Yale Psychiatry Grand Rounds: October 27, 2023

October 27, 2023

Information

"Brain Circuit Mapping and Modulation to Optimize Brain Stimulation Therapies in Psychiatry"

Speaker: Desmond Oathes, PhD, Associate Professor of Psychiatry, Perelman School of Medicine, University of Pennsylvania

ID10913

To CiteDCA Citation Guide

00:00Be here and share my work with you guys.
00:04Let's see. Does that still look OK?
00:11It does, yes. All right, Great.
00:13Thank you. So I think we'll start out
00:18with some problems and hurdles and
00:21the neuroimaging field in psychiatry,
00:23I think this is probably relevant if we
00:27think of the clinicians in the audience
00:30deciding whether there's ever going
00:32to be any horizon in which imaging is
00:35actually useful in their clinical practice.
00:38I would argue that it isn't typically.
00:41And so some of the hurdles and
00:45and problems in the field include
00:49a lot of various issues, right.
00:52Some of them have to do with really finding
00:56no clear neurobiological evidence that
00:59you know fits with the DSM categories.
01:02We have correlations with symptoms
01:04and other behavioral scales tend
01:07to be difficult to replicate.
01:11We don't use imaging and clinical
01:14decision making on the the reliability
01:16of many of the imaging measures we
01:19use are suspect and need improvement.
01:22So we have all these recent publications
01:25right in the last few years that
01:27that are really causing us to re
01:30evaluate what we're doing and and what
01:32kind of horizon we have for making
01:34imaging more useful in psychiatry.
01:39Even though as we are able to share more
01:41data with one another and try to look at
01:44big scale approaches with typically large
01:46and studies when you combine them this way.
01:50There have been some hits to finding
01:53biomarkers and biotypes in recent years,
01:57including this paper and many hundreds
01:59of patients finding minimal evidence for
02:02depression abnormality using structural MRI,
02:05DTI, task resting state,
02:07not being able to find a clear signature
02:11that hears our depression imaging marker.
02:15All right, so that's that's problematic.
02:18But this may be more familiar with the
02:21clinician for the clinicians who don't
02:23typically pay as much attention to imaging,
02:25which is that the diagnosis itself
02:28in a lot of cases is not optimal.
02:31And so if you feed in something that's
02:34kind of nebulous and not very precise
02:36and then you try to create a precise
02:39measurement of that with an imaging marker,
02:42of course you know there's there's
02:44going to be a a real difficulty there.
02:47We can't even agree amongst one another
02:50from clinician to clinician what the
02:53right diagnosis is for a patient.
02:56So these are hurdles.
02:57I don't have answers for all these,
02:59but I I feel like it's it's important
03:02to bring up some of the struggles
03:05and the challenges.
03:06I'll say on the neuroscience
03:08side with imaging.
03:09There are other issues when we think
03:12about making bridges to patients
03:15centered decision making.
03:16One of them is that you can have.
03:20So this is a paper by my
03:21friend John Medallia,
03:23who was saying that as neuroscientists,
03:26we have these average brains and we've
03:29all seen pictures of these and they
03:31have features that in aggregate have
03:33never been observed in any single patient.
03:36And and so that's problematic if
03:38you're looking at an average brain
03:40image and you're thinking about,
03:42oh,
03:42OK,
03:42how can I make the use of this for applying
03:45to this patient who's in front of me?
03:47This is problematic.
03:48Reinforcing this idea is a paper by
03:52Deanna Barch from many years ago,
03:55more than 10 years ago,
03:57and there have been other instances of this.
03:59On the left side you see something
04:02that's used very widely in
04:04cognitive neuroscience which is in
04:06designed to capture working memory,
04:08other attentional kind of factors.
04:10So this is an N back task where you have
04:13more working memory load compared to less.
04:16What areas pop up in the brain,
04:18which ones are strongly active And
04:20on the left, the left set of images
04:23are the average brain maps, right?
04:25This is what we normally report
04:27in my own work as well, right?
04:29This is what we usually show
04:30in an imaging experiment.
04:31This is the output.
04:32If on the other hand,
04:34instead of taking the average
04:35from the same contrast,
04:36if instead on the right side,
04:38you pay more attention to how
04:40many individuals in that group
04:42are showing strong activation,
04:44the map looks a little bit different there.
04:45There's some overlaps,
04:47but there's also some differences.
04:48If you look closely right,
04:50it's it's much more sparse.
04:51There's some areas that look
04:53like they have a lot more going
04:54on than on the left side maps.
04:56And I would argue something on
04:58the right side is more relevant
05:00to the individual patients.
05:02On the left side,
05:03especially with small end studies
05:04which are typical in imaging
05:06because it's so expensive.
05:07You can you can throw off the average
05:10map by having a few individuals
05:12showing lots of activation.
05:14Whereas on the right side we're
05:15probably looking for something
05:17that's very reliable in say a patient
05:19group and we want to know like is
05:21the typical patient going to show a
05:23bunch of activation in this spot.
05:24So these are ideas about forming
05:27bridges between what we normally do
05:29in imaging and thinking about how
05:32imaging can be applied more to individuals.
05:37Another thing to bring up since I'm
05:40doing TMS depression is there's a lot
05:43of excitement building especially
05:45from Nolan Williams work at Stanford
05:48that left led to an FDA approval
05:51for a new way of doing TMS for
05:54treatment resistant depression.
05:55And so we have the distressed patient
05:58or a we apply even a really amazing
06:02clinically effective stimulation
06:05protocol in studies seeing like 80%
06:09remission in treatment resistant depression.
06:11Obviously a really important tool
06:12right for for adding for that very
06:15ill patient group that doesn't
06:17respond to medication.
06:18So you do the stimulation protocol,
06:22you measure the treatment response.
06:23A bunch of the patients do well.
06:25Some of the patients don't change very much,
06:27some of the patients do worse.
06:29And you're left struggling saying,
06:31well, what do we do about that?
06:33What do we do about the patients who
06:35don't do well, The ones that do great,
06:37like, OK, problem solved,
06:38but what about for all the patients
06:41that don't do especially well?
06:42I would argue that you stimulated
06:46based on an imaging marker.
06:49You don't know what TMS actually did
06:52to that imaging marker and that may
06:54be critical in figuring out why patients,
06:57some patients don't respond.
06:58But if we don't do brain imaging,
07:01we don't do any brain based measurement,
07:03then it's gonna be really hard to
07:05unpack that and further refine
07:07the treatment and optimize it at
07:09the individual patient level.
07:13So we'll enter TMS, FM, RI
07:18where I think it's especially
07:20relevant and appropriate to think
07:23of how imaging may be relevant
07:25to the practice of psychiatry.
07:28We have this very straightforward
07:31brain based intervention with TMS.
07:34You might argue, oh,
07:35all of our interventions are brain based,
07:37but when it comes to making a
07:40very specific hypothesis about a
07:43particular brain area or circuit that
07:46you think is critical for patient
07:48alleviation of symptoms with TMS,
07:50you have to choose something, right?
07:52So really linking that brain area
07:54to a clinical outcome is very sort
07:57of required with TMS and I and I
07:59would argue since you have that that
08:02understanding or that background
08:04and the relevance of of the brain
08:07for this particular intervention,
08:08this may be the the most straightforward
08:13reasonable proving ground for putting
08:16imaging in a treatment context in
08:18psychiatry and showing that there is
08:21some utility of the imaging for for
08:25the actual treatment or intervention.
08:28All right.
08:29So more about TMSF MRI fMRI BOLD
08:32response takes a little while to
08:35really show a strong signal when
08:37you have some kind of psychological
08:40event which you know is is one of
08:42its shortcomings if you want to
08:44capture things moving really quickly.
08:46But it has a major advantage for me
08:49delivering pulses of TMS in the scanner,
08:52because I can send a pulse of TMS
08:54through the circuit and I can turn
08:57on the scanner without correcting
08:59the image and capture a really nice
09:02evokes response in the rest of the
09:05brain that follows from the causal
09:08stimulation through that pathway
09:10in a way that traditional imaging
09:13doesn't have within its toolbox.
09:16So we do that.
09:18We started this was work that I did
09:20with the media and back at Stanford said,
09:24all right, well,
09:24we have these canonical resting
09:26state networks.
09:26They're all based on correlations.
09:28Let's throw some little like causal
09:30pings into this situation by stimulating
09:33ostensible nodes of a resting state network.
09:37And we want to prove a couple
09:38different things.
09:39We want to say, well,
09:39if you hit one node of a
09:42network with TMS at a time,
09:45right?
09:45So if we ping that with a pulse of TMS,
09:48can we actually engage the network,
09:49can we do network level circuit engagement
09:52just by hitting one spot And we found
09:55evidence that we could in a couple
09:57of different task positive networks.
09:58So that's really reassuring.
10:00I suggest they we we are engaging
10:02networks even though we're stimulating
10:04a single brain area at a time.
10:06The other thing that we wanted
10:08to do had more to do with turning
10:10the correlations from resting
10:12state into more causal maps.
10:15And so we we sought to ping the task
10:17positive networks and based on the
10:19correlations in the past people thought OK,
10:22well there's this antagonistic
10:24relationship between the test
10:26positive networks and the DMN but
10:28it's not easy to causally test
10:30that non invasively in a human.
10:32So we pinned some of these
10:34test positive networks and
10:35looked at the evoke response in the
10:37default mode network and we supported the,
10:39you know the idea in the field that they had.
10:42There are some antagonistic
10:43relationships between these networks.
10:45The DMN turns off in response to
10:47a ping of a test positive network.
10:49So we're adding this causal argument to
10:52what's traditional been traditionally
10:54been just time series correlations.
10:59When I arrived at Penn about eight years
11:02ago and this priority to focus on some of
11:05the deep rain regions that we thought were
11:08most relevant for anxiety and depression,
11:11starting with the subtennial cingular
11:13cortex and the amygdala and said,
11:15well, these are deeper in the brain.
11:17You can't stimulate them directly with TMS,
11:19but through those these network approaches,
11:21can you stimulate one of the nodes in the
11:24cortical surface and show evidence that
11:26you can engage these deeper brain regions?
11:29And if so how do we how do we think
11:34that this happens at the circuit level?
11:37How do you kind of prioritize
11:39which brain areas to stimulate?
11:41And so we we collect baseline resting
11:44connectivity from individuals and
11:46we choose stimulation sites on the
11:49cortex and try stimulating them and
11:52evoking responses deeper in the brain
11:55what the imaging sequence looks like.
11:57We have these interleaved kind of gaps
11:59in between the F MRI recordings where
12:01we can put in a ping of the circuit
12:03and this is not neuromodulation,
12:05this is not repetitive TMS.
12:06This is just sending individual pings
12:09through that circuit a bunch of times just
12:12like any other task evoked brain response.
12:14It's also similar in our minds
12:17to motor evoke potential.
12:19So you're just engaging the circuit
12:21and the strength of engagement
12:23of the circuit is measured.
12:24Instead of in a finger twitch,
12:26it's measured in an fMRI BOLD response.
12:28But conceptually we see them very
12:31similarly that if the circuit is really
12:33intact for an individual through the this
12:36cortical node that we're stimulating,
12:39then your evoked response deeper in
12:41the brain should be especially strong.
12:43So that's how we measure it and what we
12:46capture is the whole brain response.
12:48So this this is not only like
12:52direct pathways,
12:53we're getting a bunch of these like
12:55downstream multi synaptic kind of responses
12:58that are downstream of where we stimulate.
13:01But you can still make a causal
13:04argument because we stimulated at
13:06a particular node cortically and
13:08generated these whole brain responses.
13:10I think we could learn a lot about
13:13how the signal kind of propagates and
13:16engages the brain from different places
13:19that we can stimulate on the surface.
13:23Also,
13:23just to say that imaging right
13:26has come a long way.
13:28There's a lot of different methods and tools,
13:30and when it comes to trying to stimulate
13:33a particular cortical location,
13:35there are a lot of variations
13:38that you could apply right you.
13:40You could choose a cortical target
13:43based on DTI, FM, RI, task, resting,
13:46ASL.
13:46Whatever your kind of pet measure is,
13:49you can look out for hypothesis about
13:52atlases and how the brain is organized
13:56into networks and you can test them
13:59like causally by picking these spots.
14:01So we will collect a a baseline
14:03imaging set of data, right?
14:05We put the patients in front of a
14:07camera and we mark some fiducial
14:09points on their scalp and then we can
14:12line up and find out exactly where
14:14we're stimulating relative to their
14:16brain. And you can also stick
14:18to your target really well by
14:20holding your TMS coil and getting
14:22this feedback from the camera on
14:24which brain area you're overlying
14:26while you do stimulation.
14:28So that's the neuro
14:30navigated part in the scanner
14:35not showing these data.
14:36But we did a smaller pilot study
14:38in 14 subjects where we looked at
14:41resting connectivity based pings
14:42and found subtennial and amygdala
14:44engagement through those pathways.
14:46I'll show you the replication 'cause
14:48they're in bigger cohorts and we
14:50explored a little bit more kind of
14:53evidence for which target is doing what.
14:56So this is in 32 healthy subjects.
14:59We did the resting fMRI
15:01guided stimulation right,
15:02based on the subgenual connectivity and
15:04we stimulated through those pathways.
15:06We had a control region and
15:08motor cortex and we say, hey,
15:10can we reliably ping this target
15:12in in through these circuits?
15:15And we found that there was evidence
15:17we could engage the subgenual singlet
15:20better than the control region,
15:22suggesting that there's some
15:24pathway specificity in choosing
15:26these individualized resting guided
15:27targets and that when we ping them,
15:29we can reliably engage that
15:31deeper brain region.
15:35All right. So another replication and
15:39extension that we tried is to say,
15:42well, all the clinical folks
15:46especially are looking at anti
15:48correlated brain stimulation targets.
15:50That's including the same
15:51protocol and there's pretty nice
15:53clinical evidence for that.
15:54You look for the subgenual negative
15:57functional connectivity partner on
15:59the brain service and you stimulate
16:01that clinically and show that
16:03there's a relationship between
16:04how patients do and the strength
16:06of connectivity to that pathway.
16:08So that that's really nice evidence.
16:11But we wanted to see if it's really
16:14important that you get the anti correlated
16:17spot or what actually happens if you
16:19look at a positively correlated spot.
16:21And there's some data from Corey
16:23Keller ET all doing some electrical
16:26stimulation and trying to map
16:28those networks from Reston State.
16:30And it looks like the positively correlated
16:33ones are a better fit for the stimulation,
16:36you know,
16:37related effects in the brain sort of thought,
16:39hey,
16:39what we should we should at least
16:41look into the positive connectivity
16:43spots and see what we get in terms of
16:46the evoked response in the subgenual.
16:48So we did that with our typical interleave,
16:50right with our single pulses,
16:51no neuromodulation,
16:52just pinging the circuit and we found
16:55that for healthy controls this is a
16:57replication again that the positive
17:00and negative connectivity spots
17:01engage the circuit pretty well.
17:03They they do pretty similarly to one another.
17:06So both of them are effective as long
17:09as you hit a high connectivity peak.
17:11It doesn't matter so much if it's anti
17:13correlated or positively correlated.
17:14They both seem to do pretty similar
17:18things and a bit smaller of a
17:21group of depressed patients.
17:22However,
17:23we found that there was a difference.
17:25The anti correlated spots still
17:28engaged the subgenual cingulant.
17:30So if the subgenual engagement is
17:33really critical for the antidepressant
17:36effects of TMS through that pathway,
17:38then this is consistent with that right.
17:39It it suggests that there is a real
17:42pathway there in depressed patients
17:44and perhaps that's why the treatments
17:46work through those pathways.
17:48But seeing that there's this difference,
17:50all right,
17:51there's like a significant difference
17:52in the strength of the evoked response
17:54depending on whether it's anti
17:55correlated or positively correlated.
17:57The positively correlated ones
17:58engage the circuit even more.
18:00So again,
18:01if we're really thinking
18:02that mechanistically,
18:03engagement of that subgenual
18:05through the cortical pathway is
18:07really clinically important,
18:09Why not start testing out the positively
18:11correlated spots and see if our
18:13clinical effects are even better?
18:19All right. So I'm going to switch
18:20over to the amygdala just briefly.
18:22We haven't done any interventions
18:24yet through the amygdala pathway,
18:25but we wanted to explore a little bit
18:27more about how the amygdala pathway
18:31works and how the TMS stimulation
18:37propagates from our stimulation site,
18:40which tended to which tended to be
18:42in the ventrilateral prefrontal
18:43cortex and engaging the amygdala.
18:45So we had a small pilot so that we
18:47can engage the amygdala in this case,
18:49we're doing that again, the TMS, fMRI,
18:51fMRI connectivity based targeting again.
18:54But we also did some DTI at the baseline
18:56and we wanted to see if there's some
18:59relationship between the evoked response,
19:01the amygdala and the DTI measure.
19:04We found some evidence that there
19:06there seems to be a pathway,
19:08a direct pathway between where we
19:10were stimulating in VLPFC and the
19:13downstream amygdala, which is useful.
19:15We also showed that the strength of the
19:19evoked response to TMS was associated
19:22with the fiber density of that pathway
19:26at the individual subject level.
19:28This supports the idea that TMS likes to
19:31flow around along white matter and that
19:34this pathway may be a direct pathway
19:37and that this may partially explain
19:40how TMS actually engages the amygdala.
19:46All right. Can you say,
19:46well, these are nice tricks.
19:49You're doing these pings of these circuits.
19:50You're showing evoked responses.
19:52That's kind of neat,
19:54but is there any like clinical
19:56relevance you're talking earlier
19:58about the SYNC protocol and how we
20:00don't know anything happening in
20:02the brain and how's that relevant
20:04for the any clinical effects.
20:07The first we're looking at TMSF MRI in
20:10this more clinically relevant context,
20:13but it's this requires a little
20:15bit of explanation.
20:16We didn't do this full clinical
20:18trial with the pings along the way.
20:21We tried to take some bit of a shortcut,
20:24which is to test the circuit
20:26hypothesis in a faster
20:31like design. Yeah.
20:33So one of the difficulties of doing
20:37a treatment with TMS is that they
20:39typically take long time like even the
20:42SYNC protocol that only takes one week,
20:44you have to do 10 sessions per day
20:46and then we have the four to six week
20:49traditional clinical TMS for depression
20:50protocol and that takes a long time.
20:52So we thought, OK,
20:53can we speed this up at all?
20:54Let's let's try to pack in a fair
20:56amount of stimulation in three days.
20:59And we thought that that's probably
21:01enough to start modulating the
21:03target and to start pushing symptoms,
21:05but it's not a full clinical trial yet.
21:09Also in in the other TMS studies,
21:11imaging has been sort of an afterthought.
21:13And the case here, we're really
21:15making a priority of how well we
21:18engage this target that we're aiming
21:20for and showing evidence that TMS of
21:22MRI can be useful here to show that
21:25there's a change in the pathway.
21:27And then usually the imaging in other
21:29TMS studies has been correlational
21:30and we want to throw in our TMS
21:33of MRI and see if if there's any
21:35utility in looking at it there.
21:37There's of course the patient provider
21:39burden of the traditional protocols.
21:41We wanna do this in a very short,
21:44like straightforward way
21:46with only a single protocol.
21:49Also throw in this little bit about sham.
21:51You can't do sham stimulation in the scanner.
21:53There isn't a commercially
21:55available stimulator for doing that.
21:58And I'll also say clinically there's
22:01at least some considerations with
22:03doing sham that you know does
22:05not reach the brain effectively.
22:07And so asking the patients to
22:09wait that out and like have these
22:12extended symptom assessments,
22:14you know that they're not getting an
22:17efficacious treatment that's that's
22:19just another hurdle to considering
22:22adding sham to TMS studies.
22:24And I'll say in this case we can
22:27still show some control conditions
22:29which is that we have a circuit
22:32specific circuit in mind.
22:33We also have a specific symptom
22:36in mind with depression.
22:37And so I'll show you some,
22:38some evidence of how well we did with
22:41the circuit and symptom specificity.
22:44All right. This is, this is our design.
22:46So we collect a baseline scan,
22:50we use that to determine the
22:52connectivity targets.
22:53So they're personalized high
22:55connectivity peaks,
22:56positive connectivity peaks with Subgenual.
23:00We also collect an amygdala seated
23:03connectivity profile for a second
23:05stimulation site and then before the
23:09intervention we pin the circuit in
23:13both kind of connectivity targets
23:15and then we do our intervention
23:18over the three days and then we
23:20ping the circuit again.
23:21So pretty straightforward, right?
23:22We do a pre and post measure and we're
23:25focusing on this subgeneral pathway
23:27to see if we can link the TMS up
23:29from Rye with some clinical change.
23:31And I call the intermittent date
23:34of birth stimulation protocol.
23:35I call it an intervention because I
23:37know it's not a full treatment protocol.
23:39I know this is not like the
23:43maximally effective dose of applying
23:45TMS to affect depression,
23:47but I was hoping that it would move
23:49it enough that we can capture this
23:51more acute response and link the TMS,
23:55HEP, MRI to a clinical change.
23:57So that's what we set out to do when
23:59we actually deliver the intervention.
24:01They're not in the scanner, right?
24:02We just do the pings before and after.
24:04So the intervention,
24:05they're sitting in front of
24:06a neuro navigation camera.
24:08We're getting 2400 pulses of
24:10intermittent date of births per
24:12day for three consecutive days and
24:13then we ping them in the scanner
24:15again the day after that.
24:19So we found evidence that there is an
24:23association between the strength of
24:25the ping the evoked response before
24:27the intervention and how well they do
24:31clinically with depression improvement.
24:33And it's very supportive of this of our
24:37hypothesis that engaging the subgenual is
24:41really relevant for depression improvement.
24:44And so we found some evidence of that.
24:48We also did the ping after all, right.
24:50So we did the pre and
24:51post change and the ping,
24:52the evoked response change was also
24:55associated with depression improvement.
24:58So showing evidence that TMS fMRI
25:01not only tells you something about
25:04circuit integrity that's relevant
25:06to improvement clinically,
25:07but it also measures a change in
25:10the communication in that pathway
25:12that we hope happens when we apply
25:14TMS and get a clinical apply.
25:20Now I'll jump back into
25:21the circuit specificities.
25:22We're like you don't have
25:23a control condition,
25:24you didn't do a sham control, right.
25:26We didn't even have another active site
25:28that we delivered the intervention to.
25:30But what we did have is two different
25:33stimulation pathways and two
25:35different downstream targets that
25:37we can measure evoked responses in.
25:39So we looked at the amygdala evoked
25:41response and the subgenual evoked
25:43response through the amygdala functional
25:45connectivity pathway and the subgenual
25:47functional connectivity pathway.
25:48And so our hypothesis was only the
25:50solid blue line that's the place
25:52where we delivered the intervention
25:54and that's our downstream target.
25:57We we thought that if if if our
25:59hypothesis is right that engaging
26:01that target and modulating that
26:03target is the is the one most
26:05relevant to depression change,
26:07then that's the only evoked response
26:09response that will be associated
26:10with depression improvement and
26:12that's indeed what we found.
26:13So we found some circuit specificity
26:16some region of interest specificity
26:21also say that anxiety improved
26:24even though we were aiming at
26:26the at the depression pathway.
26:27Anxiety improvement was not associated
26:30with change in subgenual evoked response
26:33only depression improvement loss.
26:35Now let's show some degree of symptom
26:38specificity and relevance to that
26:41pathway with the subgenual stimulant.
26:46Also say that we cast a pretty
26:48wide net we took in any patients.
26:50Actually we wanted to prioritize
26:52our medicated patients,
26:54which are not the difficult patients
26:55in the TMS clinical studies because
26:57we wanted to kind of clean our brain
27:00response as our first stab at linking
27:02TMS up from my clinical outcome.
27:03But we did get some treatment resistant
27:06patients that have not responded to
27:08multiple rounds of antidepressant
27:10medication and they tended to have
27:13stronger higher levels of depression
27:15which is the blue bar on the left
27:17compared to the non treatment resistant.
27:20But their clinical response to the
27:22intervention was very similar, right?
27:25You can see the non TRD and the TRD ones,
27:28they respond equally well
27:30to this brief intervention.
27:36I will say I mentioned that we
27:38collected whole brain data, right?
27:40And so I'm talking all about
27:42the subgenual cingulate and a
27:44little bit about the amygdala.
27:45Maybe the subgenual cingulate
27:47is not even a hotspot,
27:49you're aiming for it, you engaged it,
27:51you showed these relationships.
27:52But if you looked at the evoked response
27:54changes through the rest of the brain,
27:57probably some other parts
27:58of the network are yes,
28:00is relevant, maybe more relevant.
28:01So we looked at the rest of the brain,
28:03we looked at the evoked response
28:05change map and the symptom improvement.
28:07So this is different from other
28:09brain images that you may have
28:11seen that are just correlational.
28:12These are the evoked response changes, right?
28:15So a very unique measurement and I
28:20will say for a depression change,
28:22the subgenual came up as a hotspot.
28:23It was, it's definitely solid,
28:25it's definitely a reasonable target.
28:26But of course,
28:27other brain areas are also changing
28:29and then are relevant to depression
28:32improvement like hippocampus,
28:33posterior singlet.
28:34A bunch of these other brain
28:36areas also come along.
28:38And then we we recognized that the
28:41evoked response in the subgenual was
28:43not relevant for anxiety improvement,
28:46even though anxiety did improve.
28:48So we looked at the other parts of the brain.
28:50We found that there's a an adjacent
28:52region of intermedial prefrontal cortex
28:54that changed in response to stimulation,
28:57some other regions and posterior
29:00cingulate orbital frontal cortex.
29:02So there are other regions that
29:04seem to have been modulated and that
29:07are relevant to anxiety change.
29:09I would say these maps could be
29:11really useful because these can
29:13generate new hypothesis.
29:15If you're, if you say OK,
29:16well we want to,
29:17we want we want another pathway that
29:20might be more relevant for anxiety.
29:22So we can say,
29:23OK,
29:23we can see these regions and look for
29:25connectivity targets at the surface
29:26or we can try to capture something
29:28that's more to the surface like
29:29maybe this orbit of frontal one,
29:30you say.
29:31All right, well,
29:31that gives us some evidence that this
29:34is a pathway we want to modulate.
29:36And so let's try a treatment or another TMS,
29:39FM,
29:40RI study focusing on one of these
29:42other cortical targets and see if
29:44that actually is more effective
29:46as a as a treatment for anxiety.
29:51All right. So I tried to demonstrate
29:54some evidence that our our positive
29:56connectivity targets for the subgenual
29:59may be especially clinically relevant,
30:02but there's the brief intervention
30:06study that may not be as similar to
30:10traditional RTMS clinical trials.
30:12So what about purely based
30:14on clinical evidence,
30:16what can we show maybe it's
30:18sort of a distraction.
30:19I can come back to it if if there
30:21are questions or people want
30:23to get into more of the sham
30:25consideration etcetera. But
30:29I will, I will say,
30:29I will say that it's harder to
30:33get to compare two active sites
30:35and get a clinical difference
30:37than it is to deliver sham where
30:39we know it's not engaging the the
30:42brain networks or modulating them.
30:44So let's say it's it's sort of a
30:46higher bar to have another active
30:48site that you think may actually
30:50help with symptoms and then your
30:52personalized fMRI guided target
30:54that you hope is even better.
30:56So we tried this out in a cohort
31:00of mixed depression and PTSD
31:03patients and we chose this positive
31:07connectivity target based on their
31:09baseline F MRI and we compared
31:12that to a six centimeter anterior
31:15motor cortex spot that's been
31:17looked at clinically in depression
31:18and seems to work decently well.
31:20So we have these two active site
31:22targets we did between subjects
31:25design on those we have two
31:27weeks of daily TMS treatment.
31:30We added in this funky element where
31:35we're trying to engage circuitry
31:37through some psychological tasks
31:39and I'm going to skip talking
31:41about that because some of the the
31:44interactions were not significant.
31:46We expected them to be with the target
31:48and what tasks they were doing there.
31:50There's maybe a little bit of signal there.
31:52We want to try to follow up on it.
31:53But the the basic design here
31:55that I'm going to give you the
31:57evidence for as the fMRI guided
31:59versus the six centimeter target
32:02and this is what the cortical
32:04sites look like in standard space.
32:07So you can see the blue ones,
32:08those are the 6 centimeter ones,
32:10they tend to cluster fairly well together.
32:12Some people's heads are longer or shorter
32:14and so you get a little bit of a, you know,
32:17spread from anterior to posterior,
32:19whereas the fMRI guided ones,
32:20those have a little bit more
32:22variability in where they land.
32:24So we're looking at a nice consistent
32:26cluster that has high positive connectivity
32:29individually guided for that subgenual
32:31and you can see there's overlap,
32:33There's definitely overlap in standard space.
32:37But we still anticipated that the
32:40personalization was going to make
32:42a difference and help the symptoms
32:45even more and this is the clinical
32:47evidence that that we found.
32:49So this from across the weeks with
32:51a longer term follow up you see on
32:54the top left the PTSD checklist.
32:56So in terms of PTSD symptoms,
32:58the scalp target and the FBI guided
33:01targets seem to work decently well.
33:03Both of them look pretty similar even
33:05in the longer term follow up that they
33:08held pretty tight with one another.
33:10There was one subscale of PCL that
33:13showed a slight difference which
33:15is the bottom left and that was
33:18the hyper arousal subscale.
33:20So we showed some clear evidence
33:23like immediately post treatment out
33:26to week 10 where the fMRI guided
33:281 tended to be more efficacious.
33:31Some of that kind of slipped back in
33:33longer term follow up where they they
33:36looked a little bit more similar where
33:38we saw the the best more striking
33:41group differences is in the PHQ 9
33:43depression scale on the top right.
33:45You can see that kind of from early
33:49on those those two kind of profiles
33:52look different.
33:53The fMRI guided continues to beat
33:55the scalp based target and even
33:57in the longer term follow up it
33:59could becomes even more pronounced.
34:01Like the scalp target,
34:02the symptoms start to kind of push
34:05back towards the baseline a lot
34:07more than the fMRI guided one that
34:09tends to stick around.
34:11So these these are significantly
34:13different even accounting for the baseline
34:17symptom differences and measures.
34:20OK.
34:20So this is,
34:22this is something of you know hope
34:25for the future which is that to
34:27mess up MRI might guide us more
34:29quickly to a more efficacious target
34:32for an individual patient.
34:34So you have a couple of different
34:36imaging based market markers,
34:37right and you say well there's
34:39a connectivity peak here,
34:40there's a DTI based target down here.
34:43I don't,
34:44I'm not sure which one is better,
34:46but I do feel like engaging the subgenual,
34:48there's good evidence for that.
34:49So put them in the scanner,
34:51you ping a couple of different
34:53potential pathways,
34:54you measure the evoked response,
34:56right for that individual patient
34:58through that pathway and you say ah,
34:59it looks much stronger at this red site.
35:02And so you carry that forward
35:04as your treatment target.
35:06You know,
35:06if this,
35:07if this evidence continues to build
35:08the way we're starting out here,
35:10that engaging the circuits is really
35:12critical and tells you something
35:14about how effective the brain
35:16stimulation treatment will be,
35:18an approach like this might
35:20be particularly valuable.
35:21Save us a lot of time make the
35:24treatment protocols work better.
35:30All right. So in conclusion, fMRI guided
35:34TMS seems to engage intended targets,
35:37at least these ones that we tried so far,
35:39the subennial singular and the amygdala.
35:41So I've seen people in talk say,
35:43oh maybe we need ultrasound a lot less
35:45developed as some other treatment
35:47because TMS can't reach the amygdala or
35:49the subennial simulant And so showing
35:52evidence that actually indirectly it can.
35:54We're we're not arguing TMS
35:55directly engages these brain areas.
35:57TMS doesn't go very deep.
35:59But building on all this,
36:00it's really great neuroscience and
36:03imaging data related to brain networks.
36:05There's a cortical representation of
36:07almost any network that you would want.
36:10And so if we can show that we can
36:12effectively engage even these deep sub
36:15critical downstream regions with TMS
36:17then that may be a a great piece of
36:19evidence to encourage more people to use it.
36:24We also showed that there's a clinical
36:26relevance that engagement at this
36:28target that how strong does this
36:30circuit respond to a pulse of TMS
36:32actually tells you something useful
36:34about how well the TMS is going
36:37to treat that person's symptoms.
36:39So I'd love to continue building on that.
36:42And then in this first initial
36:45stab with this clinical trial,
36:46we found that there's at least
36:49some evidence that the fMRI guided
36:51is more clinically effective than
36:53a stout based target.
36:55I'm not sure if I mentioned,
36:55but the fMRI guided is like moving the
37:00PHQ like 60% improvement on average and
37:03the scale based target is like 52 percent,
37:0751% something like that.
37:09So significant difference,
37:11is it worth the time trouble expertise
37:13of doing the fMRI guided target
37:15like that would still be an open
37:18question I I'd say and is this the
37:20best fMRI guided target that we can
37:22come up with more PTSD impression,
37:24I'd say no,
37:25but probably not.
37:26But let's continue building on
37:28that and see if we can do the
37:31circuit based specific symptom
37:33kind of mappings and continue to
37:36improve our targeting and dosing
37:40and hopefully more of these fantastic
37:44clinical studies will add on imaging
37:46of of any kind Functional imaging
37:48would be better than just holding
37:50on to this black box where we don't
37:52know why some patients respond,
37:54We don't know what happened to the
37:56circuits in response to TMS which I think
37:59is really critical for pushing the field
38:01forward and treating patients better.
38:07All right. So this this works really
38:09well with some NIH funding priorities.
38:12We have a pending R61R33 that I think if
38:15you're talking about target engagement.
38:17However this is a very straightforward
38:19way of showing that you can engage with
38:22particular target and then build on that
38:24to do a more definitive clinical trial.
38:27So it's a very good fit I think
38:29with some objectives of of some
38:31of the funders out there.
38:34So these are my team, the,
38:36the people in my center and my closest
38:41collaborators see that we have a little time.
38:43So I have some extra slides that are
38:47based on questions that I be asked in
38:51manuscripts and in talks as just giving
38:54you a a brief response to some of these.
38:57So you'll say all right well you
38:59you take these unmedicated patients,
39:01those are not really a typical
39:03So what happens in the medicated
39:04patients and totally agree we want to
39:07replicate in a medicated patients.
39:08So the pending new grant starting
39:10in December, we're gonna,
39:12we're gonna allow for that and check it out.
39:14I'll say,
39:15well you did this brief 3 day intervention,
39:17maybe that's not exactly what happens
39:19in the brain with a higher dose of
39:23of more stimulation in the sync
39:25protocol or even the old original
39:2710 minutes for depression.
39:28I totally agree.
39:29Let's check it out with a higher dose.
39:31Now that we have the evidence
39:33linking these measurements together,
39:34I mean it's worthwhile
39:36exploring that in a higher dose.
39:39The imaging aficionados you may say that's
39:41a region with low signal noise ratio.
39:43So you shouldn't use it.
39:45And I would say well we have this
39:48evidence nevertheless that we're
39:49getting significant about responses
39:51and differences and clinical
39:53relevance with our TMS, FM, RI data.
39:56But that being said, I think we can do,
39:59we can collect higher fidelity images
40:02for example we have an 8 channel volume
40:05coil coming that we're gonna start
40:07using in our new studies say well
40:09depression is a network it's not just
40:11a subgenual you shouldn't be focusing
40:13on single brain areas like that and.
40:15I agree.
40:16I'd say if you have a network that you feel
40:19is a better fit for TMS depression outcomes,
40:23like happy to consider pulling
40:24it out of our data.
40:25Having a look at it,
40:26we did another grad student in my lab,
40:28I did an amygdala,
40:30found an amygdala change in fMRI and its
40:33meta analysis for depression treatment.
40:35We pulled that out of our data
40:37and didn't find an association
40:39with the interventions outcome,
40:41but there there are probably other
40:43ones that that are better in
40:46terms of network responses.
40:47So yeah,
40:49even improving the imaging we
40:51do at baseline to make a better,
40:53more precise,
40:54more personalized target for
40:57doing stimulation,
40:58absolutely you can do better.
41:00We try to keep up with the imaging field.
41:01We're gonna do some multi echo
41:03collect more fMRI data to make
41:05a more reliable target
41:06for the individual patients.
41:08So definitely up for you know further
41:11improvements in the imaging protocol.
41:14All right. Then there's the a lot of
41:16papers that are showing this anti
41:19correlated like spots really seem to
41:21be relevant to depression outcome.
41:24But that there is a a recent paper from
41:27Connor Liston suggesting that there's
41:29a a subgroup of patients that are
41:31anomalous that are driving that but it.
41:33But I'll also just say that once the field
41:35sort of focuses on something they're
41:37like oh look at that there's evidence
41:39everywhere for the anti correlated spot.
41:41They some sometimes we might get
41:43a like we might have a propensity
41:45to put blinders on and chase the
41:48same targets in everybody's labs.
41:49But at least for me,
41:51I feel like this basic brain measurement
41:53data of of the positive connectivity
41:56sites makes it worth considering.
41:59Like if if people are wearing blinders,
42:01maybe we can like open up the field
42:03a little bit more and and look
42:05for the possibility of positively
42:07correlated spots being relevant.
42:12And then again,
42:12since we have a little bit of time,
42:14I just want to mention some other
42:15things that we're working on.
42:17So we're doing a lot of TMS up MRI,
42:19closed loop things where we're doing
42:21different stimulation frequencies,
42:22trying them out on working memories,
42:24so personalizing not just the target
42:26but also the stimulation parameters.
42:28So testing this out and worry and
42:31rumination Also different targeting
42:32methods based on DTI or resting some
42:35different ways of splitting up the brain
42:38and personalizing target with network
42:40control theory and deep learning.
42:42We're doing some basic methods things
42:44like single pulse TMS with stereo EEG and
42:47epilepsy patients trying to get that going,
42:50some really cool stuff with KC
42:52help partners and neurosurgeon
42:54here on personalizing DBS for OCD.
42:59Things that I'm looking for collaborators
43:00on these will be new things that I,
43:02I, I do start to pilot.
43:04There's a controllable TMS
43:06system commercially available.
43:08I'm showing it down there on the left.
43:10We want to play with that pulse
43:12width and shape can be potentially
43:15even more efficacious and changing
43:17some of the stimulation protocols.
43:19Also if you if you have a clinic
43:21where you're doing TMS,
43:23we should take every single patient that
43:24comes in and do some kind of study with them.
43:27Like it doesn't actually cost
43:28anything to just try a brain state
43:31manipulation and seeing how that
43:32contributes to patient outcomes.
43:34So that's pretty straightforward
43:36one that we're starting with
43:39a couple of collaborators,
43:40I'll say we're not the,
43:41we're not the only ones that think
43:44circuit engagement with brain
43:45stimulation using an imaging marker
43:47may be clinically really interesting.
43:50So this is from Andres Lozano's
43:52group in Toronto and showing
43:54an association that DBS FM RI.
43:56It also tells you something
43:57about circuit engagement that's
43:59relevant to depression improvement.
44:01So we definitely agree with this.
44:03We want to build on this ourselves in
44:05a variety of ways that I described.
44:07I think we can learn a lot about causal
44:10connections in the brain writ large,
44:11but also specifically with these
44:14intervention tools that I think
44:16is really important for building
44:18this bridge between imaging,
44:20making it clinically useful and you know,
44:25optimizing the the stimulation parameters
44:28and locations going into the future.
44:32So look at that.
44:33Thanks for your attention.
44:38Yes, thank you, Des.