Yale Psychiatry Grand Rounds: "Brain Dynamics and Flexible Behaviors"

May 17, 2024

Ribicoff Lecture. "Brain Dynamics and Flexible Behaviors"

Speaker: Lucina Uddin, PhD, UCLA Brain Research Institute

Information

ID: 11688
To Cite: DCA Citation Guide

Download Transcript

00:00This really kind introduction and for
00:02the invitation to visit here it's I will
00:05say it's it's always fun to visit Yale.
00:08The the first time I came here
00:09was in 2012 and then I was here
00:12again in 2018 and now it's 2024.
00:15So I hope to see you all again in
00:172030 because it seems like every six
00:19years I get the opportunity to come
00:21and and and meet and just benefit from
00:24these wonderful environment in the
00:27outstanding discussions and just feel
00:29the collegiality every time I'm here.
00:31So it's really nice to be back.
00:34Thank you again, Denise,
00:35for this wonderful introduction.
00:36Thank you, John and and everyone
00:38else for for the warm welcome.
00:40Now I suppose I need to figure out
00:42where the slide advancement is.
00:43So. Oh, here we go. OK All right.
00:47So it's just it's an honor to be
00:51here following up this introduction
00:53of of RIBICOFF.
00:55So I'll talk to you today about our
00:58work on brain dynamics and flexible
01:01behaviors and I think it's it's of
01:03interest for our lab but I think
01:05for a lot of people in psychiatry
01:07to think about flexibility.
01:09We are we focus a lot on children
01:11with autism in our own research and in
01:14this particular clinical condition we
01:16often see this insistence on sameness,
01:18behavioral inflexibility.
01:19For example,
01:20a child might want to wear a particular
01:23pair of socks or eat as only yellow
01:25foods or you know these kinds of
01:28repetitive behaviors that sometimes can,
01:30you know this insistence on sameness
01:32can sometimes cause challenges
01:34for caregivers and and difficulty
01:36in in day-to-day life activities.
01:39And so I used to kind of introduce
01:40my talk saying you know flexibility
01:42is important.
01:43But during the pandemic I think
01:44we all had a a big dose of what
01:47it means to be flexible.
01:48So we went from our everyday lives and and
01:51routines to something completely different.
01:53We started giving talks over zoom instead of
01:57giving them at at lecture podiums like this.
02:00So I'm myself readjusting
02:02to being standing up again.
02:06So.
02:06So we've all sort of experienced this
02:08like how changes and and be having
02:11difficulties with flexible behaviors can
02:12kind of impact our our day-to-day lives.
02:15So we try to focus on the lab on
02:17thinking about the neural basis of
02:19flexibility and we do kind of take a
02:22developmental look at these questions.
02:23So we think about how brain networks
02:26develop and mature in ways that can
02:28support increasingly sophisticated
02:30types of cognition like flexibility
02:32across the lifespan.
02:33So we think about how brain
02:35networks develop from childhood,
02:36adolescence into adulthood,
02:38how that underlies cognitive development.
02:40And you know basically how can compromised
02:43connectivity within and between large
02:46scale brain networks be related to
02:48developmental neuropathologies.
02:49And we try to think about how we
02:53can use this basic information
02:55about brain structure and function
02:58to inform diagnosis and design
03:00interventions ultimately.
03:01And so I'm really excited to be here in
03:04the in the Ribicoff series because
03:06we do try to do in our lab a sort of
03:10combination of basic neuroscience
03:12developmental work and clinical clinically
03:14translatable kind of research in the
03:17same umbrella or under the same lab.
03:20So we use structural and
03:22functional neuroimaging.
03:23So that includes task functional
03:25magnetic resonance imaging,
03:27resting state fMRI,
03:28diffusion weighted imaging to look
03:30at structural connections and also
03:32some causal modeling approaches
03:34to look at relationships among and
03:36between brain regions as they relate
03:39to things like executive function,
03:40flexible behaviors.
03:41And although we we love neuroimaging,
03:44we're a network neuroscience lab,
03:46we loved all this sort of cutting
03:47edge tools that are now available.
03:49I think it's important to remember
03:51how much of this work just dates back
03:54to before we had these fancy tools.
03:57So the foundations of what we now
03:59call network neuroscience were
04:00planted a long time ago.
04:02You know before MRI was really
04:04very much in vogue.
04:05I I love some of these papers from
04:08Marcel Maslam who is who has you know
04:10is a is a neurologist who were and
04:13a neuropsychologist who had these
04:16great ideas about brain function.
04:18That that really came from looking
04:20at patients and lesions and figuring
04:22out what the deficits were in
04:25in patients with focal lesions.
04:27And the interesting conclusions
04:28that are nicely summarized in one
04:31of his Seminole papers.
04:32Or that cognition is served by
04:35interconnected neural networks and
04:37any complex behaviour is mapped at
04:39the level of multifocal neural systems
04:41rather than specific anatomical sites.
04:43And I, I love these these papers,
04:46They're kind of just a history of how
04:48the brain works and they really hold up.
04:50I think to this day,
04:51if you read this,
04:51you'll say, hey everything.
04:52And he was right.
04:54He was right about everything.
04:56And I think this is interesting because
04:58there was the trend at one point to say,
05:00well, OK,
05:00we have a lesion to this brain
05:01area and that causes this deficit,
05:03that brain,
05:03you know,
05:04there must be a one to one mapping
05:06between that function and that structure.
05:07But even you can use the same data to
05:09make the opposite conclusion is that,
05:11you know,
05:12brain regions can be relatively
05:14specialized and not sort of
05:16specifically tied to to one function.
05:18And so I think this is some
05:21really nice foundational work.
05:22There's also one of my favorite papers
05:25is from Randy McIntosh in 2004 where
05:27he introduced the concept of neural context.
05:30The idea that relevant functional
05:32relevance of any given brain area
05:34depends on the status of other
05:36connected areas to that brain region.
05:38So kind of not just focusing on the one,
05:41one or two areas at a time,
05:42but the whole kind of context
05:44in which brain function is,
05:45is occurring.
05:47And we also think a lot
05:48nowadays about the time domain,
05:50so about how networks need to be
05:52understood in terms of interactions
05:54as they unfold temporally between
05:56between multiple brain regions
05:57as they unfold temporarily.
05:59A nice paper from Luis Pasoa 10 years
06:02ago really drives this point home.
06:05So there are some things that many
06:07of you are very familiar with.
06:09And so back in 1995,
06:12Broad Bizwell published a paper
06:14showing that you can find these
06:16coherent spontaneous fluctuations
06:17in different parts of the brain.
06:19You focused on the motor cortex.
06:21But since that 1995 paper there have been,
06:24you know, thousands of studies showing
06:26that at this very low frequency the brain
06:29seems to recapitulate in the resting state.
06:32All of the the brain networks that we
06:34can see engaged in tasks like memory,
06:36attention, vision, motor processing.
06:38At these point O1 to .1 Hertz
06:42is low frequency.
06:43We can see these oscillations
06:45spontaneously occurring.
06:46And this is just a ongoing,
06:49you know,
06:49fMRI of somebody doing nothing
06:50at all in the scanner,
06:51just kind of asked to just lay still.
06:54But it's not sort of random
06:56spontaneous activity,
06:57but really coherent in these systems
06:59that we've been studying for many years.
07:01If you wait around a while,
07:02you'll see a language network or a vision,
07:04a visual cortices.
07:06So these kind of spontaneous fluctuations,
07:09many labs have been exploiting now for
07:11a number of years to try to understand
07:13the functional organization of the brain.
07:17And as I,
07:17as I mentioned before,
07:18we all know what flexibility is
07:20and we've had to be very flexible
07:22across this pandemic period.
07:24During that period,
07:24I had the time to sit down and
07:27write a bunch of invited reviews.
07:29So that was my my COVID,
07:31my my pandemic was review writing.
07:33But it gave me time to really think about
07:35what we mean when we study flexibility,
07:37what we're what we're using
07:38that term to refer to.
07:40And in in human sort of cognitive
07:43neuroscience and psychology we tend
07:44to use things like the Wisconsin
07:46cards sort tasks and we ask people
07:49to make categorizations and then you
07:51know switch the criteria and and we
07:53use that as an index of flexibility.
07:55But I also think we have all these
07:57parallels in behavioral neuroscience.
08:00In reversal learning paradigms for
08:01example you're just asking animals
08:03to make different kinds of stimulus
08:05outcome mappings and we we call
08:07that behavioral flexibility and
08:08I think mainly because we can't
08:10guess what animals are thinking.
08:12So we have them do a behavior and
08:14we see how how they flexibly behave.
08:17When we look at developmental psychology,
08:19we often come up with things
08:21that kids can do like,
08:23you know,
08:23tell me you know which color is this fish
08:25or which direction is the fish pointing.
08:26But we do things like task switching
08:28to engage the development of
08:30flexible behaviors in in children.
08:32But all of these kinds of paradigms tend
08:36to engage frontal parietal networks.
08:38They tend to engage singular insular
08:41cortices and frontostriatal systems and
08:43and many of you have been, you know,
08:47working in this field for years.
08:48So I'm not telling you anything new.
08:50What I do think is interesting
08:52about flexibility is that
08:53you can see difficulties,
08:54oops, across many different
08:56kinds of clinical conditions.
08:58So we focus a lot on autism and early
09:00life conditions like ADHD where you get
09:02things like and autism specifically,
09:04you get restricted and repetitive behaviors.
09:06But then you have things
09:07like anxiety and depression,
09:08which come along more during
09:10adolescence that are associated
09:11with repetitive negative thinking.
09:13Worry in the case of anxiety or
09:15rumination in the case of depression
09:17are kind of inflexible thought patterns.
09:19So it's another way of thinking about
09:22flexibility and and difficulties
09:24with flexibility.
09:24And of course,
09:26OCD is another great example of
09:28a case where you really have
09:30difficulty coming out of a routine.
09:32And then even later in life,
09:33things like Alzheimer's and
09:35and Parkinson's are associated
09:36with cognitive rigidity.
09:38So, you know,
09:38these are maybe not all the same thing
09:40we're seeing at these different stages,
09:42but there's something about getting
09:44stuck in a particular behavioral
09:46pattern or thought pattern that we
09:48can see perhaps as a, you know,
09:50transdiagnostic, you know, difficulty.
09:52And so for us, we think,
09:55well, let's try to figure out,
09:56you know,
09:56what in the brain is allowing
09:58flexibility in the first place.
10:00And so a lot of the work that
10:02I won't talk about today is on
10:03thinking about development of brain
10:05networks involved in flexibility.
10:06So we'll do things like look at MRI
10:09data collected from participants
10:10between the age of 6 and 85 S really
10:13kind of lifespan data sets and
10:15look at changes in brain signals
10:17as they associate with, you know,
10:19like flexibility of brain
10:20networks across the lifespan.
10:22We'll look at things like differences
10:24in children versus adults in the
10:26strength of certain circuits or the
10:28effective connections between brain
10:29regions and how they change with development.
10:31And we'll get how network interactions
10:33again change across the lifespan,
10:35for example,
10:36between age of 6 and 85.
10:38There's a nice Nathan Klein
10:40Institute data set that that
10:42provides a lot of these data and
10:44they're publicly available now.
10:45So lots of things are are you know
10:48analysis are available and we're
10:50we've we've shown kind of how these
10:52differences in brain organization
10:54across the lifespan differentially
10:56relate to executive function.
10:58So like some of these brain variability
11:01metrics you know are associated
11:03with better executive function and
11:05adolescence but worse executive
11:06function and old age for example.
11:08And I know you can't see that slide,
11:10so I apologize.
11:11I'll make it bigger next time, I promise.
11:13But what we do do in this scanner
11:16is sometimes we just test.
11:18You've tried to develop a paradigm
11:19that can be used across children,
11:21across adults, across clinical populations.
11:24And we took this one paradigm from the
11:27developmental psychology literature
11:28called the flexible item selection task.
11:30And what you do here is you ask
11:32somebody to pick two things that go
11:34together and they might say, well,
11:35the 2nd and the 4th card are both blue,
11:38so they go together on color.
11:40And then you might say, OK,
11:41now pick another two things that go
11:42together and they might say, well,
11:44the 3rd and the 4th are both rabbits.
11:46I think that's right. And so it's OK.
11:48Pick another set of things
11:49that go together and say, well,
11:50the 1st and the 4th, they both have one
11:52item so that they're both showing one.
11:54So that's something that goes together.
11:56And this requires, you know,
11:57working memory, inhibition,
11:59all of these processes, attention,
12:00but also of course flexibility.
12:02The further you get along these dimensions,
12:04the harder it is to come up
12:06with things that go together.
12:07So we tried to put this in
12:09the scanner and we said, OK,
12:10here's just a control version of it,
12:12just hit the two buttons that
12:13are highlighted.
12:14So OK, just follow along.
12:16So we spent some years validating this task,
12:20having adults do it,
12:21putting people in the scanners,
12:22looking at their accuracy
12:24inside of the scanner,
12:26outside the scanner,
12:27all the typical validation stuff come by,
12:30created an efficiency metric to look at
12:32how good people are at this task and
12:34how much better they get at it over time.
12:36So Dejani did this work in our lab
12:39and found as one might expect really
12:42robust brain activation and flexibility
12:45trials in lateral prefrontal cortex,
12:48parietal cortices,
12:49cerebellum and basal ganglia and the
12:52anterior cingulate and the anterior
12:55ansula favorite brain region of mine
12:58that will come up again and really,
13:01really strong and robust
13:02activation across these regions.
13:04So,
13:04so we know as as from lots of
13:06literature that these are all
13:08very much involved in in flexible
13:09thinking and flexible behaviors.
13:11What I think is interesting to
13:13note though is for those fans
13:15of neurosynth and meta analysis.
13:17If you look at if you put a term
13:19like shifting or working memory
13:20for example into the neurosynth to
13:22look at automated across studies,
13:24what are the brain regions that are
13:26involved in shifting or I'm sorry,
13:28flexibility, you'll see these regions,
13:30right, the the same ones I just mentioned,
13:32frontal parietal cingulance or
13:33if you go into neurosynth,
13:35then you type in updating or
13:37you know working memory,
13:38whatever phrase you want to use,
13:40you'll see similar kinds of activation.
13:43And then if you type in inhibition
13:44again you'll see a lot of these
13:46overlapping brain regions.
13:47So they're really kind of broadly
13:50involved in these different
13:51components of executive function.
13:53It's hard to,
13:54I guess the point is it's hard to
13:57pull out of flexibility specific
13:59activations when almost all the
14:01tasks that try to tap flexible
14:03thinking also involve attention,
14:04working memory, inhibition and you know,
14:06everything else.
14:07So it's it's all not as clean as,
14:10as you know,
14:11we like to pretend,
14:12but you can do things like there's
14:14connectivity modelling approaches that
14:16let you get at some of these questions.
14:18One of them we've worked on
14:20here with Katie Gates at UNC,
14:22it's called group iterative
14:24multiple model estimation or Gimme.
14:26It's kind of a an iterative search
14:28algorithm that tries to fit a
14:30structural equation model to describe
14:32an individual connectome using user
14:34specified regions of interest.
14:36So if you take for example of all
14:38of these nodes that are activated
14:40in the flexible item selection task,
14:42you can look at OK,
14:43which sort of nodes kind of are most
14:46directly activated by this paradigm and
14:48which are sort of secondarily activated.
14:51So here we found that inferior
14:53frontal junction is sort of directly
14:56activated by flexible thinking
14:57or flexible item selection.
14:59And then there's information flow
15:01to other regions including the
15:03dorsilateral prefrontal cortex,
15:05anterior cingulate and others.
15:07And so there's kind of some regions
15:10that are more influential than others
15:12in this in this type of flexibility.
15:15So there's kind of ways of of teasing
15:18apart some of these activations and
15:20looking at more specific information flow.
15:22There's also a great deal of
15:24individual variability,
15:25like even though all subjects tend to
15:28activate these brain nodes during this test,
15:30there's group level paths or
15:32connections that are consistent,
15:34but a lot of sub sub sub group level
15:38paths meaning like some subjects engage
15:40these connections and others don't.
15:42So there's a lot of kind of more nuance
15:46to these blobs and activation patterns
15:49than we might initially realize.
15:51So we think about,
15:53for us at least,
15:54how cognitive flexibility involves
15:56the coordination among multiple
15:57brain regions that are all known to
15:59play a role in executive function,
16:01adults and children.
16:02I didn't show the maps from children,
16:04but children between the ages of
16:058 to 12 also show these kind of
16:08similar activations of lateral
16:10frontoparrietal and singular insular
16:12networks during flexible item selection
16:14and task modulated connectivity.
16:15The inferior frontal junction seems
16:17to be particularly important for
16:19this type of flexible behavior.
16:21And I won't,
16:22I'll show you at the end of the
16:23talk some of the work we're doing
16:25now to bring this task to kids with
16:27autism and look at the brains of
16:28children between the ages of 8:00
16:30and 12:00 as they're doing this
16:32type of of flexible behavior task.
16:34And we, well, I'll show you,
16:36I don't remember if I included
16:38this slide but I'll I'll come back
16:40to this towards the end.
16:41So as we kind of develop these tasks,
16:44try to use them in clinical populations
16:47and developmental populations at the
16:49same time we try to focus on these
16:51big data sets that can help us look
16:53at sort of the adults neurotypical brain.
16:56And then the nice thing about some
16:58of these publicly shared large
16:59data sets is they're, you know,
17:01hundreds of subjects large.
17:02There are lots more data points
17:04than we typically can collect
17:06in the clinical population.
17:07So we can do a lot more with the
17:10the methods and the inferences and
17:12the replication here because for
17:14example in human connection project
17:15you actually have one hour of resting
17:17state F MRI data from each person.
17:20So a lot more,
17:21you know, signal in there so we can
17:25do things like look at moment to
17:26moment functional connectivity or
17:28dynamic functional connectivity.
17:29So in this process,
17:30instead of saying what areas are
17:32connected to each other on average,
17:34we're saying if you break down
17:36this data into 45 second chunks
17:39or 62nd windows and you look at
17:42reoccurring patterns of whole brain
17:43functional connectivity and you use
17:44some kind of clustering to say, OK,
17:46here's some different brain states.
17:48You can then start to quantify
17:50dynamic metrics like the frequency
17:52of occurrence of a particular
17:53brain state or the dwell time,
17:55which means how long does this
17:57brain state persist once it comes
17:59along and the state transitions,
18:01how much switching between
18:02brain states can you quantify?
18:04So in the human connection project is
18:06that this is just showing you a couple
18:08100 subjects do split half replication,
18:10various things.
18:11You can use independent component analysis
18:13to breakdown the brain into little regions.
18:16And then the nice thing is that
18:17all there's all kinds of other
18:19data on these same subjects,
18:21information about their processing speed,
18:23inhibition, cognitive flexibility,
18:24fluid intelligence,
18:25working memory,
18:26lots of scores to play around with.
18:29And So what we found is that typically
18:31the the brain enters different states,
18:34but it's got most of its time,
18:35about 36% of its time in this loosely
18:39connected kind of flexible state.
18:41And about 10% of the time it's really
18:43in this more cohere or really tightly
18:46correlated state that you see on the right.
18:49And it turns out that those individuals
18:51who do better on tests of working memory
18:53and tests of cognitive flexibility
18:55like the Wisconsin Card sort of test,
18:57those are the ones who are spending
18:58sort of or showing more frequent
19:00occurrence of states one and two,
19:01the loose connectivity states with
19:03more variability actually in their
19:06connection patterns and less time in the
19:08tight sort of inflexible states on the right.
19:11And so this is just,
19:13you know,
19:14showing us how we can get at sort of
19:16the basic neuroscience of of brain
19:18flexibility using these large data sets.
19:21And here we're just showing an example
19:22of how greater cognitive flexibility,
19:24which is measured outside of
19:25the scanner in this case,
19:26are associated with the propensity to
19:29occupy these more frequently occurring
19:31brain configurations characterized by
19:33attenuated correlations and greater
19:35functional connectivity variability.
19:37And it's just showing how we can
19:39think about using dynamic functional
19:42connectivity approaches to reveal
19:44relationships between brain
19:46dynamics and flexible cognition.
19:48So,
19:48OK,
19:48So what I talked about in the last
19:51study was really a whole brain kind of
19:55agnostic approach to thinking about the
19:58how flexibility might be implemented.
20:01But there's also of course,
20:02decades of cognitive neuroscience
20:04literature that that we all sort of
20:07rely on to to think about what are the
20:10more specific brain networks that might
20:11be involved in some of these processes.
20:13I already sort of mentioned
20:15the central executive or sort
20:17of lateral frontal parietal systems,
20:19the salience or mid single
20:21insular systems anchored in the
20:23cingulate and insular cortex,
20:25and the default mode network which has
20:27key nodes in the medial prefrontal
20:29and posterior parietal cortices.
20:31A number of years ago,
20:32we started to notice some interesting
20:34patterns of interrelationships
20:35among these three networks.
20:37They show up a lot in cognitive neuroscience.
20:39They show up a lot in psychiatry.
20:40But it turns out there's
20:42interrelatedness between them, right?
20:44So you can have usually when signals
20:46in the default mode network go up,
20:48the lateral front file goes
20:50down and vice versa.
20:51But it turns out you can often
20:53predict what's going to happen in
20:55these networks based on signals
20:57from the anterior insular cortex
20:58which we think of almost as a
21:00causal outflow hub like driving or
21:03orchestrating the changes between
21:04these other large scale networks.
21:06So it's a a model we've been playing
21:08around with now for a number of years.
21:10And if we think about the development
21:13of some of these networks and
21:15development of cognitive flexibility,
21:16again I mentioned this nice data
21:18set that includes several 100
21:20participants between 6 and 85
21:22resting to data from Rai data.
21:23And we can actually look at relationships
21:26between their brain dynamics and some
21:28other executive function measures.
21:29In this particular data set,
21:31the Dallas Kaplan executive
21:33function test was conducted.
21:35So participants are kind of going
21:37from letters to numbers in sequential
21:40order and we were interested here in
21:42the relationship between the brain
21:44dynamics of those three systems I
21:46mentioned and performance on this task.
21:48And so we used a different instead
21:50of sliding window,
21:51we used a Co activation pattern
21:53analysis approach,
21:54which also allows you to kind
21:56of cluster time frames based on
21:58their spatial similarity and look
22:00at functionally relevant patterns
22:02at the whole brain,
22:04not using a sliding window.
22:05But you can still get things like dwell
22:07time transitions between States and
22:09frequency of occurrence of states.
22:11And we found for example that there's
22:13some patterns like involving the
22:15lateral front to parietal executive
22:17and the medial front to parietal
22:19default mode that show AU shaped kind
22:21of trajectory across the lifespan.
22:23So you see kind of Co activation of
22:26these networks a lot more in midlife
22:28than in young individuals or in older age.
22:30And also this transitions or the
22:32switching between brain states
22:34seems to be linked with individual
22:36differences in the behavior on
22:38that executive function test.
22:39So in in middle age essentially you
22:42know between 25 and 45 or thereabouts,
22:45you actually you have pretty high
22:48flexibility performance regardless
22:49of what's happening in terms of
22:51the brain state transitions.
22:53But in children and in older adults,
22:55greater number of transitions between
22:57brain states is associated with better
22:59levels of cognitive flexibility.
23:01So there's like really different
23:02things going on at different points
23:04in life where at some some junctures
23:06it's really important to have
23:07a lot more of these brain state
23:09transitions to support flexible behaviors.
23:11But at other points they're not really as
23:15dependent on this type of brain flexibility.
23:19So this is an example where we used Co
23:22activation patterns to show that executive
23:25and default networks change in terms of
23:28their representation across the lifespan,
23:30in terms of their frequency of of occurrence.
23:33And that the brain state transitions
23:35between these networks seem to be related
23:38to cognitive flexibility in different
23:40ways at different stages of the lifespan.
23:42So if you if you don't mind now
23:44we'll go straight basic neuroscience
23:46for a little while.
23:47Because I think part of you know
23:49when you have these findings
23:50you see different developmental
23:52differences or clinical differences.
23:53It's tempting to, you know,
23:55go forth and, you know,
23:58make inferences and go straight
23:59to intervention or, you know,
24:01whatever the case may be.
24:02But I also think it's it's nice
24:03to to take a step back and say,
24:05OK, what are these brain regions
24:06doing in a broader context?
24:08What do we know about the anatomy
24:10and function of some of these areas
24:12that we seem to be implicated,
24:14you know, trans diagnostically
24:17across flexibility deficits.
24:19So the ancillar cortex in particular
24:21has kind of caught my attention.
24:23If you look at the, you know,
24:25neuroimaging literature,
24:26it often shows up in studies of affect,
24:29empathy, pain, emotion.
24:30You'll see often a talk of an empathy
24:33network that involves these brain regions.
24:36You'll also see a subject of awareness,
24:38Introception,
24:38somatic sensory processes.
24:40I I forgot to put disgust in here.
24:43But you know a lot of things happen
24:45in the insular cortex that are a lot
24:47to do with basic sensory processing,
24:49but also very high level cognitive
24:53processing, things like inhibition,
24:55attention switching and conflict
24:57executive function as we've been
24:59talking about all of those processes
25:01also activate the insular cortex.
25:02So it's it's one of those areas
25:05where everybody has their favorite
25:06thing to say about it.
25:08And so there,
25:09there tends to be a lot of literature
25:11that doesn't really talk to each
25:12other when we talk about the insular
25:14cortex because it's a little bit
25:16siloed in these different fields.
25:19But if you look at just a question of like,
25:21are there subdivisions within the insula?
25:23There's anywhere between 2 and 27
25:25depending on what study you look at.
25:27But it's true,
25:28there's a lot of subdivisions in the insula.
25:31If you look at just resting state
25:33F MRI and ask the question which
25:35voxels in the insular cortex
25:36have similar patterns of whole
25:38brain functional connectivity.
25:39So you're clustering the voxels
25:40based on their whole brain patterns.
25:42There's two studies from Bendine
25:44and Luke Chang that suggest you
25:46can at least find evidence for a
25:48dorsal anterior insula subdivision,
25:49a ventral anterior insula,
25:51a posterior insula.
25:52Like I said,
25:53maybe up to 27 depending on
25:55which Atlas you look at.
25:56So right now let's go with three for now.
26:00So if you look at these subdivisions there,
26:02they do seem to be some kind of
26:05structure to their their their
26:07division of of Labor in the sense.
26:10So in Luke Chang's meta analysis he
26:13found that the ventral anterior insula
26:15seems to be more involved in studies
26:17in the using the terms that are in red.
26:20So emotion, face,
26:21anxiety sort of affective types
26:23of terminology tend to go along
26:25with the ventral anterior insula
26:27in terms of activation.
26:28On the right is a meta analytic
26:31connectivity modeling study that
26:32we conducted some 10 years ago,
26:34which was just asking the question
26:36across many, many F MRI studies.
26:38When you see ventral anterior insula active,
26:40what else in the brain tends to
26:43coactive coactivate with that?
26:44It does tend to be the sort of limbic
26:46kinds of regions that are coactive.
26:49If you go ahead then to look at
26:51dorsal anterior insula,
26:51that's the one that seems to show up
26:54in studies with the term switching
26:56error processing inhibition.
26:58Those are the sort of higher cognitive
27:01executive function types of terms.
27:03And if you look on the right,
27:05that's the coactivation map of dorsal,
27:07anterior and slow which tends
27:08to coactivate with frontal,
27:09parietal cortices and temporal as well.
27:12And then if you go back a little
27:14bit to the posterior insula,
27:15that tends to be the one for pain,
27:17somatosensory,
27:18that kind of more sensory based
27:21kind of cognition.
27:22And it also seems to coactivate
27:25with somatosensory cortices.
27:26So on first glance,
27:28there are the patterns of
27:30coactivation and meta analytic types
27:33of analysis give us some evidence
27:35that the insular cortex can be
27:37subdivided a little bit even still,
27:39like all of these subdivisions are
27:40actually active across all of these
27:42cognitive domains I mentioned.
27:43So there's there's kind of a convergence
27:46and divergent at the same time.
27:48If you go ahead and look at
27:50the dynamics of the functional
27:51connectivity of these subdivisions,
27:53there's four here instead of three.
27:55It's independent component analysis,
27:56it's not too important.
27:58But if you again if you look
27:59at the dorsal anterior insula,
28:01it's doing something different here.
28:02That's the one in red in this top figure.
28:05And if you look at the pattern of
28:07the these subdivisions and how they
28:09connect with other parts of the brain,
28:11there's sometimes this is
28:12resting state up from rye.
28:13Again the there's some states in which
28:15all of the subdivisions are very
28:17similar like state three in the middle.
28:19But other states like state 2
28:21which only occur 5% of the time,
28:23which are much more divergent,
28:25which means the insular subdivisions
28:26are acting differently during that
28:28particular connectivity state.
28:30And the dorsal anterior insula is
28:32the most functionally flexible
28:33of all these subdivisions.
28:34So that's the one where if you
28:36look at different states,
28:37it's interacting with different brain region,
28:39has more connectivity partners
28:40with the rest of the cortex than
28:43the other subdivisions.
28:44And that's actually the same
28:46dorsal anterior insula as that
28:47same subdivision where we found
28:48that it you can use
28:50signals from that area to estimate
28:52what's going to happen at later time
28:53points in other parts of the brain.
28:55So that's that causal outflow hub that
28:57we identified many years ago using
29:00things like Granger causal analysis.
29:01So there's something curious
29:03about the dorsal anterior insula.
29:04I'll leave it at that.
29:07And again, there's some interesting
29:08structural connections there that you
29:10don't see in other parts of the insulin.
29:11There's some frontal and and subcortical
29:14projections that are detectable from
29:16the dorsal anterior insula that we
29:19don't see for the other subdivisions.
29:21So I I wanted to take that little segue
29:24into anatomy and and you know just
29:26the basic architecture of some brain
29:28areas because sometimes I think we'll
29:31see something like insula activation,
29:32but we may not look very carefully at which
29:35subdivision right or left what you know,
29:37what are we actually talking about.
29:38So I think being a little more precise
29:41on on where these things are going on
29:43really helps us you know hone in on
29:46what what the functional or dysfunction
29:48might be in a particular population.
29:50And so the ancillary cortex in adults can
29:53be divided into at least dorsal anterior,
29:55posterior and ventral anterior subdivision,
29:58probably more with the dorsal seeming
30:00to be more involved in the high level
30:02cognitive control or executive function.
30:04Things we've been talking about with the
30:07ventral a little bit more involved in
30:09the affective and emotional processing and
30:11the posterior more involved in somatic
30:13sensation and the dorsal anterior insulin.
30:15To me that's the special one which shows
30:17the most variable functional connections,
30:20kind of the greatest level of functional
30:22dynamics if you will and the greatest
30:25diversity in terms of the the brain
30:27regions it interacts with and and the
30:29test domains in which it's engaged and
30:32has some unique structural connections
30:34that we think might underlie some
30:36of that functional flexibility.
30:37So I was very heartened to see this
30:39paper from insulin and cut birth a
30:42while ago where you know they start
30:44talking about R doc and dimensional
30:46models and thinking about breaking
30:48down symptom based categories in
30:51psychiatry into a process where we
30:53where we're now all very familiar with.
30:55We're thinking about using genetic risk,
30:56brain activity,
30:57other markers to stratify samples
30:59into more data-driven categories
31:01which with the idea that this will
31:03eventually be good for us to identify
31:06subgroups that are more amenable to
31:08treatment on in one form or the other.
31:10The only reason I show this slide though
31:12is because in that figure from that paper,
31:14they have the insular cortex
31:16and I didn't plant that there.
31:18I had nothing to do with this,
31:20and I think it's probably just
31:21a random coincidence.
31:22But if I were them,
31:24I would also focus on the insular cortex
31:26in in all of my studies because it it
31:29does seem to be very much you know,
31:31whether you're interested in schizophrenia,
31:33anxiety, addiction,
31:34I mean autism,
31:36it doesn't matter you'll you
31:37can find papers that talk
31:39about the insula structure and function
31:41and connectivity that are aberrant or
31:43atypical or what have you in that disorder.
31:45So I I think part of my sort of crusade
31:48nowadays is to try to get people to read
31:51outside of their favorite disorder and
31:53and think about bigger picture models
31:55of brain function that can help us
31:57understand in a more domain general way.
31:59Like what could be the consequence of
32:02of something going wrong in this system
32:04and how would that look in early life?
32:05How would that look in late life?
32:07Which subdivision are we talking
32:09about and what how consequential
32:10is that for a particular behavior?
32:12So I love this figure had nothing
32:15to do with it.
32:16But I promise so now just to to get back
32:21to the the autism arm of of things we do.
32:24If I have time, OK, I do.
32:27I I like to think about flexibility and
32:29autism because I know there's so much
32:31great work done on social cognition and
32:33social communication and language and autism.
32:35But but I think a little bit
32:37less is known about flexibility,
32:39repetitive behaviors and their
32:41brain bases for various reasons.
32:43And now we're thinking more about this.
32:45And I think when you get into this
32:48transition to adulthood in in autism,
32:50you see some of these outcomes
32:53that are surprisingly not optimal.
32:55So you see things like 80% of
32:57individuals who are diagnosed with
32:59autism don't live outside the home
33:00or live independently as they,
33:02you know, become 18 and older.
33:04You also see these kind of
33:06surprisingly low employment rates.
33:08So you know kind of 80% unemployment
33:11in in young adults with autism and
33:13that's that's really shocking and
33:14and not what we would have hoped for
33:16especially if you look at some of
33:19these other early life disabilities
33:21where you see better outcomes.
33:23It's it's surprising to see these
33:25kinds of stats and and I think
33:28there are treatments,
33:29evidence based programs,
33:30school based programs that help to
33:32train up executive function and help
33:35to train cognitive flexibility.
33:37And I think some of that is is what
33:39we have to start thinking about
33:41like when and how to deploy these
33:44kinds of interventions.
33:45And we do have a lot of things like
33:47social skills trading and a lot of
33:50stuff that's really effective at at
33:52ameliorating some of those difficulties.
33:54But there's I think a lot more work
33:55to be done here in in flexibility.
33:57And the reason I think that
33:59flexibility in particular is,
34:01is what hinders people in that transition
34:03to independence is because that's
34:05sort of a a lot of what you do when
34:07you're like moving out of the home,
34:09getting a new job, making friends,
34:11social relationships,
34:11a lot of that is kind of moment
34:14to moment adjustment of behavior
34:16and and flexible cognition.
34:18Really it's not just social
34:20cognition but like you know,
34:21if the bus doesn't show up
34:22today and I need to get to work,
34:24I better scramble to find a different
34:26way to get to work. Otherwise,
34:27you know you're going to get fired.
34:29So. So things like that I think
34:32really do come into Stark.
34:34You know, we, we kind of noticed them more
34:38during these transition phases of life.
34:39But there's also these issues,
34:41not a clinician,
34:42but those who are will tell me that there's,
34:45you know, clearly heterogeneity and autism.
34:47So not everyone has flexibility,
34:48deficits, not everyone even shows,
34:50you know, executive function on problems,
34:53on these sort of parent reports,
34:55classic measures like the Behavioral
34:57Rating Inventory of Executive function.
34:59This is a data set from Stuart
35:01Mostofsky's group at Kennedy Krieger,
35:03and he's got kids between 8:00 and 12:00.
35:06Some of them are diagnosed with autism,
35:07some with ADHD,
35:08some with autism and ADHD and
35:10some typically developing.
35:12And if you do a latent profile
35:14analysis across these behavioral
35:15measures of executive function,
35:17you'll I like to focus here on the
35:19middle pie just to show you that a lot
35:21of kids with these diagnosis have just
35:23totally average executive function.
35:25They're not impaired.
35:26They're doing as well as a
35:28typically developing kid.
35:29In fact,
35:30the impaired kids tend to be
35:32those ASDADHD Comorbid kids.
35:33Perhaps not surprisingly,
35:35that double diagnosis really
35:37impairs or is associated with
35:39the impaired executive function.
35:41We found this kind of mixed bag
35:43as well in our own data set,
35:45typically developing kids and
35:46kids with autism.
35:47This was an honors thesis student
35:49from a student, Adriana Baez,
35:51who then went on to Med school.
35:53So I had a lot of, just incidentally,
35:54a lot of great undergrads who
35:56write papers as first author,
35:58and so that's been really exciting to see.
36:00But anyway,
36:01this was her honors thesis,
36:02and she showed again that a lot of kids with
36:06autism do have average executive functions.
36:08So it's there's heterogeneity there,
36:11as one might have expected.
36:12Not everyone needs the same
36:13kinds of interventions,
36:14right?
36:15So what if we could figure out what are
36:19the signatures of this heterogeneity?
36:20Is it something as simple as
36:22brain state transitions that are
36:24related to flexible behaviors?
36:26Is that something that's a marker
36:28of flexibility and can we see
36:30that in in our data sets?
36:32So the first thing we tested was,
36:33do kids or individuals who are
36:35diagnosed with autism show a
36:37reduction in the number of brain
36:39state transitions compared to the
36:41typically developing individuals?
36:42And so I know you're probably familiar
36:45with the Autism Brain Imaging Data Exchange.
36:47So this was something that Adriana
36:49Dimartino started a while back
36:51trying to get all of us autism
36:54researchers to share our data,
36:55make them publicly available.
36:57And this has resulted in hundreds of data
37:00sets being now available for download.
37:01So you can do things like
37:03independent component analysis,
37:05lighting,
37:05widow connectivity.
37:06And initially in this data set we
37:09do find some evidence for small but
37:11significant reduction in the number
37:13of brain state transitions for those
37:15diagnosed with autism compared to the
37:18typically developing individuals.
37:19This was another Emily Marshall in our
37:22lab as an undergrad honors thesis wrote.
37:25A paper is now a Med student doing great.
37:28So she looked at the dynamics of this
37:30salience or cingular insular network
37:32and found again a reduction in the
37:35frequency of brain states involving
37:37this network and its Co activation with
37:41others in a study she did as part of
37:43her honors thesis a few years back.
37:45So, so you know we are starting
37:47to find these signatures in the
37:49brain imaging data as well.
37:51And if again if you look at a task
37:53that was the last one was resting
37:54state but this is a task where you've
37:56got kids doing a a set shifting task,
37:58it's actually pretty easy.
38:00This was I think BJ Casey first came
38:02up with this paradigm but it's you
38:04know you see circles and squares and
38:06you have to pick the odd one out.
38:07Sometimes it's the odd one out
38:09is different based on shapes,
38:10sometimes it's color and in mixed
38:12blocks it could be shape or color.
38:13So that's the set shifting kind
38:16of part of it.
38:17It's a cognitive flexibility test.
38:20It's relatively easy to do and in fact
38:21all of our kids can do it easily.
38:23So behaviorally this isn't too challenging.
38:26But what we find is that if you look
38:28at Co activation patterns between the
38:30executive and salience and default
38:32three kind of networks I've been focusing on,
38:34kids with autism can do the task
38:36and their brains are doing the task.
38:38Similarly to typically developing
38:39kids as they go on and on and get
38:41into the fourth run of the task
38:43they actually have to engage more of
38:44this front of pride all there's more
38:47frequent patterns of that network occurring.
38:49So it's like they they can do things
38:51but the strategies might be different
38:52or the brain networks involved might
38:54change over time in ways that we can't
38:56see it with a whole brain activation
38:58analysis and we actually can't even see
39:00with a functional connectivity analysis.
39:02But we can see with these like
39:04smaller wind approaches that look at
39:06dynamics and don't look at averages
39:09but look at like changes across time.
39:11So that's something we've we found
39:13and are sort of following up on.
39:15So this is kind of a full circle about
39:19thinking about dynamic functional
39:20connectivity approaches to reveal a
39:23typical patterns of brain dynamics in
39:25prevalent neurodevelopmental conditions
39:26characterized by cognitive inflexibility
39:28like autism as I mentioned here.
39:31And in the extent to which these
39:33individual differences in brain dynamics
39:35underlie individual differences in
39:37flexible behaviors is is something
39:39we're we're actively investigating as
39:40well as as I'll mention in a moment
39:43other moderators or mediators of these
39:46relationships between brain and behavior.
39:48So this whole time before 2021 we
39:52were conducting this work in Miami,
39:54which has a huge bilingual population.
39:57So most of our participants actually
39:59were Spanish,
40:00English, bilingual and so one one thing
40:02we did which was completely unplanned.
40:05Was that we were able to test the
40:08relationship between bilingual
40:08exposure and executive function
40:10in our children in the study.
40:12So the RO one was focused
40:13on cognitive flexibility.
40:14Not just nothing to do with language,
40:16nothing to do with bilingualism,
40:18but just the way the the sampling occurred.
40:20We had a lot of this
40:22population in our sample.
40:23So it turns out there's a lot of controversy
40:26about whether or not raising a child
40:29with developmental language delays,
40:30whether raising them in a bilingual
40:32home will result in slowing down of
40:34language and cognitive development.
40:36But there's been a lot more work
40:39recently suggesting not only are there
40:42no negative consequences of either
40:44in cognitive or language development
40:45for being raised in a bilingual home,
40:47but there might even be mitigation
40:49of some flexibility deficits.
40:50And in fact,
40:51some folks are starting to report fewer
40:54executive function problems in dual
40:56language households or autism kids with
40:58autism raised in dual language household.
41:00So this is to me a really exciting
41:03kind of untapped potential to explore.
41:05It's almost like a natural intervention.
41:08So in the earlier days,
41:10I think social or sorry,
41:12communication disorders,
41:13clinicians would say, OK,
41:15if your child is having language delay,
41:17don't expose them to multiple languages,
41:18just stick with one.
41:20We don't want to confuse them.
41:21But it turns out that wasn't the best wisdom.
41:24I mean, there's really no negative effect.
41:26There might even be these boosts in executive
41:28function that we we can encourage with this.
41:30Not to mention all the other
41:32benefits of bilingualism.
41:33Right.
41:33So we found, you know,
41:34in our initial study with Celia Romero,
41:37she's just about to hopefully have this
41:40paper accepted just initial evidence.
41:42When you get to typically developing kids,
41:44there's almost a ceiling effect if
41:45they're good at executive function.
41:46The bilingualism doesn't really change that.
41:50It sort of, you know,
41:50doesn't affect anything at all.
41:53But if you look here,
41:53it's just one scale of the behavior
41:55rating inventory of executive function.
41:57If you look at it,
41:57the kids with autism being from
42:00a bilingual home was associated
42:02with better inhibitory control
42:04than being from a monolingual home
42:07in in that particular population.
42:09So there may be some boosts or
42:11protective effects or what have you
42:13in certain conditions that we don't
42:14see in a typically developing case
42:16because they may already be at ceiling
42:18or they may not be affected in such a way.
42:21So this is something we finally
42:23got an R21 to do and believe me,
42:25nobody wanted to fund this for some reason.
42:28But we're finally starting to
42:29collect data to really explore
42:31the links between bilingualism,
42:33executive function and brain
42:34development in 8 to 12 year old
42:36children with autism now at UCLA.
42:38And we put in an RO one to look
42:40at this longitudinally.
42:42So we're doing the data collection
42:43on the first set
42:44and then we want to follow them
42:45up across that adolescent phase,
42:47keep keep them going into
42:48young adulthood if possible.
42:49But we want to see how, you know,
42:52the initial bilingual exposure kind
42:54of plays out across this interesting
42:56period of adolescence when things are
42:59changing executive function, you know,
43:01deficits might be becoming more prevalent.
43:03We want to see if there
43:04are protective effect.
43:05Is there a bilingual advantage?
43:06Is there a boost?
43:08Nobody knows.
43:08And there's like no studies at all
43:10about bilingualism and the influence
43:12on brain development and autism.
43:14There's nothing that exists.
43:16So that's something we're really,
43:19as far as data collection goes,
43:20that's where we're going
43:21in the next few years.
43:23The last thing I wanted to
43:25mention was this is like,
43:26and now for something completely different.
43:29So we have so many of you know I've
43:31I've become super involved with the
43:33adolescent brain cognitive development study.
43:35So I'm the one of the associate
43:37directors now for justice,
43:39equity, diversity,
43:40inclusion for the entire consortium.
43:42In addition to the 20 hours of
43:44meetings a week that entails also
43:46a lot of knowledge about the ABCD
43:48data set that now we can use as a
43:51lab to to further our own goals.
43:52So this is the study of, you know,
43:5410,000 youth, Yale as a site,
43:56UCLA as a site.
43:58They're started in 2016 and now the
44:00kids were 9 and 10 at the first
44:02wave and now they're about 16/17/18.
44:05So it's 10 year longitudinal
44:08study imaging neurocog,
44:10substance abuse substance use data.
44:13So and I won't.
44:14I won't try to embarrass myself by
44:15pretending to know anything about addiction.
44:17This crowd knows about addiction.
44:19I don't.
44:19But what I do want to do is
44:21think about salience,
44:22network dynamics and how they might be
44:25related to substance use initiation.
44:27And I swear this wasn't my idea.
44:29A program officer asked me to do this study.
44:31So we're trying to do is is look
44:33at precursors of substance use
44:35initiation using this ABCD data set.
44:38Cause at 9:00 and 10:00 the
44:39kids aren't really doing much.
44:40They might,
44:41a few of them report sipping alcohol,
44:42but as you follow them up over
44:44time you can see, you know,
44:46they start to do things like try out smoking,
44:48try out this and that, and things change.
44:50And that's what this study has meant to
44:52track like mental health trajectories
44:53and substance use initiation trajectories.
44:55And what we want to do is use
44:57predictive modeling.
44:58Of course,
44:58connectome based predictive
44:59modeling was invented here at Yale,
45:01you all know about it.
45:02But we want to use dynamic connectome
45:04based predictive modeling to try to
45:07see if anything about the salience
45:08network and its initial state
45:10predicts vulnerability to later
45:12substance use initiation using the
45:14machine learning and some of the
45:17connectivity approaches that I mentioned.
45:19And we think this will be interesting
45:21because as far as I know about addiction,
45:23most of a lot of the studies
45:24that are done
45:25in adults aren't able to really tease
45:27apart whether the brain difference
45:28is seen are the consequence or
45:30the cause of the of the addiction.
45:33So you see kind of what happens
45:35after years of of, you know,
45:37being a user of some substance,
45:38but we don't know necessarily if
45:41there's like some vulnerability
45:42in some circuits or things like
45:44that that pre preceded the onset.
45:46And so the adolescent brain cognitive
45:48development study actually allows
45:49you to do that 'cause you'll,
45:50you might, you know,
45:51be able to predict what some of
45:53these vulnerabilities are related
45:55to inhibition and and you know,
45:57the networks that are involved in in
45:59flexibility and things like that.
46:00So this is something we're trying
46:02to get funding to do as well.
46:04Just pure data analysis of the ABCD
46:07study and we'll see how that goes.
46:10But basically as I start out talking
46:12about why we're interested in flexibility
46:15deficits and flexible behaviors can
46:18be impact day-to-day activities,
46:20forming and maintaining relationships,
46:22employment and independent living.
46:24So we think that understanding the neural
46:27mechanisms underlying flexibility deficits,
46:28you know across autism is our big focus.
46:30But across lots of clinical
46:32conditions we think can be important
46:35for tailoring therapies,
46:36predicting responses,
46:37thinking about risk and resilience for
46:40for various things like addiction.
46:43And so eventually,
46:44you know our kids can can focus
46:47more about their interactions in
46:49the world and less less on the
46:51specific repetitive behaviors and
46:53routines that might hold them back.
46:55And so this is the lab at UCLA who does
46:58all the work and my collaboration at
47:00work on the right some of the funding
47:02and the people who have donated time data,
47:06code of resources,
47:08mentorship and collaboration
47:09and friendship over the years.
47:11And this is the not where I do my work,
47:14but this is, this is UCLA.
47:16And I'm really happy to take any
47:18questions and have discussions.
47:19Thanks again.