722 --> 02:51:18,282
responsible for dopamine
synthesis, such as

3452
02:51:18,282 --> 02:51:21,315
tyrosine hydroxylase and AADC.

3453
02:51:22,432 --> 02:51:25,772
Furthermore, the amphetamine
induced dopamine release

3454
02:51:26,612 --> 02:51:29,022
also increases dopamine synthesis.

3455
02:51:29,022 --> 02:51:30,882
And this is also true
the other way around.

3456
02:51:30,882 --> 02:51:33,902
If dopamine synthesis is decreased,

3457
02:51:33,902 --> 02:51:36,722
then also the amphetamine
induced dopamine release

3458
02:51:36,722 --> 02:51:38,055
will be decreased.

3459
02:51:39,092 --> 02:51:42,972
So I think this is strong
support showing that stimulus

3460
02:51:42,972 --> 02:51:46,672
induced changes in dopamine synthesis

3461
02:51:46,672 --> 02:51:50,565
probably reflect dopamine
release to some extent.

3462
02:51:52,842 --> 02:51:55,962
And we use this model in
functional PET imaging

3463
02:51:55,962 --> 02:52:00,092
while subjects completed the
monetary incentive delay task.

3464
02:52:00,092 --> 02:52:02,932
This is a reward task basically where

3465
02:52:04,350 --> 02:52:08,092
you can win or lose money based on

3466
02:52:08,092 --> 02:52:10,062
your individual reaction time.

3467
02:52:10,062 --> 02:52:13,932
And exactly this reaction
time was manipulated to

3468
02:52:14,842 --> 02:52:18,732
create blocks with high
probability of gain or high

3469
02:52:18,732 --> 02:52:20,025
probability of loss.

3470
02:52:21,782 --> 02:52:24,292
We also corrected for
radioactive metabolites

3471
02:52:24,292 --> 02:52:26,822
as published previously.

3472
02:52:26,822 --> 02:52:30,142
Task identification was done
with a general linear model

3473
02:52:30,142 --> 02:52:33,752
as also as described
before, including regresses

3474
02:52:33,752 --> 02:52:36,252
for baseline, gain blocks, loss blocks,

3475
02:52:36,252 --> 02:52:38,042
and also head motion.

3476
02:52:38,042 --> 02:52:41,672
And with this GLM approach,
again, we could separate

3477
02:52:41,672 --> 02:52:46,672
baseline effects from task
specific effects here shown for

3478
02:52:46,842 --> 02:52:48,452
the gain condition.

3479
02:52:48,452 --> 02:52:50,752
And quantification was then carried out

3480
02:52:51,952 --> 02:52:55,402
with the Patlak plot, specifically
for the ventral striatum

3481
02:52:56,432 --> 02:52:59,825
as this was our target region
involved in reward processing.

3482
02:53:02,092 --> 02:53:06,542
So as a result, we saw
that task specific changes

3483
02:53:06,542 --> 02:53:08,372
in dopamine synthesis

3484
02:53:09,222 --> 02:53:12,612
were really higher in men

3485
02:53:12,612 --> 02:53:16,092
for monetary gain as
compared to monetary loss.

3486
02:53:16,092 --> 02:53:20,202
Interestingly, for women, we
saw the exact opposite pattern.

3487
02:53:20,202 --> 02:53:22,842
Mainly that dopamine synthesis was higher

3488
02:53:22,842 --> 02:53:25,385
during monetary loss than monetary gain.

3489
02:53:26,802 --> 02:53:30,732
We also saw that these
changes in dopamine synthesis

3490
02:53:30,732 --> 02:53:34,762
were correlated with parameters
of reward sensitivity

3491
02:53:34,762 --> 02:53:38,625
in men, but punishment
sensitivity in women.

3492
02:53:40,442 --> 02:53:43,312
We also used the fMRI
for direct comparison.

3493
02:53:43,312 --> 02:53:46,742
And as published previously,
we saw a very strong,

3494
02:53:46,742 --> 02:53:51,012
robust activation of the
contrast gain versus loss

3495
02:53:51,012 --> 02:53:52,322
in the BOLD signal.

3496
02:53:52,322 --> 02:53:55,572
However, there was no
significant difference

3497
02:53:56,662 --> 02:53:59,922
between men and women, and
this is very much in line

3498
02:53:59,922 --> 02:54:02,592
what has been published previously.

3499
02:54:02,592 --> 02:54:05,612
And we are aware of the
low sample size, but still

3500
02:54:05,612 --> 02:54:09,282
if this holds true, this
would have some important

3501
02:54:09,282 --> 02:54:13,222
implications because
the findings here from

3502
02:54:13,222 --> 02:54:16,122
the dopamine system would
explain the well known behavioral

3503
02:54:16,122 --> 02:54:20,132
differences between men and
women during reward processing.

3504
02:54:20,132 --> 02:54:22,862
And it's also important for
psychiatric disorders because

3505
02:54:22,862 --> 02:54:25,702
there are several disorders
showing alterations

3506
02:54:25,702 --> 02:54:28,502
in the dopamine system
in reward processing,

3507
02:54:28,502 --> 02:54:30,782
and also differences in
the prevalence between

3508
02:54:30,782 --> 02:54:31,615
men and women.

3509
02:54:33,872 --> 02:54:38,872
So this approach was
sensitive to task changes

3510
02:54:39,702 --> 02:54:42,732
of about 100% compared to baseline,

3511
02:54:42,732 --> 02:54:45,062
but also when using
specific control conditions,

3512
02:54:45,062 --> 02:54:47,165
such as monetary gain versus loss.

3513
02:54:48,842 --> 02:54:53,262
If we now assume that
these changes solely are

3514
02:54:53,262 --> 02:54:56,392
cost or related to changes in K3,

3515
02:54:56,392 --> 02:54:59,215
thereby reflecting AADC activity,

3516
02:55:00,132 --> 02:55:01,952
this would result in rather large changes.

3517
02:55:01,952 --> 02:55:04,442
However, we have to keep in mind

3518
02:55:04,442 --> 02:55:08,082
that only 50% are available
for dopamine synthesis

3519
02:55:08,082 --> 02:55:10,422
of the FDOPA in humans.

3520
02:55:10,422 --> 02:55:14,672
And further, 75% are only
stored in the vesicles.

3521
02:55:14,672 --> 02:55:19,172
So this will result in an
estimation of dopamine synthesized

3522
02:55:19,172 --> 02:55:22,422
by the task of about 50% to 100%.

3523
02:55:22,422 --> 02:55:24,232
This is rather crude estimate,

3524
02:55:24,232 --> 02:55:27,292
but still it is perfectly in line with

3525
02:55:27,292 --> 02:55:29,142
animal studies doing micro dialysis

3526
02:55:30,474 --> 02:55:31,739
and showing exactly this

3527
02:55:34,585 --> 02:55:36,659
release of dopamine during reward.

3528
02:55:40,692 --> 02:55:43,342
So to summarize: with functional PET,

3529
02:55:43,342 --> 02:55:47,072
you can image metabolism
and neurotransmitter action

3530
02:55:48,362 --> 02:55:51,332
during baseline and task in a single scan.

3531
02:55:51,332 --> 02:55:53,025
As I said, it's perfect,

3532
02:55:54,752 --> 02:55:56,292
a very promising approach

3533
02:55:57,432 --> 02:56:00,452
for simultaneous imaging with PET-MR.

3534
02:56:00,452 --> 02:56:04,652
It's regionally specific for
both simple and complex tasks.

3535
02:56:04,652 --> 02:56:07,452
It's also sensitive because
we can obtain differences

3536
02:56:07,452 --> 02:56:10,162
between task loads reliably.

3537
02:56:10,162 --> 02:56:14,655
And it also has some good
test, re-test reproducibility.

3538
02:56:17,352 --> 02:56:20,192
So I hope I could make the point

3539
02:56:20,192 --> 02:56:22,742
that these tasks specific changes

3540
02:56:23,772 --> 02:56:26,772
between the different imaging
parameters really carry

3541
02:56:26,772 --> 02:56:28,732
complementary information.

3542
02:56:28,732 --> 02:56:32,022
So you can obtain different information

3543
02:56:32,022 --> 02:56:35,992
on basic energy demands,
such as glucose metabolism,

3544
02:56:35,992 --> 02:56:39,082
cerebral blood flows, also
BOLD imaging, and of course

3545
02:56:39,082 --> 02:56:40,445
connectivity,

3546
02:56:41,482 --> 02:56:43,882
for instance, to derive
directional connectivity

3547
02:56:45,441 --> 02:56:48,442
with the combination of glucose metabolism

3548
02:56:48,442 --> 02:56:50,882
and functional connectivity.

3549
02:56:50,882 --> 02:56:53,602
You could also go for
neurotransmitter estimation.

3550
02:56:53,602 --> 02:56:57,482
We tried to explain the value observed,

3551
02:56:57,482 --> 02:57:01,832
deactivations seen with the
BOLD signal on a metabolic basis

3552
02:57:01,832 --> 02:57:03,862
and many more

3553
02:57:03,862 --> 02:57:07,315
hopefully promising applications
coming up in the future.

3554
02:57:08,312 --> 02:57:10,595
Of course, there are
also some limitations.

3555
02:57:12,452 --> 02:57:15,822
As with every PET scan,
there is radioactive

3556
02:57:17,332 --> 02:57:18,165
burden.

3557
02:57:18,165 --> 02:57:20,582
But this is just basically
how it is with PET.

3558
02:57:20,582 --> 02:57:22,822
There is not much way around this.

3559
02:57:22,822 --> 02:57:26,192
But it might come up soon with

3560
02:57:26,192 --> 02:57:27,872
more sensitive scanners.

3561
02:57:27,872 --> 02:57:30,152
So basically this would
allow us to decrease

3562
02:57:30,152 --> 02:57:32,152
the radiation dosage.

3563
02:57:32,152 --> 02:57:34,502
The biggest limitation
probably for functional PET now

3564
02:57:34,502 --> 02:57:38,362
is still the temporary
resolution, because it's related

3565
02:57:38,362 --> 02:57:41,622
to the radioactive counts
that can be obtained

3566
02:57:41,622 --> 02:57:42,992
in a certain frame.

3567
02:57:42,992 --> 02:57:46,172
And we are still working
on this to further

3568
02:57:46,172 --> 02:57:47,135
improve this.

3569
02:57:48,542 --> 02:57:51,552
As an outlook, it would be
very interesting to assess

3570
02:57:51,552 --> 02:57:54,962
all of these effects in
disease cohorts because,

3571
02:57:54,962 --> 02:57:59,062
as I said, with just
one scan you can obtain

3572
02:57:59,062 --> 02:58:00,935
two or three conditions.

3573
02:58:01,882 --> 02:58:05,822
And if you run an FDG scan
anyway in your patients,

3574
02:58:05,822 --> 02:58:09,692
it might be worth to add a
little bit of extra effort

3575
02:58:09,692 --> 02:58:12,642
and get more information out of it,

3576
02:58:12,642 --> 02:58:15,082
because really, it's not
so difficult to implement.

3577
02:58:15,082 --> 02:58:16,982
All you need basically is an infusion pump

3578
02:58:16,982 --> 02:58:18,762
and stimulation equipment.

3579
02:58:18,762 --> 02:58:22,022
Of course, you can come up with some

3580
02:58:22,022 --> 02:58:22,855
very

3581
02:58:23,722 --> 02:58:25,162
sophisticated paradigms.

3582
02:58:25,162 --> 02:58:29,542
You could also try to adapt
these of fMRI that are already

3583
02:58:29,542 --> 02:58:31,922
existing and established.

3584
02:58:31,922 --> 02:58:35,215
And then for the analysis,
basically, you only need GLM.

3585
02:58:36,722 --> 02:58:37,922
So it's really feasible.

3586
02:58:39,992 --> 02:58:42,372
Our current work deals with, for instance,

3587
02:58:42,372 --> 02:58:44,462
substitution of anterior blood sampling

3588
02:58:44,462 --> 02:58:45,822
and also, of course, as I said,

3589
02:58:45,822 --> 02:58:48,612
increasing temporal resolution

3590
02:58:48,612 --> 02:58:51,372
to make this approach basically

3591
02:58:51,372 --> 02:58:54,272
even more feasible and more applicable

3592
02:58:54,272 --> 02:58:57,062
also in different settings.

3593
02:58:57,062 --> 02:59:00,122
And we are very happy to
support you if you want to

3594
02:59:00,122 --> 02:59:01,215
implement this.

3595
02:59:03,466 --> 02:59:04,982
We have been doing this

3596
02:59:04,982 --> 02:59:06,512
all over the place.

3597
02:59:06,512 --> 02:59:08,172
Supporting people with the design,

3598
02:59:08,172 --> 02:59:10,372
with the implementation of the setting

3599
02:59:10,372 --> 02:59:12,922
and the infusion setting
with the analysis.

3600
02:59:12,922 --> 02:59:14,612
So if you have any questions to this,

3601
02:59:14,612 --> 02:59:16,965
please do not hesitate to contact us.

3602
02:59:18,722 --> 02:59:22,412
I would like to conclude
with this and thank

3603
02:59:22,412 --> 02:59:25,882
the neuroimaging labs,
especially the head,

3604
02:59:25,882 --> 02:59:28,602
Professor Rupert Lanzenberger,
and of course the team

3605
02:59:28,602 --> 02:59:30,762
for making all of this possible.

3606
02:59:30,762 --> 02:59:33,832
I would like to thank the
funding institutions for

3607
02:59:33,832 --> 02:59:36,092
providing support and also, of course,

3608
02:59:36,092 --> 02:59:38,172
the corporation partners.

3609
02:59:38,172 --> 02:59:39,112
And finally,

3610
02:59:39,112 --> 02:59:42,152
I would like to thank you
for listening to the talk.

3611
02:59:42,152 --> 02:59:43,995
Thank you very much and goodbye.

3612
02:59:50,715 --> 02:59:53,115
- Hi, and welcome to this presentation.

3613
02:59:53,115 --> 02:59:55,315
Today, I'll be speaking
about the recent developments

3614
02:59:55,315 --> 02:59:58,035
in the application of simultaneous PET/MR

3615
02:59:58,035 --> 03:00:00,225
to the study of human brain function.

3616
03:00:00,225 --> 03:00:01,515
My name is Shama Jamadar,

3617
03:00:01,515 --> 03:00:02,905
and I'm a senior research fellow

3618
03:00:02,905 --> 03:00:05,425
at the Turner Institute
for Brain and Mental Health

3619
03:00:05,425 --> 03:00:06,985
and Monash Biomedical Imaging

3620
03:00:06,985 --> 03:00:09,178
at Monash University in Australia.

3621
03:00:11,565 --> 03:00:14,175
In this presentation, I'll
be talking about PET/MR

3622
03:00:14,175 --> 03:00:17,665
from the perspective of
cognitive neuroscience

3623
03:00:17,665 --> 03:00:19,905
and how the method can
improve our understanding

3624
03:00:19,905 --> 03:00:22,195
of the neural basis of cognition.

3625
03:00:22,195 --> 03:00:23,925
I'll start with a very brief discussion

3626
03:00:23,925 --> 03:00:27,365
of how we can measure brain
function using PET and MRI,

3627
03:00:27,365 --> 03:00:29,555
and then I'll introduce the fPET approach,

3628
03:00:29,555 --> 03:00:30,835
which will be discussed at length

3629
03:00:30,835 --> 03:00:32,935
in the rest of the presentation.

3630
03:00:32,935 --> 03:00:35,675
Then, I'll discuss briefly
some of the early work we did

3631
03:00:35,675 --> 03:00:38,155
in designing experimental protocols

3632
03:00:38,155 --> 03:00:40,825
for simultaneous fPET and fMRI,

3633
03:00:40,825 --> 03:00:42,445
and then the application of PET/MR

3634
03:00:42,445 --> 03:00:44,595
to the study of brain connectivity.

3635
03:00:44,595 --> 03:00:46,215
Lastly, I'll offer a critique

3636
03:00:46,215 --> 03:00:48,435
of where I think we are in applying PET/MR

3637
03:00:48,435 --> 03:00:50,215
to the study of brain function

3638
03:00:50,215 --> 03:00:52,288
and discuss some future directions.

3639
03:00:53,165 --> 03:00:54,598
So let's get started.

3640
03:00:56,225 --> 03:00:59,775
What's the benefit of PET and
MRI in cognitive neuroscience?

3641
03:00:59,775 --> 03:01:01,705
From a neuroimaging perspective,

3642
03:01:01,705 --> 03:01:05,015
MRI and PET have complimentary
strengths and weaknesses.

3643
03:01:05,015 --> 03:01:08,525
While MRI provides excellent
spatial resolution,

3644
03:01:08,525 --> 03:01:11,815
the spatial resolution of PET
systems have a natural limit.

3645
03:01:11,815 --> 03:01:14,305
On the other hand, PET
is remarkably flexible

3646
03:01:14,305 --> 03:01:17,185
in the range of molecular
targets that can be imaged,

3647
03:01:17,185 --> 03:01:18,235
providing the capacity

3648
03:01:18,235 --> 03:01:20,465
to measure different
physiological properties

3649
03:01:20,465 --> 03:01:23,415
of energy metabolism,
neurotransmitter dynamics,

3650
03:01:23,415 --> 03:01:26,495
protein and inflammatory
markers, and so on.

3651
03:01:26,495 --> 03:01:29,295
So, from the perspective
of functional neuroimaging,

3652
03:01:29,295 --> 03:01:31,055
some of the markers of neuronal activity

3653
03:01:31,055 --> 03:01:33,628
can be physiologically closer
to the process of interest

3654
03:01:33,628 --> 03:01:36,125
than fMRI signals.

3655
03:01:36,125 --> 03:01:38,685
Both modalities benefit significantly

3656
03:01:38,685 --> 03:01:40,645
from a large body of literature.

3657
03:01:40,645 --> 03:01:43,785
So a PubMed search for
the term functional MRI

3658
03:01:43,785 --> 03:01:45,785
yields over 700,000 hits

3659
03:01:45,785 --> 03:01:48,915
with 16,000 in 2021 alone.

3660
03:01:48,915 --> 03:01:50,975
And, similarly, a search for PET

3661
03:01:50,975 --> 03:01:53,495
yields over 140,000 hits

3662
03:01:53,495 --> 03:01:56,125
with 12,000 in 2021 alone.

3663
03:01:56,125 --> 03:01:59,235
And so investigators using both methods

3664
03:01:59,235 --> 03:02:02,388
have got a vast body of knowledge
upon which to build upon.

3665
03:02:04,195 --> 03:02:06,835
Now, some of the following will
be old news to some of you,

3666
03:02:06,835 --> 03:02:08,955
but I'm going to very
briefly cover the basics

3667
03:02:08,955 --> 03:02:10,795
of each signal and each method,

3668
03:02:10,795 --> 03:02:12,285
both to provide context

3669
03:02:12,285 --> 03:02:14,165
but to also highlight how each approach

3670
03:02:14,165 --> 03:02:16,335
is complementary to the other.

3671
03:02:16,335 --> 03:02:18,165
So, when we're talking about fMRI,

3672
03:02:18,165 --> 03:02:20,935
we're usually talking about BOLD fMRI,

3673
03:02:20,935 --> 03:02:23,098
or blood oxygenation level dependent fMRI.

3674
03:02:24,165 --> 03:02:26,915
The BOLD response is used
to infer neuronal activity

3675
03:02:26,915 --> 03:02:30,085
indirectly from changes
in blood oxygenation.

3676
03:02:30,085 --> 03:02:31,785
The method relies upon the concept

3677
03:02:31,785 --> 03:02:33,495
of neurovascular coupling.

3678
03:02:33,495 --> 03:02:35,535
Importantly, it's not
a quantitative measure

3679
03:02:35,535 --> 03:02:37,015
of brain function.

3680
03:02:37,015 --> 03:02:40,195
The fMRI signal cannot be
compared across brain regions,

3681
03:02:40,195 --> 03:02:42,725
subjects, in the same subject across time,

3682
03:02:42,725 --> 03:02:44,158
or between scanners.

3683
03:02:45,745 --> 03:02:47,755
BOLD relies on the ratio of oxy

3684
03:02:47,755 --> 03:02:50,445
and deoxygenated hemoglobin in the blood.

3685
03:02:50,445 --> 03:02:53,495
Oxygenated hemoglobin has
little magnetic properties,

3686
03:02:53,495 --> 03:02:56,405
and so it's got little
effect on the magnetic field.

3687
03:02:56,405 --> 03:02:59,997
In contrast, deoxygenated
hemoglobin is paramagnetic,

3688
03:02:59,997 --> 03:03:02,415
and so it disrupts the relaxation of spins

3689
03:03:02,415 --> 03:03:04,305
in the transverse plane.

3690
03:03:04,305 --> 03:03:07,765
If the concentration of
deoxygenated hemoglobin changes,

3691
03:03:07,765 --> 03:03:09,245
the BOLD signal will change,

3692
03:03:09,245 --> 03:03:11,975
and this is the basis of fMRI.

3693
03:03:11,975 --> 03:03:13,525
When neurons are not active,

3694
03:03:13,525 --> 03:03:15,435
oxygenated hemoglobin is converted

3695
03:03:15,435 --> 03:03:18,675
to deoxygenated hemoglobin
at a constant rate,

3696
03:03:18,675 --> 03:03:20,395
but, when they do become active,

3697
03:03:20,395 --> 03:03:22,255
they use oxygen in the hemoglobin,

3698
03:03:22,255 --> 03:03:24,695
and so it becomes deoxygenated.

3699
03:03:24,695 --> 03:03:26,495
This changes the ratio of oxy

3700
03:03:26,495 --> 03:03:29,085
and deoxygenated hemoglobin in a region

3701
03:03:29,085 --> 03:03:31,495
such that increase in neuronal activity

3702
03:03:31,495 --> 03:03:35,268
will also increase the T2
star signal of the MRI.

3703
03:03:36,965 --> 03:03:38,915
So there's a complex cascade of events

3704
03:03:38,915 --> 03:03:41,305
that contributes to the BOLD response.

3705
03:03:41,305 --> 03:03:43,685
In essence, following the
presentation of stimulus,

3706
03:03:43,685 --> 03:03:46,075
which elicits a neuronal response,

3707
03:03:46,075 --> 03:03:48,768
there's an increase in the
flow of oxygenated blood,

3708
03:03:50,155 --> 03:03:52,175
and so there's also a
decrease in the amount

3709
03:03:52,175 --> 03:03:54,795
of deoxygenated blood in that region.

3710
03:03:54,795 --> 03:03:57,675
This causes the BOLD signal to increase.

3711
03:03:57,675 --> 03:03:59,445
When the neuronal activity ceases,

3712
03:03:59,445 --> 03:04:02,165
there's a decrease in
blood flow and volume.

3713
03:04:02,165 --> 03:04:04,295
When the blood volume returns to baseline,

3714
03:04:04,295 --> 03:04:07,295
the BOLD signal also returned to baseline.

3715
03:04:07,295 --> 03:04:10,125
So it's important to note
just how indirect fMRI

3716
03:04:10,125 --> 03:04:12,615
is a measure of neuronal activity.

3717
03:04:12,615 --> 03:04:14,885
The indirect nature of the fMRI response

3718
03:04:14,885 --> 03:04:17,118
has implications for how we can use it.

3719
03:04:19,215 --> 03:04:21,725
The vascular and indirect
nature of the BOLD response

3720
03:04:21,725 --> 03:04:24,115
has got significant implications.

3721
03:04:24,115 --> 03:04:25,875
In just one example that allows me

3722
03:04:25,875 --> 03:04:28,225
to shamelessly plug some of our work,

3723
03:04:28,225 --> 03:04:30,995
we simply looked at how estimates
of functional connectivity

3724
03:04:30,995 --> 03:04:34,915
are influenced by individual
differences in hemoglobin.

3725
03:04:34,915 --> 03:04:37,645
So simply splitting the
group into males and females

3726
03:04:37,645 --> 03:04:40,245
and then into low and
high hemoglobin groups,

3727
03:04:40,245 --> 03:04:41,805
we found that the functional connectome

3728
03:04:41,805 --> 03:04:43,595
differed significantly between high

3729
03:04:43,595 --> 03:04:45,675
and low hemoglobin groups.

3730
03:04:45,675 --> 03:04:49,125
Most concerningly, this
influenced the correlation

3731
03:04:49,125 --> 03:04:51,195
between connectivity and cognition

3732
03:04:51,195 --> 03:04:53,485
across multiple cognitive tests.

3733
03:04:53,485 --> 03:04:55,075
So, for all the cognitive tests

3734
03:04:55,075 --> 03:04:56,715
that we tested in this study,

3735
03:04:56,715 --> 03:04:58,645
despite no differences in scores

3736
03:04:58,645 --> 03:05:00,955
between low and high hemoglobin groups,

3737
03:05:00,955 --> 03:05:02,575
there was a significant decrease

3738
03:05:02,575 --> 03:05:05,365
in the cognition-connectivity
relationship.

3739
03:05:05,365 --> 03:05:06,395
So, in other words,

3740
03:05:06,395 --> 03:05:08,645
functional connectivity measured with fMRI

3741
03:05:08,645 --> 03:05:10,725
is highly confounded by simple differences

3742
03:05:10,725 --> 03:05:13,308
in hemoglobin levels between individuals.

3743
03:05:15,195 --> 03:05:19,205
Secondly, the way that we use
fMRI is, in and of itself,

3744
03:05:19,205 --> 03:05:22,605
fundamentally affected by the
nature of the BOLD response,

3745
03:05:22,605 --> 03:05:25,845
so much so that the people
who work regularly with fMRI,

3746
03:05:25,845 --> 03:05:27,235
I'm one of those people,

3747
03:05:27,235 --> 03:05:29,425
sometimes forgets that our modus operandi

3748
03:05:29,425 --> 03:05:32,195
is actually quite idiosyncratic.

3749
03:05:32,195 --> 03:05:33,065
So, for example,

3750
03:05:33,065 --> 03:05:35,275
we know that the BOLD
response can't be compared

3751
03:05:35,275 --> 03:05:37,855
between brain regions
and between individuals,

3752
03:05:37,855 --> 03:05:39,425
and so we just don't do this.

3753
03:05:39,425 --> 03:05:43,475
We just don't design
experiments that do this.

3754
03:05:43,475 --> 03:05:46,365
Similarly, direct
comparison of fMRI responses

3755
03:05:46,365 --> 03:05:49,245
between groups just often isn't done.

3756
03:05:49,245 --> 03:05:52,505
So, for example, comparing
the BOLD response to a task

3757
03:05:52,505 --> 03:05:54,285
between older and younger people

3758
03:05:54,285 --> 03:05:55,705
often isn't performed

3759
03:05:55,705 --> 03:05:58,585
because we know that
cerebrovascular disease

3760
03:05:58,585 --> 03:06:01,068
is more common in older
than in younger people.

3761
03:06:02,165 --> 03:06:03,565
So all of this is old news.

3762
03:06:03,565 --> 03:06:04,848
So you can see from the dates on this

3763
03:06:04,848 --> 03:06:07,605
that this is stuff that
we've known for a long time,

3764
03:06:07,605 --> 03:06:09,865
but it's sometimes worthwhile
to stop and take stock

3765
03:06:09,865 --> 03:06:12,235
of how our dominant method has influenced

3766
03:06:12,235 --> 03:06:15,198
the type of questions that we
are asking in a discipline.

3767
03:06:17,105 --> 03:06:19,795
Okay, so let's now move on to PET.

3768
03:06:19,795 --> 03:06:21,605
When we consider the cascade of events

3769
03:06:21,605 --> 03:06:23,785
that occur during neuronal activity,

3770
03:06:23,785 --> 03:06:25,765
there are numerous
indices that we could use

3771
03:06:25,765 --> 03:06:27,625
to infer neuronal activity.

3772
03:06:27,625 --> 03:06:28,615
As we've just seen,

3773
03:06:28,615 --> 03:06:31,695
BOLD measures an indirect
outcome from the activity,

3774
03:06:31,695 --> 03:06:33,675
namely the change in the concentration

3775
03:06:33,675 --> 03:06:35,495
of deoxygenated hemoglobin

3776
03:06:35,495 --> 03:06:38,905
that occurs as a result of
changes in oxygen consumption

3777
03:06:38,905 --> 03:06:40,745
and cerebral blood flow.

3778
03:06:40,745 --> 03:06:44,895
We can use PET traces to
measure specific components

3779
03:06:44,895 --> 03:06:47,255
of the oxygen consumption process,

3780
03:06:47,255 --> 03:06:51,215
like blood flow, volume,
oxygen, metabolism, and so on.

3781
03:06:51,215 --> 03:06:52,595
It's also possible to measure

3782
03:06:52,595 --> 03:06:55,825
specific neurotransmitter
molecules, like dopamine,

3783
03:06:55,825 --> 03:06:58,545
and measure its synthesis,
uptake, and release.

3784
03:06:58,545 --> 03:06:59,695
So, as you can appreciate,

3785
03:06:59,695 --> 03:07:01,525
there's a very wide range of targets

3786
03:07:01,525 --> 03:07:03,205
to measure neuronal activity,

3787
03:07:03,205 --> 03:07:05,075
and the capacity of PET to image them

3788
03:07:05,075 --> 03:07:08,655
is limited only by the range
of the radio traces to do so.

3789
03:07:08,655 --> 03:07:11,065
So, today, we'll focus on
one of the primary methods

3790
03:07:11,065 --> 03:07:14,538
to image neuronal activity using PET, FDG.

3791
03:07:16,175 --> 03:07:17,875
Glucose is one of the primary sources

3792
03:07:17,875 --> 03:07:19,185
of energy in the brain,

3793
03:07:19,185 --> 03:07:22,425
and regional glucose uptake
primarily reflects activity

3794
03:07:22,425 --> 03:07:24,615
at the excitatory synapses.

3795
03:07:24,615 --> 03:07:27,315
So glucose uptake can
be considered an index

3796
03:07:27,315 --> 03:07:28,665
of neuronal activity

3797
03:07:28,665 --> 03:07:31,345
that's physiologically closer
to the biological process

3798
03:07:31,345 --> 03:07:33,825
of interest in the BOLD response.

3799
03:07:33,825 --> 03:07:36,545
You'll recall that one of the
disadvantages of BOLD fMRI

3800
03:07:36,545 --> 03:07:39,015
and its semi-quantitative nature

3801
03:07:39,015 --> 03:07:42,095
is that BOLD can't be
compared across brain regions

3802
03:07:42,095 --> 03:07:45,285
within the same individual across times

3803
03:07:45,285 --> 03:07:46,945
or between individuals.

3804
03:07:46,945 --> 03:07:49,155
So there's a whole bunch
of neurovascular confounds

3805
03:07:49,155 --> 03:07:51,495
that limits these comparisons,

3806
03:07:51,495 --> 03:07:56,425
but, in contrast, FDG-PET
doesn't have these disadvantages

3807
03:07:56,425 --> 03:07:58,028
because it can be quantified.

3808
03:08:00,365 --> 03:08:03,555
PET imaging produces quantitative
radioactivity measurements

3809
03:08:03,555 --> 03:08:04,725
throughout the brain.

3810
03:08:04,725 --> 03:08:06,695
And, when it comes to glucose uptake,

3811
03:08:06,695 --> 03:08:09,455
there are a few different
ways that it might be used.

3812
03:08:09,455 --> 03:08:11,145
I'll briefly mention two now,

3813
03:08:11,145 --> 03:08:13,215
but I'll cover the third in greater detail

3814
03:08:13,215 --> 03:08:15,195
later in the talk.

3815
03:08:15,195 --> 03:08:17,275
A single static PET
image might be acquired

3816
03:08:17,275 --> 03:08:19,985
at a specific time point post injection,

3817
03:08:19,985 --> 03:08:22,615
or the full time course of
radioactivity can be measured

3818
03:08:22,615 --> 03:08:26,185
and the cerebral metabolic
rate of glucose calculated.

3819
03:08:26,185 --> 03:08:28,575
So, at this point, I think
it's necessary to stop

3820
03:08:28,575 --> 03:08:30,475
and define exactly what I mean

3821
03:08:30,475 --> 03:08:33,155
when I use the term static and dynamic.

3822
03:08:33,155 --> 03:08:35,775
So, as an MRI person, I
probably use these terms

3823
03:08:35,775 --> 03:08:38,005
in a different way to a PET person.

3824
03:08:38,005 --> 03:08:39,565
As such, I find it really important

3825
03:08:39,565 --> 03:08:42,628
to define these terms
whenever I use them nowadays.

3826
03:08:45,045 --> 03:08:46,545
When I use the term static,

3827
03:08:46,545 --> 03:08:48,645
what I refer to is a measure

3828
03:08:48,645 --> 03:08:51,635
that essentially takes a
snapshot of glucose uptake

3829
03:08:51,635 --> 03:08:53,235
over the course of a scan.

3830
03:08:53,235 --> 03:08:56,605
So that is resulting in one
or a small number of images

3831
03:08:56,605 --> 03:08:59,405
with a frame duration
in the order of minutes.

3832
03:08:59,405 --> 03:09:01,825
Where it gets confusing
is where dynamic PET

3833
03:09:01,825 --> 03:09:04,678
has been acquired in order to
calculate a metabolic rate.

3834
03:09:05,535 --> 03:09:09,005
However, if that metabolic
rate is calculated

3835
03:09:09,005 --> 03:09:10,765
either over the scan duration

3836
03:09:10,765 --> 03:09:13,425
or over a frame duration
of several minutes,

3837
03:09:13,425 --> 03:09:15,095
then it's assuming that the uptake

3838
03:09:15,095 --> 03:09:17,025
during that time is stationary.

3839
03:09:17,025 --> 03:09:18,535
So, therefore, I use the term static

3840
03:09:18,535 --> 03:09:21,505
or static snapshot to describe this.

3841
03:09:21,505 --> 03:09:23,135
This contrasts with what we would get

3842
03:09:23,135 --> 03:09:25,895
in fMRI, EEG, and infusion PET,

3843
03:09:25,895 --> 03:09:28,135
where you have a very short frame duration

3844
03:09:28,135 --> 03:09:29,605
or repetition time,

3845
03:09:29,605 --> 03:09:31,485
and you're able to estimate a time course

3846
03:09:31,485 --> 03:09:33,305
of uptake or activity.

3847
03:09:33,305 --> 03:09:35,555
I often use the word dynamic in that case.

3848
03:09:35,555 --> 03:09:38,005
However, depending on the analysis method,

3849
03:09:38,005 --> 03:09:40,935
that too could also be
a stationary metric.

3850
03:09:40,935 --> 03:09:45,165
So I'll be careful to define
the terms on each use,

3851
03:09:45,165 --> 03:09:48,005
but I recommend that you
take a look at these papers,

3852
03:09:48,005 --> 03:09:51,975
particularly this one by
Liegeois et al. in 2017,

3853
03:09:51,975 --> 03:09:54,105
where they have a nice
mathematical exploration

3854
03:09:54,105 --> 03:09:56,638
of these concepts with regards to fMRI.

3855
03:10:00,965 --> 03:10:02,745
Lastly, it's important to consider

3856
03:10:02,745 --> 03:10:04,645
what brain function is being measured.

3857
03:10:06,427 --> 03:10:09,595
In an fMRI scanner, we can
put people in the scanner

3858
03:10:09,595 --> 03:10:10,985
and measure their brain activity

3859
03:10:10,985 --> 03:10:15,015
to multiple instances of
the same or similar stimuli.

3860
03:10:15,015 --> 03:10:17,725
So take, for example, a faces experiment,

3861
03:10:17,725 --> 03:10:20,645
where we might be interested in activation

3862
03:10:20,645 --> 03:10:22,775
of the fusiform face area.

3863
03:10:22,775 --> 03:10:24,645
The assumption is that the neural response

3864
03:10:24,645 --> 03:10:27,275
to a face stimulus at the
start of the experiment

3865
03:10:27,275 --> 03:10:29,825
is essentially the same as
a face stimulus presented

3866
03:10:29,825 --> 03:10:31,705
at the end of the experiment.

3867
03:10:31,705 --> 03:10:33,735
The multiple presentations of the stimulus

3868
03:10:33,735 --> 03:10:35,785
allows us to use signal averaging,

3869
03:10:35,785 --> 03:10:37,325
which assumes that the neural response

3870
03:10:37,325 --> 03:10:39,345
to multiple presentations of stimuli

3871
03:10:39,345 --> 03:10:41,375
is statistically linear.

3872
03:10:41,375 --> 03:10:45,435
Rough linearity of the BOLD
response has been demonstrated.

3873
03:10:45,435 --> 03:10:48,145
However, PET's a
completely different beast.

3874
03:10:48,145 --> 03:10:50,885
First off, standard
approaches need to wait

3875
03:10:50,885 --> 03:10:52,655
until the radio tracer has been taken up

3876
03:10:52,655 --> 03:10:54,965
and transported throughout the body

3877
03:10:54,965 --> 03:10:56,278
before it can be measured.

3878
03:10:57,455 --> 03:11:00,175
So people are injected
with the radio tracer,

3879
03:11:00,175 --> 03:11:01,805
they might wait in an uptake room

3880
03:11:01,805 --> 03:11:04,105
for, say, 10, 20, 30 minutes,

3881
03:11:04,105 --> 03:11:05,815
and then they're taken into a scanner.

3882
03:11:05,815 --> 03:11:08,235
They're positioned, given
instructions on what to do,

3883
03:11:08,235 --> 03:11:10,335
and then the scan starts.

3884
03:11:10,335 --> 03:11:11,475
So, if we were then to present them

3885
03:11:11,475 --> 03:11:13,245
with a face stimulus in the scanner,

3886
03:11:13,245 --> 03:11:15,625
the measured response
is actually the integral

3887
03:11:15,625 --> 03:11:17,135
of all the brain activity

3888
03:11:17,135 --> 03:11:20,075
that's occurred during the
uptake and the scan periods.

3889
03:11:20,075 --> 03:11:22,855
And, unlike fMRI, the signal
at the beginning of the scan

3890
03:11:22,855 --> 03:11:25,695
is different to the signal
at the end of the scan.

3891
03:11:25,695 --> 03:11:27,585
So it's worth bearing these issues in mind

3892
03:11:27,585 --> 03:11:30,155
when interpreting the FDG-PET scan

3893
03:11:30,155 --> 03:11:32,438
as an index of brain function.

3894
03:11:34,325 --> 03:11:35,935
So, from a cognitive neuroscience,

3895
03:11:35,935 --> 03:11:39,185
what are the benefits of the
simultaneous acquisition?

3896
03:11:39,185 --> 03:11:41,185
Well, I'm guessing, as a
participant in this workshop,

3897
03:11:41,185 --> 03:11:42,805
you've got a pretty good idea,

3898
03:11:42,805 --> 03:11:45,415
but let's just briefly take stock about

3899
03:11:45,415 --> 03:11:47,085
what are the specific advantages

3900
03:11:47,085 --> 03:11:48,985
for the study of human brain function?

3901
03:11:50,995 --> 03:11:54,555
The true benefit of the
simultaneous acquisition

3902
03:11:54,555 --> 03:11:58,675
is that you can examine multiple
physiological components

3903
03:11:58,675 --> 03:12:01,615
of the same neural
function at the same time.

3904
03:12:01,615 --> 03:12:04,655
So, rather than having
differences in state or behavior

3905
03:12:04,655 --> 03:12:06,305
between testing sessions,

3906
03:12:06,305 --> 03:12:09,695
you're measuring the
exact same neural function

3907
03:12:09,695 --> 03:12:11,785
from multiple perspectives.

3908
03:12:11,785 --> 03:12:14,142
So, in my case, I use it
to measure hemodynamic

3909
03:12:14,142 --> 03:12:17,705
and glucose dynamic components
of neuronal activity,

3910
03:12:17,705 --> 03:12:18,715
but it's also possible

3911
03:12:18,715 --> 03:12:20,625
to look at other physiological targets,

3912
03:12:20,625 --> 03:12:24,135
like neurotransmitter
systems, oxygen, and so forth.

3913
03:12:24,135 --> 03:12:25,875
The simultaneous acquisition,

3914
03:12:25,875 --> 03:12:29,195
just like with other
systems, like EEG-fMRI,

3915
03:12:29,195 --> 03:12:30,555
ensures that the brain activity

3916
03:12:30,555 --> 03:12:32,495
is measured in each modality,

3917
03:12:32,495 --> 03:12:35,388
significantly improving
your inferential power.

3918
03:12:37,195 --> 03:12:40,658
So let's move on to the fPET approach.

3919
03:12:42,645 --> 03:12:46,295
One method that's been developed
with simultaneous PET/MR

3920
03:12:46,295 --> 03:12:49,745
is the so-called
functional FDG-PET method.

3921
03:12:49,745 --> 03:12:51,995
It's called functional PET, or fPET,

3922
03:12:51,995 --> 03:12:54,786
to highlight the similarities with fMRI.

3923
03:12:54,786 --> 03:12:57,485
fPET uses the constant infusion technique,

3924
03:12:57,485 --> 03:12:59,975
where FDG is administered by infusion

3925
03:12:59,975 --> 03:13:02,015
over the course of the scan period.

3926
03:13:02,015 --> 03:13:06,125
And the goal here is to maintain
a constant plasma supply

3927
03:13:06,125 --> 03:13:07,855
of radioactive glucose

3928
03:13:07,855 --> 03:13:09,795
so that dynamic changes in uptake

3929
03:13:09,795 --> 03:13:12,718
in response to a stimulus
or task can be measured.

3930
03:13:14,195 --> 03:13:16,595
So I won't spend time discussing

3931
03:13:16,595 --> 03:13:18,385
the acquisition procedure here,

3932
03:13:18,385 --> 03:13:19,635
but we recently published

3933
03:13:19,635 --> 03:13:22,125
a Journal of Visualized
Experiments video article

3934
03:13:22,125 --> 03:13:23,805
demonstrating the procedure,

3935
03:13:23,805 --> 03:13:26,515
and you can find that
either on the JoVE website

3936
03:13:26,515 --> 03:13:28,958
or on the Monash
Biomedical Imaging website.

3937
03:13:31,045 --> 03:13:35,155
In 2014, the MGH Group published
a proof-of-concept study

3938
03:13:35,155 --> 03:13:37,965
demonstrating the potential
of the fPET method.

3939
03:13:37,965 --> 03:13:40,865
They showed that a constant
plasma supply of FDG

3940
03:13:40,865 --> 03:13:43,765
allows measurement of dynamic
changes in glucose uptake

3941
03:13:43,765 --> 03:13:46,365
in response to a checkerboard stimulus.

3942
03:13:46,365 --> 03:13:49,915
And here, you can see the model
fit within the visual cortex

3943
03:13:49,915 --> 03:13:51,715
in response to alternating blocks

3944
03:13:51,715 --> 03:13:53,355
of flickering checkerboard,

3945
03:13:53,355 --> 03:13:55,405
which they measured with FDG-PET

3946
03:13:55,405 --> 03:13:57,505
in one-minute frame durations.

3947
03:13:57,505 --> 03:14:01,965
They didn't acquire simultaneous
fMRI, just structural MRI.

3948
03:14:01,965 --> 03:14:04,075
So this landmark study
really set the stage

3949
03:14:04,075 --> 03:14:06,475
for a number of studies that
have continued to develop

3950
03:14:06,475 --> 03:14:09,015
and refine the fPET methodology.

3951
03:14:09,015 --> 03:14:12,415
Much of this work has
been conducted in PET/MRI,

3952
03:14:12,415 --> 03:14:15,095
although the fPET approach itself

3953
03:14:15,095 --> 03:14:18,195
can be deployed on other PET systems.

3954
03:14:18,195 --> 03:14:19,555
A lot of work is focused

3955
03:14:19,555 --> 03:14:21,425
on the additional information
that can be gained

3956
03:14:21,425 --> 03:14:25,365
by acquiring fPET
simultaneously with fMRI.

3957
03:14:25,365 --> 03:14:26,995
And Andreas will speak a lot more

3958
03:14:26,995 --> 03:14:30,575
about task-related fPET-fMRI,

3959
03:14:30,575 --> 03:14:34,545
particularly in regards
to modeling the data.

3960
03:14:34,545 --> 03:14:38,435
So I'm gonna focus more on the
application of fPET and fMRI

3961
03:14:38,435 --> 03:14:39,718
in the resting state.

3962
03:14:41,135 --> 03:14:44,485
But, first, I'll just
highlight this first study

3963
03:14:44,485 --> 03:14:46,815
that my group did in this area,

3964
03:14:46,815 --> 03:14:48,745
just to see if we could design a protocol

3965
03:14:48,745 --> 03:14:52,055
that would yield simultaneous
BOLD and FDG signal

3966
03:14:52,055 --> 03:14:53,935
in the visual cortex.

3967
03:14:53,935 --> 03:14:55,915
We designed this embedded block design,

3968
03:14:55,915 --> 03:14:59,365
which is based on the standard
block design used in fMRI.

3969
03:14:59,365 --> 03:15:02,475
Essentially, there are two
block designs in the protocol,

3970
03:15:02,475 --> 03:15:05,225
a slow design with an on/off period

3971
03:15:05,225 --> 03:15:07,465
of between 5 to 10 minutes,

3972
03:15:07,465 --> 03:15:10,065
which was intended to
give an fPET contrast

3973
03:15:10,065 --> 03:15:12,095
between task and rest.

3974
03:15:12,095 --> 03:15:14,055
Embedded within each on period

3975
03:15:14,055 --> 03:15:19,055
was a standard 32 16-second
on/off block design for fMRI,

3976
03:15:19,215 --> 03:15:23,555
which was there to provide
contrast between task and rest.

3977
03:15:23,555 --> 03:15:26,555
We assume that the fast on/off
period would be too fast

3978
03:15:26,555 --> 03:15:29,055
to influence the fPET signal too much,

3979
03:15:29,055 --> 03:15:31,785
and indeed, in this experiment, it was.

3980
03:15:31,785 --> 03:15:33,955
We used a one-minute frame duration,

3981
03:15:33,955 --> 03:15:36,535
and even the five-minute
duration was a bit too fast

3982
03:15:36,535 --> 03:15:39,365
to distinguish between task and rest.

3983
03:15:39,365 --> 03:15:42,045
Down here, you can see
the joint ICA component

3984
03:15:42,045 --> 03:15:43,905
that has a component from each modality

3985
03:15:43,905 --> 03:15:46,228
that is maximally related to the other.

3986
03:15:47,785 --> 03:15:50,475
Also, it's important to note
that the embedded design

3987
03:15:50,475 --> 03:15:54,185
isn't the only approach to
task design for stimulation,

3988
03:15:54,185 --> 03:15:56,875
simultaneous fPET and fMRI.

3989
03:15:56,875 --> 03:15:58,545
This design is fully constrained,

3990
03:15:58,545 --> 03:16:00,185
so we know exactly when people

3991
03:16:00,185 --> 03:16:02,735
are performing each component of the task.

3992
03:16:02,735 --> 03:16:06,405
It's probably a bit of a
holdover from my training at EEG,

3993
03:16:06,405 --> 03:16:08,965
where everything has to
be millisecond accurate.

3994
03:16:08,965 --> 03:16:12,575
Other groups, however, have
used much less constrained tasks

3995
03:16:12,575 --> 03:16:15,178
to great success with fPET and fMRI.

3996
03:16:17,285 --> 03:16:20,545
One question that's been explored
since that original paper

3997
03:16:20,545 --> 03:16:23,025
is how we can improve
the temporal resolution

3998
03:16:23,025 --> 03:16:25,595
from the one minute
reported in that paper.

3999
03:16:25,595 --> 03:16:28,835
So the Vienna Group published
this nice paper in 2018,

4000
03:16:28,835 --> 03:16:31,485
focusing on manipulating
the experimental design

4001
03:16:31,485 --> 03:16:33,055
and task duration.

4002
03:16:33,055 --> 03:16:34,285
On the basis of this study,

4003
03:16:34,285 --> 03:16:37,765
they concluded that an optimal
task duration of five minutes

4004
03:16:37,765 --> 03:16:41,208
with 30-second frame rates was best.

4005
03:16:43,335 --> 03:16:46,095
We've also looked at whether or not

4006
03:16:46,095 --> 03:16:48,075
a bolus and infusion protocol

4007
03:16:48,075 --> 03:16:50,655
might provide an improved
temporal resolution.

4008
03:16:50,655 --> 03:16:54,665
So we compared bolus infusion
and bolus plus infusion

4009
03:16:54,665 --> 03:16:56,815
in a proof-of-concept study.

4010
03:16:56,815 --> 03:16:57,655
So, down the bottom here,

4011
03:16:57,655 --> 03:16:59,515
you can see the experimental design

4012
03:16:59,515 --> 03:17:00,348
and at the top

4013
03:17:00,348 --> 03:17:03,525
is the hypothesized plasma
radioactivity levels.

4014
03:17:03,525 --> 03:17:05,595
Note that, for the infusion protocols,

4015
03:17:05,595 --> 03:17:08,195
we turned the pump off at 55 minutes,

4016
03:17:08,195 --> 03:17:10,545
assuming that plasma
supply would remain stable

4017
03:17:10,545 --> 03:17:12,705
for a few minutes after.

4018
03:17:12,705 --> 03:17:14,225
In this initial study,

4019
03:17:14,225 --> 03:17:17,985
we used a 50% bolus and a 50% infusion.

4020
03:17:17,985 --> 03:17:20,925
Although, note that the
previous study that I discussed

4021
03:17:20,925 --> 03:17:23,455
on the previous slide by Rischka et al.

4022
03:17:23,455 --> 03:17:25,268
used a 20% bolus.

4023
03:17:27,980 --> 03:17:30,335
And so here are the plasma
radioactivity curves.

4024
03:17:30,335 --> 03:17:32,805
They basically resemble our hypotheses

4025
03:17:32,805 --> 03:17:34,625
except that there was an obvious drop

4026
03:17:34,625 --> 03:17:37,345
when the infusion pump was switched off.

4027
03:17:37,345 --> 03:17:39,815
And so our first lesson from this study

4028
03:17:39,815 --> 03:17:43,185
was that we need to infuse
for as long as we scan.

4029
03:17:43,185 --> 03:17:45,305
Down the bottom, you can
see the gray matter signal

4030
03:17:45,305 --> 03:17:47,275
for each of the three protocols.

4031
03:17:47,275 --> 03:17:49,555
You can see, qualitatively at least,

4032
03:17:49,555 --> 03:17:53,075
that the infusion protocol
has got a really low signal.

4033
03:17:53,075 --> 03:17:55,325
It never reaches the
levels of signals obtained

4034
03:17:55,325 --> 03:17:57,165
in the other two protocols.

4035
03:17:57,165 --> 03:18:00,125
The bolus approach has got
the fastest increase in signal

4036
03:18:00,125 --> 03:18:01,905
at the beginning of the scan,

4037
03:18:01,905 --> 03:18:05,825
but then the slope of the
uptake is relatively shallow

4038
03:18:05,825 --> 03:18:07,705
over the course of the scan.

4039
03:18:07,705 --> 03:18:09,575
The bolus plus infusion protocol

4040
03:18:09,575 --> 03:18:12,135
shows a larger signal at the beginning

4041
03:18:12,135 --> 03:18:14,245
relative to the infusion protocol,

4042
03:18:14,245 --> 03:18:16,255
but it's not as high as the bolus only.

4043
03:18:16,255 --> 03:18:18,185
Again, this is as expected

4044
03:18:18,185 --> 03:18:21,885
'cause we're only providing
50% of the dose as the bolus.

4045
03:18:21,885 --> 03:18:24,875
The slope of the uptake for the
bolus plus infusion protocol

4046
03:18:24,875 --> 03:18:27,005
is larger over the course of the scan,

4047
03:18:27,005 --> 03:18:29,325
and, for at least four
of the five subjects,

4048
03:18:29,325 --> 03:18:32,195
it ultimately shows an
overall larger signal

4049
03:18:32,195 --> 03:18:34,235
than both the bolus and infusion only

4050
03:18:34,235 --> 03:18:35,578
by the end of the scan.

4051
03:18:36,925 --> 03:18:40,325
Finally, on the bottom left,
you can see the fMRI response

4052
03:18:40,325 --> 03:18:42,345
averaged across the three protocols.

4053
03:18:42,345 --> 03:18:44,465
Again, there's nothing
groundbreaking here.

4054
03:18:44,465 --> 03:18:47,555
A flickering checkerboard
is activating visual cortex

4055
03:18:47,555 --> 03:18:48,948
and oculomotor network.

4056
03:18:50,585 --> 03:18:53,495
When we look at the fPET
results for the three protocols,

4057
03:18:53,495 --> 03:18:55,595
the first thing that we noticed
was that the first block

4058
03:18:55,595 --> 03:18:57,218
was virtually inestimable,

4059
03:18:58,723 --> 03:19:01,585
unestimable for the infusion protocol

4060
03:19:01,585 --> 03:19:03,715
because the signal was so low.

4061
03:19:03,715 --> 03:19:04,685
And, by the fifth block,

4062
03:19:04,685 --> 03:19:07,315
the signal for the bolus
group was all over the place

4063
03:19:07,315 --> 03:19:10,735
because it was also starting to get low.

4064
03:19:10,735 --> 03:19:13,868
So, here, we're looking at
the three middle blocks.

4065
03:19:14,725 --> 03:19:16,785
The bolus actually performed not too badly

4066
03:19:16,785 --> 03:19:17,775
at this time point,

4067
03:19:17,775 --> 03:19:21,405
showing activity in at least
the left visual cortex.

4068
03:19:21,405 --> 03:19:23,735
Possibly, there's some
artifact or false positive

4069
03:19:23,735 --> 03:19:27,425
in the medial parietal
area around the precuneus.

4070
03:19:27,425 --> 03:19:30,795
The infusion protocol also
showed visual cortex activity,

4071
03:19:30,795 --> 03:19:33,165
but, again, there were some
potential false positives

4072
03:19:33,165 --> 03:19:35,785
around the rim of the frontal cortex.

4073
03:19:35,785 --> 03:19:40,785
Lastly, the fPET signal in
the bolus infusion protocol

4074
03:19:41,595 --> 03:19:44,185
showed bilateral activity
in the visual cortex

4075
03:19:44,185 --> 03:19:46,715
with some hints that it
might also be coming apparent

4076
03:19:46,715 --> 03:19:48,855
in the intraparietal sulcus,

4077
03:19:48,855 --> 03:19:51,655
although this is probably a
little too inferior for IPS.

4078
03:19:52,825 --> 03:19:55,125
Lastly, note that across
all three protocols,

4079
03:19:55,125 --> 03:19:56,795
the activation pattern for fPET

4080
03:19:56,795 --> 03:19:59,285
is more focal than for fMRI.

4081
03:19:59,285 --> 03:20:00,118
So, first off,

4082
03:20:00,118 --> 03:20:02,565
because we're not comparing
apples to apples here,

4083
03:20:02,565 --> 03:20:04,602
the fMRI is across 15 subjects

4084
03:20:04,602 --> 03:20:09,602
and each of these PET protocols
is in five subjects each.

4085
03:20:09,945 --> 03:20:12,475
Secondly, the finding
that fPET tends to provide

4086
03:20:12,475 --> 03:20:14,745
more focal activity than fMRI

4087
03:20:14,745 --> 03:20:18,055
is actually a consistent
finding across studies.

4088
03:20:18,055 --> 03:20:21,315
So I interpret this as reflecting
less vascular spreading

4089
03:20:21,315 --> 03:20:23,195
or a more strongly localized signal

4090
03:20:23,195 --> 03:20:26,128
to the presynaptic
neuron in PET than fMRI.

4091
03:20:27,375 --> 03:20:29,265
And, on the basis of these results,

4092
03:20:29,265 --> 03:20:32,785
the plasma radioactivity, the
gray matter signal, and maps,

4093
03:20:32,785 --> 03:20:36,405
we decided to use a bolus
infusion protocol from now on.

4094
03:20:36,405 --> 03:20:38,465
However, note that we've
never actually optimized

4095
03:20:38,465 --> 03:20:42,055
which combination of bolus
and infusion is best.

4096
03:20:42,055 --> 03:20:45,965
We've used a 50/50 bolus and
infusion for simplicity's sake,

4097
03:20:45,965 --> 03:20:49,728
but, as I noted, others
are using a 20/80 ratio.

4098
03:20:52,335 --> 03:20:53,465
So, in this final slide,

4099
03:20:53,465 --> 03:20:55,255
I'll briefly note that my collaborators

4100
03:20:55,255 --> 03:20:57,905
at Monash Biomedical
Imaging have also explored

4101
03:20:57,905 --> 03:21:00,995
whether using MR-informed
PET reconstruction

4102
03:21:00,995 --> 03:21:05,035
can be useful for improving the
temporal resolution of fPET.

4103
03:21:05,035 --> 03:21:08,045
They've used my low-dose
PET with 100 megabecquerels

4104
03:21:08,045 --> 03:21:10,585
as well as some artificially degraded data

4105
03:21:10,585 --> 03:21:12,845
to simulate ultra low-dose fPET.

4106
03:21:14,045 --> 03:21:16,765
They've then used a number
of down-sampling factors

4107
03:21:16,765 --> 03:21:19,525
to simulate a faster temporal resolution.

4108
03:21:19,525 --> 03:21:22,115
This work suggests that faster
sampling could be possible

4109
03:21:22,115 --> 03:21:24,478
with signal optimization
strategies like this.

4110
03:21:26,555 --> 03:21:30,335
So, in summary, FDG-fPET
allows measurement

4111
03:21:30,335 --> 03:21:33,455
of glucose uptake with a
high temporal resolution,

4112
03:21:33,455 --> 03:21:36,085
and the bolus plus infusion approach

4113
03:21:36,085 --> 03:21:39,418
appears to have a better
SNR than infusion only.

4114
03:21:40,955 --> 03:21:44,305
So let's move on to the
next part of the talk,

4115
03:21:44,305 --> 03:21:47,465
PET/MR for the study
of brain connectivity.

4116
03:21:47,465 --> 03:21:49,205
And here, I'll just pause and let you note

4117
03:21:49,205 --> 03:21:52,215
that the first ever resting-state
connectivity analysis

4118
03:21:52,215 --> 03:21:56,138
was actually conducted almost
40 years ago using FDG-PET.

4119
03:21:59,285 --> 03:22:01,965
Network neuroscience has become
one of the dominant models

4120
03:22:01,965 --> 03:22:04,775
for understanding how human
cognition and function

4121
03:22:04,775 --> 03:22:08,345
arises from the interaction
of multiple neural systems,

4122
03:22:08,345 --> 03:22:11,825
including brain structure,
anatomical connections,

4123
03:22:11,825 --> 03:22:15,305
systems interactions, and
coordinated neuronal activity.

4124
03:22:15,305 --> 03:22:17,435
And, at least in its
first conceptualization

4125
03:22:18,316 --> 03:22:21,705
almost 20 years ago, the human
connectome was considered

4126
03:22:21,705 --> 03:22:25,755
as a comprehensive map
of anatomical connections

4127
03:22:25,755 --> 03:22:27,908
at micro and meso scales.

4128
03:22:29,825 --> 03:22:32,295
So, while the term connectome
was originally conceived

4129
03:22:32,295 --> 03:22:34,815
as a map of anatomical connections,

4130
03:22:34,815 --> 03:22:38,415
it's grown to encompass multiple
dimensions of connectivity.

4131
03:22:38,415 --> 03:22:39,615
The most prominent one,

4132
03:22:39,615 --> 03:22:42,305
and the one that we'll
talk about the most today,

4133
03:22:42,305 --> 03:22:44,235
is functional connectivity,

4134
03:22:44,235 --> 03:22:47,325
which, by and large, is usually used

4135
03:22:47,325 --> 03:22:49,745
to infer connectivity measured using fMRI.

4136
03:22:52,477 --> 03:22:55,795
At its simplest level, functional
connectivity simply refers

4137
03:22:55,795 --> 03:22:58,705
to the statistical
dependencies from a time series

4138
03:22:58,705 --> 03:23:01,845
and is often calculated from
a simple Pearson correlation

4139
03:23:01,845 --> 03:23:04,375
of time series between brain regions.

4140
03:23:04,375 --> 03:23:06,538
Can we calculate a
similar measure for PET?

4141
03:23:08,025 --> 03:23:09,365
Well, the short answer is yes.

4142
03:23:09,365 --> 03:23:12,305
We've already seen that
this was done 40 years ago

4143
03:23:13,185 --> 03:23:15,495
by Horwitz and colleagues.

4144
03:23:15,495 --> 03:23:17,575
Network-based measures of PET

4145
03:23:17,575 --> 03:23:19,485
actually predate the fMRI literature

4146
03:23:19,485 --> 03:23:21,305
by at least two decades.

4147
03:23:21,305 --> 03:23:24,405
And here, you can see the
very early network analysis

4148
03:23:24,405 --> 03:23:26,225
by Horwitz and colleagues.

4149
03:23:26,225 --> 03:23:28,095
And what's really cool
about these matrices

4150
03:23:28,095 --> 03:23:29,685
is that you can see some of the structure

4151
03:23:29,685 --> 03:23:32,425
that we're used to seeing
in functional connectomes,

4152
03:23:32,425 --> 03:23:35,585
including a predominance of
connections along the diagonal,

4153
03:23:35,585 --> 03:23:37,545
indicating that regions
that are close together

4154
03:23:37,545 --> 03:23:39,398
are sharing a common variance.

4155
03:23:41,565 --> 03:23:43,575
So this concept of metabolic connectivity

4156
03:23:43,575 --> 03:23:46,935
is being proposed as a
punitive biomarker for disease.

4157
03:23:46,935 --> 03:23:49,095
So, for example, it's
been shown to be useful

4158
03:23:49,095 --> 03:23:52,065
for distinguishing between
different types of dementia,

4159
03:23:52,065 --> 03:23:55,115
between different subtypes
of Alzheimer's disease,

4160
03:23:55,115 --> 03:23:57,855
and also between a
relatively rare condition,

4161
03:23:57,855 --> 03:23:59,155
Meige syndrome,

4162
03:23:59,155 --> 03:24:01,625
which is a type of dystonia
of the neck and jaw,

4163
03:24:01,625 --> 03:24:03,188
compared to controls.

4164
03:24:05,255 --> 03:24:07,045
However, the vast
majority of these studies

4165
03:24:07,045 --> 03:24:08,845
have used a static acquisition

4166
03:24:08,845 --> 03:24:11,455
that acquires a single image per subject.

4167
03:24:11,455 --> 03:24:14,195
So this means that the
measure of connectivity

4168
03:24:14,195 --> 03:24:17,335
is really a measure of
across-subject covariance.

4169
03:24:17,335 --> 03:24:18,565
So these measures are useful

4170
03:24:18,565 --> 03:24:21,775
for distinguishing between
groups of subjects,

4171
03:24:21,775 --> 03:24:23,705
but, for a measure to be a biomarker,

4172
03:24:23,705 --> 03:24:25,775
I'd argue that it needs to be estimable

4173
03:24:25,775 --> 03:24:27,288
at the individual level.

4174
03:24:28,985 --> 03:24:30,935
So, what do we mean by this?

4175
03:24:30,935 --> 03:24:32,045
If we first consider

4176
03:24:32,045 --> 03:24:34,395
how we calculate a functional connectome,

4177
03:24:34,395 --> 03:24:37,225
you can see that we're acquiring
multiple functional images

4178
03:24:37,225 --> 03:24:39,145
per subject over time.

4179
03:24:39,145 --> 03:24:42,505
So, here, we have a
temporal resolution, or TR,

4180
03:24:42,505 --> 03:24:47,075
of 2.45 seconds, but this time
course could be sub-second

4181
03:24:47,075 --> 03:24:50,175
with a multi-band sequence or
in the range of milliseconds

4182
03:24:50,175 --> 03:24:52,525
if we were to use a
technique like EEG and MEG.

4183
03:24:55,245 --> 03:24:57,985
We then parcellate the
brain into some regions,

4184
03:24:57,985 --> 03:25:01,595
extract the time series of
fMRI signal for each region,

4185
03:25:01,595 --> 03:25:04,635
and then correlate those
time series across regions.

4186
03:25:04,635 --> 03:25:06,915
We do this for every
subject in our sample,

4187
03:25:06,915 --> 03:25:08,195
and then we get a group average

4188
03:25:08,195 --> 03:25:09,948
and illustrate it in some way.

4189
03:25:12,025 --> 03:25:14,535
Now, if we take the scenario of PET

4190
03:25:14,535 --> 03:25:17,135
for what I'm calling a static acquisition

4191
03:25:17,135 --> 03:25:18,605
or an acquisition where we end up

4192
03:25:18,605 --> 03:25:20,578
with a single image per subject,

4193
03:25:21,705 --> 03:25:24,365
here, we also parcellate the brain,

4194
03:25:24,365 --> 03:25:26,762
but then we calculate
covariance across subjects

4195
03:25:26,762 --> 03:25:29,005
for each pair of regions.

4196
03:25:29,005 --> 03:25:30,575
So, unlike the case for fMRI,

4197
03:25:30,575 --> 03:25:33,875
we're not getting a connectome
for each individual.

4198
03:25:33,875 --> 03:25:36,505
So, if we consider the argument
that metabolic covariance

4199
03:25:36,505 --> 03:25:39,575
is a biomarker for disease,
this can't be the case

4200
03:25:39,575 --> 03:25:43,325
because, to be useful, a
biomarker needs to be estimable

4201
03:25:43,325 --> 03:25:45,065
at the individual level.

4202
03:25:45,065 --> 03:25:46,995
So, while it's telling
us something interesting

4203
03:25:46,995 --> 03:25:49,215
about the group itself, as we've seen

4204
03:25:49,215 --> 03:25:52,485
even from those very early
PET connectivity papers,

4205
03:25:52,485 --> 03:25:55,688
it's just not as useful as a biomarker.

4206
03:25:57,855 --> 03:25:59,625
As an example of this approach

4207
03:25:59,625 --> 03:26:03,505
in standalone PET without
simultaneous PET/MR,

4208
03:26:03,505 --> 03:26:05,285
Arnemann and colleagues examined

4209
03:26:05,285 --> 03:26:08,645
metabolic covariance
networks using static PET.

4210
03:26:08,645 --> 03:26:10,965
They found that the young
group differed profoundly

4211
03:26:10,965 --> 03:26:11,865
from the older group,

4212
03:26:11,865 --> 03:26:15,115
showing widespread, high
metabolic correlation strength

4213
03:26:15,115 --> 03:26:17,395
compared to younger adults.

4214
03:26:17,395 --> 03:26:19,895
In addition, the authors also examined

4215
03:26:19,895 --> 03:26:22,185
between and within network connectivity.

4216
03:26:22,185 --> 03:26:26,095
So, for reference, in fMRI,
it's an established finding

4217
03:26:26,095 --> 03:26:29,565
that within-network
connectivity, along the diagonal,

4218
03:26:29,565 --> 03:26:31,765
decreases with increasing age,

4219
03:26:31,765 --> 03:26:34,245
and between-network connectivity,

4220
03:26:34,245 --> 03:26:36,795
or those regions off the
diagonal in the matrix,

4221
03:26:36,795 --> 03:26:39,155
increases with increasing age.

4222
03:26:39,155 --> 03:26:40,285
So this is interpreted

4223
03:26:40,285 --> 03:26:42,995
as reflecting reduced network segregation

4224
03:26:42,995 --> 03:26:45,915
and increased network
integration with age.

4225
03:26:45,915 --> 03:26:48,695
So, in other words, a
loss of specialization

4226
03:26:48,695 --> 03:26:52,497
and an increase in generalization
of network function.

4227
03:26:52,497 --> 03:26:54,735
Arnemann and colleagues found
that metabolic covariance

4228
03:26:54,735 --> 03:26:57,895
showed uniformly high
metabolic correlation strength

4229
03:26:57,895 --> 03:27:00,605
both within and between networks.

4230
03:27:00,605 --> 03:27:03,715
And, in some respect, this
can be considered an update

4231
03:27:03,715 --> 03:27:05,505
of that very early work that we saw

4232
03:27:05,505 --> 03:27:08,178
from Horwitz and colleagues in the 1980s.

4233
03:27:09,955 --> 03:27:13,245
So, what about PET/MR measures
of brain connectivity?

4234
03:27:13,245 --> 03:27:15,705
Some of the first
simultaneous MR/PET studies

4235
03:27:15,705 --> 03:27:18,245
were conducted in the resting state.

4236
03:27:18,245 --> 03:27:21,565
So Riedl and colleagues
examined eyes open versus closed

4237
03:27:21,565 --> 03:27:22,805
in the resting state,

4238
03:27:22,805 --> 03:27:25,305
and they found that there
was a remarkable overlap

4239
03:27:25,305 --> 03:27:29,645
between increased FDG uptake
in eyes open versus closed

4240
03:27:29,645 --> 03:27:31,815
and functional connectivity in V1

4241
03:27:31,815 --> 03:27:35,105
and all of the major
salience network regions.

4242
03:27:35,105 --> 03:27:37,855
Importantly, at least
for the discussion today,

4243
03:27:37,855 --> 03:27:39,565
while this was a PET/MR study

4244
03:27:39,565 --> 03:27:41,285
and they used functional connectivity

4245
03:27:41,285 --> 03:27:43,905
estimated from fMRI data,

4246
03:27:43,905 --> 03:27:45,375
the authors did not calculate

4247
03:27:45,375 --> 03:27:49,465
a metabolic connectivity
measure based on the FDG data.

4248
03:27:49,465 --> 03:27:51,325
In a follow-up of this paper,

4249
03:27:51,325 --> 03:27:53,785
the authors used an ICA decomposition

4250
03:27:53,785 --> 03:27:56,425
of the static FDG and fMRI data

4251
03:27:56,425 --> 03:27:58,885
to look at resting state networks.

4252
03:27:58,885 --> 03:28:02,015
They found that the FDG networks
showed some similarities

4253
03:28:02,015 --> 03:28:03,385
with fMRI networks,

4254
03:28:03,385 --> 03:28:05,075
but there were also a lot of networks

4255
03:28:05,075 --> 03:28:07,315
that were unique to each modality.

4256
03:28:07,315 --> 03:28:11,185
The results are compatible
with FDG ICA decomposition

4257
03:28:11,185 --> 03:28:13,535
published previously by
Biswal and colleagues

4258
03:28:13,535 --> 03:28:15,045
from the ADNI data,

4259
03:28:15,045 --> 03:28:17,835
where they reported that FDG components

4260
03:28:17,835 --> 03:28:21,015
typically show more left-right
homolog connectivity

4261
03:28:21,015 --> 03:28:23,608
and less anterior-posterior connectivity.

4262
03:28:24,925 --> 03:28:27,765
So data from the simultaneous MR/PET

4263
03:28:27,765 --> 03:28:29,365
with static acquisitions

4264
03:28:29,365 --> 03:28:31,495
is suggesting that metabolic connectivity

4265
03:28:31,495 --> 03:28:33,315
may show some similarities

4266
03:28:33,315 --> 03:28:36,322
but also differences in
functional connectivity.

4267
03:28:38,715 --> 03:28:41,285
One cool way that the
simultaneous PET/MR approach

4268
03:28:41,285 --> 03:28:43,725
has been used to estimate connectivity

4269
03:28:43,725 --> 03:28:47,645
is by using the metabolic
connectivity mapping approach.

4270
03:28:47,645 --> 03:28:50,595
This approach essentially
uses the FDG data

4271
03:28:50,595 --> 03:28:54,145
to derive a directional
functional connectivity,

4272
03:28:54,145 --> 03:28:57,375
which is also known as
effective connectivity.

4273
03:28:57,375 --> 03:29:00,255
Under the assumption that
most FDG-related signal

4274
03:29:00,255 --> 03:29:04,355
is occurring post-synaptically,
i.e., at target neurons,

4275
03:29:04,355 --> 03:29:07,195
they assume that an
increase in local metabolism

4276
03:29:07,195 --> 03:29:10,875
indicates an increase in
afferent effective connectivity

4277
03:29:10,875 --> 03:29:12,815
from source regions.

4278
03:29:12,815 --> 03:29:14,355
So, using this approach,

4279
03:29:14,355 --> 03:29:17,635
Riedl et al. studied
eyes open versus closed

4280
03:29:17,635 --> 03:29:20,655
in the same data presented
in the previous slide.

4281
03:29:20,655 --> 03:29:22,655
They found that metabolic
connectivity mapping

4282
03:29:22,655 --> 03:29:26,765
reliably detected stable and
bidirectional communication

4283
03:29:26,765 --> 03:29:29,385
between earlier and higher visual regions

4284
03:29:29,385 --> 03:29:31,305
as well as stable top-down signaling

4285
03:29:31,305 --> 03:29:32,995
from the frontoparietal network

4286
03:29:32,995 --> 03:29:35,655
encompassing the frontal eye fields.

4287
03:29:35,655 --> 03:29:38,795
The MCM approach has since
been used quite a bit

4288
03:29:38,795 --> 03:29:42,475
in task-related PET/MR by
Andreas Hahn and colleagues.

4289
03:29:42,475 --> 03:29:45,045
So this is a nice example
of an additional way

4290
03:29:45,045 --> 03:29:46,585
to use metabolic information

4291
03:29:46,585 --> 03:29:48,385
in functional connectivity analysis.

4292
03:29:50,995 --> 03:29:53,915
So far we've seen approaches
to PET/MR estimates

4293
03:29:53,915 --> 03:29:55,995
of brain connectivity that have relied

4294
03:29:55,995 --> 03:29:59,055
upon the across-subject
covariance approach.

4295
03:29:59,055 --> 03:30:01,035
However, it's increasingly
being acknowledged

4296
03:30:01,035 --> 03:30:03,495
that individual level metric is required

4297
03:30:03,495 --> 03:30:07,395
in order to use metabolic
connectivity as a biomarker.

4298
03:30:07,395 --> 03:30:10,245
One approach to estimating
metabolic connectivity

4299
03:30:10,245 --> 03:30:13,915
uses a similarity estimation
to obtain a surrogate measure

4300
03:30:13,915 --> 03:30:16,975
of metabolic connectivity
between brain regions.

4301
03:30:16,975 --> 03:30:19,345
Studies that use this
approach are recognizing

4302
03:30:19,345 --> 03:30:21,415
that the individual-level connectomes

4303
03:30:21,415 --> 03:30:25,565
are necessary to make inferences
on individual variability.

4304
03:30:25,565 --> 03:30:27,085
This approach was originally developed

4305
03:30:27,085 --> 03:30:30,575
by Wang and colleagues
and tested on ADNI data.

4306
03:30:30,575 --> 03:30:33,515
In this study, they used
intraregional similarities

4307
03:30:33,515 --> 03:30:35,875
calculated from KLSE

4308
03:30:35,875 --> 03:30:37,625
to estimate the strength of connections

4309
03:30:37,625 --> 03:30:39,725
to obtain individual level,

4310
03:30:39,725 --> 03:30:42,928
undirected, weighted metabolic
connectivity matrices.

4311
03:30:43,995 --> 03:30:45,805
Using the method, Wang and colleagues

4312
03:30:45,805 --> 03:30:49,221
were able to predict
conversion from MCI to AD

4313
03:30:49,221 --> 03:30:52,255
in the ADNI data.

4314
03:30:52,255 --> 03:30:53,965
And, in fact, the surrogate measure

4315
03:30:53,965 --> 03:30:55,925
of individual-level connectivity

4316
03:30:55,925 --> 03:30:59,035
outperformed the across-subject
covariance approach,

4317
03:30:59,035 --> 03:31:01,475
which has previously been
highlighted as useful

4318
03:31:01,475 --> 03:31:03,968
as a predictive biomarker from this data.

4319
03:31:05,985 --> 03:31:08,905
And because we're specifically
interested in PET/MR here,

4320
03:31:08,905 --> 03:31:11,365
the KLSE approach has also been applied

4321
03:31:11,365 --> 03:31:13,917
to simultaneously acquired PET/MR.

4322
03:31:13,917 --> 03:31:14,935
Albeit, in this case,

4323
03:31:14,935 --> 03:31:18,995
functional MRI was not
obtained with the PET-FDG data.

4324
03:31:18,995 --> 03:31:22,425
Mertens and colleagues first
parcellated the FDG data

4325
03:31:22,425 --> 03:31:25,595
using the Schaefer atlas,
which is based on fMRI,

4326
03:31:25,595 --> 03:31:27,955
and then sorted the regions
of interest into networks

4327
03:31:27,955 --> 03:31:29,775
based on that atlas.

4328
03:31:29,775 --> 03:31:32,645
Metabolic connectivity
strength was estimated

4329
03:31:32,645 --> 03:31:35,915
as the similarity metric
to obtain an undirected,

4330
03:31:35,915 --> 03:31:38,845
weighted metabolic connectivity matrix.

4331
03:31:38,845 --> 03:31:41,635
They then applied nodal metrics
to the subsequent networks

4332
03:31:41,635 --> 03:31:44,465
to examine metabolic
connectivity in aging,

4333
03:31:44,465 --> 03:31:46,625
and they found that metabolic
connectivity strength

4334
03:31:46,625 --> 03:31:49,445
decreased with age in predefined networks

4335
03:31:49,445 --> 03:31:54,395
as well as increased with
age between networks.

4336
03:31:54,395 --> 03:31:56,385
So this is compatible with what we've seen

4337
03:31:56,385 --> 03:31:58,395
in functional connectivity,

4338
03:31:58,395 --> 03:32:00,255
but it is the opposite of what we've seen

4339
03:32:00,255 --> 03:32:03,028
in network-based measures
of across-subject variance,

4340
03:32:04,031 --> 03:32:05,775
as discussed with reference

4341
03:32:05,775 --> 03:32:08,088
to Arnemann and colleagues in 2018.

4342
03:32:10,975 --> 03:32:12,955
So we've seen a number of approaches

4343
03:32:12,955 --> 03:32:15,075
to estimating metabolic connectivity

4344
03:32:15,075 --> 03:32:16,728
using the static PET data.

4345
03:32:17,755 --> 03:32:20,515
We've seen across-subject
metabolic covariance,

4346
03:32:20,515 --> 03:32:24,025
which is analogous to
cortical thickness networks,

4347
03:32:24,025 --> 03:32:26,615
which estimate covariance
between brain regions

4348
03:32:26,615 --> 03:32:28,505
across subjects.

4349
03:32:28,505 --> 03:32:31,415
In such networks, positive
correlations indicate

4350
03:32:31,415 --> 03:32:33,845
that both ROIs are either increase

4351
03:32:33,845 --> 03:32:37,485
or decreasing consistently
across subjects,

4352
03:32:37,485 --> 03:32:39,425
and a negative correlation indicates

4353
03:32:39,425 --> 03:32:41,545
that one ROI is increasing

4354
03:32:41,545 --> 03:32:44,818
and the other is decreasing
consistently across subjects.

4355
03:32:46,045 --> 03:32:48,695
The strength of the positive
or negative correlation

4356
03:32:48,695 --> 03:32:50,725
indicates the consistency of the pattern

4357
03:32:50,725 --> 03:32:53,225
across subjects in the samples.

4358
03:32:53,225 --> 03:32:56,135
This allows inferences to be
drawn about the homogeneity

4359
03:32:56,135 --> 03:32:57,645
of change between groups,

4360
03:32:57,645 --> 03:32:59,995
e.g., between younger and older adults,

4361
03:32:59,995 --> 03:33:04,606
and also homogeneity of
associations between regions.

4362
03:33:04,606 --> 03:33:08,685
ICA static PET data, as in
Savio and Riedl's paper,

4363
03:33:08,685 --> 03:33:10,635
offer a similar interpretation,

4364
03:33:10,635 --> 03:33:14,185
the covariance of metabolic
activity across subject.

4365
03:33:14,185 --> 03:33:18,035
Again, there's no time
dimension in that analysis.

4366
03:33:18,035 --> 03:33:21,095
Metabolic connectivity mapping
is a little bit different,

4367
03:33:21,095 --> 03:33:22,745
and it's one of the few approaches

4368
03:33:22,745 --> 03:33:26,785
for studying the relationship
between BOLD and FDG data.

4369
03:33:26,785 --> 03:33:28,715
It's valuable for
understanding the relationship

4370
03:33:28,715 --> 03:33:32,065
between uptake over a scan
period and its relationship

4371
03:33:32,065 --> 03:33:35,105
with dynamically changing
estimates of activity

4372
03:33:35,105 --> 03:33:37,735
derived from the hemodynamic response.

4373
03:33:37,735 --> 03:33:38,568
The nice thing about

4374
03:33:38,568 --> 03:33:42,155
the metabolic connectivity
mapping approach highlighted here

4375
03:33:42,155 --> 03:33:43,735
is that it uses an assumption

4376
03:33:43,735 --> 03:33:47,105
based on the known physiology
of neuronal activity

4377
03:33:47,105 --> 03:33:49,235
to derive an estimate of directionality

4378
03:33:49,235 --> 03:33:51,045
and functional connectivity.

4379
03:33:51,045 --> 03:33:52,705
And, at least in the context of the review

4380
03:33:52,705 --> 03:33:54,235
that I'm presenting today,

4381
03:33:54,235 --> 03:33:56,315
this is a complimentary approach

4382
03:33:56,315 --> 03:33:58,288
to metabolic connectivity analysis.

4383
03:33:59,705 --> 03:34:02,765
Lastly, we also saw that the
metabolic similarity mapping

4384
03:34:02,765 --> 03:34:04,745
is the closest thing that we have so far

4385
03:34:04,745 --> 03:34:06,075
that we've discussed

4386
03:34:06,075 --> 03:34:08,665
that is an individual-level connectome.

4387
03:34:08,665 --> 03:34:12,305
These are similar to fMRI-based measures

4388
03:34:12,305 --> 03:34:13,955
of functional connectivity

4389
03:34:13,955 --> 03:34:16,305
in that there is an
individual-level estimate

4390
03:34:16,305 --> 03:34:17,595
of connectivity.

4391
03:34:17,595 --> 03:34:19,665
They're also promising because
their network properties

4392
03:34:19,665 --> 03:34:23,175
are showing similarities
to functional connectivity.

4393
03:34:23,175 --> 03:34:24,055
However, they're different

4394
03:34:24,055 --> 03:34:25,895
in that they're a similarity metric

4395
03:34:25,895 --> 03:34:30,695
and not a temporal time series
or a temporal dependency,

4396
03:34:30,695 --> 03:34:33,658
as is the definition of
functional connectivity.

4397
03:34:35,935 --> 03:34:39,945
And, at this point, we're
just going to step aside

4398
03:34:39,945 --> 03:34:43,095
from the primary argument of this talk

4399
03:34:43,095 --> 03:34:47,865
and just remember that
fMRI is not ground truth.

4400
03:34:47,865 --> 03:34:50,135
And so this is something that we all know.

4401
03:34:50,135 --> 03:34:52,085
It's not controversial.

4402
03:34:52,085 --> 03:34:55,375
We all know that fMRI
is not the best method

4403
03:34:55,375 --> 03:34:57,498
of brain function that we have.

4404
03:34:58,335 --> 03:35:00,405
But knowing that,

4405
03:35:00,405 --> 03:35:02,975
in the development of
metabolic connectivity,

4406
03:35:02,975 --> 03:35:05,155
it's very, very easy to fall into the trap

4407
03:35:05,155 --> 03:35:08,865
of using what is known from
fMRI and fMRI connectivity

4408
03:35:08,865 --> 03:35:10,655
as ground truth.

4409
03:35:10,655 --> 03:35:14,785
It's tempting because we just
know so much more about it.

4410
03:35:14,785 --> 03:35:17,485
It does remain possible
that metabolic connectivity

4411
03:35:17,485 --> 03:35:19,785
shows a different pattern in aging

4412
03:35:19,785 --> 03:35:23,785
to functional connectivity,
as we saw in Arnemann et al.,

4413
03:35:23,785 --> 03:35:28,665
and the fact that Mertens'
metabolic similarity approach

4414
03:35:28,665 --> 03:35:31,255
yielded information consistent with fMRI

4415
03:35:31,255 --> 03:35:33,325
is indicative that we're
on the right track,

4416
03:35:33,325 --> 03:35:34,285
but it's only that.

4417
03:35:34,285 --> 03:35:35,955
It's only indicative

4418
03:35:35,955 --> 03:35:38,348
because we don't know
what the ground truth is.

4419
03:35:39,705 --> 03:35:43,485
Also, it can be really tempting
when doing these analyses

4420
03:35:43,485 --> 03:35:47,385
to constrain the PET analysis
by the fMRI in some way.

4421
03:35:47,385 --> 03:35:51,195
So either to use fMRI
patterns of activity as a mask

4422
03:35:51,195 --> 03:35:55,715
or just using an fMRI-based
functional parcellation.

4423
03:35:55,715 --> 03:35:57,055
So I wanna be clear

4424
03:35:57,055 --> 03:35:58,735
that there's nothing
wrong with this per se.

4425
03:35:58,735 --> 03:36:00,065
It's just important to remember

4426
03:36:00,065 --> 03:36:02,245
how this influences the inferences

4427
03:36:02,245 --> 03:36:04,905
that we can draw from the analysis.

4428
03:36:04,905 --> 03:36:07,085
We've got a fairly good
reason for expecting the fMRI

4429
03:36:07,085 --> 03:36:10,435
and fPET to be highly correlated,

4430
03:36:10,435 --> 03:36:12,385
but, at the same time, we also expect

4431
03:36:12,385 --> 03:36:13,815
that they will be different.

4432
03:36:13,815 --> 03:36:15,655
So, first off, because
they're just indexing

4433
03:36:15,655 --> 03:36:18,405
different physiological
properties of the brain.

4434
03:36:18,405 --> 03:36:23,405
And, secondly, we're not trying
to develop a new fMRI here.

4435
03:36:23,825 --> 03:36:25,095
We have fMRI.

4436
03:36:25,095 --> 03:36:26,985
It's relatively straightforward, used,

4437
03:36:26,985 --> 03:36:28,715
and lots of people like it.

4438
03:36:28,715 --> 03:36:31,425
What we're interested
in is the new knowledge

4439
03:36:31,425 --> 03:36:34,065
that can be gained from the FDG-PET,

4440
03:36:34,065 --> 03:36:38,135
in this case, from connectomic
analysis of glucose uptake.

4441
03:36:38,135 --> 03:36:40,055
So it's for this reason

4442
03:36:40,055 --> 03:36:43,215
that I try not to
constrain my fPET analyses

4443
03:36:43,215 --> 03:36:46,335
either by simultaneously
acquired fMRI data

4444
03:36:46,335 --> 03:36:50,498
or by atlases derived from fMRI data.

4445
03:36:52,255 --> 03:36:55,655
Okay, so, after that aside,
let's get back to the data.

4446
03:36:55,655 --> 03:36:59,165
So we've said that, while the
static PET covariance approach

4447
03:36:59,165 --> 03:37:00,955
gives us some interesting information

4448
03:37:00,955 --> 03:37:05,385
about the covariance of
metabolism across individuals,

4449
03:37:05,385 --> 03:37:08,175
it isn't really the same
as functional connectivity,

4450
03:37:08,175 --> 03:37:11,125
which examines relationships
between time series

4451
03:37:11,125 --> 03:37:13,818
across brain regions within individuals.

4452
03:37:15,325 --> 03:37:18,085
This is where the fPET
approach comes into handy.

4453
03:37:18,085 --> 03:37:20,915
fPET provides a time series of FDG uptake

4454
03:37:20,915 --> 03:37:22,725
over the course of the scan.

4455
03:37:22,725 --> 03:37:24,535
So this means that we can analyze it

4456
03:37:24,535 --> 03:37:26,228
in a similar way to fMRI.

4457
03:37:29,035 --> 03:37:30,645
In our first exploration

4458
03:37:30,645 --> 03:37:33,055
of within-subject metabolic connectivity,

4459
03:37:33,055 --> 03:37:36,285
we ran an ICA on resting state fPET.

4460
03:37:36,285 --> 03:37:39,455
And the goal of that paper
wasn't to compare the results

4461
03:37:39,455 --> 03:37:41,420
to sPET or to fMRI,

4462
03:37:41,420 --> 03:37:44,135
so we just have the fPET components here.

4463
03:37:44,135 --> 03:37:46,735
I'm showing the sPET results
from Savio and colleagues

4464
03:37:46,735 --> 03:37:49,075
on the side here for reference.

4465
03:37:49,075 --> 03:37:53,855
So, in this analysis, just
like in the previous in sPET,

4466
03:37:53,855 --> 03:37:58,485
there were fPET networks
that resembled fMRI networks,

4467
03:37:58,485 --> 03:38:01,925
including default mode network,

4468
03:38:01,925 --> 03:38:05,755
primary and secondary visual
networks, and the cerebellum.

4469
03:38:05,755 --> 03:38:07,005
However, there were other components

4470
03:38:07,005 --> 03:38:10,105
that were not easily mapped
onto fMRI components,

4471
03:38:10,105 --> 03:38:13,355
including a very frontal
component and a medial component

4472
03:38:13,355 --> 03:38:17,618
that comes up quite frequently
in sPET decompositions.

4473
03:38:18,795 --> 03:38:22,185
So, in summary, fPET appears
to yield complimentary

4474
03:38:22,185 --> 03:38:26,095
and unique components during
resting state compared to fMRI.

4475
03:38:26,095 --> 03:38:29,275
And while the components in Li
et al. are analogous to fMRI

4476
03:38:29,275 --> 03:38:31,265
in that they're estimated across subjects

4477
03:38:31,265 --> 03:38:33,355
with temporal dimension,

4478
03:38:33,355 --> 03:38:36,095
they are indicating,
at least qualitatively,

4479
03:38:36,095 --> 03:38:39,038
a similar story to the
components estimated from sPET.

4480
03:38:42,565 --> 03:38:45,895
So let's have a deeper dive
into fPET connectivity.

4481
03:38:45,895 --> 03:38:48,045
This time, if we analyze the fPET

4482
03:38:48,045 --> 03:38:50,635
and the fMRI in the same way.

4483
03:38:50,635 --> 03:38:51,468
In this study,

4484
03:38:51,468 --> 03:38:54,335
we used a constant infusion
administration technique.

4485
03:38:54,335 --> 03:38:56,055
We didn't have a bolus here.

4486
03:38:56,055 --> 03:38:58,545
And then, we parcellated
the brain into 82 ROIs.

4487
03:38:59,655 --> 03:39:02,205
We estimated time courses
for the fMRI and the fPET

4488
03:39:03,205 --> 03:39:06,485
with a 2.45 second TR for MRI

4489
03:39:06,485 --> 03:39:09,695
and a 16-second frame duration for fPET.

4490
03:39:09,695 --> 03:39:12,705
We then correlated the time
courses between regions

4491
03:39:12,705 --> 03:39:15,005
and constructed a connectome matrix

4492
03:39:15,005 --> 03:39:17,305
for each modality separately.

4493
03:39:17,305 --> 03:39:21,255
And then, just to compare
the fPET metabolic connectome

4494
03:39:21,255 --> 03:39:23,445
to the sPET covariance matrix,

4495
03:39:23,445 --> 03:39:24,758
we also calculated that.

4496
03:39:27,205 --> 03:39:28,665
And this is what we found.

4497
03:39:28,665 --> 03:39:32,405
So static PET showed strong
temporo subcortical covariance

4498
03:39:32,405 --> 03:39:34,575
within and between hemispheres

4499
03:39:34,575 --> 03:39:36,695
and strong fronto subcortical,

4500
03:39:36,695 --> 03:39:40,485
frontotemporal, and parieto
occipital covariance,

4501
03:39:40,485 --> 03:39:41,925
albeit with a smaller magnitude

4502
03:39:41,925 --> 03:39:43,875
than the temporal occipital covariance.

4503
03:39:45,339 --> 03:39:47,365
fPET connectivity was dominated

4504
03:39:47,365 --> 03:39:49,355
by frontoparietal connections,

4505
03:39:49,355 --> 03:39:51,725
both within and between hemispheres,

4506
03:39:51,725 --> 03:39:54,305
and the low signal that's inherent

4507
03:39:54,305 --> 03:39:57,618
in the constant infusion fPET
here is really noticeable.

4508
03:39:58,685 --> 03:40:00,545
fMRI showed strong connectivity

4509
03:40:00,545 --> 03:40:02,685
within anatomical subdivisions

4510
03:40:02,685 --> 03:40:04,785
as well as a number of
long-range connections

4511
03:40:04,785 --> 03:40:08,115
between the frontoparietal,
parietal occipital,

4512
03:40:08,115 --> 03:40:09,635
and temporoparietal regions.

4513
03:40:12,555 --> 03:40:14,898
We then sorted the networks into regions.

4514
03:40:16,335 --> 03:40:19,515
Sorry, we then sorted the
regions into networks.

4515
03:40:19,515 --> 03:40:21,635
And I'll note that these are networks

4516
03:40:21,635 --> 03:40:24,985
that have been previously
identified using fMRI.

4517
03:40:24,985 --> 03:40:27,275
So, in this way, the result
is a bit more constrained

4518
03:40:27,275 --> 03:40:29,335
by fMRI than the previous result

4519
03:40:29,335 --> 03:40:32,518
because we are using
that fMRI-derived atlas.

4520
03:40:33,705 --> 03:40:35,195
Static PET network graphs

4521
03:40:35,195 --> 03:40:37,305
showed strong widespread connectivity

4522
03:40:37,305 --> 03:40:38,795
between the dorsal attention,

4523
03:40:38,795 --> 03:40:41,645
frontoparietal subcortical sensory motor,

4524
03:40:41,645 --> 03:40:44,835
and default mode network.

4525
03:40:44,835 --> 03:40:47,965
fPET network graphs showed
strongest connectivity

4526
03:40:47,965 --> 03:40:51,165
between frontoparietal and
dorsal attention networks

4527
03:40:51,165 --> 03:40:52,865
and between default mode network,

4528
03:40:52,865 --> 03:40:54,495
dorsal attention, frontoparietal,

4529
03:40:54,495 --> 03:40:56,823
and visual attention networks.

4530
03:40:56,823 --> 03:40:58,975
fMRI graphs showed strongest connectivity

4531
03:40:58,975 --> 03:41:01,725
between the dorsal attention
and visual networks.

4532
03:41:01,725 --> 03:41:02,955
The ventral attention network

4533
03:41:02,955 --> 03:41:05,605
showed high interconnectedness
with frontoparietal

4534
03:41:05,605 --> 03:41:07,485
and sensory motor network,

4535
03:41:07,485 --> 03:41:10,235
and the default mode network

4536
03:41:10,235 --> 03:41:12,865
showed intermediate connectedness

4537
03:41:12,865 --> 03:41:15,668
with frontoparietal and
dorsal attention networks.

4538
03:41:17,795 --> 03:41:21,069
We looked at the relationship
between fPET and fMRI

4539
03:41:21,069 --> 03:41:21,902
and sPET and fPET.

4540
03:41:22,795 --> 03:41:25,395
And, in the paper, we looked at residuals,

4541
03:41:25,395 --> 03:41:29,035
but I quite like this early
analysis that we did too.

4542
03:41:29,035 --> 03:41:32,355
So that you can see that across
the brain, across regions,

4543
03:41:32,355 --> 03:41:35,045
fPET and fMRI showed many regions

4544
03:41:35,045 --> 03:41:38,695
where the correlation between
modalities was non zero.

4545
03:41:38,695 --> 03:41:41,665
The strongest fPET-fMRI
correlation was obtained

4546
03:41:41,665 --> 03:41:43,935
for superior cortex,

4547
03:41:43,935 --> 03:41:47,155
so superior frontal,
superior parietal regions,

4548
03:41:47,155 --> 03:41:51,565
intermediate in temporal
cortex, and lower subcortically.

4549
03:41:51,565 --> 03:41:54,065
However, for the sPET
and fPET correlation,

4550
03:41:54,065 --> 03:41:56,785
there were no regions above zero,

4551
03:41:56,785 --> 03:42:00,215
even before correcting
for multiple comparisons.

4552
03:42:00,215 --> 03:42:02,635
This is consistent with
our residual analysis,

4553
03:42:02,635 --> 03:42:04,995
suggesting that the metabolic covariance

4554
03:42:04,995 --> 03:42:07,838
was a poor predictor of
metabolic connectivity.

4555
03:42:09,765 --> 03:42:12,725
So fPET and fMRI matrices
showed a greater level

4556
03:42:12,725 --> 03:42:15,285
of shared variance than fPET and sPET,

4557
03:42:15,285 --> 03:42:18,455
and supports a contention that
fPET metabolic connectivity

4558
03:42:18,455 --> 03:42:21,175
is complimentary to fMRI.

4559
03:42:21,175 --> 03:42:23,915
It's important to note
that direct comparison

4560
03:42:23,915 --> 03:42:26,365
between the methods was challenging

4561
03:42:26,365 --> 03:42:28,955
because there are such
substantial SNR differences

4562
03:42:28,955 --> 03:42:30,505
between the methods.

4563
03:42:30,505 --> 03:42:33,045
It's also challenging
because we do want to avoid

4564
03:42:33,045 --> 03:42:36,305
using fMRI as the ground truth here.

4565
03:42:36,305 --> 03:42:39,705
Each method should provide a
unique, but likely correlated,

4566
03:42:39,705 --> 03:42:41,625
index of neural activity.

4567
03:42:41,625 --> 03:42:44,658
So work in this area is still continuing.

4568
03:42:46,765 --> 03:42:48,555
And while we expected at some level

4569
03:42:48,555 --> 03:42:49,995
that the static PET covariance

4570
03:42:49,995 --> 03:42:52,495
might be a less than
satisfactory reflection

4571
03:42:52,495 --> 03:42:55,265
of within-subject PET connectivity,

4572
03:42:55,265 --> 03:42:57,725
we really didn't expect to
see such a stark difference

4573
03:42:57,725 --> 03:42:59,055
between the metrics,

4574
03:42:59,055 --> 03:43:01,435
nor did we expect fPET connectivity

4575
03:43:01,435 --> 03:43:05,118
to be more similar to fMRI
than to sPET covariance.

4576
03:43:06,885 --> 03:43:09,605
This led us down the path
of exploring the concept

4577
03:43:09,605 --> 03:43:12,355
of ergodicity in neuroimaging.

4578
03:43:12,355 --> 03:43:15,225
An ergodic process occurs
when a group-level result

4579
03:43:15,225 --> 03:43:19,795
is generalizable to the
individuals within the sample.

4580
03:43:19,795 --> 03:43:23,235
Simpson's paradox is a special
case of non-ergodicity,

4581
03:43:23,235 --> 03:43:25,625
where the within-individual correlation

4582
03:43:25,625 --> 03:43:27,055
is in the opposite direction

4583
03:43:27,055 --> 03:43:29,555
to the group-level correlation.

4584
03:43:29,555 --> 03:43:32,685
In reality, ergodic
processes are quite rare

4585
03:43:32,685 --> 03:43:35,525
because two strict criteria must be met.

4586
03:43:35,525 --> 03:43:37,875
Firstly, the process has to be homogenous

4587
03:43:37,875 --> 03:43:40,375
across individuals within a sample,

4588
03:43:40,375 --> 03:43:42,405
and, secondly, the statistical parameters

4589
03:43:42,405 --> 03:43:45,175
that describe the process must be constant

4590
03:43:45,175 --> 03:43:47,258
or stationary over time.

4591
03:43:50,325 --> 03:43:52,235
In a recently published commentary

4592
03:43:52,235 --> 03:43:55,495
on our static versus functional
PET connectivity paper,

4593
03:43:55,495 --> 03:43:57,805
Sala and colleagues raised
some important points

4594
03:43:57,805 --> 03:44:01,295
about the concept of ergodicity
for metabolic connectivity.

4595
03:44:01,295 --> 03:44:03,535
They argued that ergodicity
should be considered

4596
03:44:03,535 --> 03:44:06,055
on a continuum rather
than being all or none,

4597
03:44:06,055 --> 03:44:07,268
which we do agree with.

4598
03:44:08,580 --> 03:44:10,725
They also noted that many parameters

4599
03:44:10,725 --> 03:44:13,905
may have influenced the
ergodicity in our experiment,

4600
03:44:13,905 --> 03:44:16,825
and the ones they noted
included demographic factors,

4601
03:44:16,825 --> 03:44:20,375
pre-processing choices,
scanner resolution, and so on.

4602
03:44:20,375 --> 03:44:21,815
Again, we do agree with this,

4603
03:44:21,815 --> 03:44:24,935
but we also feel that the
argument is demonstrating

4604
03:44:24,935 --> 03:44:28,615
that the parameters for achieving
ergodicity are so strict

4605
03:44:28,615 --> 03:44:30,685
that it's questionable
whether true ergodicity

4606
03:44:30,685 --> 03:44:34,035
could ever be obtained
using neuroimaging data.

4607
03:44:34,035 --> 03:44:35,335
Therefore, we argue that,

4608
03:44:35,335 --> 03:44:38,575
given that ergodicity is
so difficult to obtain,

4609
03:44:38,575 --> 03:44:39,955
we should, at the very least,

4610
03:44:39,955 --> 03:44:42,872
avoid to use metrics
like metabolic covariance

4611
03:44:44,035 --> 03:44:46,765
if we have another metric
that can be estimable

4612
03:44:46,765 --> 03:44:48,418
at the individual level.

4613
03:44:51,275 --> 03:44:53,575
In the last experiment
that I'd like to discuss,

4614
03:44:53,575 --> 03:44:56,385
we've recently explored
how metabolic connectivity

4615
03:44:56,385 --> 03:44:58,135
relates to cognition.

4616
03:44:58,135 --> 03:45:01,925
And this work was led by my
new postdoc Katharina Voigt

4617
03:45:01,925 --> 03:45:06,755
and used the data that I
just presented earlier.

4618
03:45:06,755 --> 03:45:10,175
We explored the question of
whether neural connectivity

4619
03:45:10,175 --> 03:45:12,975
maps to cognition in a domain-general way

4620
03:45:12,975 --> 03:45:15,635
or in a domain-specific way.

4621
03:45:15,635 --> 03:45:19,245
In a PLS analysis of the
functional metabolic connectomes,

4622
03:45:19,245 --> 03:45:21,105
we found that both modalities

4623
03:45:21,105 --> 03:45:25,015
yielded a single cognition
connectome latent pair.

4624
03:45:25,015 --> 03:45:28,155
The functional connectome
loaded strongest on depression,

4625
03:45:28,155 --> 03:45:30,735
inhibitory control, and memory retention

4626
03:45:30,735 --> 03:45:31,885
with the strongest loadings

4627
03:45:31,885 --> 03:45:34,315
within the frontoparietal cortex.

4628
03:45:34,315 --> 03:45:35,765
And here, you can see that we found

4629
03:45:35,765 --> 03:45:38,055
that both long-range and
short-range connections

4630
03:45:38,055 --> 03:45:41,578
were significantly
related to the cognition.

4631
03:45:42,535 --> 03:45:45,095
In fPET, we found something similar.

4632
03:45:45,095 --> 03:45:48,155
Metabolic connectivity in
the frontoparietal cortex

4633
03:45:48,155 --> 03:45:50,725
loaded strongly on measures
of executive function,

4634
03:45:50,725 --> 03:45:53,895
including inhibitory
control and verbal fluency.

4635
03:45:53,895 --> 03:45:56,615
While the metabolic connectome
in the 99th percentile

4636
03:45:56,615 --> 03:45:59,215
was faster than the fMRI connectome,

4637
03:45:59,215 --> 03:46:01,955
you can see that both long
and short-range connections

4638
03:46:01,955 --> 03:46:04,675
were related to cognition.

4639
03:46:04,675 --> 03:46:06,395
So we concluded that both metabolic

4640
03:46:06,395 --> 03:46:09,315
and functional connectivity
is supporting the global view

4641
03:46:09,315 --> 03:46:11,245
of cognition connectome mapping

4642
03:46:11,245 --> 03:46:15,745
because we did just see that
one cognition connectome pair

4643
03:46:15,745 --> 03:46:17,665
for each modality.

4644
03:46:17,665 --> 03:46:21,775
We also concluded that metabolic
and functional connectivity

4645
03:46:21,775 --> 03:46:23,785
are explaining unique variance

4646
03:46:23,785 --> 03:46:26,585
in the cognition-connectome relationship.

4647
03:46:26,585 --> 03:46:28,615
So, in other words, the two modalities

4648
03:46:28,615 --> 03:46:31,105
are showing complimentary
and unique information

4649
03:46:31,105 --> 03:46:33,348
about connectivity in the brain.

4650
03:46:35,435 --> 03:46:37,348
So, where to from here?

4651
03:46:39,395 --> 03:46:41,865
First off, I'd like to
return to the concept of time

4652
03:46:41,865 --> 03:46:43,265
for PET/MR.

4653
03:46:43,265 --> 03:46:44,905
The terms dynamic and static

4654
03:46:44,905 --> 03:46:46,565
are so fraught with this methodology

4655
03:46:46,565 --> 03:46:48,735
that I'm sure that I've
offended half of the audience

4656
03:46:48,735 --> 03:46:50,415
with my use of them.

4657
03:46:50,415 --> 03:46:53,745
As I mentioned earlier, when
I use the term static PET,

4658
03:46:53,745 --> 03:46:55,995
the static is really just
referring to the fact

4659
03:46:55,995 --> 03:46:58,405
that there is an assumption
that glucose uptake

4660
03:46:58,405 --> 03:47:01,605
is stationary across
the acquisition period.

4661
03:47:01,605 --> 03:47:04,525
The image may have been
or probably was acquired

4662
03:47:04,525 --> 03:47:06,225
with dynamic acquisition in order

4663
03:47:06,225 --> 03:47:09,228
to determine the cerebral
metabolic rate of glucose.

4664
03:47:10,443 --> 03:47:12,505
While fPET is useful
in that it's providing

4665
03:47:12,505 --> 03:47:14,935
a within-subject time course of uptake,

4666
03:47:14,935 --> 03:47:18,665
and we're using that time series
to calculate a connectome,

4667
03:47:18,665 --> 03:47:22,125
the analysis itself, the ones
that I've presented today,

4668
03:47:22,125 --> 03:47:24,895
have assumed stationarity as well.

4669
03:47:24,895 --> 03:47:27,955
So, if we compare this to a
dynamic connectivity measure,

4670
03:47:27,955 --> 03:47:30,442
as used in fMRI and EEG,

4671
03:47:30,442 --> 03:47:32,715
our analysis would be called static

4672
03:47:32,715 --> 03:47:35,195
by these people in that field.

4673
03:47:35,195 --> 03:47:38,295
So, in order to move towards what an fMRI

4674
03:47:38,295 --> 03:47:40,755
or a different type of neuroscientist

4675
03:47:40,755 --> 03:47:42,735
would call a dynamic analysis,

4676
03:47:42,735 --> 03:47:43,995
we'd have to employ something

4677
03:47:43,995 --> 03:47:46,538
like a sliding window
approach to our data.

4678
03:47:47,395 --> 03:47:49,485
As an aside, we actually
did try to do this

4679
03:47:49,485 --> 03:47:52,805
with our early checkerboard
data, but we did have no luck.

4680
03:47:52,805 --> 03:47:54,475
But we also had a very low dose

4681
03:47:54,475 --> 03:47:56,015
and a frame duration of one minute.

4682
03:47:56,015 --> 03:47:58,925
So, possibly, this might be possible

4683
03:47:58,925 --> 03:48:02,898
with the bolus infusion
and the faster frame rates.

4684
03:48:04,615 --> 03:48:08,885
Secondly, the signal-to-noise
ratio of fPET is low.

4685
03:48:08,885 --> 03:48:10,385
It's really low.

4686
03:48:10,385 --> 03:48:12,855
And, at least in our
cerebral cortex paper,

4687
03:48:12,855 --> 03:48:14,665
it was difficult to compare the fPET

4688
03:48:14,665 --> 03:48:16,485
to the fMRI connectomes

4689
03:48:16,485 --> 03:48:19,345
because the signal was just
so much smaller in fPET

4690
03:48:19,345 --> 03:48:21,305
compared to fMRI.

4691
03:48:21,305 --> 03:48:23,305
We're hoping that the
bolus infusion approach

4692
03:48:23,305 --> 03:48:24,875
will help here.

4693
03:48:24,875 --> 03:48:26,915
Unfortunately, we've been delayed by COVID

4694
03:48:26,915 --> 03:48:29,295
in acquiring this data,

4695
03:48:29,295 --> 03:48:33,715
but possibly other approaches,
like signal optimization,

4696
03:48:33,715 --> 03:48:36,305
like the MR-informed PET reconstruction,

4697
03:48:36,305 --> 03:48:39,635
and deep neural networks, might help here.

4698
03:48:39,635 --> 03:48:42,425
Thirdly, something that is
a little bit of a challenge

4699
03:48:42,425 --> 03:48:45,015
is that standardized imaging processes

4700
03:48:45,015 --> 03:48:47,265
don't yet exist for fPET.

4701
03:48:47,265 --> 03:48:49,485
So fMRI and functional connectivity

4702
03:48:49,485 --> 03:48:52,415
benefit from very large research bodies

4703
03:48:52,415 --> 03:48:55,315
that give us an idea how
best to analyze the data,

4704
03:48:55,315 --> 03:48:57,895
but fPET is still in
its infancy in this area

4705
03:48:57,895 --> 03:48:59,608
and much work is required.

4706
03:49:01,825 --> 03:49:03,145
Another thing I want to explore

4707
03:49:03,145 --> 03:49:04,765
is whether or not we're getting the most

4708
03:49:04,765 --> 03:49:07,635
out of our simultaneous acquisition.

4709
03:49:07,635 --> 03:49:09,145
So, if we consider the benefits

4710
03:49:09,145 --> 03:49:11,925
of a simultaneous multimodal acquisition,

4711
03:49:11,925 --> 03:49:14,195
there are three levels of benefit.

4712
03:49:14,195 --> 03:49:15,875
First off, we get a simple benefit

4713
03:49:15,875 --> 03:49:18,085
of spatial registration.

4714
03:49:18,085 --> 03:49:20,865
So PET and MRI being
registered on acquisition

4715
03:49:20,865 --> 03:49:22,258
is a very useful thing.

4716
03:49:23,185 --> 03:49:26,435
At the second level, we have
an asymmetric integration.

4717
03:49:26,435 --> 03:49:29,305
Here, we might use one
method to inform the other.

4718
03:49:29,305 --> 03:49:33,245
So we might use the spatial
resolution of MRI to inform PET,

4719
03:49:33,245 --> 03:49:35,765
or the more direct and
quantitative nature of PET

4720
03:49:35,765 --> 03:49:36,935
to inform fMRI.

4721
03:49:38,145 --> 03:49:40,065
Similarly, we might also consider

4722
03:49:40,065 --> 03:49:42,475
using the information from fMRI,

4723
03:49:42,475 --> 03:49:46,435
e.g., the fMRI-derived functional atlases,

4724
03:49:46,435 --> 03:49:48,505
or fMRI activity masks

4725
03:49:48,505 --> 03:49:51,695
to inform or constrain the PET analysis.

4726
03:49:51,695 --> 03:49:54,925
The third level of simultaneous
multimodal acquisition

4727
03:49:54,925 --> 03:49:59,555
is starting to get to a truly
synergistic integration,

4728
03:49:59,555 --> 03:50:00,765
where we're learning much more

4729
03:50:00,765 --> 03:50:02,555
about the physiology of the brain

4730
03:50:02,555 --> 03:50:05,505
by capitalizing on the
benefits of each method.

4731
03:50:05,505 --> 03:50:07,945
This might be a technical integration

4732
03:50:07,945 --> 03:50:10,188
or a physiological interaction.

4733
03:50:12,665 --> 03:50:14,395
I think, at the moment, at best,

4734
03:50:14,395 --> 03:50:16,295
we're currently at level two

4735
03:50:16,295 --> 03:50:19,435
in terms of the data that
I've discussed today,

4736
03:50:19,435 --> 03:50:20,975
and it'll be great to see how we can get

4737
03:50:20,975 --> 03:50:23,658
to that third level of
benefit in the near future.

4738
03:50:25,835 --> 03:50:29,515
I'm also super excited by the
possibility of using the fPET

4739
03:50:29,515 --> 03:50:32,905
to measure the dynamics of
other physiological targets.

4740
03:50:32,905 --> 03:50:35,615
I think the ability to track
neurotransmitter dynamics

4741
03:50:35,615 --> 03:50:38,955
in response to tasks is such
an exciting future direction

4742
03:50:38,955 --> 03:50:40,305
with this method.

4743
03:50:40,305 --> 03:50:42,375
And if we take the models
that we're used to seeing

4744
03:50:42,375 --> 03:50:45,285
in cognitive psychology,
it's really common to see

4745
03:50:45,285 --> 03:50:46,985
that these models do make predictions

4746
03:50:46,985 --> 03:50:49,955
about the role of different
neurotransmitter systems

4747
03:50:49,955 --> 03:50:53,045
in behavior or in psychiatric illness.

4748
03:50:53,045 --> 03:50:56,155
These models are often
based off indirect evidence,

4749
03:50:56,155 --> 03:50:58,865
including known disruptions
of neurotransmitter systems

4750
03:50:58,865 --> 03:51:00,475
in certain disorders,

4751
03:51:00,475 --> 03:51:03,965
or it might be on the basis of
preclinical challenge models.

4752
03:51:03,965 --> 03:51:06,145
I've just screenshotted
two examples of this

4753
03:51:06,145 --> 03:51:07,455
that I just had in my files

4754
03:51:07,455 --> 03:51:09,555
from back when I was doing my thesis,

4755
03:51:09,555 --> 03:51:12,775
but this is a very common
approach in cognitive psychology.

4756
03:51:12,775 --> 03:51:15,595
I think the fPET approach
offers the ability

4757
03:51:15,595 --> 03:51:18,875
to directly measure some
of these dynamics in vivo

4758
03:51:18,875 --> 03:51:21,345
in healthy humans, and
I think that's so cool,

4759
03:51:21,345 --> 03:51:22,485
and I can't wait to see

4760
03:51:22,485 --> 03:51:25,018
where we go to with this in the future.

4761
03:51:28,065 --> 03:51:29,745
Lastly, I just wanna highlight

4762
03:51:29,745 --> 03:51:32,095
that the international
community is finding value

4763
03:51:32,095 --> 03:51:35,015
in early PET/MR data releases.

4764
03:51:35,015 --> 03:51:38,725
So my group have made three
such datasets available.

4765
03:51:38,725 --> 03:51:41,225
And here, you can see the
references to them here.

4766
03:51:42,495 --> 03:51:45,149
One nice thing about the visfPET

4767
03:51:45,149 --> 03:51:48,765
and the DaCRA-fPET datasets

4768
03:51:48,765 --> 03:51:52,285
is that we've made our
list mode data available.

4769
03:51:52,285 --> 03:51:54,455
The technical validation for each datasets

4770
03:51:54,455 --> 03:51:59,255
is published in either
Scientific Data or GigaScience.

4771
03:51:59,255 --> 03:52:02,435
And those papers demonstrate
how the list mode data

4772
03:52:02,435 --> 03:52:06,168
can be reconstructed using
open source algorithms.

4773
03:52:07,765 --> 03:52:09,315
I just wanna finish off by saying that,

4774
03:52:09,315 --> 03:52:11,615
while this is a little bit
of shameless publicity,

4775
03:52:11,615 --> 03:52:13,295
I do urge you in considering

4776
03:52:13,295 --> 03:52:15,845
making your data publicly available.

4777
03:52:15,845 --> 03:52:18,315
The international neuroimaging
community has shown

4778
03:52:18,315 --> 03:52:21,995
that great gains in knowledge
can be made by making unique

4779
03:52:21,995 --> 03:52:26,645
and difficult to acquire
data publicly available.

4780
03:52:26,645 --> 03:52:29,055
When we started this, we
were a bit in the dark

4781
03:52:29,055 --> 03:52:31,845
because PET BIDS was
still in its draft form,

4782
03:52:31,845 --> 03:52:33,835
but we now do have the standard.

4783
03:52:33,835 --> 03:52:36,095
And plus, we've developed a sharing format

4784
03:52:36,095 --> 03:52:38,505
for list mode data that
allows it to be used

4785
03:52:38,505 --> 03:52:41,825
without access to the
proprietary software.

4786
03:52:41,825 --> 03:52:44,015
And I, at least, feel really lucky

4787
03:52:44,015 --> 03:52:45,955
to be able to use this
cool new technology,

4788
03:52:45,955 --> 03:52:47,135
and so I think it's valuable

4789
03:52:47,135 --> 03:52:48,948
to share it with others as well.

4790
03:52:51,045 --> 03:52:52,835
We've already started
to see some cool papers

4791
03:52:52,835 --> 03:52:55,125
coming out on the basis of our data,

4792
03:52:55,125 --> 03:52:57,405
and I'll very briefly discuss them here.

4793
03:52:57,405 --> 03:53:00,305
In the first, Guo and
colleagues capitalized

4794
03:53:00,305 --> 03:53:02,035
on the simultaneous acquisition

4795
03:53:02,035 --> 03:53:04,735
to show that functional
connectivity within white matter

4796
03:53:04,735 --> 03:53:09,248
is related to glucose uptake
and so is unlikely to be noise.

4797
03:53:10,505 --> 03:53:13,175
And, in this pre-print,
Wang and colleagues

4798
03:53:13,175 --> 03:53:16,565
compared effective connectivity
approaches of fPET data,

4799
03:53:16,565 --> 03:53:19,535
the metabolic connectivity
mapping approach,

4800
03:53:19,535 --> 03:53:22,645
which does not require
a temporal dimension

4801
03:53:22,645 --> 03:53:24,555
for the PET data,

4802
03:53:24,555 --> 03:53:27,085
and Granger causality, which does.

4803
03:53:27,085 --> 03:53:29,725
They found that the two approaches
to effective connectivity

4804
03:53:29,725 --> 03:53:31,995
differed in their directed connections,

4805
03:53:31,995 --> 03:53:34,785
and concluded that each method
showed its own strengths

4806
03:53:34,785 --> 03:53:37,645
in estimating effective connectivity.

4807
03:53:37,645 --> 03:53:39,925
I look forward to see what
other new discoveries are made

4808
03:53:39,925 --> 03:53:42,668
on the basis of this open PET/MR data.

4809
03:53:44,795 --> 03:53:47,095
Okay, so thank you,
everybody, for your time.

4810
03:53:47,095 --> 03:53:49,225
I just want to acknowledge
the significant input

4811
03:53:49,225 --> 03:53:53,335
of all these people that
have made to the work

4812
03:53:53,335 --> 03:53:55,115
that I presented today.

4813
03:53:55,115 --> 03:53:57,595
I'm lucky to work with a
great team of neuroscientists,

4814
03:53:57,595 --> 03:53:59,855
biomedical engineers,
clinical radiographers,

4815
03:53:59,855 --> 03:54:01,885
and nuclear medicine technologists,

4816
03:54:01,885 --> 03:54:04,365
as well as informatics officers.

4817
03:54:04,365 --> 03:54:07,085
This work was supported by the
Australian Research Council

4818
03:54:07,085 --> 03:54:08,245
and a National Health

4819
03:54:08,245 --> 03:54:11,385
and Medical Research
Council of Australia grant.

4820
03:54:11,385 --> 03:54:13,945
The work was conducted at
Monash Biomedical Imaging

4821
03:54:13,945 --> 03:54:17,345
on the lands of the
indigenous Kulin Nations,

4822
03:54:17,345 --> 03:54:20,795
and I acknowledge the Elders
past, present, and emerging.

4823
03:54:20,795 --> 03:54:21,628
Thank you.

4824
03:54:26,768 --> 03:54:27,968
- My name is Christin Sander,

4825
03:54:27,968 --> 03:54:29,398
and I'm delighted to be a part

4826
03:54:29,398 --> 03:54:31,621
of this NIMH PET/MR Symposium.

4827
03:54:32,498 --> 03:54:33,645
The title of my talk is

4828
03:54:33,645 --> 03:54:36,708
"From Neuroreceptor
Binding to Brain Function

4829
03:54:36,708 --> 03:54:38,495
with PET and Functional MRI."

4830
03:54:41,298 --> 03:54:44,888
Networks in the brain
exist at various scales.

4831
03:54:44,888 --> 03:54:48,298
Functional networks span
the whole-brain level,

4832
03:54:48,298 --> 03:54:50,958
and functional MRI is used to investigate

4833
03:54:50,958 --> 03:54:55,171
related underlying neuronal
networks across the whole-brain.

4834
03:54:56,778 --> 03:55:00,348
Now, PET can be used to inform
us about brain chemistry

4835
03:55:00,348 --> 03:55:02,978
at the synaptic or protein level.

4836
03:55:02,978 --> 03:55:06,528
And in today's talk, I will
focus on how interventions

4837
03:55:06,528 --> 03:55:09,478
such as drug treatment, stimulation,

4838
03:55:09,478 --> 03:55:11,268
or behavioral modulation

4839
03:55:11,268 --> 03:55:15,178
may alter or change
these kinds of networks.

4840
03:55:15,178 --> 03:55:16,768
And I will focus specifically

4841
03:55:16,768 --> 03:55:20,368
on how we can use combined PET and fMRI

4842
03:55:20,368 --> 03:55:24,288
to understand relationships
across networks

4843
03:55:24,288 --> 03:55:26,481
and across scales of these networks.

4844
03:55:27,818 --> 03:55:30,428
While we want to reach
across spatial scales

4845
03:55:30,428 --> 03:55:33,228
in order to enhance our
understanding of the brain,

4846
03:55:33,228 --> 03:55:35,741
we also want to integrate temporal scales.

4847
03:55:36,668 --> 03:55:38,038
And here it's really important

4848
03:55:38,038 --> 03:55:39,938
to understand strengths and limitations

4849
03:55:39,938 --> 03:55:42,628
of each of the modalities
that we're looking at,

4850
03:55:42,628 --> 03:55:44,998
as well as understand

4851
03:55:44,998 --> 03:55:47,811
the biological phenomena
that we want to study.

4852
03:55:49,438 --> 03:55:54,438
EEG and fMRI can really
resolve second time scales

4853
03:55:55,398 --> 03:55:57,198
for whole-brain functional networks,

4854
03:55:57,198 --> 03:55:58,798
and then look at circuit dynamics

4855
03:55:58,798 --> 03:56:01,458
or neuronal information flow.

4856
03:56:01,458 --> 03:56:04,588
PET, on the other hand,
is somewhat slower,

4857
03:56:04,588 --> 03:56:08,438
and so we're looking at
changes that we can resolve

4858
03:56:08,438 --> 03:56:11,328
on the order of minutes or hours or days.

4859
03:56:11,328 --> 03:56:13,108
And so, depending

4860
03:56:13,108 --> 03:56:16,618
on the biological question
we are looking at,

4861
03:56:16,618 --> 03:56:20,758
each modality can add their
own strength and limitation

4862
03:56:20,758 --> 03:56:23,901
to this type of understanding.

4863
03:56:26,828 --> 03:56:30,178
Now, neuroreceptor PET
imaging is the gold standard

4864
03:56:30,178 --> 03:56:35,178
for determining in vivo drug
binding to a specific target.

4865
03:56:35,778 --> 03:56:38,568
In this image, you can
see a parametric map

4866
03:56:38,568 --> 03:56:42,288
of a D2/D3 radiotracer

4867
03:56:42,288 --> 03:56:46,318
that is in competition
with an antipsychotic drug.

4868
03:56:46,318 --> 03:56:49,348
The drug is administered
at different doses.

4869
03:56:49,348 --> 03:56:51,858
And so at each dose, we can determine

4870
03:56:51,858 --> 03:56:54,941
receptor occupancy using PET imaging.

4871
03:56:57,638 --> 03:56:59,868
Functional MRI allows us to image

4872
03:56:59,868 --> 03:57:01,928
the brain's functional response

4873
03:57:01,928 --> 03:57:04,958
or activation due to a stimulus.

4874
03:57:04,958 --> 03:57:08,938
And in pharmacological MRI,
the stimulus is a drug.

4875
03:57:08,938 --> 03:57:12,908
And so we can also use functional MRI

4876
03:57:12,908 --> 03:57:17,411
to look at an in vivo drug
response, similar to PET imaging.

4877
03:57:18,658 --> 03:57:21,418
Of course, it gives us a
different set of information.

4878
03:57:21,418 --> 03:57:25,988
And so here you can see a
change in cerebral blood flow

4879
03:57:25,988 --> 03:57:28,448
due to the psychostimulant amphetamine

4880
03:57:28,448 --> 03:57:32,831
that has been shown using functional MRI.

4881
03:57:36,458 --> 03:57:38,578
And in our lab, we are really interested

4882
03:57:38,578 --> 03:57:43,578
in understanding how a stimulus,
such as a drug or a task,

4883
03:57:43,978 --> 03:57:45,988
changes receptor binding,

4884
03:57:45,988 --> 03:57:50,028
which then in turn
modulates neuronal activity,

4885
03:57:50,028 --> 03:57:52,378
and how we can record that

4886
03:57:52,378 --> 03:57:56,008
using vascular response as a readout,

4887
03:57:56,008 --> 03:57:58,358
by understanding that
there is a connection

4888
03:57:58,358 --> 03:58:00,591
through energy consumption and metabolism.

4889
03:58:01,518 --> 03:58:05,958
And I will cover a set of
studies in this talk today

4890
03:58:05,958 --> 03:58:09,918
that will touch upon this
flow of information chain here

4891
03:58:09,918 --> 03:58:12,168
and how we can use PET and MRI

4892
03:58:12,168 --> 03:58:17,168
to dissect some of these
questions along this pathway.

4893
03:58:18,918 --> 03:58:20,631
So this is the outline of my talk.

4894
03:58:21,518 --> 03:58:23,458
I will touch upon four main topics,

4895
03:58:23,458 --> 03:58:26,068
the first one being
pharmacological PET/MRI

4896
03:58:26,068 --> 03:58:30,038
and how we can use it to
classify drugs in vivo.

4897
03:58:30,038 --> 03:58:33,698
Second, I will show some examples
of how we can use PET/MRI

4898
03:58:33,698 --> 03:58:36,621
to measure endogenous
neurotransmitter release.

4899
03:58:37,938 --> 03:58:40,698
I will then go into
receptor quantification

4900
03:58:40,698 --> 03:58:42,588
and potential sources of bias,

4901
03:58:42,588 --> 03:58:45,108
and how we can use PET/MRI

4902
03:58:45,108 --> 03:58:47,628
to address these kinds of challenges.

4903
03:58:47,628 --> 03:58:50,708
And finally, I will show an example

4904
03:58:50,708 --> 03:58:54,168
of where we can get insight
into receptor adaptations

4905
03:58:54,168 --> 03:58:55,421
using PET/MRI.

4906
03:58:57,668 --> 03:59:00,051
So we'll start with
pharmacological PET/MRI.

4907
03:59:01,088 --> 03:59:03,008
Pharmacodynamics describes the effects

4908
03:59:03,008 --> 03:59:05,488
of a drug acting at a receptor.

4909
03:59:05,488 --> 03:59:08,268
It is characterized by
dose-response relationships,

4910
03:59:08,268 --> 03:59:10,888
which is what you can
see in this graph below

4911
03:59:10,888 --> 03:59:14,898
of response plotted versus
concentration of drug.

4912
03:59:14,898 --> 03:59:16,808
And depending on the response,

4913
03:59:16,808 --> 03:59:19,828
we can then classify drugs

4914
03:59:19,828 --> 03:59:23,541
into full agonists or
antagonists, for example.

4915
03:59:24,698 --> 03:59:26,688
This behavior is typically determined

4916
03:59:26,688 --> 03:59:28,821
using functional assays in vitro.

4917
03:59:29,728 --> 03:59:33,648
In vivo, our best measurement for this

4918
03:59:33,648 --> 03:59:36,691
is to use receptor
occupancy from PET imaging.

4919
03:59:38,628 --> 03:59:42,008
And it is typically assumed
that receptor occupancy

4920
03:59:42,008 --> 03:59:44,328
is related to tissue function,

4921
03:59:44,328 --> 03:59:47,591
or is really representative
of tissue function.

4922
03:59:48,838 --> 03:59:51,918
However, we also know
that receptor binding

4923
03:59:51,918 --> 03:59:55,448
does not always equate receptor function.

4924
03:59:55,448 --> 03:59:57,088
And so one of the strengths

4925
03:59:57,088 --> 04:00:00,428
of using PET and fMRI as a readout,

4926
04:00:00,428 --> 04:00:03,068
or pharmacological MRI as a readout,

4927
04:00:03,068 --> 04:00:05,558
is that we can have independent readouts

4928
04:00:05,558 --> 04:00:08,368
of receptor occupancy

4929
04:00:08,368 --> 04:00:12,088
and potentially tissue function from fMRI,

4930
04:00:12,088 --> 04:00:13,958
and to understand better

4931
04:00:13,958 --> 04:00:16,571
what this relationship
actually looks like.

4932
04:00:18,388 --> 04:00:19,658
In a series of studies,

4933
04:00:19,658 --> 04:00:22,588
we looked at a set of
pharmacological challenges

4934
04:00:22,588 --> 04:00:24,428
in non-human primates

4935
04:00:24,428 --> 04:00:27,948
to evaluate different types of drugs

4936
04:00:27,948 --> 04:00:31,238
that bind to the D2/D3 dopamine system.

4937
04:00:31,238 --> 04:00:34,368
And we used the radiotracer C11 raclopride

4938
04:00:34,368 --> 04:00:37,241
together with functional MRI as a readout.

4939
04:00:38,078 --> 04:00:42,308
Specifically, we looked
at two antagonists,

4940
04:00:42,308 --> 04:00:45,338
a partial agonist, and two agonists,

4941
04:00:45,338 --> 04:00:47,751
quinpirole and Ropinirole.

4942
04:00:51,628 --> 04:00:53,048
Our study design was to do

4943
04:00:53,048 --> 04:00:56,718
a pharmacological
intervention during the scan,

4944
04:00:56,718 --> 04:00:59,868
and simultaneously acquire functional MRI

4945
04:00:59,868 --> 04:01:01,941
together with dynamic PET.

4946
04:01:03,088 --> 04:01:07,288
Because we are looking at
imaging non-human primates,

4947
04:01:07,288 --> 04:01:09,938
we are also injecting an
iron oxide contrast agent

4948
04:01:09,938 --> 04:01:14,088
that allows us to determine
fMRI signal changes

4949
04:01:14,088 --> 04:01:16,031
in terms of cerebral blood volume.

4950
04:01:18,718 --> 04:01:20,638
Here are more details
of our image acquisition

4951
04:01:20,638 --> 04:01:22,488
and analysis methods.

4952
04:01:22,488 --> 04:01:24,758
We acquire PET images in list mode,

4953
04:01:24,758 --> 04:01:26,558
and then reconstruct into time bins

4954
04:01:26,558 --> 04:01:28,361
of 30 seconds to one minute.

4955
04:01:29,238 --> 04:01:32,048
All images are quantified
using kinetic modeling

4956
04:01:32,048 --> 04:01:34,058
and a reference tissue model

4957
04:01:34,058 --> 04:01:36,341
with cerebellum as the reference region.

4958
04:01:37,448 --> 04:01:40,628
To compare with dynamic
functional MR measurements,

4959
04:01:40,628 --> 04:01:43,528
we also include a
time-varying binding term

4960
04:01:43,528 --> 04:01:46,038
that allows us to compute
dynamic binding potential

4961
04:01:46,038 --> 04:01:49,631
and also plot receptor
occupancies over time.

4962
04:01:51,288 --> 04:01:53,248
Functional MR images are acquired

4963
04:01:53,248 --> 04:01:56,698
using PET-compatible non-human
primate receive coil,

4964
04:01:56,698 --> 04:01:59,128
and we acquire gradient
echo-echo plan imaging

4965
04:01:59,128 --> 04:02:01,541
with a timing resolution of three seconds.

4966
04:02:02,378 --> 04:02:04,078
We then use a general linear model

4967
04:02:04,078 --> 04:02:07,578
to analyze the fMRI images

4968
04:02:07,578 --> 04:02:10,128
and model the pharmacological challenge

4969
04:02:10,128 --> 04:02:12,771
with a gamma-variate function
or sigmoidal function.

4970
04:02:13,788 --> 04:02:15,318
I also want to point out again

4971
04:02:15,318 --> 04:02:17,648
that we use an iron oxide contrast agent,

4972
04:02:17,648 --> 04:02:20,438
which allows us to convert
fMRI signal changes

4973
04:02:20,438 --> 04:02:21,971
to cerebral blood volume.

4974
04:02:23,368 --> 04:02:26,448
When using the D2/D3
antagonist raclopride,

4975
04:02:26,448 --> 04:02:29,168
as both the challenge
and the radio tracer,

4976
04:02:29,168 --> 04:02:32,008
we get a robust decrease

4977
04:02:32,008 --> 04:02:34,721
in C11 raclopride binding potential.

4978
04:02:35,838 --> 04:02:39,148
We also see an increase
in cerebral blood volume

4979
04:02:39,148 --> 04:02:41,588
due to the drug challenge itself.

4980
04:02:41,588 --> 04:02:45,708
And in the time course that
you can see below the images,

4981
04:02:45,708 --> 04:02:49,958
you can see that the blue curve
shows PET-specific binding

4982
04:02:49,958 --> 04:02:52,758
that is similar in its time course

4983
04:02:52,758 --> 04:02:55,638
to the increase in cerebral blood volume.

4984
04:02:55,638 --> 04:02:57,668
And these similar temporal responses

4985
04:02:57,668 --> 04:03:00,678
are concordant with a
classical occupancy model

4986
04:03:00,678 --> 04:03:03,378
in which binding directly relates

4987
04:03:03,378 --> 04:03:05,691
to receptor functional changes.

4988
04:03:08,558 --> 04:03:12,188
When using a D2/D3 agonist,
in this case quinpirole,

4989
04:03:12,188 --> 04:03:15,208
we again get an increase
in receptor occupancy.

4990
04:03:15,208 --> 04:03:20,091
And now we get a localized
decrease in CBV in the striatum.

4991
04:03:20,998 --> 04:03:24,058
In the temporal responses, we find

4992
04:03:24,058 --> 04:03:28,328
that there's diverging time
courses between PET and fMRI,

4993
04:03:28,328 --> 04:03:32,308
which may indicate receptor
internalization mechanisms.

4994
04:03:32,308 --> 04:03:35,568
And we will go a little bit
more into details on this

4995
04:03:35,568 --> 04:03:37,291
in the last section of my talk.

4996
04:03:39,248 --> 04:03:41,758
By investigating a variety
of dopaminergic drugs,

4997
04:03:41,758 --> 04:03:45,808
we find that our results
generally hold true.

4998
04:03:45,808 --> 04:03:48,348
At baseline, there's a
high specific binding

4999
04:03:48,348 --> 04:03:51,148
of C11 raclopride in the striatum.

5000
04:03:51,148 --> 04:03:55,898
And with both preclinical as
well as a clinical antagonist

5001
04:03:55,898 --> 04:03:58,628
that's specific to D2/D3 receptors,

5002
04:03:58,628 --> 04:04:01,208
we find that there's high occupancy

5003
04:04:01,208 --> 04:04:04,498
coupled with an increase
in cerebral blood volume

5004
04:04:04,498 --> 04:04:06,351
that's localized to the striatum.

5005
04:04:07,918 --> 04:04:10,115
With agonists, in this case quinpirole

5006
04:04:10,115 --> 04:04:12,418
and the clinical agonist ropinirole,

5007
04:04:12,418 --> 04:04:17,418
we find that there is again a
decrease in receptor occupancy

5008
04:04:17,488 --> 04:04:22,368
that is related to a decrease
in cerebral blood volume,

5009
04:04:22,368 --> 04:04:25,948
which is the functional response
and expected to be negative

5010
04:04:25,948 --> 04:04:28,641
with D2 receptors being inhibitory.

5011
04:04:30,668 --> 04:04:32,788
We also looked at partial agonists,

5012
04:04:32,788 --> 04:04:36,178
and this is the third
generation antipsychotic

5013
04:04:36,178 --> 04:04:37,998
aripiprazole here.

5014
04:04:37,998 --> 04:04:41,518
At about 90% occupancy, which is matched

5015
04:04:41,518 --> 04:04:44,908
to the same occupancy as the antagonists,

5016
04:04:44,908 --> 04:04:49,031
we find that there's
still a positive response

5017
04:04:49,031 --> 04:04:51,918
in the cerebral blood volume,

5018
04:04:51,918 --> 04:04:54,488
but it's at a lower magnitude

5019
04:04:54,488 --> 04:04:57,638
compared to the full acting antagonists.

5020
04:04:57,638 --> 04:05:01,458
And this area of partial agonists

5021
04:05:01,458 --> 04:05:03,768
is really an interesting area to explore

5022
04:05:03,768 --> 04:05:06,468
in how we can add functional MRI

5023
04:05:06,468 --> 04:05:11,108
to understand better about
the functional effects

5024
04:05:11,108 --> 04:05:13,521
when bound to the receptor.

5025
04:05:15,718 --> 04:05:17,058
The effects of partial agonists

5026
04:05:17,058 --> 04:05:19,228
have also been explored
at the serotonin system

5027
04:05:19,228 --> 04:05:21,551
in this study led by Dr. Hansen.

5028
04:05:22,778 --> 04:05:26,238
In this case, a serotonin
receptor partial agonist,

5029
04:05:26,238 --> 04:05:29,598
AZ, was given as a within scan challenge

5030
04:05:29,598 --> 04:05:32,571
and showed to reduce binding potential.

5031
04:05:33,978 --> 04:05:37,078
At the same time, the
cerebral blood volume response

5032
04:05:37,078 --> 04:05:38,968
was shown to be biphasic,

5033
04:05:38,968 --> 04:05:42,881
with an initial negative
followed by a positive component,

5034
04:05:43,738 --> 04:05:46,298
suggesting that this is really a drug

5035
04:05:46,298 --> 04:05:48,598
that's acting as a partial agonist

5036
04:05:48,598 --> 04:05:52,081
by having both a negative
and a positive response.

5037
04:05:54,278 --> 04:05:56,538
So pharmacological PET/MRI allows us

5038
04:05:56,538 --> 04:06:00,838
to gain more insight into drug studies

5039
04:06:00,838 --> 04:06:05,288
beyond just simple target
engagement and receptor occupancy.

5040
04:06:05,288 --> 04:06:10,288
With PET, we can ensure
and directly get evidence

5041
04:06:10,708 --> 04:06:13,358
for the fact that a drug
binds to a receptor,

5042
04:06:13,358 --> 04:06:16,378
as shown here for
examples at the dopamine,

5043
04:06:16,378 --> 04:06:19,458
the opioid, and the serotonin system.

5044
04:06:19,458 --> 04:06:23,858
When additionally acquiring
fMRI, and in this case,

5045
04:06:23,858 --> 04:06:26,638
these are all cerebral
blood volume responses,

5046
04:06:26,638 --> 04:06:29,938
we can then evaluate the drug function

5047
04:06:29,938 --> 04:06:34,798
by looking at its time courses,
its signaling properties,

5048
04:06:34,798 --> 04:06:39,778
and gain more insight beyond
occupancy measurements.

5049
04:06:39,778 --> 04:06:42,578
And it is a really interesting area

5050
04:06:42,578 --> 04:06:45,908
to further explore time courses

5051
04:06:45,908 --> 04:06:49,338
and compare both PET time activity curves

5052
04:06:49,338 --> 04:06:51,248
and specific binding reductions

5053
04:06:51,248 --> 04:06:56,248
with changes in cerebral blood
volume or fMRI time courses,

5054
04:06:56,268 --> 04:06:58,968
because there's a lot of insights

5055
04:06:58,968 --> 04:07:02,278
that can be gained from understanding

5056
04:07:02,278 --> 04:07:05,081
concordant or diverging time courses.

5057
04:07:07,208 --> 04:07:09,898
This brings me to the
second topic of my talk,

5058
04:07:09,898 --> 04:07:12,678
which is on measuring endogenous
neurotransmitter release

5059
04:07:12,678 --> 04:07:13,711
with PET/MRI.

5060
04:07:15,548 --> 04:07:17,838
I've touched upon how PET and fMRI

5061
04:07:17,838 --> 04:07:21,588
are complementary tools for
dynamic functional imaging,

5062
04:07:21,588 --> 04:07:24,048
but this table more specifically outlines

5063
04:07:24,048 --> 04:07:26,858
some of the parameters
that are of interest

5064
04:07:26,858 --> 04:07:30,861
when we are doing combined
PET and fMRI experiments.

5065
04:07:32,408 --> 04:07:34,258
And I'd like to outline a couple

5066
04:07:34,258 --> 04:07:37,258
of these parameters more specifically.

5067
04:07:37,258 --> 04:07:38,818
PET imaging of course is known

5068
04:07:38,818 --> 04:07:41,478
for its molecular
sensitivity to its target.

5069
04:07:41,478 --> 04:07:44,798
And this allows us to
determine pharmacokinetics

5070
04:07:44,798 --> 04:07:48,321
and then receptor occupancies,
all at tracer-level doses.

5071
04:07:49,628 --> 04:07:53,938
PET, compared to fMRI, is slower

5072
04:07:53,938 --> 04:07:57,218
and has a temporal resolution
on the order of minutes,

5073
04:07:57,218 --> 04:08:02,218
whereas functional MRI can
resolve images at a second level.

5074
04:08:04,838 --> 04:08:07,388
But the pure temporal
resolution is different

5075
04:08:07,388 --> 04:08:10,908
from what are potentially
observable dynamic changes

5076
04:08:10,908 --> 04:08:12,598
when related to biology.

5077
04:08:12,598 --> 04:08:16,068
And in PET, this is really
radiotracer-dependent,

5078
04:08:16,068 --> 04:08:19,368
as it depends on the
kinetics of the tracer.

5079
04:08:19,368 --> 04:08:23,178
And in functional MRI, this
also depends on biology

5080
04:08:23,178 --> 04:08:25,551
and several other factors.

5081
04:08:26,868 --> 04:08:31,058
Another key parameter that's
important to keep in mind

5082
04:08:31,058 --> 04:08:35,568
is the detectable signal
changes from baseline.

5083
04:08:35,568 --> 04:08:40,278
This really depends, one,
on how fast the changes are,

5084
04:08:40,278 --> 04:08:45,038
but also, two, on how reliable
the baseline measurement is

5085
04:08:45,038 --> 04:08:49,538
and whether you can, for
example, repeat certain stimuli

5086
04:08:49,538 --> 04:08:51,428
as is typically done in fMRI.

5087
04:08:52,548 --> 04:08:54,668
And so a lot of these parameters

5088
04:08:54,668 --> 04:08:58,268
influence the kinds of robust measurements

5089
04:08:58,268 --> 04:09:02,988
that you can do with PET or fMRI.

5090
04:09:02,988 --> 04:09:07,568
And really understanding the parameters

5091
04:09:07,568 --> 04:09:09,088
between those two modalities,

5092
04:09:09,088 --> 04:09:11,908
in order to get the
best experimental design

5093
04:09:11,908 --> 04:09:15,591
to understand underlying
biology, is really important.

5094
04:09:16,478 --> 04:09:21,018
And I'm gonna outline a couple
of examples of studies now

5095
04:09:21,018 --> 04:09:23,738
that have really made use

5096
04:09:23,738 --> 04:09:27,568
of some of these features in PET and fMRI

5097
04:09:27,568 --> 04:09:32,081
to understand more about
brain networks in general,

5098
04:09:33,548 --> 04:09:36,048
In this study led by Dr. Hansen,

5099
04:09:36,048 --> 04:09:40,248
serotonin release was measured
during visual attention.

5100
04:09:40,248 --> 04:09:45,198
And here, autobiographical
images with positive valence

5101
04:09:45,198 --> 04:09:48,211
were shown during simultaneous
PET/MRI acquisition.

5102
04:09:49,478 --> 04:09:53,268
They showed that there's
changes in binding potential

5103
04:09:53,268 --> 04:09:55,808
during the stimulus session,

5104
04:09:55,808 --> 04:09:59,258
and that there's increases
in cerebral blood flow

5105
04:09:59,258 --> 04:10:02,251
during the viewing of these images.

5106
04:10:04,068 --> 04:10:07,238
Most interestingly,
there was a correlation

5107
04:10:07,238 --> 04:10:09,468
between the change in cerebral blood flow

5108
04:10:09,468 --> 04:10:11,338
and the change in binding potential

5109
04:10:11,338 --> 04:10:15,458
from the 5-HT1B receptor radioligand.

5110
04:10:15,458 --> 04:10:17,218
And so this really suggests a link

5111
04:10:17,218 --> 04:10:19,388
between serotonin signaling

5112
04:10:19,388 --> 04:10:21,628
and visual processing and attention.

5113
04:10:21,628 --> 04:10:25,338
And so overall, this
PET/fMRI study allowed us

5114
04:10:25,338 --> 04:10:29,578
to get a better insight
into the role of 5-HT

5115
04:10:29,578 --> 04:10:33,151
in overall sensory and
behavioral processing.

5116
04:10:35,428 --> 04:10:36,838
In another example of looking

5117
04:10:36,838 --> 04:10:39,258
at endogenous neurotransmitter release,

5118
04:10:39,258 --> 04:10:42,258
Dr. Wey et al. looked at pain-induced

5119
04:10:42,258 --> 04:10:44,558
PET and fMRI activations.

5120
04:10:44,558 --> 04:10:47,198
Here, pain pressure was used to induce

5121
04:10:47,198 --> 04:10:49,858
endogenous new opioid release,

5122
04:10:49,858 --> 04:10:53,438
and increases in BOLD fMRI were found,

5123
04:10:53,438 --> 04:10:55,528
together with changes

5124
04:10:55,528 --> 04:10:59,201
in diprenorphine PET
receptor availability.

5125
04:11:00,248 --> 04:11:03,118
And in the thalamus,
a positive correlation

5126
04:11:03,118 --> 04:11:05,908
between BOLD and changes
in binding potential

5127
04:11:05,908 --> 04:11:10,908
suggests that there's a
direct opioid modulation.

5128
04:11:11,418 --> 04:11:14,918
However, in a different
area, in the striatum,

5129
04:11:14,918 --> 04:11:16,518
this correlation was not found.

5130
04:11:16,518 --> 04:11:18,718
And so this suggests that maybe

5131
04:11:18,718 --> 04:11:20,298
other neurotransmitter systems

5132
04:11:20,298 --> 04:11:24,611
drive the BOLD response in the striatum.

5133
04:11:27,178 --> 04:11:30,688
In this PET/MR study, the
aim was to look at glutamate

5134
04:11:30,688 --> 04:11:34,068
release using MR spectroscopy
and a radiotracer

5135
04:11:34,068 --> 04:11:37,791
that binds to the metabotropic
glutamate receptor five.

5136
04:11:38,668 --> 04:11:42,278
And indeed, an acute decrease
in striatal glutamate

5137
04:11:42,278 --> 04:11:44,928
was observed using spectroscopy

5138
04:11:44,928 --> 04:11:48,571
after stimulation with an acetylcysteine.

5139
04:11:49,798 --> 04:11:51,348
However, no significant changes

5140
04:11:51,348 --> 04:11:54,378
in MGluR5 availability were observed.

5141
04:11:54,378 --> 04:11:57,888
Nevertheless, the relationship
between changes in glutamate

5142
04:11:57,888 --> 04:12:01,698
and a baseline binding
potential was found,

5143
04:12:01,698 --> 04:12:05,498
which is an interesting relationship

5144
04:12:05,498 --> 04:12:10,011
that I'm sure is of interest
to explore in future studies.

5145
04:12:12,368 --> 04:12:14,928
We've been interested to image
neurotransmitter release,

5146
04:12:14,928 --> 04:12:18,541
and specifically in a model
of deep brain stimulation.

5147
04:12:20,058 --> 04:12:21,998
The ventral tegmental area is known

5148
04:12:21,998 --> 04:12:25,348
as the primary dopamine-producing nucleus,

5149
04:12:25,348 --> 04:12:28,711
and plays a role in reinforcement
learning and motivation.

5150
04:12:29,568 --> 04:12:32,378
So it's been shown in the literature

5151
04:12:32,378 --> 04:12:35,718
that VTA stimulation
leads to dopamine release,

5152
04:12:35,718 --> 04:12:39,081
but it's not known what are
the receptor contributions,

5153
04:12:39,948 --> 04:12:42,878
what may be the effect of
stimulation parameters,

5154
04:12:42,878 --> 04:12:46,588
and what are the in vivo circuit effects

5155
04:12:46,588 --> 04:12:50,391
or dynamic changes that
occur to this stimulation.

5156
04:12:51,968 --> 04:12:56,968
So we looked at a monkey model
of deep brain stimulation

5157
04:12:59,008 --> 04:13:01,708
in order to determine
neurochemical dynamics

5158
04:13:01,708 --> 04:13:05,291
of microstimulation and
the functional networks.

5159
04:13:06,398 --> 04:13:11,308
You can see here an
image of the DBS implant,

5160
04:13:11,308 --> 04:13:16,308
together with the actual
microstimulation or DBS electrode.

5161
04:13:17,128 --> 04:13:20,638
And our experimental paradigm then looked

5162
04:13:20,638 --> 04:13:23,938
at dynamic acquisition of fMRI

5163
04:13:23,938 --> 04:13:27,148
and C11 raclopride PET
during a stimulation,

5164
04:13:27,148 --> 04:13:30,868
which allowed us to look
at different parameter

5165
04:13:30,868 --> 04:13:31,938
of the stimulation,

5166
04:13:31,938 --> 04:13:34,681
and also at different
experimental designs.

5167
04:13:37,158 --> 04:13:40,938
Overall, we found a really
robust fMRI activation

5168
04:13:40,938 --> 04:13:45,416
with increases in CBV
on the ipsilateral side,

5169
04:13:45,416 --> 04:13:48,938
that was mainly shown in
ventral striatum and caudate,

5170
04:13:48,938 --> 04:13:51,471
and caused by the microstimulation.

5171
04:13:53,168 --> 04:13:55,098
The PET data proved to be

5172
04:13:55,098 --> 04:13:57,418
somewhat more challenging to analyze.

5173
04:13:57,418 --> 04:14:02,418
And so we looked at various
new models to implement,

5174
04:14:03,538 --> 04:14:06,688
and we, in the end, were able to show

5175
04:14:06,688 --> 04:14:09,018
that a forward model implementation

5176
04:14:09,018 --> 04:14:11,335
of the full reference tissue model

5177
04:14:11,335 --> 04:14:14,318
localizes decreases in binding potential

5178
04:14:14,318 --> 04:14:18,808
in the ventral stratum that
match the fMRI response.

5179
04:14:18,808 --> 04:14:22,478
This model specifically uses
a faster rate of convergence

5180
04:14:22,478 --> 04:14:25,338
for estimating changes
in binding potential,

5181
04:14:25,338 --> 04:14:30,028
and really allowed for
accounting for very fast changes

5182
04:14:30,028 --> 04:14:33,218
like microstimulation in the data.

5183
04:14:33,218 --> 04:14:38,218
When using a standard
approach with an MRTM2 model,

5184
04:14:38,768 --> 04:14:42,578
we mainly saw an apparent
increase in binding potential

5185
04:14:42,578 --> 04:14:47,578
in low-binding areas that
are surrounding the stratum.

5186
04:14:50,088 --> 04:14:52,208
So we've seen that PET is highly sensitive

5187
04:14:52,208 --> 04:14:54,748
to imaging molecular processes,

5188
04:14:54,748 --> 04:14:57,918
but the sensitivity for
imaging dynamic changes

5189
04:14:57,918 --> 04:14:59,688
is really important.

5190
04:14:59,688 --> 04:15:01,328
And it's important to keep in mind

5191
04:15:01,328 --> 04:15:04,888
that the test-retest variability in PET

5192
04:15:04,888 --> 04:15:07,088
can be on the order of five to 10%.

5193
04:15:07,088 --> 04:15:09,438
And so that sometimes effect sizes

5194
04:15:09,438 --> 04:15:12,448
need to be quite large
over a sustained period

5195
04:15:12,448 --> 04:15:15,421
in order to reliably measure changes.

5196
04:15:16,428 --> 04:15:20,858
And so this brings me to
the next topic of my talk,

5197
04:15:20,858 --> 04:15:23,658
which touches upon receptor quantification

5198
04:15:23,658 --> 04:15:26,051
and potential sources of bias.

5199
04:15:28,718 --> 04:15:31,148
In this recent paper by Dr. Levine,

5200
04:15:31,148 --> 04:15:34,708
motion and model bias were investigated

5201
04:15:34,708 --> 04:15:36,428
with a set of behavioral challenges

5202
04:15:36,428 --> 04:15:39,231
using the radiotracer C11 raclopride.

5203
04:15:40,528 --> 04:15:42,788
There's both human task data

5204
04:15:42,788 --> 04:15:46,158
as well as simulations
that were investigated.

5205
04:15:46,158 --> 04:15:50,008
And he showed that motion
bias can account for changes

5206
04:15:50,008 --> 04:15:52,581
in binding potential of more than 10%.

5207
04:15:53,448 --> 04:15:55,898
And it's really important
to do correction methods,

5208
04:15:55,898 --> 04:15:58,008
such as frame-based motion correction

5209
04:15:58,008 --> 04:16:00,558
or reconstruction-based motion correction

5210
04:16:00,558 --> 04:16:04,701
in order to account for these
sources of bias due to motion.

5211
04:16:06,668 --> 04:16:09,338
Model bias can also affect the data,

5212
04:16:09,338 --> 04:16:11,548
and it can differentially affect

5213
04:16:11,548 --> 04:16:14,748
different anatomical
regions, with positive bias

5214
04:16:14,748 --> 04:16:17,761
specifically in low
binding potential areas.

5215
04:16:18,928 --> 04:16:21,418
The magnitude of the bias is dependent

5216
04:16:21,418 --> 04:16:23,318
on the choice of your model,

5217
04:16:23,318 --> 04:16:26,768
and when using a two parameter
reference tissue model,

5218
04:16:26,768 --> 04:16:29,911
is dependent on fixing K2 prime.

5219
04:16:31,498 --> 04:16:34,178
But he also showed that de-biasing

5220
04:16:34,178 --> 04:16:36,138
the contribution of the
initial uptake period

5221
04:16:36,138 --> 04:16:38,941
can reduce dependence on this parameter.

5222
04:16:40,958 --> 04:16:42,398
Furthermore, he went on to show

5223
04:16:42,398 --> 04:16:45,318
that the magnitude of
neurotransmitter release

5224
04:16:45,318 --> 04:16:49,258
can really affect some of these biases.

5225
04:16:49,258 --> 04:16:52,978
And sources of bias in
general pose a great challenge

5226
04:16:52,978 --> 04:16:56,188
when the elicited
task-response effect is small.

5227
04:16:56,188 --> 04:16:59,028
Often in behavioral paradigms,

5228
04:16:59,028 --> 04:17:03,688
a task-based change is less than 5%,

5229
04:17:03,688 --> 04:17:05,758
and this is really on the order

5230
04:17:05,758 --> 04:17:08,858
of what motion or model bias

5231
04:17:08,858 --> 04:17:12,308
can also be affected by.

5232
04:17:12,308 --> 04:17:13,888
And so it's important to be aware

5233
04:17:13,888 --> 04:17:15,911
of some of these sources of bias.

5234
04:17:17,998 --> 04:17:20,478
We have been interested in how physiology

5235
04:17:20,478 --> 04:17:24,258
may affect PET or fMRI data,

5236
04:17:24,258 --> 04:17:27,258
and specifically the effects of blood flow

5237
04:17:27,258 --> 04:17:29,658
on PET radiotracer kinetics

5238
04:17:29,658 --> 04:17:33,028
is something that's been really a question

5239
04:17:33,028 --> 04:17:35,251
in the PET literature for a long time.

5240
04:17:37,918 --> 04:17:40,668
Some radiotracers have even been suggested

5241
04:17:40,668 --> 04:17:43,958
to be surrogate markers
of cerebral blood flow,

5242
04:17:43,958 --> 04:17:47,248
but there's not been
experimental data out there

5243
04:17:47,248 --> 04:17:51,068
to specifically investigate this.

5244
04:17:51,068 --> 04:17:54,898
So we wanted to use combined PET and fMRI

5245
04:17:54,898 --> 04:17:56,888
to measure the CBF effects

5246
04:17:56,888 --> 04:18:00,751
on PET radiotracers that
bind to neuroreceptors.

5247
04:18:02,348 --> 04:18:06,058
Using combined PET and
arterial spin labeling,

5248
04:18:06,058 --> 04:18:08,678
this allowed us to look at both

5249
04:18:08,678 --> 04:18:13,108
tracer binding to receptors
together with flow.

5250
04:18:13,108 --> 04:18:17,838
And we could then also induce
changes in flow artificially

5251
04:18:17,838 --> 04:18:20,108
using a hypercapnia challenge,

5252
04:18:20,108 --> 04:18:23,478
where you breathe in
higher CO2 than normal,

5253
04:18:23,478 --> 04:18:27,411
and where you really
induce large flow changes.

5254
04:18:28,268 --> 04:18:30,068
It's possible to control and vary

5255
04:18:30,068 --> 04:18:32,408
the timing of the
challenge very precisely,

5256
04:18:32,408 --> 04:18:34,888
so it's a really good
model to investigate.

5257
04:18:34,888 --> 04:18:37,368
And so all of these study
results that I'm gonna show you

5258
04:18:37,368 --> 04:18:40,138
were done here in non-human primates

5259
04:18:40,138 --> 04:18:42,311
during a hypercapnia challenge.

5260
04:18:44,858 --> 04:18:48,188
Here are results from C11 raclopride.

5261
04:18:48,188 --> 04:18:50,358
This is a typical time activity curve

5262
04:18:50,358 --> 04:18:52,401
that you see from C11 raclopride.

5263
04:18:53,556 --> 04:18:55,818
And this is really the time activity curve

5264
04:18:55,818 --> 04:18:57,845
where we did apply hypercapnia,

5265
04:18:57,845 --> 04:19:00,468
and we did not see measurable changes

5266
04:19:00,468 --> 04:19:03,981
in the time activity curve
during the hypercapnea challenge.

5267
04:19:05,058 --> 04:19:08,488
We also did not see any residuals

5268
04:19:08,488 --> 04:19:13,488
that map with the hypercapnea
pattern from kinetic modeling.

5269
04:19:13,908 --> 04:19:17,358
And at the same time,
we have very robust data

5270
04:19:17,358 --> 04:19:22,265
showing that flow increases
by more than twofold

5271
04:19:23,268 --> 04:19:25,981
based on our arterial spin
labeling measurements.

5272
04:19:28,688 --> 04:19:30,568
We repeated a similar paradigm

5273
04:19:30,568 --> 04:19:33,718
with the radiotracer F18 fallypride,

5274
04:19:33,718 --> 04:19:36,608
which is thought to have a
high extraction fraction,

5275
04:19:36,608 --> 04:19:39,358
therefore would be more
susceptible to flow.

5276
04:19:39,358 --> 04:19:41,778
But even here we don't
see measurable changes

5277
04:19:41,778 --> 04:19:44,058
in the time activity curve,

5278
04:19:44,058 --> 04:19:47,968
and no specific pattern that relate

5279
04:19:47,968 --> 04:19:50,931
to large increases in flow.

5280
04:19:52,508 --> 04:19:54,838
These data were also repeated

5281
04:19:54,838 --> 04:19:57,438
with changing flow at the very beginning,

5282
04:19:57,438 --> 04:19:59,888
during the uptake period
of the radiotracer,

5283
04:19:59,888 --> 04:20:02,818
and we found very similar results.

5284
04:20:02,818 --> 04:20:05,408
So overall, we used PET and MRI

5285
04:20:05,408 --> 04:20:07,358
to enable simultaneous measurement

5286
04:20:07,358 --> 04:20:11,178
of receptor-specific radiotracer
kinetics and blood flow.

5287
04:20:11,178 --> 04:20:13,648
And we showed that they did not affect

5288
04:20:13,648 --> 04:20:16,001
fallypride or raclipride binding.

5289
04:20:18,728 --> 04:20:22,788
We also evaluated a set
of human imaging studies

5290
04:20:22,788 --> 04:20:26,861
in chronic pain patients
and the radiotracer PBR28,

5291
04:20:28,258 --> 04:20:30,391
which looks at neuroinflammation.

5292
04:20:31,648 --> 04:20:34,648
And here we find that
SUV outcome measurements

5293
04:20:34,648 --> 04:20:37,468
were not correlated with
cerebral blood flow,

5294
04:20:37,468 --> 04:20:39,461
as determined from pCASL.

5295
04:20:41,138 --> 04:20:45,428
We also found that PBR28 signal increases

5296
04:20:45,428 --> 04:20:48,648
were not accompanied by increases in CBF,

5297
04:20:48,648 --> 04:20:52,538
and that areas where
there was lower perfusion

5298
04:20:52,538 --> 04:20:54,971
were not accompanied by decreases in CBF.

5299
04:20:58,108 --> 04:21:02,298
When we did a set of measurements
in non-human primates,

5300
04:21:02,298 --> 04:21:05,148
where we again used a
hypercapnea challenge,

5301
04:21:05,148 --> 04:21:08,248
we showed that K1 and K2 were not elevated

5302
04:21:08,248 --> 04:21:11,578
with higher blood flow,
and outcome measures

5303
04:21:11,578 --> 04:21:13,791
were also not correlated with blood flow.

5304
04:21:14,998 --> 04:21:18,998
This makes sense when looking
at the direct relationships

5305
04:21:18,998 --> 04:21:23,078
between K1 and the extraction fraction

5306
04:21:23,078 --> 04:21:26,828
in comparison to increases in blood flow.

5307
04:21:26,828 --> 04:21:31,828
And we show that K1 is not
expected to have large changes,

5308
04:21:32,558 --> 04:21:34,341
even with large changes in flow.

5309
04:21:37,108 --> 04:21:39,218
We also more generally
looked at the effects

5310
04:21:39,218 --> 04:21:41,641
of kinetic properties and quantification.

5311
04:21:43,348 --> 04:21:47,158
Changes in binding potential
were quantified here

5312
04:21:47,158 --> 04:21:49,468
due to changes in blood flow.

5313
04:21:49,468 --> 04:21:52,728
And overall, we found that radiotracers

5314
04:21:52,728 --> 04:21:55,858
with large K1 or small K2 values

5315
04:21:55,858 --> 04:21:58,918
may be most sensitive to
changes in blood flow,

5316
04:21:58,918 --> 04:22:03,918
but that K3 and K4 are really robust

5317
04:22:04,258 --> 04:22:05,911
to changes in blood flow.

5318
04:22:08,118 --> 04:22:12,118
And for CBF increases
up to more than twofold,

5319
04:22:12,118 --> 04:22:15,308
binding potential values were
affected by less than 5%,

5320
04:22:15,308 --> 04:22:19,828
which is really within the
test-retest variability

5321
04:22:19,828 --> 04:22:21,405
of most radiotracers.

5322
04:22:23,898 --> 04:22:26,058
This brings me to the
fourth topic of my talk,

5323
04:22:26,058 --> 04:22:29,398
which is how we can use PET and MRI

5324
04:22:29,398 --> 04:22:34,218
to understand more about
changes in receptor signaling

5325
04:22:34,218 --> 04:22:36,071
and receptor adaptations over time.

5326
04:22:38,758 --> 04:22:40,478
I'll start this section
with just a little bit

5327
04:22:40,478 --> 04:22:43,758
of background on receptor adaptations.

5328
04:22:43,758 --> 04:22:46,558
Receptor internalization is a homeostatic

5329
04:22:46,558 --> 04:22:49,058
neuroadaptation mechanism at the synapse

5330
04:22:49,058 --> 04:22:51,178
that can modulate cell signaling.

5331
04:22:51,178 --> 04:22:53,008
And it generally occurs in response

5332
04:22:53,008 --> 04:22:57,001
to G-protein-coupled receptor
activation by agonist binding.

5333
04:23:00,068 --> 04:23:02,838
Initially, a receptor
would be desensitized

5334
04:23:02,838 --> 04:23:04,968
by binding to its agonist,

5335
04:23:04,968 --> 04:23:07,858
and can then undergo internalization,

5336
04:23:07,858 --> 04:23:11,761
all within the span of seconds to minutes.

5337
04:23:13,188 --> 04:23:16,328
And there may be short-term
as well as long-term tolerance

5338
04:23:16,328 --> 04:23:21,328
or adaptations that follow
from these kinds of mechanisms.

5339
04:23:21,388 --> 04:23:23,178
And our understanding of this phenomena

5340
04:23:23,178 --> 04:23:27,598
is mainly gained from in vitro evidence.

5341
04:23:27,598 --> 04:23:31,198
And in vivo, we've not
had direct measurements

5342
04:23:31,198 --> 04:23:33,318
of receptor internalization,

5343
04:23:33,318 --> 04:23:36,838
so its dynamics are relatively unknown.

5344
04:23:36,838 --> 04:23:39,748
Nevertheless, it does
affect neuroimaging data,

5345
04:23:39,748 --> 04:23:43,088
and that has been shown
in several publications,

5346
04:23:43,088 --> 04:23:47,391
which is the case both for
fMRI as well as for PET data.

5347
04:23:49,428 --> 04:23:52,048
So a non-invasive in vivo measurement

5348
04:23:52,048 --> 04:23:56,708
of any parts of these receptor
internalization mechanisms

5349
04:23:56,708 --> 04:23:59,878
could enable the evaluation
of drug dynamics,

5350
04:23:59,878 --> 04:24:01,558
and allow us to understand

5351
04:24:01,558 --> 04:24:05,021
receptor dysfunction in
disease in more detail.

5352
04:24:08,748 --> 04:24:11,728
Here you can see a dopamine synapse

5353
04:24:11,728 --> 04:24:15,038
and receptors at the
postsynaptic membrane.

5354
04:24:15,038 --> 04:24:17,078
In a classical occupancy model,

5355
04:24:17,078 --> 04:24:22,078
a drug can wash in and
bind to these receptors,

5356
04:24:22,428 --> 04:24:24,198
eliciting a downstream response

5357
04:24:24,198 --> 04:24:28,028
that we can read out with an fMRI signal.

5358
04:24:28,028 --> 04:24:31,038
Because the drug is now
bound to the receptors,

5359
04:24:31,038 --> 04:24:35,088
we would expect a decrease
in our PET imaging signal.

5360
04:24:35,088 --> 04:24:38,648
And in this classical occupancy model,

5361
04:24:38,648 --> 04:24:41,151
binding directly follows function.

5362
04:24:44,318 --> 04:24:47,128
In an internalization model, though,

5363
04:24:47,128 --> 04:24:50,368
the drug also washes in, it binds,

5364
04:24:50,368 --> 04:24:55,238
but after it's bound, it can desensitize

5365
04:24:55,238 --> 04:24:58,238
and then internalize receptors,

5366
04:24:58,238 --> 04:25:02,651
which then may attenuate the
functional signaling response,

5367
04:25:03,548 --> 04:25:06,418
yet at the same time,
the PET signal would show

5368
04:25:06,418 --> 04:25:10,238
a similarly decreased PET
signal as in the left hand case,

5369
04:25:10,238 --> 04:25:15,038
because only available
receptors are similarly bound

5370
04:25:15,038 --> 04:25:16,488
at the postsynaptic membrane.

5371
04:25:18,168 --> 04:25:21,398
And in vitro measurements have shown

5372
04:25:21,398 --> 04:25:23,768
that internalization does occur

5373
04:25:23,768 --> 04:25:26,431
in response to agonist binding.

5374
04:25:28,388 --> 04:25:30,588
So we've looked at this

5375
04:25:30,588 --> 04:25:32,518
from a more theoretical point of view,

5376
04:25:32,518 --> 04:25:36,128
where we looked at how
internalization may affect

5377
04:25:36,128 --> 04:25:38,758
both PET and fMRI imaging signals.

5378
04:25:38,758 --> 04:25:43,758
And we can express the total
number of receptors, Bmax,

5379
04:25:44,538 --> 04:25:46,868
as the sum of available receptors,

5380
04:25:46,868 --> 04:25:50,378
receptors bound by agonist,
and receptors bound by DA.

5381
04:25:50,378 --> 04:25:52,948
Both of these compartments are in flux

5382
04:25:52,948 --> 04:25:55,791
with the free drug concentrations.

5383
04:25:57,098 --> 04:25:59,198
If we then add an additional compartment

5384
04:25:59,198 --> 04:26:01,158
that describes internalized receptors,

5385
04:26:01,158 --> 04:26:05,881
we can simulate its effects on
both a PET and fMRI signals.

5386
04:26:07,438 --> 04:26:10,428
And in simulations without
receptor trafficking,

5387
04:26:10,428 --> 04:26:13,288
we see that there is a
classical occupancy model

5388
04:26:13,288 --> 04:26:17,168
where CBV and PET occupancy
time courses are matched.

5389
04:26:17,168 --> 04:26:20,048
And this is what I've shown you in data

5390
04:26:20,048 --> 04:26:23,821
with the D2/D3 antagonist
in the beginning of my talk.

5391
04:26:25,668 --> 04:26:27,598
When internalization does occur,

5392
04:26:27,598 --> 04:26:32,378
we find that PET and
MR imaging time courses

5393
04:26:32,378 --> 04:26:36,598
start to diverge, with fMRI
time courses being shortened

5394
04:26:36,598 --> 04:26:39,418
and PET displacement being prolonged.

5395
04:26:39,418 --> 04:26:44,418
And again, this is the
kind of experimental data

5396
04:26:44,478 --> 04:26:48,388
that we also see in our experiments

5397
04:26:48,388 --> 04:26:52,451
when using a D2/D3 agonist drug challenge.

5398
04:26:55,038 --> 04:26:58,028
We've been interested in
exploring receptor internalization

5399
04:26:58,028 --> 04:27:01,631
in the context of high doses
of endogenous dopamine.

5400
04:27:03,828 --> 04:27:06,578
Amphetamine type
stimulants acutely increase

5401
04:27:06,578 --> 04:27:08,585
extracellular dopamine levels.

5402
04:27:08,585 --> 04:27:12,088
And internalization is an
immediate synaptic response

5403
04:27:12,088 --> 04:27:15,111
to adapting to such high
concentrations of dopamine.

5404
04:27:17,218 --> 04:27:20,238
In vitro internalization of D2 receptors

5405
04:27:20,238 --> 04:27:22,318
can occur very fast, within minutes,

5406
04:27:22,318 --> 04:27:25,478
and receptors can stay internalized

5407
04:27:25,478 --> 04:27:28,901
for more than an hour and up to 24 hours.

5408
04:27:30,358 --> 04:27:33,138
It's also been shown that
there may be differences

5409
04:27:33,138 --> 04:27:36,428
between D1 and D2
receptor internalizations,

5410
04:27:36,428 --> 04:27:39,058
with D1 receptors recycling quickly

5411
04:27:39,058 --> 04:27:41,201
back to the cell membrane surface.

5412
04:27:43,028 --> 04:27:46,958
Overall, this phenomena has
really only been explored

5413
04:27:46,958 --> 04:27:49,038
at the D2/D3 receptors,

5414
04:27:49,038 --> 04:27:51,488
and we have some PET imaging data

5415
04:27:51,488 --> 04:27:53,328
that show that receptor availability

5416
04:27:53,328 --> 04:27:56,318
may stay depressed for up to two days

5417
04:27:56,318 --> 04:27:58,211
due to receptor internalization.

5418
04:28:00,298 --> 04:28:03,638
So we designed a study
to look at the effects

5419
04:28:03,638 --> 04:28:07,498
of repeated amphetamine and
receptor internalization.

5420
04:28:07,498 --> 04:28:09,428
The goal of the study was to determine

5421
04:28:09,428 --> 04:28:12,538
dopamine receptor
adaptation longitudinally

5422
04:28:12,538 --> 04:28:14,778
due to repeated amphetamine in vivo,

5423
04:28:14,778 --> 04:28:16,488
and how this would be reflected

5424
04:28:16,488 --> 04:28:19,251
in a modulation of PET and fMRI signals.

5425
04:28:20,738 --> 04:28:24,198
Our hypothesis for this
was that stimulant drugs

5426
04:28:24,198 --> 04:28:27,988
cause rapid internalization
of dopamine receptors in vivo,

5427
04:28:27,988 --> 04:28:30,198
but that D1 and D2 receptors

5428
04:28:30,198 --> 04:28:33,378
may have differential recovery times.

5429
04:28:33,378 --> 04:28:35,628
So in order to run this study,

5430
04:28:35,628 --> 04:28:39,438
we used the simultaneous
PET/MR acquisitions

5431
04:28:39,438 --> 04:28:41,311
with the radiotracer C11 raclopride.

5432
04:28:42,828 --> 04:28:45,588
This study was done in non-human primates,

5433
04:28:45,588 --> 04:28:47,788
and amphetamine was injected

5434
04:28:47,788 --> 04:28:51,028
about 40 minutes after
radiotracer injection,

5435
04:28:51,028 --> 04:28:53,451
and fMRI was simultaneously recorded.

5436
04:28:54,788 --> 04:28:56,678
After first amphetamine injection,

5437
04:28:56,678 --> 04:28:58,978
we then injected amphetamine a second time

5438
04:28:58,978 --> 04:29:02,388
at either three or 24 hours,

5439
04:29:02,388 --> 04:29:06,711
allowing us to evaluate the
effects of a second injection.

5440
04:29:07,828 --> 04:29:12,828
And then, in order to look
at D1 versus D2 receptors,

5441
04:29:12,958 --> 04:29:17,788
we blocked the D1 receptors
using Schering 23390

5442
04:29:17,788 --> 04:29:21,328
in an equivalent set of experiments

5443
04:29:21,328 --> 04:29:23,078
that allowed us to differentiate

5444
04:29:23,078 --> 04:29:25,925
between D1 and D2 receptor signaling.

5445
04:29:27,768 --> 04:29:29,998
Our results show that
dopamine release occurred

5446
04:29:29,998 --> 04:29:32,481
with each amphetamine
challenge in the striatum.

5447
04:29:33,628 --> 04:29:36,068
Second injection showed a slightly reduced

5448
04:29:36,068 --> 04:29:40,031
dopamine release capacity at
both three hours and 24 hours.

5449
04:29:41,168 --> 04:29:44,308
Simultaneously acquired cerebral
blood volume measurements

5450
04:29:44,308 --> 04:29:47,098
showed a negative cerebral
blood volume response

5451
04:29:47,098 --> 04:29:50,128
with the first injection at zero hours.

5452
04:29:50,128 --> 04:29:54,878
But when given amphetamine
for a second time

5453
04:29:54,878 --> 04:29:58,538
at three hours, this negative CBV response

5454
04:29:58,538 --> 04:30:01,711
converted to a positive CBV response.

5455
04:30:02,568 --> 04:30:06,058
And at 24 hours, we find
that there is a mixed

5456
04:30:06,058 --> 04:30:08,678
negative and positive CBV component

5457
04:30:08,678 --> 04:30:12,401
that is evident on the
cerebral blood volume maps.

5458
04:30:13,988 --> 04:30:16,608
Time activity curves from C11 raclopride

5459
04:30:16,608 --> 04:30:18,171
are shown in this slide here.

5460
04:30:19,548 --> 04:30:21,118
With each amphetamine challenge

5461
04:30:21,118 --> 04:30:23,128
there's a decrease in binding,

5462
04:30:23,128 --> 04:30:25,838
suggesting a release of dopamine

5463
04:30:25,838 --> 04:30:28,051
for each of these amphetamine injections.

5464
04:30:29,384 --> 04:30:32,208
You can also see that binding
potential stays decreased

5465
04:30:32,208 --> 04:30:34,548
for the three hour time point,

5466
04:30:34,548 --> 04:30:37,618
but returns back to baseline by 24 hours.

5467
04:30:37,618 --> 04:30:39,038
And that's also reflected

5468
04:30:39,038 --> 04:30:41,681
in the corresponding
occupancy measurements.

5469
04:30:43,358 --> 04:30:45,978
The corresponding cerebral
blood volume time courses

5470
04:30:45,978 --> 04:30:50,098
show that there's a short
lasting negative CBV response

5471
04:30:50,098 --> 04:30:52,418
for the first amphetamine injection,

5472
04:30:52,418 --> 04:30:54,808
but that this response
is markedly different

5473
04:30:54,808 --> 04:30:58,378
at the second injection at
the three hour time point,

5474
04:30:58,378 --> 04:31:00,538
where the CBV response now shows

5475
04:31:00,538 --> 04:31:03,948
a longer lasting positive response.

5476
04:31:03,948 --> 04:31:07,468
At 24 hours, there's a mix of a positive

5477
04:31:07,468 --> 04:31:09,851
and negative response
that you can see here.

5478
04:31:11,328 --> 04:31:14,958
We can also visualize the
occupancy time courses

5479
04:31:14,958 --> 04:31:16,918
together with the CBV time courses,

5480
04:31:16,918 --> 04:31:19,088
as shown in these graphs here.

5481
04:31:19,088 --> 04:31:21,998
And it is clearly visible
that D2 receptor occupancy

5482
04:31:21,998 --> 04:31:24,398
remains elevated at three hours,

5483
04:31:24,398 --> 04:31:27,911
but then returns back to
baseline after 24 hours.

5484
04:31:29,908 --> 04:31:33,358
When using a D1 receptor blocker,

5485
04:31:33,358 --> 04:31:36,801
we see that dopamine release
was slightly attenuated.

5486
04:31:37,778 --> 04:31:39,968
But the most interesting part here

5487
04:31:39,968 --> 04:31:44,118
is that the CBV response
at the three hour mark

5488
04:31:44,118 --> 04:31:48,608
was effectively blocked due
to the D1 receptor blocking,

5489
04:31:48,608 --> 04:31:51,898
suggesting that this positive
response at three hours

5490
04:31:51,898 --> 04:31:55,048
is mediated by D1 receptors.

5491
04:31:55,048 --> 04:31:58,908
The negative responses
at zero and 24 hours

5492
04:31:58,908 --> 04:32:02,921
were hardly affected by
Schering 23390 blocking.

5493
04:32:04,218 --> 04:32:07,788
The corresponding C11
raclopride time activity curves

5494
04:32:07,788 --> 04:32:12,208
also show that Schering did
not significantly affect

5495
04:32:12,208 --> 04:32:13,948
C11 raclopride binding

5496
04:32:13,948 --> 04:32:16,281
compared to the non-blocking condition.

5497
04:32:17,958 --> 04:32:21,448
And finally, we quantified
all binding potentials

5498
04:32:21,448 --> 04:32:23,448
for each of the time points,

5499
04:32:23,448 --> 04:32:26,568
again, showing that Schering had

5500
04:32:26,568 --> 04:32:28,798
a somewhat reduced dopamine release,

5501
04:32:28,798 --> 04:32:33,661
but did not affect D2
receptor binding otherwise.

5502
04:32:35,428 --> 04:32:37,428
We can also clearly visualize the effect

5503
04:32:37,428 --> 04:32:40,451
of blocking D1 receptors
on Voxel-Wise maps.

5504
04:32:42,318 --> 04:32:45,058
Even with a D1 receptor blocker on board,

5505
04:32:45,058 --> 04:32:47,338
we still get dopamine release

5506
04:32:47,338 --> 04:32:48,991
with each amphetamine injection.

5507
04:32:49,878 --> 04:32:51,798
The main striking part here

5508
04:32:51,798 --> 04:32:55,098
is that the positive cerebral
blood volume component

5509
04:32:55,098 --> 04:32:59,701
seems to be entirely eliminated
by blocking D1 receptors.

5510
04:33:01,828 --> 04:33:06,018
We find that striatal
responses are all very similar

5511
04:33:06,018 --> 04:33:07,918
in terms of the functional signaling

5512
04:33:07,918 --> 04:33:09,661
with each amphetamine injection.

5513
04:33:10,838 --> 04:33:13,148
However, we find that the thalamus

5514
04:33:13,148 --> 04:33:14,658
behaves a little bit different,

5515
04:33:14,658 --> 04:33:18,258
in that the sharing compound was not able

5516
04:33:18,258 --> 04:33:21,908
to block a positive
component in the thalamus,

5517
04:33:21,908 --> 04:33:24,368
whereas it was effectively blocking

5518
04:33:24,368 --> 04:33:27,518
all of the positive
signaling in the striatum.

5519
04:33:27,518 --> 04:33:31,378
So this part is really
an interesting phenomena

5520
04:33:31,378 --> 04:33:34,791
that would be of value
to be explored further.

5521
04:33:36,318 --> 04:33:37,718
In this final section of my talk,

5522
04:33:37,718 --> 04:33:41,018
I've shown you how we can
use simultaneous PET and fMRI

5523
04:33:41,018 --> 04:33:44,228
to decipher the role of
subtype receptor contributions,

5524
04:33:44,228 --> 04:33:47,158
and to learn about their desensitization

5525
04:33:47,158 --> 04:33:49,301
and internalization mechanisms.

5526
04:33:50,688 --> 04:33:53,198
We've learned that PET decreases

5527
04:33:53,198 --> 04:33:55,998
can sometimes represent a combination

5528
04:33:55,998 --> 04:34:00,298
of dopamine release and D2/D3
receptor internalization.

5529
04:34:00,298 --> 04:34:02,078
And this is important to be aware of

5530
04:34:02,078 --> 04:34:06,098
when interpreting both
PET and fMRI signals.

5531
04:34:06,098 --> 04:34:08,648
The main striking finding
in this amphetamine study

5532
04:34:08,648 --> 04:34:10,818
was that after three hours,

5533
04:34:10,818 --> 04:34:13,211
receptor internalization
was still persistent,

5534
04:34:14,378 --> 04:34:17,598
but D1 receptors were again available

5535
04:34:17,598 --> 04:34:19,391
for functional reactivation.

5536
04:34:21,258 --> 04:34:25,588
Overall, this allowed us to
look at first in vivo evidence

5537
04:34:25,588 --> 04:34:27,018
for different recycling rates

5538
04:34:27,018 --> 04:34:29,288
between dopamine receptor subtypes.

5539
04:34:29,288 --> 04:34:32,688
And this was enabled by using
simultaneous PET and fMRI.

5540
04:34:34,268 --> 04:34:36,008
To summarize, I've shown
you that we can use

5541
04:34:36,008 --> 04:34:39,518
simultaneous PET and fMRI to
image neuroreceptor binding

5542
04:34:39,518 --> 04:34:42,488
together with brain function
in the living brain.

5543
04:34:42,488 --> 04:34:44,718
We can use pharmacological PET/MRI

5544
04:34:44,718 --> 04:34:47,688
to classify drugs in vivo,

5545
04:34:47,688 --> 04:34:50,618
and we can look at endogenous
neurotransmitter release

5546
04:34:50,618 --> 04:34:54,358
at various receptor
systems using PET and fMRI.

5547
04:34:54,358 --> 04:34:55,798
At the same time, it's important

5548
04:34:55,798 --> 04:34:58,898
to really understand
receptor quantification

5549
04:34:58,898 --> 04:35:01,778
and the sources of bias that come with it,

5550
04:35:01,778 --> 04:35:05,728
and how we can use PET/MRI to validate

5551
04:35:05,728 --> 04:35:09,378
or characterize some of
these sources of bias.

5552
04:35:09,378 --> 04:35:10,798
And finally, I've shown you

5553
04:35:10,798 --> 04:35:15,468
that PET/MRI can give us insight
into receptor adaptations

5554
04:35:15,468 --> 04:35:17,948
and further understand dopamine

5555
04:35:17,948 --> 04:35:20,511
and other receptor systems in more detail.

5556
04:35:22,388 --> 04:35:25,018
Overall, the value of simultaneous PET/MRI

5557
04:35:25,018 --> 04:35:27,218
really lies in integrating

5558
04:35:27,218 --> 04:35:30,378
systems biology and systems neuroscience

5559
04:35:30,378 --> 04:35:32,918
in order to understand
biological mechanisms

5560
04:35:32,918 --> 04:35:35,861
in vivo and at the whole
brain circuit level.

5561
04:35:36,708 --> 04:35:39,578
We can use novel MR and PET techniques

5562
04:35:39,578 --> 04:35:42,898
together with novel
pharmacological challenges,

5563
04:35:42,898 --> 04:35:45,598
as well as models and analysis approaches,

5564
04:35:45,598 --> 04:35:50,598
in order to probe receptor
systems and their interactions.

5565
04:35:51,188 --> 04:35:53,868
And I'd like to close with this slide,

5566
04:35:53,868 --> 04:35:58,038
showing that there's really
a world of opportunities here

5567
04:35:58,038 --> 04:36:02,478
that allow us to look at
different targets in PET imaging

5568
04:36:02,478 --> 04:36:04,558
for different receptor classes.

5569
04:36:04,558 --> 04:36:08,528
And in combination with
different contrasts of fMRI

5570
04:36:08,528 --> 04:36:11,521
or MR techniques in general,

5571
04:36:12,728 --> 04:36:16,418
we can really try to understand the brain

5572
04:36:16,418 --> 04:36:17,971
at a more holistic level.

5573
04:36:19,018 --> 04:36:20,428
Finally, I'd like to thank

5574
04:36:20,428 --> 04:36:23,468
the symposium organizers at the NIMH,

5575
04:36:23,468 --> 04:36:25,888
and I'd like to thank my collaborators

5576
04:36:25,888 --> 04:36:28,228
as well as the funding sources.

5577
04:36:28,228 --> 04:36:31,158
If you have any questions
about our studies,

5578
04:36:31,158 --> 04:36:32,878
please feel free to contact me,

5579
04:36:32,878 --> 04:36:34,648
and I'd be happy to discuss further

5580
04:36:34,648 --> 04:36:37,301
in the live Q&A session as well.

5581
04:36:43,771 --> 04:36:48,771
- Hello, my name is Dardo
Tomasi, and I am a physicist

5582
04:36:48,791 --> 04:36:50,941
with experience in PET and MRI

5583
04:36:51,861 --> 04:36:53,651
and a staff scientist at the NIH.

5584
04:36:54,541 --> 04:36:57,091
I will talk about how the speed

5585
04:36:57,091 --> 04:37:00,001
of dopamine increases
modulates brain function

5586
04:37:00,001 --> 04:37:04,201
and reward in humans, and
I will present results

5587
04:37:04,201 --> 04:37:07,074
from our simultaneous PET/fMRI study.

5588
04:37:10,091 --> 04:37:15,001
I will show you how we are
using simultaneous PET/fMRI

5589
04:37:15,001 --> 04:37:17,451
to assess dopamine neurotransmission,

5590
04:37:17,451 --> 04:37:19,134
brain function, and behavior.

5591
04:37:20,571 --> 04:37:23,021
The overarching goal of our study

5592
04:37:23,021 --> 04:37:25,831
was to assess the dynamic association

5593
04:37:25,831 --> 04:37:28,661
between dopamine D2 receptor occupancy

5594
04:37:28,661 --> 04:37:32,311
in the striatum meshed with
carbon 11-raclopride base

5595
04:37:33,261 --> 04:37:37,011
and brain activity inferred
by pharmacological fMRI

5596
04:37:37,011 --> 04:37:38,041
in the human brain

5597
04:37:39,201 --> 04:37:41,334
and also to assess the
relative sensitivity

5598
04:37:41,334 --> 04:37:44,841
and specificity of the
neurovascular coupling

5599
04:37:44,841 --> 04:37:47,211
for a slow drug delivery achieved

5600
04:37:47,211 --> 04:37:51,101
with oral administration
of methylphenidate versus

5601
04:37:51,101 --> 04:37:54,611
rapid drug delivery using intravenous

5602
04:37:54,611 --> 04:37:56,374
methylphenidate administration.

5603
04:37:57,831 --> 04:38:01,731
I will describe how we are
extracting the time course

5604
04:38:01,731 --> 04:38:05,191
of extracellular dopamine increases elicit

5605
04:38:05,191 --> 04:38:10,191
by methylphenidate from PET
data and how we are using it

5606
04:38:11,011 --> 04:38:15,201
for modeling pharmacodynamic
fMRI responses

5607
04:38:15,201 --> 04:38:17,611
to methylphenidate challenges.

5608
04:38:17,611 --> 04:38:21,031
And for linking the
methylphenidate related functional

5609
04:38:21,031 --> 04:38:24,651
connectivity changes to the
rate of dopamine increases

5610
04:38:24,651 --> 04:38:25,818
in a striatum.

5611
04:38:29,161 --> 04:38:32,241
We want understand why some routes

5612
04:38:32,241 --> 04:38:34,701
of drug administration are more likely

5613
04:38:34,701 --> 04:38:36,374
to be addictive than others.

5614
04:38:37,601 --> 04:38:40,271
Data from priority studies suggests

5615
04:38:40,271 --> 04:38:43,231
that drug pharmacokinetics and the route

5616
04:38:43,231 --> 04:38:45,844
of administration played
huge roles in this regard.

5617
04:38:47,481 --> 04:38:51,101
For instance, oral methylphenidate doses

5618
04:38:51,101 --> 04:38:54,231
are not addictive and
used in the treatment

5619
04:38:54,231 --> 04:38:56,031
of attention hyperactivity disorder.

5620
04:38:57,151 --> 04:39:00,418
However, intravenous
methylphenidate is addictive.

5621
04:39:02,131 --> 04:39:04,361
The routes of abuse are often taken

5622
04:39:04,361 --> 04:39:09,361
by intravenous, smoked,
oral, or intranasal routes.

5623
04:39:11,161 --> 04:39:16,001
The route of drug administration
determines both how fast

5624
04:39:16,001 --> 04:39:19,799
and how much drug reaches
the general circulation

5625
04:39:19,799 --> 04:39:21,604
and ultimately the brain.

5626
04:39:23,011 --> 04:39:25,571
The faster drug reaches the brain,

5627
04:39:25,571 --> 04:39:28,038
the more likely it will be addictive.

5628
04:39:30,591 --> 04:39:33,211
The figure on the left was extracted

5629
04:39:33,211 --> 04:39:35,161
from a review by Allain and colleagues.

5630
04:39:36,091 --> 04:39:39,581
It show pharmacokinetic profiles
of plasma cocaine levels

5631
04:39:39,581 --> 04:39:44,581
in humans for different
drug administration routes.

5632
04:39:46,521 --> 04:39:51,271
Plasma cocaine levels rise
sharply and decline rapidly

5633
04:39:51,271 --> 04:39:54,234
when cocaine is injected
intravenously or smoked.

5634
04:39:55,381 --> 04:39:58,941
In contrast, plasma cocaine levels rise

5635
04:39:58,941 --> 04:40:03,941
and decline more slowly
peaking one to two hours

5636
04:40:05,471 --> 04:40:08,654
following intranasal
or oral administration.

5637
04:40:10,451 --> 04:40:13,741
The figure on the right was
extracted from a review by

5638
04:40:13,741 --> 04:40:18,741
Volkow and Swanson and shows
in purple a time activity curve

5639
04:40:19,571 --> 04:40:23,611
demonstrating the first
uptake of cocaine in the brain

5640
04:40:23,611 --> 04:40:27,311
which peaks four to six minutes

5641
04:40:27,311 --> 04:40:32,311
after the intravenous
injection and in blue

5642
04:40:34,971 --> 04:40:39,441
the perceived high, which
parallels the fast cocaine uptake

5643
04:40:39,441 --> 04:40:40,524
in the striatum.

5644
04:40:44,231 --> 04:40:49,231
In products, faster cocaine
injection results in higher

5645
04:40:51,101 --> 04:40:55,521
mesolimbic c-fos expression
and metabolic activity

5646
04:40:55,521 --> 04:40:59,051
and also faster dopamine
transporter blockade

5647
04:40:59,051 --> 04:41:02,021
and more cocaine self-administration.

5648
04:41:02,021 --> 04:41:07,021
For instance, if cocaine was delivered

5649
04:41:07,231 --> 04:41:12,231
in over a 92nd period,
animals did not escalate their

5650
04:41:15,881 --> 04:41:18,051
cocaine self-administration as a function

5651
04:41:18,051 --> 04:41:19,714
of cession.

5652
04:41:21,361 --> 04:41:26,361
Differently, self-administration
increased gradually

5653
04:41:26,671 --> 04:41:30,744
if cocaine was administered
in five seconds periods.

5654
04:41:35,421 --> 04:41:39,451
In humans, however, we do
not know much about why

5655
04:41:39,451 --> 04:41:44,141
some routes of administrations
are more addictive

5656
04:41:44,141 --> 04:41:45,004
than others.

5657
04:41:45,991 --> 04:41:50,991
For instance, we do not
know which brain circuits

5658
04:41:51,301 --> 04:41:54,461
are sensitive to the
speed of dopamine release

5659
04:41:54,461 --> 04:41:58,601
and how this relates to
the subjective experience

5660
04:41:58,601 --> 04:42:03,351
of the drug, which seems
to be very relevant

5661
04:42:03,351 --> 04:42:05,894
for the misuse potential of a drug.

5662
04:42:07,671 --> 04:42:11,891
Drugs like cocaine or
methylphenidate block dopamine

5663
04:42:11,891 --> 04:42:15,801
and epinephrine
transporters in the synapse,

5664
04:42:16,891 --> 04:42:21,431
which cause some changes
in brain function that lead

5665
04:42:21,431 --> 04:42:23,854
to the behaviors that we care about.

5666
04:42:26,101 --> 04:42:28,692
But the truth is that we do not know much

5667
04:42:28,692 --> 04:42:30,284
about this medium part.

5668
04:42:31,841 --> 04:42:35,891
We want to uncover the brain
circuitry that is sensitive

5669
04:42:35,891 --> 04:42:37,791
to the speed of dopamine release

5670
04:42:37,791 --> 04:42:42,791
because we think this may
explain the subjective experience

5671
04:42:43,211 --> 04:42:44,044
of the drug.

5672
04:42:47,901 --> 04:42:50,341
These are some of the
questions we want to address

5673
04:42:50,341 --> 04:42:51,234
in our study.

5674
04:42:52,271 --> 04:42:55,881
We wanted to study the
dynamics of dopamine release

5675
04:42:55,881 --> 04:42:59,191
and identify brain
circuits that are sensitive

5676
04:42:59,191 --> 04:43:00,914
to drug administration route.

5677
04:43:02,001 --> 04:43:05,841
Also, we aim to study decoupling

5678
04:43:05,841 --> 04:43:09,891
between dopamine
neurotransmission, brain activation

5679
04:43:09,891 --> 04:43:13,691
and functional connectivity,
and the relations

5680
04:43:13,691 --> 04:43:15,871
with the subjective perception of reward

5681
04:43:15,871 --> 04:43:17,024
from methylphenidate.

5682
04:43:18,611 --> 04:43:21,491
We wonder if these associations differ

5683
04:43:21,491 --> 04:43:24,794
for intravenous and oral administrations.

5684
04:43:26,851 --> 04:43:30,561
Simultaneous PET/fMRI is
the ideal imaging tool

5685
04:43:30,561 --> 04:43:34,091
for this because PET can
measure molecular changes

5686
04:43:34,091 --> 04:43:36,161
with high sensitivity and specificity

5687
04:43:37,719 --> 04:43:39,521
and fMRI can assist brain function

5688
04:43:39,521 --> 04:43:41,564
with high spatial temporal resolution.

5689
04:43:42,410 --> 04:43:45,370
And also because PET
and MRI data collection

5690
04:43:45,370 --> 04:43:47,084
occurs simultaneously.

5691
04:43:50,841 --> 04:43:52,571
we just completed this study

5692
04:43:53,461 --> 04:43:55,394
and this is its timeline.

5693
04:43:56,691 --> 04:43:58,811
Subjects first received an oral pill

5694
04:43:59,691 --> 04:44:02,821
60 milligrams of
methylphenidate or placebo,

5695
04:44:02,821 --> 04:44:05,244
and had a baseline MRI.

5696
04:44:07,321 --> 04:44:12,291
Then carbon-11 raclopride,
the tracer was injected

5697
04:44:12,291 --> 04:44:14,204
30 minutes after the oral pill.

5698
04:44:15,418 --> 04:44:17,771
And the intravenous medication

5699
04:44:17,771 --> 04:44:22,771
either 25, 0.25 milligram per
kilogram of methylphenidate

5700
04:44:23,821 --> 04:44:27,764
or saline was injected 30
minutes after the tracer.

5701
04:44:29,810 --> 04:44:34,701
Simultaneous fMRI data was
collected uninterruptedly

5702
04:44:34,701 --> 04:44:39,701
for 90 minutes while subjects
rated their feelings of high

5703
04:44:40,261 --> 04:44:41,244
during the study.

5704
04:44:42,831 --> 04:44:47,831
We selected oral doses of
methylphenidate and IV doses

5705
04:44:50,421 --> 04:44:51,391
of methylphenidate that
left to roughly equivalent

5706
04:44:53,531 --> 04:44:55,774
levels of dopamine transport occupancy,

5707
04:44:57,081 --> 04:44:59,351
such that the main difference

5708
04:44:59,351 --> 04:45:04,081
between the IV and oral
conditions was the speed

5709
04:45:04,081 --> 04:45:07,714
not the magnitude of drug brain delivery.

5710
04:45:09,231 --> 04:45:12,951
The study had a
double-blind placebo control

5711
04:45:12,951 --> 04:45:14,861
within subjects design.

5712
04:45:14,861 --> 04:45:16,701
There were three sessions

5713
04:45:16,701 --> 04:45:20,234
on separate days for each participant.

5714
04:45:21,081 --> 04:45:24,651
The active dose was delivered orally

5715
04:45:24,651 --> 04:45:29,651
in one session and intravenously
in another session.

5716
04:45:30,131 --> 04:45:35,131
And additionally we had a
double placebo control session.

5717
04:45:35,501 --> 04:45:38,071
Session order was counterbalanced.

5718
04:45:38,071 --> 04:45:40,221
Participant and researchers were blind

5719
04:45:40,221 --> 04:45:42,434
to the nature of the medication.

5720
04:45:44,821 --> 04:45:48,481
We studied 20 healthy adults

5721
04:45:48,481 --> 04:45:52,901
from the D.C. area who had prior
experience with alcohol use

5722
04:45:52,901 --> 04:45:57,531
or stimulant drugs, but did
not have drug dependence.

5723
04:45:57,531 --> 04:46:00,511
Exclusion criteria included allergy to

5724
04:46:00,511 --> 04:46:04,061
methylphenidate,
cardiovascular abnormalities,

5725
04:46:04,061 --> 04:46:07,941
use of medications that could
interact with methylphenidate,

5726
04:46:07,941 --> 04:46:10,764
and positive drug toxicology test results.

5727
04:46:13,051 --> 04:46:16,641
For each PET study we collected data

5728
04:46:16,641 --> 04:46:21,531
in a Siemens biograph scanner
using this mode acquisition

5729
04:46:21,531 --> 04:46:24,941
after the bolas injection of 16 millicurie

5730
04:46:24,941 --> 04:46:26,384
of carbon 11-raclopride.

5731
04:46:28,411 --> 04:46:32,121
For PET imagery reconstruction
into 48-time windows

5732
04:46:32,121 --> 04:46:35,441
we used the three order subset expectation

5733
04:46:35,441 --> 04:46:39,034
maximization algorithm
within the Siemens platform.

5734
04:46:40,551 --> 04:46:44,481
Because this is a PET
MRI, not a PET CT scanner,

5735
04:46:44,481 --> 04:46:48,601
We used two echo UTE images

5736
04:46:48,601 --> 04:46:53,121
and a convolution network
to estimate the coefficients

5737
04:46:53,121 --> 04:46:54,564
for attenuation correction.

5738
04:46:56,481 --> 04:47:01,481
The standardized and take
values, SUVs, compute

5739
04:47:01,511 --> 04:47:04,551
and calculated after
normalization for body weight

5740
04:47:04,551 --> 04:47:07,463
and injected dose were
especially normalized

5741
04:47:07,463 --> 04:47:09,814
to MNI template.

5742
04:47:11,991 --> 04:47:14,491
We computed relative SUV
time series in a striatum

5743
04:47:16,501 --> 04:47:19,641
by normalizing each SUV volume

5744
04:47:19,641 --> 04:47:23,654
by its mean in the cerebellum
using the three for ROIs.

5745
04:47:28,531 --> 04:47:32,601
Our goal was to study dynamic associations

5746
04:47:32,601 --> 04:47:35,764
between stable dopamine
release and brain activity.

5747
04:47:36,731 --> 04:47:38,841
For this purpose, we are using the rate

5748
04:47:39,741 --> 04:47:43,151
of dopamine increases in a striatum

5749
04:47:43,151 --> 04:47:45,931
as a regressor to understand

5750
04:47:45,931 --> 04:47:49,751
if fMRI signals elicit by metal rate

5751
04:47:49,751 --> 04:47:51,624
using a general linear model.

5752
04:47:52,991 --> 04:47:57,151
We focus on the rate of dopamine increases

5753
04:47:57,151 --> 04:48:00,981
because faster dopamine
release could overwhelm

5754
04:48:00,981 --> 04:48:04,691
this slow neuro adaptation
of the dopamine system

5755
04:48:04,691 --> 04:48:07,624
and be associated with
drug-related euphoria.

5756
04:48:09,021 --> 04:48:12,451
Overall, our approach
is analogous to the use

5757
04:48:12,451 --> 04:48:14,411
of thermodynamic response function

5758
04:48:14,411 --> 04:48:16,574
in a standard fMRI studies.

5759
04:48:18,801 --> 04:48:23,201
Now changes in binding potential following

5760
04:48:23,201 --> 04:48:25,271
a pharmacological challenge

5761
04:48:25,271 --> 04:48:27,591
with the methylphenidate transporter block

5762
04:48:27,591 --> 04:48:30,331
like methylphenidate or cocaine

5763
04:48:30,331 --> 04:48:33,164
are frequently interpreted
as dopamine release.

5764
04:48:34,621 --> 04:48:39,051
We found that changes in a
standardized uptake value

5765
04:48:39,051 --> 04:48:43,971
ranges between placebo and
methylphenidate conditions

5766
04:48:44,861 --> 04:48:49,758
were highly correlated
across subjects independently

5767
04:48:50,821 --> 04:48:53,951
for IV and oral conditions

5768
04:48:56,571 --> 04:48:59,804
with the corresponding
changes in binding potential.

5769
04:49:01,612 --> 04:49:04,261
This suggests that
methylphenidate related changes

5770
04:49:04,261 --> 04:49:08,431
in SUVr could be interpreted

5771
04:49:08,431 --> 04:49:12,830
as dopamine increases
average dopamine increases

5772
04:49:12,830 --> 04:49:14,247
during the study.

5773
04:49:15,081 --> 04:49:17,291
This made us think of the difference

5774
04:49:17,291 --> 04:49:22,291
in SUVr between placebo and
methylphenidate condition

5775
04:49:22,921 --> 04:49:26,764
as a dynamic marker of dopamine increases.

5776
04:49:27,951 --> 04:49:31,811
We are testing this
hypothesis using simulations

5777
04:49:31,811 --> 04:49:35,434
of raclopride binding with
and without methylphenidate.

5778
04:49:37,431 --> 04:49:42,431
Specifically, we are
simulating the SUVr difference

5779
04:49:42,621 --> 04:49:45,721
between placebo and
methylphenidate using Alpert's

5780
04:49:46,841 --> 04:49:51,391
linear extension of the
simplified reference model.

5781
04:49:51,391 --> 04:49:55,031
We are using a gamma feed, the red curve

5782
04:49:55,031 --> 04:50:00,031
in this plot as a surrogate
for the dopamine increases

5783
04:50:02,051 --> 04:50:03,301
which is the black curve.

5784
04:50:04,371 --> 04:50:08,994
We quantify the speed of
the SUVr changes by the time

5785
04:50:12,021 --> 04:50:16,951
to peak of the dopamine
rate, which is the blue curve

5786
04:50:18,191 --> 04:50:19,894
in this plot.

5787
04:50:23,351 --> 04:50:26,621
We interpret the time varying SUVr changes

5788
04:50:26,621 --> 04:50:31,411
in terms of extracellular dopamine changes

5789
04:50:31,411 --> 04:50:35,101
which reflect accumulation
of dopamine released

5790
04:50:35,101 --> 04:50:38,751
due to the dopamine transporter
blockade the methylphenidate

5791
04:50:39,761 --> 04:50:44,761
and also dopamine clearance via
block dopamine transporters.

5792
04:50:47,990 --> 04:50:52,328
In simulations, including
noise, and also within subjects

5793
04:50:54,671 --> 04:50:59,671
physiological variability, we
found that increasingly lower

5794
04:51:00,901 --> 04:51:04,811
the methylphenidate doses were associated

5795
04:51:04,811 --> 04:51:09,811
with lower dopamine
increases the red curve here

5796
04:51:11,791 --> 04:51:14,284
and a smaller SUVr changes.

5797
04:51:15,691 --> 04:51:20,071
And this suggests that Delta
SUVr is highly sensitive to

5798
04:51:20,071 --> 04:51:22,371
dopamine increases induced
by methylphenidate.

5799
04:51:26,601 --> 04:51:30,091
In simulations, we also found that errors

5800
04:51:30,091 --> 04:51:33,574
in protocol execution may
not be a serious problem.

5801
04:51:35,421 --> 04:51:39,031
However, variability across individuals

5802
04:51:39,031 --> 04:51:41,981
in raclopride kinetic
parameters could result

5803
04:51:41,981 --> 04:51:44,794
in significant variants in time to peak.

5804
04:51:46,701 --> 04:51:51,171
Our simulations also suggest
that time to peak estimated

5805
04:51:51,171 --> 04:51:56,171
from SUVr changes is highly
sensitive to the time to peak

5806
04:51:57,661 --> 04:52:01,438
of the true rate of
dopamine increases produced

5807
04:52:01,438 --> 04:52:02,624
by methylphenidate.

5808
04:52:06,691 --> 04:52:09,274
Now, this is our real data.

5809
04:52:10,311 --> 04:52:14,904
These figure show the
average SUVr difference

5810
04:52:18,251 --> 04:52:21,901
between placebo and
methylphenidate experiments

5811
04:52:21,901 --> 04:52:25,264
across the 20 participants in our study.

5812
04:52:26,651 --> 04:52:28,691
The figures show good agreement

5813
04:52:28,691 --> 04:52:33,390
between the gamma feeds
and the SUVr changes

5814
04:52:33,390 --> 04:52:38,381
across individuals independently for IV

5815
04:52:38,381 --> 04:52:40,264
and oral methylphenidate.

5816
04:52:41,781 --> 04:52:46,781
This approach allowed us to
estimate the average dopamine

5817
04:52:47,241 --> 04:52:51,831
grade time courses we are
using for the analysis

5818
04:52:51,831 --> 04:52:56,451
of the fMRI signals as well as individual

5819
04:52:56,451 --> 04:52:58,231
time-to-peak values

5820
04:52:59,531 --> 04:53:02,894
which were of course shorter for IV

5821
04:53:02,894 --> 04:53:04,964
than for oral methylphenidate.

5822
04:53:08,904 --> 04:53:10,754
And we are observing
an association between

5823
04:53:10,754 --> 04:53:12,344
time-to-peak and reward.

5824
04:53:14,261 --> 04:53:17,731
During the scans participants
use a bottom box to

5825
04:53:17,731 --> 04:53:21,159
record their responses
to the question how high

5826
04:53:21,159 --> 04:53:25,234
you feel right now in
a scale from one to 10?

5827
04:53:27,141 --> 04:53:31,191
The figure on the right shows that shorter

5828
04:53:31,191 --> 04:53:36,191
time-to-peak estimated
from individual feeds

5829
04:53:36,331 --> 04:53:39,824
was associated with more
intense feelings of high,

5830
04:53:41,301 --> 04:53:44,864
both for IV and for oral methylphenidate.

5831
04:53:45,901 --> 04:53:48,971
We think that this could be
important to understand why

5832
04:53:48,971 --> 04:53:52,181
some routes of drug
administration are more addictive

5833
04:53:52,181 --> 04:53:53,024
than others.

5834
04:53:56,831 --> 04:54:01,831
From the simple noninvasive
approach to assess the dynamics

5835
04:54:01,851 --> 04:54:04,201
of dopamine increases with methylphenidate

5836
04:54:04,201 --> 04:54:09,201
we learn that time value SUVr change

5837
04:54:09,441 --> 04:54:12,631
induced by methylphenidate
reflects the time course

5838
04:54:12,631 --> 04:54:16,961
of extracellular dopamine increases.

5839
04:54:16,961 --> 04:54:21,111
Modeling the time course
of SUVr allowed us to

5840
04:54:21,111 --> 04:54:25,221
assess the rate of dopamine
increases and its time-to-peak

5841
04:54:26,471 --> 04:54:28,741
which was associated with the high

5842
04:54:28,741 --> 04:54:30,344
elicited by methylphenidate.

5843
04:54:32,231 --> 04:54:35,851
In the next slides I will
describe how we are using

5844
04:54:35,851 --> 04:54:39,821
the average rate of dopamine
increases in containment

5845
04:54:39,821 --> 04:54:44,821
as a regressor to understand
fMRI signals elicit

5846
04:54:47,191 --> 04:54:49,764
by methylphenidate using
the general linear.

5847
04:54:54,591 --> 04:54:57,851
Only one group has
investigated the association

5848
04:54:57,851 --> 04:55:00,691
between dopamine activity
and brain activation

5849
04:55:00,691 --> 04:55:05,691
using simultaneous MRI with
pharmacological challenges.

5850
04:55:07,041 --> 04:55:10,854
These studies were carried
out in monkeys, not in humans.

5851
04:55:12,101 --> 04:55:15,531
The MGH group did this
first study to assess

5852
04:55:15,531 --> 04:55:18,381
the neurovascular coupling
to dopamine receptor

5853
04:55:18,381 --> 04:55:23,078
occupancy using raclopride
pharmacological challenge.

5854
04:55:25,211 --> 04:55:29,111
And the study shows a dynamic association

5855
04:55:29,111 --> 04:55:32,841
between receptor occupancy using

5856
04:55:35,033 --> 04:55:37,051
raclopride pharmacological challenge

5857
04:55:38,248 --> 04:55:42,024
and cerebral blood volume in a striatum.

5858
04:55:42,941 --> 04:55:46,281
This study, however, was
limited by anesthesia

5859
04:55:46,281 --> 04:55:48,731
and cannot be repeated
in humans due to the

5860
04:55:48,731 --> 04:55:51,681
toxicity of the nanoparticles
used to push the fMRI signal.

5861
04:55:56,641 --> 04:56:00,451
Simultaneous fMRI can be
used to assess effects

5862
04:56:00,451 --> 04:56:04,661
of drug administration
route on brain activation.

5863
04:56:04,661 --> 04:56:08,781
Specifically, we wonder if
oral NIV methylphenidate

5864
04:56:08,781 --> 04:56:11,564
activate different brain circuits.

5865
04:56:13,571 --> 04:56:16,881
We hypothesize that
methylphenidate-related fMRI

5866
04:56:16,881 --> 04:56:20,131
signal changes would
correlate with the rate

5867
04:56:20,131 --> 04:56:23,611
of dopamine increases in
striatum as a function of time

5868
04:56:23,611 --> 04:56:28,254
and that this would differ
between oral and IV routes.

5869
04:56:32,671 --> 04:56:36,541
As I said simultaneously,
with PET acquisition

5870
04:56:36,541 --> 04:56:41,001
we collected fMRI data using a 12-channel

5871
04:56:41,001 --> 04:56:45,861
radiofrequency coil for fMRI.

5872
04:56:45,861 --> 04:56:49,091
We monitor blood pressure, heart rate,

5873
04:56:49,091 --> 04:56:52,291
blood oxygen level during
simultaneous PET MRI

5874
04:56:52,291 --> 04:56:54,532
for safety reasons.

5875
04:56:54,532 --> 04:56:58,961
And we collected one millimeter
isotropic MRI structure

5876
04:56:58,961 --> 04:57:01,281
for free surface analysis.

5877
04:57:01,281 --> 04:57:06,281
And for bolus MRI we collected
1800 volumes at three seconds

5878
04:57:08,081 --> 04:57:12,201
temporal resolution using
standard whole brain

5879
04:57:12,201 --> 04:57:13,151
ecoplainer imaging.

5880
04:57:14,551 --> 04:57:18,541
The pipelines of the human
connections project were used

5881
04:57:20,351 --> 04:57:22,011
for image per processing

5882
04:57:22,011 --> 04:57:27,011
and aCompCor was used to
remove artifacts associated

5883
04:57:27,601 --> 04:57:30,834
with physiological
signals and head motion.

5884
04:57:32,441 --> 04:57:37,141
Linear regression was used
to access the association

5885
04:57:37,141 --> 04:57:42,141
between both signals of
dopamine rate extracted

5886
04:57:42,811 --> 04:57:44,294
from PET data.

5887
04:57:48,731 --> 04:57:52,421
To identify brain regions
that were stimulated

5888
04:57:52,421 --> 04:57:55,121
by the speed of dopamine release,

5889
04:57:55,121 --> 04:57:58,822
as I said, Pete Manza is using
the rate of extracellular

5890
04:57:58,822 --> 04:58:00,428
dopamine increases.

5891
04:58:00,428 --> 04:58:05,428
And has a regressor for a
statistical paramedic mapping.

5892
04:58:05,611 --> 04:58:10,611
And he found that intravenous
but not oral methylphenidate

5893
04:58:11,291 --> 04:58:14,151
strongly activated the insulin.

5894
04:58:16,821 --> 04:58:21,821
And also the MP singular
here, you see the 10 courses

5895
04:58:24,091 --> 04:58:27,024
which are the central notes
of the resiliency network.

5896
04:58:30,231 --> 04:58:34,861
He also found that both for
oral and IV methylphenidate

5897
04:58:37,231 --> 04:58:38,771
there was a deactivation

5898
04:58:40,302 --> 04:58:43,971
in the medial prefrontal
cortex, which is hyperactive

5899
04:58:43,971 --> 04:58:46,944
in individuals with cocaine use disorder.

5900
04:58:50,311 --> 04:58:54,878
Similarly, Pete found that
fMRI responses in ventral

5901
04:58:56,451 --> 04:59:00,321
striatum decreased and also
dorsal is there to increase

5902
04:59:00,321 --> 04:59:02,601
after intravenous methylphenidate.

5903
04:59:02,601 --> 04:59:05,081
But the signals in these
regions were weaker

5904
04:59:05,081 --> 04:59:08,811
than those in cortical
regions where fMRI signal

5905
04:59:08,811 --> 04:59:10,284
to noise is higher.

5906
04:59:12,641 --> 04:59:17,461
To summarize, the rate of
extracellular dopamine increases

5907
04:59:17,461 --> 04:59:20,071
in use by methylphenidate in striatum,

5908
04:59:20,071 --> 04:59:24,441
modulated the amplitude
of fMRI signals in dorals

5909
04:59:24,441 --> 04:59:27,471
anterior simulate insula striatum

5910
04:59:27,471 --> 04:59:30,541
and intermediate prefrontal cortex.

5911
04:59:30,541 --> 04:59:32,761
Our results are consistent with those

5912
04:59:32,761 --> 04:59:37,589
from prior PET MRI studies
with nanoparticles in monkeys

5913
04:59:37,589 --> 04:59:41,141
that show that pharmacological doses

5914
04:59:41,141 --> 04:59:44,404
of raclopride can increase
the blood volume in striatum.

5915
04:59:46,131 --> 04:59:50,511
We found in all subjects that
time varying fMRI responses

5916
04:59:50,511 --> 04:59:52,741
in cortical regions were coupled

5917
04:59:52,741 --> 04:59:56,561
to the rate of extracellular
dopamine increases in use

5918
04:59:56,561 --> 04:59:59,521
by methylphenidate in striatum.

5919
04:59:59,521 --> 05:00:00,941
These findings are consistent

5920
05:00:00,941 --> 05:00:04,261
with our hypothesis that
methylphenidate-related

5921
05:00:04,261 --> 05:00:07,391
fMRI signal changes in
the salience network

5922
05:00:07,391 --> 05:00:11,371
and limbic regions would
correlate with the rate

5923
05:00:11,371 --> 05:00:13,851
of dopamine increases in striatum,

5924
05:00:13,851 --> 05:00:17,071
and that piece effects
would differ between oral

5925
05:00:17,071 --> 05:00:18,224
and IV methylphenidate.

5926
05:00:20,451 --> 05:00:24,671
This shows how simultaneous
PET MRI can be used to

5927
05:00:24,671 --> 05:00:28,824
assess the influence of drugs
of abuse on brain activity.

5928
05:00:33,791 --> 05:00:38,351
This slide highlights the only
pharmacological PET MRI study

5929
05:00:38,351 --> 05:00:40,268
on functional connectivity.

5930
05:00:41,121 --> 05:00:45,531
This preclinical study by
Rasmus Birn and colleagues

5931
05:00:45,531 --> 05:00:48,981
from University of Wisconsin Madison

5932
05:00:48,981 --> 05:00:52,401
investigated the effect of methylphenidate

5933
05:00:52,401 --> 05:00:55,861
on functional connectivity
on three awakes monkeys

5934
05:00:57,431 --> 05:00:59,294
using the Fallypride tracer.

5935
05:01:00,891 --> 05:01:05,271
The study reports an association
between methylphenidate

5936
05:01:05,271 --> 05:01:08,711
related changes in binding
potential in striatum

5937
05:01:10,561 --> 05:01:15,561
and increase cortical
functional connectivity

5938
05:01:17,541 --> 05:01:18,744
in these monkeys.

5939
05:01:21,801 --> 05:01:24,931
To study the association
between a spontaneous

5940
05:01:24,931 --> 05:01:28,291
brain activity in the low frequency band

5941
05:01:29,461 --> 05:01:33,328
and the rate of dopamine
increases in striatum,

5942
05:01:36,201 --> 05:01:40,791
we first bypass filter the
pre-processed time series to

5943
05:01:42,301 --> 05:01:46,841
remove ultra low frequency components

5944
05:01:46,841 --> 05:01:50,871
of methylphenidate effects
on both fMRI signals.

5945
05:01:54,851 --> 05:01:57,621
Then we created time series of voxelwise

5946
05:02:02,061 --> 05:02:07,061
voxel correlations using the
cluster that demonstrated

5947
05:02:07,831 --> 05:02:12,801
significant association with
dopamine rate using a sliding

5948
05:02:12,801 --> 05:02:14,124
window approach.

5949
05:02:15,791 --> 05:02:19,641
Based on our findings on the activation

5950
05:02:19,641 --> 05:02:22,701
of the resiliency network and the findings

5951
05:02:22,701 --> 05:02:26,421
of increased stato cortical
connectivity in monkeys

5952
05:02:26,421 --> 05:02:28,568
after IV methylphenidate,

5953
05:02:30,035 --> 05:02:32,691
we hypothesized that IV methylphenidate

5954
05:02:32,691 --> 05:02:36,411
would increase functional
connectivity between the

5955
05:02:36,411 --> 05:02:38,914
caldate and the anterior singular.

5956
05:02:40,881 --> 05:02:43,831
When analyzing this data, Pete Manza

5957
05:02:43,831 --> 05:02:48,451
found that IV methylphenidate
indeed increased significantly

5958
05:02:48,451 --> 05:02:52,701
the connectivity between the covate

5959
05:02:52,701 --> 05:02:57,121
and the anterior singulum
which is consistent

5960
05:02:57,121 --> 05:03:00,051
with the findings in awake
monkeys from Rasmus Birn

5961
05:03:00,051 --> 05:03:00,884
and colleagues.

5962
05:03:01,871 --> 05:03:03,861
Differently, oral methylphenidate

5963
05:03:03,861 --> 05:03:07,201
did not change the functional connectivity

5964
05:03:07,201 --> 05:03:10,444
within the dorsal striatum
and the anterior singulum.

5965
05:03:13,991 --> 05:03:18,581
Simultaneously MRI can also
be used to study the effect

5966
05:03:18,581 --> 05:03:20,451
of drug administration route

5967
05:03:20,451 --> 05:03:22,564
on a spontaneous brain activity.

5968
05:03:23,821 --> 05:03:26,341
Specifically, we wondered if oral and

5969
05:03:26,341 --> 05:03:27,911
intravenous administration

5970
05:03:27,911 --> 05:03:31,811
of methylphenidate would
alter the fractional amplitude

5971
05:03:31,811 --> 05:03:34,424
of low-frequency
fluctuations in the brain.

5972
05:03:35,661 --> 05:03:40,431
We hypothesized that oral
methylphenidate would stimulate

5973
05:03:40,431 --> 05:03:44,634
inhibitory D2 receptors and
reduce spontaneous activity.

5974
05:03:45,931 --> 05:03:49,531
In this case, we use the
sliding window approach to

5975
05:03:49,531 --> 05:03:53,301
compute FR fractional
amplitude of low-frequency

5976
05:03:53,301 --> 05:03:58,301
fluctuations to which we
fitted our dopamine regressor.

5977
05:04:05,371 --> 05:04:07,761
To identify brain regions
in which methylphenidate

5978
05:04:07,761 --> 05:04:10,331
alters spontaneous brain activity

5979
05:04:10,331 --> 05:04:12,831
we compare the fractional amplitude

5980
05:04:12,831 --> 05:04:17,131
of low-frequency fluctuations
in the last 30 minutes

5981
05:04:17,131 --> 05:04:20,404
against that of the first
30 minutes of this study.

5982
05:04:21,461 --> 05:04:23,891
We found that intravenous methylphenidate

5983
05:04:23,891 --> 05:04:28,891
strongly reduced spontaneous
brain activity in

5984
05:04:29,691 --> 05:04:34,471
occipital, parietal, and
and temporal cortices

5985
05:04:34,471 --> 05:04:36,434
and also in the motor strength.

5986
05:04:39,531 --> 05:04:43,051
Also for oral methylphenidate

5987
05:04:43,051 --> 05:04:46,811
the decreases were weaker than those

5988
05:04:46,811 --> 05:04:48,411
for intravenous methylphenidate.

5989
05:04:54,161 --> 05:04:57,481
This spatial pattern of
decreases in a spontaneous brain

5990
05:04:57,481 --> 05:05:02,161
activity is consistent with
the distribution of a straight

5991
05:05:02,161 --> 05:05:05,501
of dopamine increases,
which was not homogeneous

5992
05:05:05,501 --> 05:05:06,801
in the brain.

5993
05:05:06,801 --> 05:05:09,931
Specifically, dopamine
increases were larger

5994
05:05:09,931 --> 05:05:13,551
in posterior regions
of the dorsal striatum

5995
05:05:13,551 --> 05:05:18,536
than in anterior, dorsal or
ventral striatum regions.

5996
05:05:18,536 --> 05:05:21,309
These posterior regions
of the dorsal striatum

5997
05:05:21,309 --> 05:05:25,644
are structurally connected
to posterior brain regions

5998
05:05:25,644 --> 05:05:28,865
that show a significant
decrease in the spontaneous

5999
05:05:28,865 --> 05:05:32,321
brain activity following
intravenous administration

6000
05:05:32,321 --> 05:05:34,071
in the present study.

6001
05:05:35,565 --> 05:05:38,853
Differently, anterior,
dorsal or ventral regions

6002
05:05:38,853 --> 05:05:42,741
of the striatum which
did not show significant

6003
05:05:42,741 --> 05:05:46,236
dopamine increases in
our study are connected

6004
05:05:46,236 --> 05:05:49,741
to anterior brain
regions that did not show

6005
05:05:49,741 --> 05:05:52,684
significant changes in a
spontaneous brain activity.

6006
05:05:53,961 --> 05:05:57,069
Therefore, the spatial
distribution of decrease

6007
05:05:57,069 --> 05:06:01,861
in cortical activity with
methylphenidate is consistent

6008
05:06:01,861 --> 05:06:04,581
with the spatial distribution of changes

6009
05:06:04,581 --> 05:06:08,791
in binding potential in striatum
caused by methylphenidate.

6010
05:06:11,321 --> 05:06:13,831
To assess the role of
dopamine in the reduction

6011
05:06:13,831 --> 05:06:18,101
of brain activity we map
the statistical significance

6012
05:06:18,101 --> 05:06:22,741
for the slope of the linear association

6013
05:06:22,741 --> 05:06:26,921
between the dynamic fractional amplitude

6014
05:06:26,921 --> 05:06:30,031
of low-frequency
fluctuations and the dopamine

6015
05:06:30,031 --> 05:06:35,031
rate regressor extracted from PET data.

6016
05:06:35,361 --> 05:06:38,691
We found a linear
association between the rate

6017
05:06:38,691 --> 05:06:41,381
of dopamine increases in the reduction

6018
05:06:41,381 --> 05:06:45,434
of spontaneous brain
activity in occipital,

6019
05:06:45,434 --> 05:06:47,158
parietal, and temporal core

6020
05:06:47,158 --> 05:06:50,094
and in the motor acro subjects.

6021
05:06:53,961 --> 05:06:57,878
Plots show for seven
traditional brain networks

6022
05:07:00,621 --> 05:07:03,951
the negative correlation
between the fractional amplitude

6023
05:07:03,951 --> 05:07:06,701
of low-frequency fluctuations and the rate

6024
05:07:06,701 --> 05:07:07,944
of dopamine increases.

6025
05:07:09,151 --> 05:07:12,071
The associations between dopamine rate

6026
05:07:12,071 --> 05:07:15,401
and the fractional amplitude
of low-frequency fluctuations

6027
05:07:15,401 --> 05:07:19,901
differentiated the networks
for oral and IV routes.

6028
05:07:19,901 --> 05:07:24,901
For instance, in somatomotor
cortex, the linear association

6029
05:07:26,031 --> 05:07:30,737
between dopamine rate and
fALFF was very different

6030
05:07:30,737 --> 05:07:33,764
for oral and IV methylphenidate.

6031
05:07:36,961 --> 05:07:40,051
We also found for the
first time an association

6032
05:07:40,051 --> 05:07:43,869
between peak high ratings and
a spontaneous brain activity

6033
05:07:43,869 --> 05:07:45,536
across participants.

6034
05:07:47,411 --> 05:07:51,441
Such that the lower the amplitude

6035
05:07:51,441 --> 05:07:55,511
of low-frequency fluctuations
in sensory motorcortices

6036
05:07:55,511 --> 05:07:59,494
the greater the peak high ratings.

6037
05:08:00,531 --> 05:08:03,611
This association was
maximal in the cuneus.

6038
05:08:04,981 --> 05:08:09,031
This is very interesting and
intriguing to us because we see

6039
05:08:09,031 --> 05:08:12,311
for the first time an
association between high ratings

6040
05:08:12,311 --> 05:08:17,034
and brain responses to
methylphenidate outside the striatum.

6041
05:08:18,801 --> 05:08:21,881
In summary, we found that methylphenidate

6042
05:08:21,881 --> 05:08:24,141
increased functional connectivity

6043
05:08:24,141 --> 05:08:28,251
between the striatum and the
resiliency network consistent

6044
05:08:28,251 --> 05:08:31,621
with prior studies in awake monkeys.

6045
05:08:31,621 --> 05:08:34,621
Methylphenidate reduces
spontaneous brain activity

6046
05:08:34,621 --> 05:08:37,601
in sensorimotor cortices
that are structurally

6047
05:08:37,601 --> 05:08:41,541
connected to posterior
regions of the striatum where

6048
05:08:41,541 --> 05:08:45,684
methylphenidate induced
significant dopamine released.

6049
05:08:46,821 --> 05:08:49,971
We also found that a spontaneous activity

6050
05:08:49,971 --> 05:08:54,351
in this cortical regions
was coupled to dopamine

6051
05:08:54,351 --> 05:08:57,781
increases induced by
methylphenidate in striatum

6052
05:08:57,781 --> 05:08:59,574
as a function of time.

6053
05:09:00,611 --> 05:09:03,011
These findings are consistent
with our hypothesis

6054
05:09:03,011 --> 05:09:07,491
that dopamine increases
will stimulate inhibitor

6055
05:09:07,491 --> 05:09:12,491
the two receptors and weight
spontaneous brain activity.

6056
05:09:12,896 --> 05:09:16,041
And with the notion that
methylphenidate enhances alertness

6057
05:09:16,041 --> 05:09:19,371
and vigilance which should decrease

6058
05:09:19,371 --> 05:09:21,724
low-frequency fluctuations in the brain.

6059
05:09:22,721 --> 05:09:25,851
This is also consistent
with the role of dopamine

6060
05:09:25,851 --> 05:09:27,451
in the promotion of wakefulness.

6061
05:09:29,211 --> 05:09:33,221
Overall, this study also
shows how simultaneous

6062
05:09:33,221 --> 05:09:36,161
PET/fMRI can be used
to study the influence

6063
05:09:36,161 --> 05:09:38,694
of drugs of abuse on
functional connectivity.

6064
05:09:41,401 --> 05:09:45,001
Our old Siemens biographic
scanner is now being replaced

6065
05:09:45,001 --> 05:09:49,731
by a new GE PET MRI scanner
in the clinical center.

6066
05:09:49,731 --> 05:09:53,301
And in the future we will use
this scanner to investigate

6067
05:09:53,301 --> 05:09:56,371
further the effect of the
rate of dopamine increases

6068
05:09:56,371 --> 05:09:59,351
on brain activation,
functional connectivity,

6069
05:09:59,351 --> 05:10:01,791
and behavior using slow and fast

6070
05:10:01,791 --> 05:10:04,224
intravenous methylphenidate infusions.

6071
05:10:06,811 --> 05:10:10,671
I want to end by thinking
everyone's contribution

6072
05:10:10,671 --> 05:10:15,244
to this study, especially
those of Pete and Nora.

6073
05:10:16,411 --> 05:10:19,801
This study was one of
the most complex studies

6074
05:10:19,801 --> 05:10:21,594
we have done at the NIH.

6075
05:10:23,221 --> 05:10:25,561
Our study wouldn't have been possible

6076
05:10:25,561 --> 05:10:30,561
without the effort and smart
contributions from post docs,

6077
05:10:31,901 --> 05:10:35,011
post bacs, and staff from the Laboratory

6078
05:10:35,011 --> 05:10:39,281
of Neuroimaging and the
significant contributions

6079
05:10:39,281 --> 05:10:42,221
from NIH Clinical Center personnel

6080
05:10:42,221 --> 05:10:45,751
and collaborators from other institutions.

6081
05:10:45,751 --> 05:10:47,401
Thank you very much for watching.