Multimodal neuroimaging markers of transdiagnostic symptom domains in youth: The role of emotion regulation

Transcript

• 00:00My lab as a PGA,
• 00:03in twenty fourteen.
• 00:05Since that time,
• 00:07Kareem completed his doctorate in
• 00:09psychology
• 00:10and a t thirty two
• 00:12post doctoral training in translational
• 00:14neuroscience.
• 00:17He is a recipient of
• 00:18the Clinical and Translational
• 00:20Science Award
• 00:22and a K23
• 00:23Career Development Award from NIMH.
• 00:27In twenty twenty two, he
• 00:29received a Rising
• 00:31star award
• 00:32from the Society of Biological
• 00:34Psychiatry.
• 00:37So what are the secret
• 00:37of his success?
• 00:39Kareem, can I tell them?
• 00:44You have to learn the
• 00:45most technical method of fMRI
• 00:48data analysis and apply it
• 00:50to the largest dataset in
• 00:51the world
• 00:53to test mechanistic hypothesis
• 00:55about transdiagnostic
• 00:56developmental psychopathology.
• 00:59So Karim will tell us
• 01:00what he learned.
• 01:08Hey. Thank you, Dennis, for
• 01:09those kind words and, generous
• 01:10introduction.
• 01:11It's,
• 01:12great to be here today
• 01:13and tell you about, my
• 01:15research program, which, as Dennis
• 01:17mentioned, is really focused on
• 01:18understanding emotion regulation
• 01:20impairments in child mental health,
• 01:22and what might go awry
• 01:23in these networks that play
• 01:25such a critical role in,
• 01:26the top down regulation of
• 01:28emotion to lead to increased
• 01:29risk for,
• 01:31child mental health,
• 01:32disorders.
• 01:35This is just,
• 01:36my disclosures,
• 01:38grants that are supporting some
• 01:39of the work I'll be
• 01:40discussing today.
• 01:42So in terms of the
• 01:44flow for today's talk, I'd
• 01:46like to start with an
• 01:46overview of what exactly motion
• 01:48regulation is,
• 01:50and focus on some of
• 01:51the work that's been really
• 01:52foundational for my lab's work
• 01:54and,
• 01:55really centering on one particular
• 01:57circuit that's been,
• 01:59consistently implicated in child mental
• 02:01health or this frontal limbic
• 02:03circuit.
• 02:04I'd I'd like to share
• 02:05with you some work regarding,
• 02:07MyLab's
• 02:09focus on understanding the functional
• 02:11connectome,
• 02:12and wrap up with some
• 02:13of our ongoing studies and,
• 02:14new work and also spotlighting
• 02:16some, potential areas for collaboration.
• 02:20So we're going to be
• 02:21talking about a lot of
• 02:22neuroscience today where my brain
• 02:24ties specifically for this as
• 02:25you can see. So,
• 02:27let's just jump right in.
• 02:29When we think about emotion
• 02:31regulation,
• 02:32there's actually,
• 02:33no consistent
• 02:34definition of emotion regulation, surprisingly.
• 02:38And if we think about,
• 02:40the work that my lab
• 02:41focuses on, it's really thinking
• 02:43about, emotion regulation as the
• 02:45top down control of emotion.
• 02:47So recruitment of, related cognitive
• 02:50control networks,
• 02:51to modulate emotion generativity.
• 02:55And, there's different ways of
• 02:57categorizing emotion regulation.
• 02:59One,
• 03:00way of understanding emotion regulation
• 03:02is,
• 03:03thinking about adaptive strategies.
• 03:06And here we might think
• 03:07about cognitive reappraisal,
• 03:09problem solving, and,
• 03:11yoga and other,
• 03:14strategies. So cognitive reappraisal is,
• 03:17one approach that we will
• 03:18be focusing on today, especially.
• 03:21Another way of,
• 03:23thinking about this is if
• 03:24there's maladaptive,
• 03:25there's also maladaptive ways of,
• 03:28that can be conceptualized with
• 03:30emotion regulation.
• 03:31So two common strategies that
• 03:33fall under this,
• 03:34construct are suppression and rumination.
• 03:37And interesting,
• 03:39actually, suppression could be considered
• 03:40a maladaptive or an adaptive
• 03:42response in terms of inhibiting,
• 03:44one's emotional,
• 03:46ex expression.
• 03:49The for me, the most
• 03:50interesting thing about emotion regulation
• 03:52is,
• 03:54how this construct is really
• 03:56a commonality
• 03:57across several child mental health
• 03:58conditions.
• 04:00When I was a postdoc,
• 04:01this was an area that
• 04:02I spent,
• 04:03quite some time,
• 04:05thinking about and really thinking
• 04:06about the commonalities that we
• 04:08were observing,
• 04:09in our different neuroimaging work.
• 04:11And and I I observed
• 04:13these same networks and circuits
• 04:15coming up across,
• 04:16different conditions in my own
• 04:17clinical work.
• 04:19I I also observed a
• 04:20lot of, commonalities with emotion
• 04:23dysregulation,
• 04:25across child mental health conditions.
• 04:27These are just a few
• 04:28to,
• 04:29to spotlight. And today, we
• 04:30will be focusing on, disruptive
• 04:32behavior disorders and autism spectrum,
• 04:34but this is, in no
• 04:35way an exhaustive list.
• 04:39And when we think about
• 04:40emotion regulation as a transdiagnostic,
• 04:42construct,
• 04:43generally, studies have shown decreased
• 04:45use of adaptive strategies,
• 04:47such as, reappraisal,
• 04:49and increased use of maladaptive
• 04:52strategies such as, rumination or,
• 04:55or suppression as I mentioned
• 04:56earlier.
• 04:58If we, think about this
• 05:00on a neural level,
• 05:02there's been several meta analyses
• 05:04that have, really sought to
• 05:06understand what are the main,
• 05:08nodes or regions or networks
• 05:09involved in emotion regulation. I
• 05:11just wanna spotlight a few
• 05:12here. Again, this is not,
• 05:14an exhaustive list, but just,
• 05:16to spotlight a few hubs
• 05:18that, are known to play
• 05:19a critical role in, modulating
• 05:21emotions.
• 05:22And this really, spans the
• 05:24lateral prefrontal cortex,
• 05:26medial,
• 05:27prefrontal cortex as well,
• 05:29and temporoparietal,
• 05:31cortex. And, so I've broken
• 05:33this down by, subregions as
• 05:35well, because the the ventral
• 05:37and dorsal,
• 05:38prefrontal cortex are are really
• 05:40large functional areas that can
• 05:42be subdivided into many,
• 05:44subregions.
• 05:45But if we take a
• 05:46step back and think about
• 05:47these regions,
• 05:48and how they work together
• 05:49in concert and,
• 05:51to form networks,
• 05:52we can think about emotion
• 05:54regulation recruiting several cognitive control
• 05:56networks. So this includes,
• 05:58networks such as frontoparietal,
• 06:00frontal limbic,
• 06:01default mode, tension, and salience.
• 06:05So the the,
• 06:06message I really want to
• 06:07emphasize here is emotion regulation,
• 06:10recruit several large scale networks
• 06:12that have overlapping functions,
• 06:14and nodes as well.
• 06:17The process of emotion regulation
• 06:19is,
• 06:21something that continues to develop
• 06:22this dynamic. It it, develops
• 06:25throughout, young childhood, throughout even
• 06:27late adolescence.
• 06:29And,
• 06:30these, skills in executive functioning,
• 06:33cognitive control more broadly really
• 06:34come online and and really
• 06:36start fine tuning around the
• 06:37age of ten,
• 06:39and mid adolescence and onward.
• 06:41So this is really a
• 06:42process that,
• 06:44behaviorally,
• 06:46is fine tuning across development
• 06:47alongside,
• 06:49development of these functional networks.
• 06:51And this is actually just
• 06:52one aspect of,
• 06:55of, the brain's architecture when
• 06:56you think about it because
• 06:57there's functional networks and then
• 06:58you have maturation taking place
• 07:00in this, structural,
• 07:02aspects of the brain as
• 07:04well. So there's a lot
• 07:04happening during, this development.
• 07:07In general, we,
• 07:09expect to see,
• 07:12a fine tune,
• 07:13synchrony across regions involved in
• 07:15emotion control,
• 07:17and in dampening, motion generative
• 07:19regions.
• 07:23The frontal limbic circuit,
• 07:25typically is, might be thought
• 07:26to involve amygdala to prefrontal,
• 07:29cortex.
• 07:30And,
• 07:31this is a circuit that's
• 07:33been consistently
• 07:34implicated in
• 07:35several child mental health disorders,
• 07:37not just disruptive behavior disorders.
• 07:41But my work here is
• 07:42really focused on,
• 07:44initially and this has really
• 07:45been the foundation of a
• 07:46lot of my work early
• 07:47on in,
• 07:49focusing on disruptive behavior disorders,
• 07:51which, really has been an
• 07:52ideal,
• 07:54model disorder to to think
• 07:55of it because there's decades
• 07:57of, fMRI research that's focused
• 07:59on this area.
• 08:01And disruptive,
• 08:03mood and behavior disorders are
• 08:05typically characterized by, impairments and
• 08:07emotion regulation.
• 08:10When we think about disruptive
• 08:11behaviors, this can,
• 08:13include different classes of symptoms
• 08:15such as maladaptive,
• 08:17aggression, noncompliance,
• 08:18irritability, or anger.
• 08:21There's different ways to,
• 08:23categorize this as well, and
• 08:24my lab's work actually takes,
• 08:27a dual approach. And we
• 08:28think about this also as
• 08:29a DSM five diagnostic classification,
• 08:32that includes disruptive behavior disorders,
• 08:35that can be unpacked to
• 08:36oppositional defiant disorder and conduct
• 08:38disorder,
• 08:39as well as a transdiagnostic
• 08:41construct.
• 08:42What I mean by this
• 08:43is, thinking about disruptive behavior
• 08:46problems,
• 08:48in terms of cutting across
• 08:49several diagnostic categories,
• 08:52and and, occurring across a
• 08:54continuum in the population.
• 08:58In our own work, we
• 08:59found,
• 09:00impairments in
• 09:03emotion regulation capacity linked to
• 09:06disruptive behavior problems. And,
• 09:08this is work actually from
• 09:10a student, summer student in
• 09:12my lab who continues to
• 09:13collaborate, Olivia, Chocha, who,
• 09:15really focused on this for
• 09:17her summer internship.
• 09:19And this was,
• 09:21you know, very, you know,
• 09:23interesting to see this to
• 09:24this,
• 09:25direct relation ship between,
• 09:28on the x axis increasing
• 09:29emotion regulation impairment and on
• 09:31the y axis increasing,
• 09:33externalizing behavior problems. So this
• 09:35is, really what we have
• 09:38hypothesized. This is in a
• 09:39relatively small sample, but I
• 09:41wanna emphasize even at a
• 09:43sample this small, we still
• 09:44see these, these moderate effects.
• 09:49Now if we think about
• 09:50on a neural level, what,
• 09:52what transdiagnostic
• 09:53markers might be implicated in
• 09:55disruptive behavior? And this was
• 09:56really one of the first
• 09:57questions that has driven a
• 09:58lot of my work,
• 10:00started during postdoc and has
• 10:01continued actually,
• 10:03today in in in leveraging
• 10:04different,
• 10:06neuroimaging approaches to understand these
• 10:08networks.
• 10:09And one of the first
• 10:10questions,
• 10:11that we really sought, to
• 10:13to to address was understanding,
• 10:16if there's transdiagnostic,
• 10:18markers in this frontal limbic
• 10:20circuit that's associated with, maladaptive
• 10:23aggression in children.
• 10:24And,
• 10:25and and here, this was
• 10:26a transdiagnostic sample of one
• 10:28hundred,
• 10:29thirty three children, a large
• 10:31portion of which had elevated
• 10:32aggressive behavior.
• 10:34And and we use the
• 10:36CBCL aggression, and and this
• 10:37will be a theme actually
• 10:38throughout today. So So CBCL,
• 10:40or child behavior checklist is
• 10:41really one of the most
• 10:42widely used measure of, transdiagnostic
• 10:45symptoms.
• 10:47And the majority of participants,
• 10:49were diagnosed with oppositional defiant
• 10:51disorder,
• 10:53and they completed a task
• 10:54of implicit emotion regulation.
• 10:57So I should mention it
• 10:58with emotion regulation. And, another
• 11:00way to categorize emotion regulation
• 11:02is thinking about explicit and
• 11:04implicit emotion regulation.
• 11:07Now with explicit emotion regulation,
• 11:09this is typically what you
• 11:10might think of with reappraisal.
• 11:12There's, there's an instruction,
• 11:14participants are typically taught to
• 11:16down regulate emotions,
• 11:17and there's some type of
• 11:18effort, cautious effort that's involved
• 11:21in this process.
• 11:22When we think about implicit
• 11:24regulation,
• 11:25this is thought to occur,
• 11:28automatically without,
• 11:30cautious effort or instruction.
• 11:32So one common example
• 11:34or task of,
• 11:36engaging implicit regulation is,
• 11:39showing a series of emotionally
• 11:40expressive faces.
• 11:43And here, we,
• 11:45we we found that in,
• 11:47children with elevated aggression,
• 11:50severity,
• 11:51showed reduced connectivity between the
• 11:52amygdala and this region of
• 11:54the lateral prefrontal cortex,
• 11:56or dorsal lateral prefrontal cortex.
• 11:58I'll just abbreviate for DLPFC,
• 12:01but,
• 12:02this this region was,
• 12:04really,
• 12:05interesting,
• 12:06when we
• 12:08observe these findings, and and
• 12:09this is also a region
• 12:10that's known as a hub
• 12:11of emotion regulation. So I
• 12:13think what this is showing
• 12:14us is,
• 12:16children with elevated,
• 12:18disruptive behavior severity are showing,
• 12:21disruptions in this,
• 12:23front Olympic
• 12:25circuit that's that's really placed
• 12:27a critical role in modulating
• 12:30emotion generativity or overreactivity.
• 12:35We also know that, social
• 12:36impairments are very common,
• 12:38as well across disruptive behavior
• 12:40problems.
• 12:41This might,
• 12:43this might include difficulties with
• 12:44interpreting social cues, which often
• 12:44itself can be a trigger
• 12:44for,
• 12:49an anger response or frustration.
• 12:53One example might be hostile
• 12:54attribution bias where children might
• 12:56interpret
• 12:57a social cue as, as
• 12:59aggressive and and respond,
• 13:01with an anger response.
• 13:03And so here we wanted
• 13:04to ask, we,
• 13:06we know that social impairments
• 13:08are common to, disruptive behavior
• 13:10problems, in in youth, and
• 13:11we wanna understand,
• 13:13is is there,
• 13:15could the same circuit also
• 13:16play a role in,
• 13:18both aggression and social impairment?
• 13:21And to answer this, one
• 13:22of the best ways to,
• 13:25that that we
• 13:26we we address is was
• 13:28really,
• 13:29in a sample of children
• 13:31with autism. And we had
• 13:32two subgroups here. We had
• 13:33a sample of children with,
• 13:35autism and elevated aggressive behavior
• 13:37and autism with,
• 13:39low levels of, aggression, which
• 13:41is is termed here ASD
• 13:43alone.
• 13:44And there was one region
• 13:45that,
• 13:46really came to the top
• 13:47and was,
• 13:49in terms of, differentiating
• 13:51children with, autism with and
• 13:53without aggression. And this was,
• 13:55connectivity. This is showing connectivity
• 13:57between the amygdala and this
• 13:59region of the lateral PFC
• 14:00called the ventral lateral prefrontal
• 14:02cortex.
• 14:03Now this was interesting because
• 14:05this region,
• 14:06for the first time, we
• 14:07found associations where children with
• 14:09autism and aggression showed reduced
• 14:11connectivity in this, circuit relative
• 14:14to children with autism,
• 14:16without aggression. And it's interesting
• 14:18that,
• 14:19within a sample that's,
• 14:21in enriched for this social,
• 14:24impairment phenotype, we still find
• 14:26this, distinction in these circuits.
• 14:30This is just another way
• 14:31of plotting these findings,
• 14:33and where where we have,
• 14:37aggressive behavior on the x
• 14:39axis and connectivity strength on
• 14:41the y axis.
• 14:45So everything I've discussed so
• 14:46far has really focused on,
• 14:49understanding,
• 14:50one particular circuit and and,
• 14:52the approach that we can
• 14:54use to understand a particular
• 14:55circuit is what we might
• 14:56call seed based connectivity. So
• 14:58we can, take the time
• 15:00course of one region such
• 15:01as amygdala and test,
• 15:03its correlation to all other
• 15:04voxels across the brain. This
• 15:05is what we call c
• 15:07based connectivity
• 15:08or,
• 15:09c to voxel.
• 15:11Now another approach is,
• 15:14a bit more extensive in
• 15:15the way that,
• 15:16you can take,
• 15:19connections from every single,
• 15:21region of the brain. And,
• 15:22of course, there's many different
• 15:23ways to parcelate the brain.
• 15:24This can be a separate,
• 15:26discussion in itself. There's many
• 15:28different atlases out there
• 15:29and different ways of functionally,
• 15:32or structurally parceulating the brain.
• 15:34And
• 15:35and here, what you can
• 15:37do is,
• 15:39take each of these regions
• 15:40and correlate them to each
• 15:41other, and this is what
• 15:43we have when when we
• 15:43think about the connectome. So
• 15:45you you you get, very
• 15:47rich information about the brain,
• 15:48about this interconnectedness of the
• 15:50brain as well.
• 15:52It's interesting to think about
• 15:53when we
• 15:54think about the connectome and
• 15:56and, I guess, where we've
• 15:57come in in neuroscience,
• 15:59just thinking and,
• 16:01thinking about the use of
• 16:03the term human connectome,
• 16:05where we see there is
• 16:06a surge, or an increasing
• 16:08surge in,
• 16:09use of this term around
• 16:10twenty ten,
• 16:12which is which is where
• 16:13the human connectome project was
• 16:15launched. So,
• 16:17and I have other approaches
• 16:18here just for reference. So
• 16:19there's, amygdala connectivity and seed
• 16:22based connectivity,
• 16:23and I just placed these
• 16:24here for reference. But,
• 16:26you you know, hopefully, you
• 16:27you can,
• 16:29really get a sense for,
• 16:31how much on the frontier
• 16:32this is and understanding the
• 16:34connectome, especially with the launch
• 16:35of the human connectome project
• 16:37at that time.
• 16:40So as I mentioned before,
• 16:42with the connectome, you get,
• 16:44really a vast amount of,
• 16:46of,
• 16:47data about the brain.
• 16:49And, so just to,
• 16:51spotlight a few terms that,
• 16:54things that might be important
• 16:55to emphasize here,
• 16:57Nodes indicate
• 16:59are indicated by these spheres,
• 17:00and we can think about
• 17:01these as as nodes or
• 17:02regions.
• 17:04And, these lines are indicative
• 17:06of what we call edges
• 17:07or connections,
• 17:09between these nodes.
• 17:10And,
• 17:11now you you might see
• 17:12you're gonna see these rendered
• 17:13brains and these, different spheres.
• 17:16Another thing to keep in
• 17:17mind is generally in connectomics,
• 17:18the,
• 17:19radius of the spear denotes
• 17:20the extent of connections. So
• 17:22spears that are larger have
• 17:24more of an extensive,
• 17:26set of connections or edges.
• 17:29And when we think about
• 17:30how we can,
• 17:31analyze,
• 17:32the connectome or data from
• 17:33the human connectome,
• 17:35and today, I'm focusing on
• 17:36the functional connectome. So I
• 17:38should make that distinction here
• 17:39that there's structural and functional
• 17:41connectome.
• 17:42Here we'll be focusing on,
• 17:44the functional connectome.
• 17:47You can take this,
• 17:49data from the connectome, or
• 17:50we can,
• 17:52combine this with a machine
• 17:53learning approach. And the exciting
• 17:55thing about this is now
• 17:56we can start predicting clinical
• 17:57phenotypes or behaviors based on,
• 18:00brain connectivity patterns.
• 18:02We can plot this to,
• 18:04to assess performance for predicted
• 18:06and observed.
• 18:07We can take a step
• 18:08back and and,
• 18:10go from the node region
• 18:11to a large scale networks
• 18:12and start understanding,
• 18:16the within and between network,
• 18:17connectivity,
• 18:19of these,
• 18:20of these patterns.
• 18:23So one one,
• 18:25one area that,
• 18:27really launched us into this,
• 18:29area of connectomas was, taking
• 18:32this method and applying it,
• 18:34to predict aggression in, in
• 18:36a clinical sample of,
• 18:38children,
• 18:39with disruptive behavior disorders.
• 18:41So for this study, we
• 18:42had a hundred twenty nine,
• 18:44children with wide age range,
• 18:46up to, adolescents,
• 18:48and again using the same
• 18:49implicit,
• 18:50emotion,
• 18:52processing task. And here our
• 18:54question was, can we take,
• 18:55the functional connectome and can
• 18:57during task, and can we
• 18:58predict,
• 19:00our clinical phenotype of interest?
• 19:02In this case, it was
• 19:02aggression severity.
• 19:05And the answer to that
• 19:05was yes. We found that,
• 19:07just based on,
• 19:09brain connectivity patterns, we can
• 19:11predict,
• 19:11aggression severity with,
• 19:14a relatively,
• 19:15moderate effect size.
• 19:17And,
• 19:19if we start looking at,
• 19:21types of, nodes or regions
• 19:23that, may have played, a
• 19:25role in this predictive,
• 19:26model, we we again find
• 19:28this region of dorsolateral prefrontal
• 19:30cortex or DLPFC,
• 19:32but also VLPFC or ventral
• 19:34lateral prefrontal cortex.
• 19:36So it it's interesting that
• 19:38on the top panel, we're
• 19:39looking at,
• 19:40hyperconnectivity,
• 19:41so overconnectivity
• 19:43across regions that predicted aggression.
• 19:45On the bottom panel, we're
• 19:46looking at under connectivity that
• 19:48predicted aggression severity.
• 19:50So this model was largely
• 19:52driven by, over connectivity of,
• 19:55these lateral prefrontal nodes, but
• 19:56we did,
• 19:58we did see some prediction
• 19:59in, these temporal parietal regions
• 20:01as well, for the negative
• 20:03networks.
• 20:04So this is not for
• 20:06interpretation, but I I like
• 20:07showing this slide because it
• 20:09I I think it really
• 20:10emphasizes the interconnectedness
• 20:12of the human connectome.
• 20:14I would not ask someone
• 20:16to interpret this slide, but
• 20:17but you can see is
• 20:19definitely, as you look to
• 20:20the right of this slide,
• 20:22you can see that these
• 20:23extent of connections as we
• 20:25increase the threshold needed
• 20:27to be included,
• 20:28you can see that these
• 20:29prefrontal edges continue to survive
• 20:32at a very relatively high
• 20:33threshold. So,
• 20:35here, I I really just
• 20:36wanna emphasize that the human
• 20:37connectome is so vast and
• 20:39interconnected.
• 20:41I think we're really just
• 20:42starting to understand,
• 20:45much about a very,
• 20:46almost a Pandora's box.
• 20:49When we think about networks
• 20:50that play a role in
• 20:51predicting aggression,
• 20:54we really see this interconnected,
• 20:57network here, these features that
• 20:59are spanning several large scale
• 21:00networks
• 21:02across cognitive control, sensory motor,
• 21:04and emotion generative,
• 21:06networks.
• 21:11The next thing we did
• 21:11is, well, we we asked
• 21:13if we can take this
• 21:13model and can we predict
• 21:15aggression in
• 21:16children with, co occurring,
• 21:19conditions that typically are comorbid
• 21:21with,
• 21:22disruptive behavior disorders. And we
• 21:23found even focusing on this,
• 21:26DLPFC
• 21:27network or node,
• 21:28we could still predict aggression
• 21:30in children with elevated internalizing
• 21:32symptoms,
• 21:33ADHD,
• 21:35elevated social impairment transdiagnostically
• 21:37as well as in a
• 21:39a relative a very small
• 21:40group of children with autism.
• 21:41So there's something very,
• 21:44this, DLPFC,
• 21:46node and edges connected to
• 21:47it,
• 21:49is really,
• 21:51emerge as a highly predictive
• 21:52feature in predicting aggression severity.
• 21:58So one of the the
• 21:59great aspects of these large
• 22:00data sets and consortia is
• 22:02they they provide a wealth
• 22:03of data that would be
• 22:05very difficult to collect across
• 22:06one's career as a neuroimager.
• 22:10One,
• 22:12one benefit of this is
• 22:13to,
• 22:14test replication of findings or
• 22:16generalization of findings,
• 22:18and and and,
• 22:19and pulling different,
• 22:21neuroimaging
• 22:22data across sites.
• 22:24And one study that,
• 22:26my lab, does a lot
• 22:28of work with is, the
• 22:29ABCD study.
• 22:31And this is a study
• 22:32across twenty one sites,
• 22:34in the US.
• 22:35Children were between the ages
• 22:36of nine to ten in
• 22:37the initial release,
• 22:39and we're we're expecting really
• 22:41six point o,
• 22:43any day now, I think.
• 22:45So,
• 22:47so this is a a
• 22:47great dataset,
• 22:49to work with for our,
• 22:51from my lab and our
• 22:52research because there's really a
• 22:54vast,
• 22:55detailed clinical phenotyping of children.
• 22:58Even though children are are
• 23:00it's more of a community
• 23:01based sample.
• 23:02The the measures are quite
• 23:04extensive,
• 23:05and the neuroimaging measures,
• 23:07are really, tapping into,
• 23:10constructs related to regulation, reward
• 23:12processing, and,
• 23:14and cognitive control more broadly.
• 23:16So this was really ideal
• 23:17for us.
• 23:18And we took two tasks
• 23:20that are related to, this
• 23:21implicit face processing task that
• 23:23we worked with,
• 23:25and the stop signal task
• 23:26in ABCD is a task
• 23:27of inhibition.
• 23:29And we found that we
• 23:29could replicate these findings,
• 23:31using this task fMRI from
• 23:33ABCD,
• 23:34using this task of inhibition
• 23:36where we found a similar
• 23:37pattern,
• 23:39of lateral prefrontal nodes,
• 23:41and core, cognitive control networks
• 23:43predicting aggression.
• 23:45Then we we can also
• 23:46replicate these findings using the,
• 23:48emotional and back task, which
• 23:50is using actually a similar,
• 23:52NIMSTIM,
• 23:54set as we have in
• 23:55our implicit task. But we
• 23:56can also replicate these findings
• 23:58and find a similar pattern
• 23:59as well with lateral PFC
• 24:01emerging.
• 24:03So this was very interesting
• 24:05because I I think one
• 24:07thing to place this into
• 24:09context is,
• 24:11traditional
• 24:12models of aggression have really
• 24:14focused on this triadic model
• 24:15and thinking about,
• 24:17over reactivity of emotion generative
• 24:19circuits,
• 24:20underactivity
• 24:21within prefrontal circuits that are,
• 24:23involved in cognitive control or
• 24:25emotion control
• 24:27and disruptions and connectivity among
• 24:29the two.
• 24:31And here, what we've based
• 24:32on our findings and and
• 24:34also current,
• 24:36and also recent work from
• 24:37other groups is really started
• 24:39to suggest more of a
• 24:40a broader network dysfunction. And
• 24:42this is really,
• 24:44you know, I I think
• 24:45something that, we've observed in
• 24:46our work that,
• 24:48in predicting aggression and, again,
• 24:50this is not necessarily unique
• 24:52to,
• 24:53maladaptive aggression or disruptive behavior
• 24:55problems, but,
• 24:56this is something that,
• 24:58in in our opinion,
• 25:00most likely involves a broader
• 25:02network dysfunction spanning large scale
• 25:04networks involving cognitive control, social
• 25:06functioning, and emotion generation.
• 25:09This is not an exhaustive
• 25:10list, but this is just,
• 25:13as as a theoretical basis
• 25:14to to think about,
• 25:16these larger scale networks and,
• 25:19really the the complexity of
• 25:20these interactions,
• 25:22in predicting this phenotype.
• 25:30So for the,
• 25:32for the second half and
• 25:33the the,
• 25:34this segment of the talk,
• 25:35I'd like to focus on
• 25:36some ongoing
• 25:37studies and some, some work
• 25:39that my lab has been
• 25:40focusing on
• 25:41and and, thinking about where
• 25:43where to take, all of
• 25:44this next.
• 25:48Everything that I've mentioned so
• 25:49far has really focused on
• 25:51understanding static connectivity,
• 25:53And that's where we can
• 25:54take,
• 25:55the bold time course across
• 25:57a scan or a run,
• 25:58and we can average the,
• 25:59brain,
• 26:02activation across
• 26:03this time course.
• 26:05And we can refer to
• 26:06this as static connectivity.
• 26:09But this might not necessarily
• 26:10be the entire picture. When
• 26:12you think about,
• 26:13a scan, sometimes the assumption
• 26:14is a participant is going
• 26:16to, remain in a particular
• 26:18brain state when they're viewing,
• 26:19faces or or completing a
• 26:21task.
• 26:22But the reality is we
• 26:24don't really know that, and
• 26:25participants can actually,
• 26:28jump in and out of
• 26:29different,
• 26:30brain states throughout a scan.
• 26:31So the assumption that, there's
• 26:34one relatively
• 26:35stable brain state occurring throughout
• 26:37our
• 26:39our task or our run.
• 26:39It it might not be
• 26:41the give us the full
• 26:42picture. And, if you go
• 26:44back to, what I mentioned
• 26:45before from the connectome, we
• 26:46can derive so much rich
• 26:48information from this and analyze
• 26:49it in different ways and
• 26:51understand different aspects of the
• 26:52brain.
• 26:55So with, dynamic functional connectivity,
• 26:57the the exciting thing here
• 26:58is, we can take these
• 27:00functional,
• 27:01brain networks,
• 27:03and we can think about
• 27:04them as
• 27:06being derived and we can
• 27:08derive,
• 27:09specific latent brain states
• 27:11through these,
• 27:13from these functional networks. And
• 27:14the exciting thing with this
• 27:16approach is that we can
• 27:17start to understand,
• 27:19not just about the,
• 27:21disruptions within cognitive control networks
• 27:23or other networks,
• 27:25that's associated with the phenotype
• 27:27of interest, but the the
• 27:29more of a nuanced detail,
• 27:32in in this connectivity patterns
• 27:34as well. So here we're
• 27:35we're starting to understand,
• 27:37the transient properties of connectivity,
• 27:39the this,
• 27:40time varying,
• 27:42changes in connectivity moment to
• 27:43moment in the brain, which
• 27:44really is extraordinary,
• 27:46for me.
• 27:49And if we think about
• 27:50static connectivity, you have,
• 27:53the this is the the
• 27:54the arrows are showing you
• 27:55have a scan or a
• 27:56run length,
• 27:57and,
• 27:59the the, circles are, nodes
• 28:01or brain regions. And,
• 28:04across a scan, you might
• 28:05have a situation where, nodes
• 28:07a and b is correlated
• 28:08for some portion of the
• 28:09scan. And then in the
• 28:10middle panel, nodes b and
• 28:12c are correlated. And in
• 28:13the,
• 28:15the last panel, nodes a
• 28:16and c are correlated. And
• 28:17if that strength of correlation
• 28:19is strong enough,
• 28:20we can call that a
• 28:21network such as default mode
• 28:22or frontoparietal.
• 28:27Okay. There's animation to go
• 28:28with it. Okay.
• 28:31But another way of, thinking
• 28:33about this, if we,
• 28:35take the same data, but
• 28:36now we think about this
• 28:38in terms of brain states,
• 28:39the correlation between nodes a
• 28:41and b, we might derive
• 28:43this as one state.
• 28:45The correlation between nodes b
• 28:47and c, this would be
• 28:48another state,
• 28:50and correlation with a and
• 28:51c would be a third
• 28:53state.
• 28:54And
• 28:55so in this case, we
• 28:56can take these brain states,
• 28:57these metrics of brain states
• 28:59and dynamics.
• 29:00We can map it onto
• 29:01behavior and symptom dimensions as
• 29:03well,
• 29:07just in case you didn't
• 29:08see it the first time.
• 29:09Okay. So
• 29:10what are some of the
• 29:11metrics we can derive? So
• 29:12time and state is a
• 29:13is a very common and
• 29:15useful metric to have. How
• 29:16long is a participant in
• 29:17a particular brain state? And,
• 29:19of course, there can be
• 29:20several brain states.
• 29:22We might refer to this
• 29:23as occupancy time,
• 29:26or the amount of time
• 29:27a participant is in a
• 29:28state before transitioning to another
• 29:30state. And this nuanced term
• 29:31we call, dwell time or
• 29:33sojourn,
• 29:34time.
• 29:35The second,
• 29:36metric that we can derive
• 29:38is related to the state
• 29:39transition probability. So what is
• 29:41the probability that a participant
• 29:43will switch in and out
• 29:45of, states,
• 29:47multiple times? And, of course,
• 29:50there could be multiple transitions,
• 29:51and participants can stay in
• 29:53these states,
• 29:54for varying time lengths.
• 29:57This is just a different
• 29:58way of depicting what I've
• 29:59shown you with,
• 30:00with the brain animation.
• 30:03But you can see the
• 30:04time course and, the different
• 30:06lengths of time that a
• 30:07participant might be in a
• 30:07particular state, and this and
• 30:07these states can be reoccurring
• 30:07at different time points.
• 30:10States can be reoccurring at
• 30:11different time points. So this
• 30:13is really fascinating way for
• 30:14me to think about the
• 30:15brain.
• 30:16There's different ways to derive
• 30:18these metrics. You can do
• 30:19a sliding window approach,
• 30:21and you can also use,
• 30:23different computational methods such as
• 30:25hidden semi mark off modeling
• 30:27also to derive these states.
• 30:31So here we we wanted
• 30:32to understand more about,
• 30:34what exactly is driving some
• 30:36of these disruptions in these,
• 30:37connectivity,
• 30:39and connectivity across these, cognitive
• 30:41controlled networks. So we went
• 30:43to the ABCD dataset.
• 30:45And and here we we,
• 30:47really work with a smaller,
• 30:49sample of children, from this
• 30:50dataset just because of the
• 30:52high computational demand and and
• 30:54cost of,
• 30:56not actual cost, but computational
• 30:58cost of this,
• 31:01approach. And, so the children
• 31:03were in the first wave
• 31:04of ABCD, so they were
• 31:05nine to ten. We had
• 31:07a relatively,
• 31:09well gender balanced sample.
• 31:11And here we're we're taking
• 31:13the similar measures from CBCL.
• 31:16And here we we also
• 31:17work with wrestling state fMRI
• 31:19data.
• 31:20We derive these
• 31:22networks using a data driven
• 31:24approach called, ICA or independent
• 31:26components analysis,
• 31:28which we then mapped onto,
• 31:30the yoparcylation.
• 31:32And,
• 31:34and and this is actually
• 31:35the,
• 31:36the independent components.
• 31:38So we have thirty three
• 31:39independent components, which mapped onto,
• 31:41ten of these large scale
• 31:42networks.
• 31:43Now I I placed this
• 31:44slide here,
• 31:47just to really spotlight the
• 31:49the fact that
• 31:51even at rest, we can
• 31:52still identify and derive these,
• 31:56large scale networks. We can
• 31:57derive a,
• 31:59visual network, sensory motor attention,
• 32:01etcetera.
• 32:02For me, that's a fascinating
• 32:03thing about the brain. Children
• 32:05are laying in a scanner.
• 32:05They're looking at a blank
• 32:07screen.
• 32:09You can decompose these signals,
• 32:11using ICA, and you can
• 32:13still find these networks. And
• 32:15it's really a truly remarkable
• 32:17facet about the brain that
• 32:18even at rest, you're no
• 32:19one's engaging these circuits necessarily
• 32:21through a task. We can
• 32:23still find these networks.
• 32:26And when we take this,
• 32:29we we derive twelve brain
• 32:30states based on prior literature,
• 32:32and this was in collaboration
• 32:34with my colleague, Heather Chapelle
• 32:35at Wake Forest University.
• 32:37And the twelve state solution
• 32:38has been used in prior
• 32:39work, and,
• 32:41and we we wanted to
• 32:42keep this consistent.
• 32:44And what we found here
• 32:45was across these twelve, brain
• 32:47states that were derived, there
• 32:48was one state that was
• 32:50specific to,
• 32:53disruptive behavior problems. And this
• 32:54was after accounting for, internalizing
• 32:57an ADHD comorbidity,
• 32:59and that was
• 33:00state twelve. And
• 33:02state twelve showed that,
• 33:04and this is the the
• 33:05plot on the,
• 33:07the plot on the right
• 33:08is showing with increasing disruptive
• 33:10behavior severity, children are,
• 33:12showed greater occupancy time in
• 33:13state twelve.
• 33:15So, stated differently,
• 33:18children with, elevated disruptive behavior
• 33:20severity were basically getting stuck
• 33:22in this state. They were
• 33:23spending longer periods of time
• 33:25in this state. They weren't
• 33:25transitioning in and out of
• 33:27the state, as you would
• 33:28expect.
• 33:29And, well, what what's character
• 33:31what's
• 33:32special about state twelve?
• 33:34And we can we,
• 33:36to answer this, we compared
• 33:38state twelve to every single
• 33:39other state.
• 33:41And,
• 33:43basically, the red color on
• 33:45this color bar denotes,
• 33:46weaker positive connectivity. So state
• 33:48twelve
• 33:49is a state that we
• 33:50can conclude from this has
• 33:52glow is globally disconnected, but
• 33:54how it's disconnected, it's, shows,
• 33:57an overall pattern of hypo
• 33:59connectivity. So participants are spending
• 34:01longer in these underconnected states
• 34:04or state twelve.
• 34:06So,
• 34:08this this was great.
• 34:10And our next question was,
• 34:12can we do this again?
• 34:13You always need a sequel.
• 34:14Right? So,
• 34:16so we we actually,
• 34:18took participants with four runs
• 34:20of resting states. So we
• 34:21had two runs, which we
• 34:23we used here as a
• 34:24discovery sample.
• 34:25We had a held out
• 34:26two runs,
• 34:28which we ask, and we
• 34:29replicate findings in these held
• 34:30out two runs for every
• 34:31participant, and we reran the
• 34:33entire analysis again.
• 34:35And,
• 34:36and we found the the
• 34:38highly similar pattern
• 34:40where, this time it was
• 34:41for two states,
• 34:43that that that mapped on
• 34:44to disruptive behavior problem, but
• 34:45it it was the, really
• 34:46the similar
• 34:48association.
• 34:50Longer time spent in these
• 34:52two states
• 34:53mapped onto greater disruptive behavior
• 34:55severity. So children getting stuck
• 34:57in certain states,
• 34:58is potentially linked to,
• 35:01increased severity of disruptive behavior
• 35:07problems. So I I wanna
• 35:09make a jump from dynamic
• 35:11networks, which,
• 35:12is really an exciting area.
• 35:14And and,
• 35:15I guess, how does
• 35:17this this is really something
• 35:18that adds, for me, nuance
• 35:20into understanding these,
• 35:22impairments in cognitive control networks
• 35:24and understanding that,
• 35:26these,
• 35:29alter connectivity patterns might be
• 35:31further
• 35:32characterized
• 35:32by,
• 35:34alterations in the time varying
• 35:36properties of state switching in
• 35:39participants.
• 35:40And it might be that
• 35:41one one way to interpret
• 35:42this in addition is, if
• 35:44children are spending longer states
• 35:46in longer time in states
• 35:47characterized by under connectivity or
• 35:49disconnectivity,
• 35:52then,
• 35:53that could also reflect,
• 35:54reduced cognitive,
• 35:56flexibility or cognitive shifting across
• 35:58states, which is really,
• 36:01really a critical part to
• 36:02maintaining this cognitive flexibility that's
• 36:04also,
• 36:06linked to successful emotion regulation.
• 36:09And,
• 36:10so so this is one
• 36:12interpretation that,
• 36:14getting stuck in these certain,
• 36:16brain states might reflect reduced
• 36:18cognitive flexibility
• 36:19and time spent in other
• 36:20states that could,
• 36:22support,
• 36:23successful emotion regulation.
• 36:26So my lab,
• 36:28is a human neuroscience lab,
• 36:29and we,
• 36:32we we we have thought
• 36:33a lot about head motion
• 36:35in,
• 36:36in children. And,
• 36:39and do completing an MRI
• 36:41scan or fMRI scan is
• 36:42really not an easy thing
• 36:44for a child, most children
• 36:45or adults.
• 36:46It's,
• 36:48it's a very tightly enclosed,
• 36:50space as many of you
• 36:51may know. And,
• 36:52but at some point,
• 36:54you know, a a lot
• 36:55of my work
• 36:57with fMRI, we really started
• 36:58asking,
• 36:59well, who are whose data
• 37:01are we analyzing? And you
• 37:03have to you have to
• 37:04also appreciate that,
• 37:06the
• 37:07imaging data needs to have
• 37:08some type of,
• 37:10minimum quality control standards. So
• 37:13we wanted to understand more
• 37:15about,
• 37:16head motion and,
• 37:18whose data we're really analyzing
• 37:20when we who what data
• 37:21really makes it to this
• 37:22final stage, of analysis.
• 37:26And and this study was
• 37:27led by, several graduate students
• 37:29in my lab,
• 37:30Kavari, Zach, and, and Eleni
• 37:33who really did amazing work
• 37:34on this. This was an
• 37:35adventure project
• 37:37that two years later is
• 37:39is now impressed,
• 37:40and and everyone,
• 37:42you know, has has really
• 37:43put so much into this.
• 37:45And here we wanted to
• 37:46ask our transdiagnostic
• 37:48symptom domains linked to head
• 37:50motion in children. This is
• 37:51such a fundamental aspect
• 37:53of our work in in,
• 37:55developmental cognitive neuroscience. We took
• 37:58everything
• 37:59that ABCD study had to
• 38:00offer. We took functional, every
• 38:02single functional,
• 38:04mean motion and and,
• 38:06pass fail quality control data
• 38:08from every single
• 38:09functional scan, resting state task,
• 38:11diffusion,
• 38:13everything. E. Right? T one,
• 38:15t two weighted. And,
• 38:17and here we wanted to
• 38:18ask,
• 38:19does
• 38:20does,
• 38:22elevated symptom severity,
• 38:24how does that impact
• 38:26quality control pass and and
• 38:28motion during a scan?
• 38:30While other studies have looked
• 38:32at this,
• 38:33no this this was really
• 38:35the the first study to
• 38:36leverage this this size of
• 38:37of a dataset and and
• 38:39look at we we really
• 38:40wanted to take this in
• 38:41very nuanced,
• 38:42details and and our understanding.
• 38:45The the
• 38:47the bottom line from,
• 38:48some of our main findings
• 38:49here were higher severity of
• 38:51attention and disruptive behavior problems
• 38:53were linked with,
• 38:55increased head motion during scanning
• 38:56and increased likelihood of failing,
• 39:00quality control checks.
• 39:02But the the inverse for
• 39:03internalizing problem severity. So children
• 39:05who had elevated internalizing problems
• 39:07actually did better during scanning
• 39:08and had,
• 39:10lower motion and were more
• 39:11likely to pass
• 39:13quality control. And the plots
• 39:15here are just, for resting
• 39:16state just, for illustration of
• 39:18of some of these effects.
• 39:21And
• 39:22these are the the full
• 39:23extent of of the plots.
• 39:25But, it it just to
• 39:26give the the spotlight that
• 39:27we we really we looked
• 39:28across, modalities, and, there were
• 39:31some differences across each modality,
• 39:33but broadly,
• 39:34the the take home messages
• 39:35that I I presented here
• 39:36were,
• 39:37some of the main salient
• 39:39points.
• 39:40Why is this important?
• 39:42One is we know that
• 39:43motion can impact functional connectivity.
• 39:45There have been studies on
• 39:46this,
• 39:47that it can impact the
• 39:48the strength of connectivity.
• 39:51We know that motion can
• 39:52impact test retest reliability
• 39:54across networks that really are
• 39:56of interest for our group
• 39:58and many other groups, especially
• 39:59cognitive control.
• 40:02There's also genetic,
• 40:04heritable feature to motion, which
• 40:05is very interesting, and this
• 40:06comes from twin studies.
• 40:09Now the the
• 40:10really,
• 40:12I guess, the priority point
• 40:14here,
• 40:15for for my group is
• 40:16really thinking about we we
• 40:17do a lot of we
• 40:18we work with clinical populations,
• 40:21and we know that there's
• 40:22differences that in terms of
• 40:23clinical populations tend to move
• 40:25more than unaffected control populations.
• 40:28The problem with this is,
• 40:30there's the it's the idea
• 40:32and the, research has has
• 40:34shown that when
• 40:35we start selecting clinical subgroups
• 40:37with low motion, these clinical
• 40:39subgroups start to lose this
• 40:40heterogeneity
• 40:41that we want in our
• 40:42clinical samples. They tend to,
• 40:44be more phenotypically
• 40:45similar to controls, in terms
• 40:48of clinical severity.
• 40:50Younger children move more than
• 40:52older children, but, interestingly, there's
• 40:54actually inverse u,
• 40:56u shape association with motion.
• 40:57So younger children
• 40:59and,
• 41:00and into later adulthood,
• 41:02these these two subgroups actually
• 41:03show similar,
• 41:05motion effects.
• 41:06Males tend to move more
• 41:07than females, and we found
• 41:09that also in in in
• 41:10our research,
• 41:11on this motion study.
• 41:14And and there's some things
• 41:15that can be done to
• 41:16really make a scan easier.
• 41:17So,
• 41:18breaks helps, splitting up a
• 41:19session helps. So we found
• 41:20actually in a b c
• 41:21d, splitting,
• 41:23scan sessions,
• 41:25increased likelihood of success,
• 41:27and real time motion correction
• 41:29during resting state does does
• 41:31work. So I think a,
• 41:32b, c, d,
• 41:34PIs would be would be
• 41:35thrilled to know that, so
• 41:36we we confirmed that.
• 41:38So there's there's real time
• 41:39algorithms that you can use
• 41:40during your scan to correct
• 41:42for head motion.
• 41:46Okay.
• 41:49And I I guess thinking
• 41:51about and we so I
• 41:52I mentioned that my lab
• 41:53does a lot of work
• 41:53with a, b, c, d
• 41:54study.
• 41:55And another question we turn
• 41:56to a, b, c, d
• 41:57study to understand more about
• 41:58is is this link with
• 42:00emotion regulation. We understand transdiagnostic
• 42:02symptoms,
• 42:05can are are typically associated
• 42:06with impairments in regulation. This
• 42:06is,
• 42:12Zach and and Eleni in
• 42:13my lab.
• 42:14And this was really a
• 42:15fundamental question that that we
• 42:17addressed using the
• 42:19the entirety of the ABCD
• 42:21dataset and,
• 42:23really,
• 42:25understanding the distinct,
• 42:26and shared relationships
• 42:28with,
• 42:29emotion regulation difficulties that's associated
• 42:32with, increased severity across each
• 42:34symptom domain.
• 42:36And,
• 42:37and and Zach and the
• 42:38lady went on to unpack
• 42:39this further in terms of
• 42:41specific domains of
• 42:43regulation, and we found that
• 42:45increased levels of suppression, which
• 42:47I mentioned earlier, we can
• 42:48consider maladaptive
• 42:51strategy
• 42:52was linked to increased severity
• 42:54across,
• 42:54each of these,
• 42:56symptom domains.
• 42:57But we did not find
• 42:58an effect for reappraisal.
• 43:00So ABCD has been extremely
• 43:02valuable to take some of
• 43:03our hypotheses and test this
• 43:04in these large datasets,
• 43:06and and provide the statistical
• 43:08power that, would be sometimes
• 43:10very difficult,
• 43:11and take
• 43:12years to collect.
• 43:14And I wanna wrap up
• 43:16with,
• 43:17work from really our flagship
• 43:19study in my lab. So,
• 43:22we call it the BRAIN
• 43:23study. It it stands for
• 43:24something here.
• 43:26I to save time, I'm
• 43:27gonna skip that, so just
• 43:28go with the acronym. But
• 43:29the BRAIN study,
• 43:32the BRAIN study is is
• 43:33really bringing together,
• 43:35a a lot of the
• 43:36the themes that I've covered
• 43:37in today's talk. So here
• 43:38we have a school age
• 43:39population of children,
• 43:41with varying levels of disruptive
• 43:43behavior with and without autism
• 43:45so we can tease apart,
• 43:47and understand more of this
• 43:48nuance detail with,
• 43:51networks that are unique to
• 43:52disruptive behavior, unique to social
• 43:54impairment, and that are overlapping
• 43:56across the two.
• 43:58We conduct multimodal imaging, functional
• 44:00and structural MRI,
• 44:02as well as the deep,
• 44:03clinical phenotyping as well.
• 44:06And our main task here
• 44:08is really a task of
• 44:09reappraisal that was developed here.
• 44:12And,
• 44:15and and and what we're
• 44:16this is actually a very
• 44:17common, task that's often used
• 44:19in in, neuroimaging research where,
• 44:22there's different conditions, and we're
• 44:24teaching children how to explicitly
• 44:25regulate emotion across different,
• 44:28conditions. So they're looking at
• 44:30disgust inducing images. They're asked
• 44:31to,
• 44:33look at neutral images, and
• 44:34then they're they're asked to
• 44:35look at the disgust inducing
• 44:37images
• 44:38and down regulate their emotions
• 44:39by pretending it's fake. And
• 44:41this type of strategy we
• 44:43call distancing.
• 44:45And,
• 44:46this is actually a very
• 44:47common approach with fMRI,
• 44:50studies using this specific reappraisal,
• 44:54strategy of distancing.
• 44:56So in this,
• 44:57you know, preliminary sample, we
• 44:59have twenty five children twenty
• 45:01seven children, excuse me, nine
• 45:02to twelve years of age.
• 45:05This is not a gender
• 45:06balanced sample. And and, and
• 45:08our recent
• 45:10recruitment actually is is focused
• 45:12on,
• 45:14on balancing this this gender
• 45:15imbalance in our sample.
• 45:17But I wanna share with
• 45:18you in in the next
• 45:19few minutes some of our
• 45:20preliminary data. I wanna spotlight,
• 45:22our our labs, PGA, Gladys,
• 45:24who,
• 45:26puts in an enormous effort
• 45:27to,
• 45:29recruit recruit children,
• 45:31con conduct the scans,
• 45:33and
• 45:34and and at regular,
• 45:35time points and follow ups.
• 45:38And and given what I
• 45:38mentioned about considerations for head
• 45:40motion,
• 45:42Gladys has really led the
• 45:43way in thinking about this
• 45:44and,
• 45:45thinking about how to bring
• 45:46these skills to our in
• 45:48house data collection to make
• 45:49these experiences fun and, accessible
• 45:52for children.
• 45:54And,
• 45:55this is just to share
• 45:56some,
• 45:57some of our early data
• 45:58on this. But, so during
• 46:00the task, what we can
• 46:00see is children are looking
• 46:01at neutral,
• 46:03images. The the y axis
• 46:05throughout these next couple slides
• 46:06will be the affect rating,
• 46:07and it's on a Likert
• 46:08scale of one to five.
• 46:09But there's a significant difference
• 46:11even in this relatively this
• 46:13small sample,
• 46:14that we see an elevated
• 46:16affect,
• 46:17rating for look gross and
• 46:18a and a and a
• 46:19decrease,
• 46:20during the down regulate condition.
• 46:21So,
• 46:23this is really just showing
• 46:24that, we teach children how
• 46:26to reappraise,
• 46:27and this is in scanner
• 46:28data that's collected during the
• 46:30task that children,
• 46:33are most likely reappraising
• 46:35it in the scanner. And
• 46:36we do have a post
• 46:37scan,
• 46:38questionnaire also.
• 46:39Another way to think about
• 46:40this data is we can
• 46:41derive scores for emotion reactivity,
• 46:44and emotion regulation success. And
• 46:46if we look at this
• 46:47in
• 46:48a different way, we we
• 46:50can still see that children,
• 46:52in in this initial sample,
• 46:54this preliminary sample are, seem
• 46:56to be regulating and using
• 46:58reappraisal,
• 47:00to down regulate.
• 47:02So this work has been,
• 47:05you know, really,
• 47:07led by undergraduate student in
• 47:09my lab, Goen Li. This
• 47:10was part of her, senior
• 47:11thesis. What what a senior
• 47:13thesis was a lot it
• 47:14was a lot of data,
• 47:16that she analyzed in a
• 47:17very short amount of time.
• 47:18So,
• 47:21and one thing we can
• 47:22do is we can take
• 47:23this metric of regulation success,
• 47:27and we can test if
• 47:27this is associated with externalizing
• 47:29or disruptive behavior problems. And
• 47:31we see here
• 47:32actually and and, it's scaled
• 47:34differently here so that way
• 47:35larger numbers is better regulation
• 47:38success just for interpretation.
• 47:40Children who are better able
• 47:41to regulate during this task,
• 47:44show,
• 47:45lower levels of destructive behavior
• 47:47problems.
• 47:49And this is,
• 47:51interesting because it's even just
• 47:53in a small sample like
• 47:54this, we're still we're we're
• 47:55seeing some of these effects.
• 47:58Alice Dyer was, an undergraduate
• 48:00student in my lab who
• 48:01just graduated,
• 48:02one semester ago.
• 48:04And, this was Alice's thesis
• 48:06project, and and Alice looked
• 48:08at functional connectivity correlates of,
• 48:11regulation during this task and
• 48:13how this is linked to
• 48:14disruptive behavior. And,
• 48:16and here, Alice found that,
• 48:18we see reduced connectivity,
• 48:21between the amygdala and the
• 48:22lateral prefrontal cortex during regulation,
• 48:25that's associated with, externalizing
• 48:28behavior. So,
• 48:29stated differently, children with,
• 48:32increasing level of disruptive behavior
• 48:34problems showed reduced connectivity in
• 48:36this cognitive control circuit,
• 48:38during explicit regulation.
• 48:42So this study,
• 48:43for me is, you know,
• 48:46exciting because well, for many
• 48:48reasons, but one, because
• 48:51the everything up until now
• 48:52I presented has been implicit
• 48:53emotion regulation.
• 48:55Here, we're we're testing an
• 48:56explicit task of regulation to
• 48:58really,
• 48:59to engage these networks involved
• 49:01in regulation and understand,
• 49:04this explicit link between,
• 49:07top down control of emotion
• 49:08and disruptive behavior problems in
• 49:10children.
• 49:13And this is of course,
• 49:15we we,
• 49:16are also,
• 49:18looking to understand this, in
• 49:20terms of resting state in
• 49:21a, b, c, d and,
• 49:23in our ongoing studies. This
• 49:24is led by Eleni,
• 49:26and the ABCD study has,
• 49:28Eleni has taken time one
• 49:30and time two points,
• 49:31for children in the a,
• 49:32b, c, d. And here
• 49:33we wanna understand about this
• 49:34longitudinal
• 49:35trajectories
• 49:36of,
• 49:37of symptom change and how
• 49:38this maps onto the connectome.
• 49:40So can we take the
• 49:41connectome? Can we predict
• 49:42change in severity,
• 49:44in behavior?
• 49:46And,
• 49:47and and this is,
• 49:50really excellent work, and it's
• 49:51showing that,
• 49:53so this is baseline. This
• 49:54is,
• 49:55some work from Eleni's, analysis
• 49:58showing,
• 49:59we can take baseline connectomes
• 50:01and predict,
• 50:03destructive behavior severity, and we
• 50:04can see a pattern of
• 50:05frontal parietal and other cognitive
• 50:07control networks,
• 50:08that are emerging as highly
• 50:10predictive.
• 50:12In terms of symptom change,
• 50:14we can still predict change
• 50:16in symptoms. So taking the
• 50:18baseline connectivity in participants, we
• 50:20can predict,
• 50:21children who are getting worse
• 50:23in terms of symptom severity
• 50:25across time.
• 50:26And even though we still
• 50:28see a pattern of cognitive
• 50:29control and and also sensory
• 50:31motor networks,
• 50:32it's interesting that the pattern,
• 50:34the the key nodes are
• 50:36a bit shifted, so we
• 50:37can still predict. And it
• 50:38might be that different networks
• 50:39or subsets of nodes,
• 50:42are predictive of,
• 50:43cross sectional versus longitudinal trajectory.
• 50:46So this is something that
• 50:46we wanna understand,
• 50:48in in greater detail in
• 50:50MyLab's work.
• 50:53And
• 50:55and another area we've been,
• 50:57exploring, and this is also
• 50:59in collaboration with, my colleagues
• 51:01at at Haskins,
• 51:03Naveen, Ken, Vince, and and
• 51:05Gladys has and,
• 51:08I think Aslan,
• 51:09and Gladys has been, leading
• 51:11a lot of this effort
• 51:12with NEARS, but we wanna
• 51:13understand, can we take these
• 51:15markers from fMRI? Can we
• 51:16cross validate them in an
• 51:18approach that's cost efficient and
• 51:19has high translational potential to
• 51:21clinics?
• 51:23And,
• 51:24when we first started, I
• 51:25thought I thought participants were
• 51:27gonna love it.
• 51:29And they did not so
• 51:30the biggest complaint is the
• 51:32the pressure on the forehead.
• 51:33And I was shocked because
• 51:35the first two,
• 51:36NEAR scans,
• 51:38I and we have participants
• 51:39who did back to back
• 51:40back fMRI and NIRS. And
• 51:41when I asked
• 51:43our participants,
• 51:44so what did you think
• 51:45of NIRS? And they said,
• 51:46I still prefer,
• 51:47fMRI.
• 51:48So,
• 51:49but they do have an
• 51:50iPad that they're playing games
• 51:51with. So it's,
• 51:53but but, again, this is,
• 51:55really, I think, just a
• 51:56minor discomfort from, from the
• 51:58cap, but it's something to
• 51:59consider.
• 52:01Another area, and this is
• 52:02in in collaboration with with
• 52:03Kieran, is thinking about,
• 52:05combining what we know from,
• 52:07the functional connectome and epigenomics
• 52:10and understanding are there specific
• 52:12epigenetic markers that are linked
• 52:13with stress and emotion regulation
• 52:15that can help,
• 52:17help us understand more about
• 52:18these,
• 52:19these impacts of,
• 52:23across different levels of analysis.
• 52:25And this is a pipeline
• 52:26and workflow that we're developing
• 52:28in collaboration and thinking about,
• 52:30different types of biospecimen collection.
• 52:33So, to wrap up in
• 52:35terms of clinical implications, where
• 52:36is this all going? So
• 52:37in there's many different
• 52:39aspects of thinking about,
• 52:42what are the implications in
• 52:43terms of,
• 52:46developing brain based biomarkers and
• 52:48and neural markers to inform
• 52:50treatment.
• 52:50For my lab's work, it's
• 52:52I I see this as
• 52:53two areas that are very
• 52:54relevant for what we do.
• 52:55One is informing treatment decisions.
• 52:57And this is also based
• 52:58on some of, my own
• 53:00clinical work. And can we
• 53:01take some of that guesswork
• 53:02out of these clinical decisions?
• 53:03Can we help,
• 53:05inform this decision of of,
• 53:07children who might do,
• 53:11respond best to a cognitive
• 53:12behavioral intervention, psychotropic medication, or
• 53:14a combination of the two.
• 53:17The second area that I
• 53:19see as a potential,
• 53:20relevance here is target engagement.
• 53:22Can we, can we use
• 53:23these validated neural markers to,
• 53:28to understand is the treatment,
• 53:30engaging and enhancing the, circuitry
• 53:32of interest?
• 53:33And if it's not, this
• 53:34could also,
• 53:35give clinicians,
• 53:37several points along this treatment
• 53:39timeline where they can shift
• 53:41treatment or modify the treatment.
• 53:43This is just to show
• 53:44some of my work on,
• 53:46look,
• 53:47understanding the neural changes to
• 53:49treatment. But the idea here
• 53:51is,
• 53:52can we understand more about
• 53:53and take the guesswork out
• 53:55of who's going to respond
• 53:56best to a particular treatment?
• 53:59So I'll I'll end with
• 54:01this slide, and this is,
• 54:03you know, just thinking about
• 54:04how to really personalize our
• 54:06treatment approach. Is how can
• 54:07we take this information from,
• 54:09from the scanner, from neuroscience,
• 54:11and, and and, and and
• 54:13inform clinical decisions. This is
• 54:15really an important part of
• 54:16what we do. This was
• 54:18a participant who,
• 54:20developed her own way of,
• 54:23of communicating emotion regulation and
• 54:26feelings charts, thermometers.
• 54:28Those were really of, no
• 54:30interest to her. So she
• 54:32came up with her own
• 54:33way of,
• 54:34thinking about emotion regulation,
• 54:36as characters from a band.
• 54:38I won't say who the
• 54:39band is, but I put
• 54:39a hint on the slide
• 54:40because I found this to
• 54:41be a very difficult question,
• 54:43for people. So I put
• 54:45a hint.
• 54:46And, these different,
• 54:48characters reflect her different emotional
• 54:50states. Ichimoda was, her state
• 54:53for negative affect, and she
• 54:54needed help with regulating.
• 54:56And the the idea here
• 54:57is can we use neuroscience
• 54:58to,
• 54:59help children along this path
• 55:01of learning to regulate,
• 55:02and and reaching their potential?
• 55:06The band is My Chemical
• 55:08Romance.
• 55:09So and they're making a
• 55:10comeback from what I hear.
• 55:12So look out for them.
• 55:14Okay. So these are I
• 55:16wanted I wanna thank all
• 55:17of the students and trainees
• 55:18in my lab. They're they're
• 55:19the heroes. They drive everything
• 55:21that we do. They bring
• 55:22all this energy, exceptional ideas.
• 55:24Thank you to to all
• 55:25of you. And,
• 55:27and mentors, collaborators,
• 55:31thank you. So,
• 55:33happy to take any questions.
• 55:44Mhmm.
• 55:55Mike, they have a question
• 55:57at least. There's another one?
• 55:58Yes. First of all, thank
• 55:59you so much for this
• 56:00really
• 56:01spectacular work. It's a lot
• 56:03of work. So congratulations. Thank
• 56:05you.
• 56:06On the clinical implications that
• 56:07you said at the end,
• 56:08I was wondering
• 56:10very early in one of
• 56:11your early slides, you put
• 56:12the breakdown of the diagnosis
• 56:14of kids,
• 56:15and that seems like so
• 56:16retro. Right? Mhmm. ODD, whatever
• 56:18that means.
• 56:19So I'm wondering whether through
• 56:21your science on these approaches
• 56:23and perhaps with our doc
• 56:26Mhmm. She didn't mention. So
• 56:27I I wonder where it
• 56:27fits here. Right. Whether that
• 56:29could be a clinical implication
• 56:30moving from an ODD, whatever
• 56:31that means Mhmm. To a
• 56:33p value or a z
• 56:34value or something. Can you
• 56:35comment on that? Yes. Thank
• 56:36you for the question.
• 56:39So I, RDoc was,
• 56:42I guess, implicit in in
• 56:43I I I right? So
• 56:45we,
• 56:45when I mentioned transdiagnostic,
• 56:47I think one thing is
• 56:48we,
• 56:49we we do use a
• 56:51lot of these principles from
• 56:52RDoc and thinking about these
• 56:55networks that are transdiagnostic.
• 56:58And
• 56:58it's interesting that sometimes,
• 57:01studies have suggested that these,
• 57:05categorical approach such as a
• 57:07DSM five has not really
• 57:09necessarily led to any type
• 57:11of,
• 57:12validated biomarkers. And and, the
• 57:15translational
• 57:16approach might add, more of
• 57:18a nuance, might help advance,
• 57:20this might map more closely
• 57:22onto, I I guess, real
• 57:24world heterogeneity. So we we
• 57:27do, actually, we we do,
• 57:30conceptualize a lot of what
• 57:32we do transdiagnostically, and we
• 57:34we don't necessarily base inclusion
• 57:36on
• 57:37diagnosis cutoffs. We do collect
• 57:39diagnostic information for our our
• 57:41analysis, but, we we really
• 57:43base it on,
• 57:45severity of of symptoms and
• 57:47we do try to capture,
• 57:49this heterogeneity.
• 57:51And really who
• 57:52who is interested in if
• 57:53children wanna complete a scan,
• 57:55we will do our best,
• 57:57to acclimate them to this
• 57:58environment and and help make
• 57:59this a success for them.
• 58:01But we have been using
• 58:02other approaches such as, CCA
• 58:04to take,
• 58:06to map these symptom dimensions
• 58:07onto,
• 58:08neural markers, which we think
• 58:10might be more of a,
• 58:12potential approach for,
• 58:15elucidating transdiagnostic
• 58:16markers. And and other groups
• 58:17have used this type of
• 58:19canonical correlation approaches as well
• 58:21and other computational approaches.
• 58:23Thank you. I believe we
• 58:25have a question. Yes. Doctor.
• 58:27Brennan, if you'd like to
• 58:28unmute.
• 58:30Yes. Thank you for a
• 58:32great,
• 58:33presentation. This was very interesting.
• 58:36My question is,
• 58:38to we can actually hear
• 58:39you first seen. Sorry. Can
• 58:41you hear me?
• 58:44Can you hear me?
• 58:47Okay.
• 58:51Can you hear me?
• 58:54I'm probably not.
• 58:57I don't know why because
• 58:59I'm unmuted.
• 59:05Christine, can you try again?
• 59:06Okay. I'll try again. There
• 59:08we go. Great. Thanks. Okay.
• 59:09Alright.
• 59:10Thank you for a great
• 59:11presentation.
• 59:13This was very interesting.
• 59:15My question is,
• 59:16do you have any information
• 59:20on which
• 59:22social environments
• 59:23how social environments
• 59:25might contribute to the development
• 59:28of differential
• 59:29functional brain connectivity?
• 59:39Yes.
• 59:40Yes. Can you hear me
• 59:42okay
• 59:43on Zoom? Yes. Yes. I
• 59:44can. Okay. Okay. Great.
• 59:46Yeah. So,
• 59:48one one area I I
• 59:49didn't have time to go
• 59:50into was,
• 59:52you know, we we really
• 59:53collect I was mentioning,
• 59:55really, a quite extensive,
• 59:58characterization
• 59:59of measures from participants. And,
• 01:00:01in addition to emotion regulation
• 01:00:03and symptoms,
• 01:00:04we also wanna understand,
• 01:00:07children's,
• 01:00:08environment and exposure to childhood
• 01:00:10adversities as well,
• 01:00:12and protective
• 01:00:13factors for that matter.
• 01:00:15And,
• 01:00:16I think this goes back
• 01:00:17to,
• 01:00:19actually Zach and and,
• 01:00:21Eleni's project looking at emotion
• 01:00:23regulation a, b, c, d
• 01:00:25where they found,
• 01:00:27some very interesting findings,
• 01:00:29linking,
• 01:00:30children's exposure to family conflict,
• 01:00:34and,
• 01:00:34exposure to negative life events,
• 01:00:37as,
• 01:00:38linked to,
• 01:00:40increased
• 01:00:41impairment and emotion regulation and
• 01:00:43increased severity across,
• 01:00:46symptom domains. And within those
• 01:00:48two,
• 01:00:50areas of childhood adversities, it
• 01:00:52was really family conflict,
• 01:00:55that,
• 01:00:56well, it was family conflict
• 01:00:57and, exposure to negative life
• 01:00:59events that were two that
• 01:01:00we focused on, and this
• 01:01:02has also been, looked at
• 01:01:04in in prior work. So
• 01:01:05I would say,
• 01:01:07exposure to childhood adversities is
• 01:01:09something that,
• 01:01:11many groups are looking at,
• 01:01:12including our groups, to understand
• 01:01:14this link and more nuance
• 01:01:16with, disruptive behavior problems as
• 01:01:18well as other symptoms as
• 01:01:20well.
• 01:01:22Thank you.