Measurement in Neuropsychiatric Biomarkers

October 17, 2022

Information

YCSC Grand Rounds October 11, 2022
Adam Naples, PhD
Assistant Professor, Yale Child Study Center

ID8174

To CiteDCA Citation Guide

00:00It just feels good
00:02to see all these people here in this room,
00:04and it feels even better to have the
00:07opportunity to introduce Doctor Adam Naples.
00:10So I have known Adam and worked
00:12closely with Adam for a long time.
00:14Adam is a is a long-term denizen
00:17of the Child Study Center.
00:19He through a kind of interesting
00:21and circuitous path.
00:22Adam started out as a musician,
00:25was admitted to the Berkeley School of
00:26Music and was there and got exposed to music
00:29therapy and found the psychology part.
00:30Of the music therapy,
00:32maybe more interesting than the music.
00:34So he transferred to Cornell,
00:35got his degree in psychology,
00:37came to Yale,
00:38got his PhD in cognitive science,
00:40and has been here in various
00:41ways until he was promoted to
00:44assistant professor this past July.
00:46As one might infer from making the pivot
00:50from guitar to statistical genetics,
00:53Adam is a person who is good
00:56at a lot of different things.
00:58Adam is very good at statistics
01:00he's knowledge about. Genetics.
01:01He's knowledge about neuroscience,
01:03cognitive science.
01:04And I like to say if if anything
01:07operates using electricity,
01:09Adam understands it better than you and
01:12can make it work better than anybody else.
01:15He is wildly creative in the work
01:18that you'll hear about today.
01:20He has built experiments,
01:21he's built machines,
01:23he's built labs,
01:24and he's equivalently creative
01:26in terms of his ideas.
01:27And so you're going to hear
01:28very creative ideas as well.
01:31Adam is
01:32a he's a a very.
01:35Prolific collaborator.
01:36He is known in in our field internationally
01:39for his expertise specifically in
01:41things like eye tracking and EEG which
01:44both these electricity and he Umm,
01:46yeah, I'm going to be quiet.
01:49I want to just highlight one thing it is,
01:51it is such a it's a such a great
01:54achievement for Adam to be here as
01:56an assistant professor and it's also
01:58such a great achievement for Yale
02:00because as we recognize one of the
02:02challenges of the academic system
02:03is being as collaborative as Adam.
02:05Can in weird ways be attention
02:08with the promotional process.
02:09And so Adam being here is
02:11an achievement for Adam,
02:12and it's also an achievement for Yale.
02:14And really embracing Team Science is
02:15what we we know that we need to do.
02:18And with that, I give you Adam,
02:20who is my closest colleague and my very,
02:22very dear friend.
02:40Thank you, Jamie, that.
02:43Is a lots to follow. It's it.
02:47It is. As Jamie mentioned,
02:50I've been here for years and
02:53it's it's pretty amazing to be
02:55giving this talk right now.
02:57I've seen a lot of very, I've seen
03:01a lot of amazing talks in this room.
03:03For maybe the past like 18 years or so,
03:07and I'm just going to disable some of
03:10these distracting things over here.
03:14So as you mentioned,
03:16I'm a cognitive psychologist by training.
03:18My research is really focused
03:20on using measurement and methods
03:22from cognitive psychology to
03:23understand individual differences.
03:25And today I'm going to talk
03:26to you about a lot of our work
03:28studying biomarkers and autism.
03:29And so if anything's unclear,
03:31ask me a question and I'll try
03:33to make it more clear. And then.
03:35All right, you can see my mouse.
03:36Great.
03:39So just a quick overview, just I'm
03:41going to tell you what biomarkers are,
03:42why they're important because we hear
03:44this word a lot and I use it a lot,
03:46but it's always good to be reminded.
03:50I'm going to talk to you about large
03:52scale biomarker research and autism that
03:54have done a lot and a lot of this work.
03:56All the work I need to
03:57talk to you about today.
03:58I've done with Jamie as a we've called
04:01collaborated very closely for years.
04:03Then I'm going to talk about some new
04:05experiments that we've designed to be
04:07more inclusive for traditionally very
04:09underserved group of kids with autism.
04:11And then finally I'm going to end
04:13with some novel biomarkers that we're
04:15exploring to study social cognition,
04:16but so, so what's a biomarker?
04:19It's a defined.
04:20Characteristic that's measured
04:21as an indicator of a process,
04:23a pathogenic process.
04:26You can look at this definition for a
04:28while and think of all kinds of things,
04:29but biomarkers are things like heart rate,
04:31blood glucose, blood pressure,
04:34their routine measurements,
04:35they're reliable measurements,
04:37and their objective measurements.
04:39So when you go to Quest Diagnostics
04:40and they take their blood,
04:41they take your blood or someone else's blood,
04:43and you get back to that e-mail
04:45with hundreds of numbers, and all of
04:46those things are biomarkers, right?
04:48And some of them are better or worse for.
04:52Different kinds of decisions you might make.
04:54And that's that's really
04:55why we need biomarkers,
04:56because we need to make decisions
04:58about all kinds of things.
05:00We need to make decisions about
05:01who gets an early intervention.
05:04And these are psychiatric
05:05biomarkers I'm talking about here.
05:06But it's not just for psychiatric biomarkers.
05:10We need biomarkers for
05:12more accurate diagnosis.
05:13We need them for treatment selection.
05:15We need them for individualized treatments,
05:17and we need them so we can understand
05:20how treatments are going to
05:21evaluate how well they're going.
05:23So now these are some of these decisions.
05:27That's sort of an abstract way of saying it.
05:29But we might think,
05:30like if you have a kid who's three years
05:31old and they're learning how to read,
05:33you might think,
05:33do I want to put that kid in an early
05:36reading intervention right now?
05:37Right. That's an expensive process.
05:38That's there's limited accessibility to them.
05:41And is that going to help your
05:43kid become a better reader?
05:44If you're worried your kids at risk, maybe.
05:48Or you might wonder, like,
05:49well, you know, somebody said,
05:50my kids not sitting still enough,
05:51should I give that kid?
05:53A stimulant medication for that,
05:55and that's a decision that
05:57many parents struggle with,
05:59many people struggle with,
06:00and it would be really great if we had good,
06:02solid, objective measurements
06:03to help us make those decisions.
06:06They can even be things like,
06:07should I eat cake?
06:08Right, for people with diabetes,
06:10there's Thanksgiving comes along,
06:12they're measuring their blood glucose level.
06:13And then the third cake comes out
06:15and they think, well,
06:15should I have this third cake?
06:17Let me check where I'm at,
06:19and I'm going to give you an example
06:21of a very concrete biomarker,
06:22and I'm going to come back to
06:24it repeatedly here.
06:25It's heart rate.
06:25And I like heart rate because I
06:28think probably half the people in
06:30this room are measuring their heart
06:32rate right now on their watch.
06:33And it's a. It's it's ubiquitous.
06:37It's so ubiquitous.
06:37You don't think about it as a
06:39biomarker anymore.
06:39We just think of it like it's a
06:41measurement and it's such a good and
06:43reliable measurement that people use
06:44it for all kinds of things that people, the,
06:46you know, initial studies of heart rate,
06:48which are in the prehistory at this point,
06:51probably never conceived of.
06:55Things about heart rate.
06:56Well, it's routine.
06:57I measure my heart rate every second.
07:00I think on my wrist.
07:02It's reliable.
07:03My heart rate is under
07:05the same circumstances,
07:06the same as it is the day before,
07:08and it's subjective.
07:09So it's signing a light through my wrist,
07:11it's measuring the blood reflectance,
07:13doing some math to those signals,
07:14and then coming back with a number.
07:17That's not to say that
07:19that process is perfect,
07:21but it always gives you the
07:22same output for the same input,
07:24and that way it's objective
07:26and very reliable.
07:27And so now that we have this
07:29biomarker that's on my wrist,
07:30we can use it for all kinds of fun things.
07:33So well,
07:34we can use it for early intervention.
07:36And you can see there that little
07:38blue line by resting heart rate
07:40goes from 55 to 80 in a day,
07:42which is a little bit unusual.
07:43And that's when I got my
07:45second COVID booster right.
07:46The heart rate predicted the
07:48onset of a pretty prolonged and
07:51exciting experience dealing with
07:53that heart rate used in many,
07:54many diagnostic tests if you've been to
07:56a doctor and they've checked your heart rate.
07:58That's it.
08:00They follow it.
08:02I'm running out of time,
08:04so it's also used for treatment selection
08:06so I can go for a run on my watch and it
08:08suggests you know what you should do today?
08:10Based on my heart rate over
08:11the past couple of months,
08:12you should go for a 27 minute
08:14run at 11-5 minute mile,
08:16which is not very impressive.
08:18No,
08:18I don't have to follow this advice,
08:19but it's at least it's giving
08:21me an individualized treatment,
08:23right?
08:25And then it evaluates how well I did,
08:28and that treatment too.
08:29So if I fail to follow that,
08:30it uses the language of an NIH grant
08:32reviewer and it tells me that my my
08:35training status was unproductive.
08:36So this is, I think, kind of a a
08:39motivating example of once you get really,
08:41really good, inexpensive,
08:42reliable and effective measurements,
08:45then they might not be used for whatever
08:47you thought they were going to be used for.
08:50But now they're tools,
08:51they're screwdrivers,
08:52they're hammers.
08:52And so people can use them for
08:54whatever their imaginations.
08:55Guide them to their hypothesis generating,
08:59but we're not talking about heart rate,
09:01we're talking about autism.
09:02So I'm going to get to that quickly.
09:05All right,
09:05so talk about autism.
09:09I think the first time I saw this slide
09:12was in this room maybe 20 years ago.
09:15Was diagnosed early in life.
09:16It's heterogeneous syndrome.
09:18It's described as defined by
09:21persistent challenges in social
09:23communication and social interaction,
09:25restricted repetitive patterns of behavior,
09:27sensory sensitivities.
09:28What's common among all diagnosis of
09:30autism is that they're associated with
09:33challenges and reciprocal social behavior.
09:36But the mechanisms for autism are unknown,
09:38and our current current tools
09:40for assessment are subjective.
09:41Report. They're subjective report
09:43from a person who's.
09:45You know,
09:45reporting on themselves from a parent,
09:47reporting on their child or
09:48caregiver on their child,
09:49someone interacting with the person.
09:51And it's not to say these
09:52aren't very good measures,
09:54but they're still not necessarily
09:55the same thing as the same exact
09:58output for the same input that we get
10:00from shining a light for your wrist.
10:03Umm. And so.
10:06With that heterogeneity,
10:08with that unknown mechanisms,
10:09there's there's many theories of what
10:12the underlying causal deficits are in autism.
10:15And I'm going to really briefly
10:16go over a couple of them.
10:17And the point here is to give
10:19you an idea of the breadth and
10:20how these theories hit different
10:22levels of abstraction rather than
10:23they go into detail about anyone,
10:25because that would be hours and hours,
10:27that would be a dissertation defense,
10:29so.
10:31I could talk about social motivation,
10:34predictive coding.
10:34These are different levels of analysis,
10:37just regulated arousal.
10:39And the imbalance or imbalance of
10:42excitation and inhibition in the brain.
10:46The social motivation which
10:47has a strong history here.
10:49Both Jamie and Cara published on it
10:51together here and at other universities.
10:54Briefly,
10:54and I'm terrified to talk
10:56about it in front of you guys,
10:58we can describe this as reduced motivation
11:01to interact and attend to other people.
11:04And this starts early in development.
11:05And so the cascading consequences of this
11:07are you pay less attention to other people,
11:10you interact with people less,
11:12and so you learn less about other people.
11:14And it's harder to develop sort of
11:16social mastery and expertise about
11:18other people if that's not the
11:20world that you're interacting with.
11:23So this is really, really.
11:24Primarily inspired by the
11:26social symptoms of autism.
11:30So. Secondly, I'm going to
11:33talk about what's called the
11:34predictive coding theory of autism.
11:36Predictive coding is a if you Google this,
11:37you get a lot of things that
11:39don't have autism to do with it.
11:40But the idea is that in the brain,
11:42the brain has worked.
11:43Brain works by continuously generating
11:45predictions about the environment,
11:46and these are implicit, not explicit.
11:48Like I'm going to make a prediction
11:50about what's going to happen today.
11:51These are things like low level
11:53sensory input all of the way up,
11:56and most of these predictions are wrong.
11:58And so we learn about the world.
12:01Based on how we deal with the amount
12:04of wrongness of these predictions,
12:06and there's some evidence that
12:08people autism learn from those
12:09prediction errors differently.
12:11And similarly to social motivation,
12:14this is happening early in
12:15development and it cascades.
12:16So starting from, you know,
12:17very young age.
12:20They're learning about the
12:21environment differently,
12:22but in this case it's not specific
12:23to social things.
12:24It's more kind of kind of about everything.
12:27And the a lot of the inspiration for
12:29these theories comes from the sort of
12:32sensory symptoms that people often
12:33report increased sensitivity sounds
12:35or people will also report that some
12:38kids have decreased sensitivity to.
12:422 sounds to.
12:45All right. I just got a high heart rate
12:48alert from my walk and so it's a I'm going
12:51to take that off before it calls 911.
12:57So that that's predictive coding,
13:00not necessarily social,
13:01still about learning and learning
13:03differently about the world.
13:04And you'll see that most of the time we
13:05were talking about anything developmental,
13:06it's about learning about the world.
13:09Alright. Now dysregulated arousal,
13:10I've put these references here at the bottom
13:13just to show that in 1961 they're doing
13:15the same experiments that we're doing now.
13:17So we think about arousal,
13:18we're talking about these overall
13:19states of brain and body.
13:21We think and describing this sort of the way
13:25people describe or things like sweating,
13:27high heart rates, you know,
13:29behavioral activation,
13:30you can't sit still.
13:31Alternatively,
13:31insensitivity, you know,
13:32you're not sensitive to sounds
13:35or external stimuli or sleep,
13:37you're.
13:40It just less sensitive.
13:42So at the extreme ends we have sleep
13:45for arousals like the low end.
13:47And this there is evidence that
13:49these are always are typical in ASD,
13:51primarily from physiological research.
13:53So disregulated Physiology and heart rate,
13:56sleep, sweating, pupil diameter,
14:00sound sensitivity, sound insensitivity,
14:02pain sensitive pain insensitivity.
14:07And then lastly,
14:08there's this theory of that autism.
14:10These symptoms emerge from an imbalance
14:13of excitation and inhibition in the brain.
14:16And the attractive thing about this
14:19theory about many of these is that.
14:22A dysregulation of excitation in the whole
14:25brain is going to impact everything.
14:28And autism is tremendously heterogeneous and
14:31their symptoms that emerge that are sensory,
14:33their cognition, their motor,
14:35which I haven't talked about.
14:37There's also an increased
14:39occurrence of seizures and autism.
14:41And so. That's the we call that
14:45the EI balance just for.
14:47So lots and lots of theories.
14:48They're hitting at different
14:49levels of abstraction.
14:50But, but I'm a cognitive psychologist.
14:52And so when I was an undergrad,
14:54somebody gave me this book
14:54and they said to read it,
14:55and I couldn't get too far
14:57and it was too hard for me.
14:59But all those theories are not
15:01really mutually exclusive.
15:03And this book came out and it's about vision.
15:07It's not about autism.
15:07But it gave sort of a rubric for
15:09thinking about lots and lots of theories.
15:11And if the idea is that, look,
15:12we can break down any information
15:14processing system and into how
15:16it's approaching its problems,
15:17there's a computational level which is just.
15:19What's the problem?
15:20It's trying to solve?
15:21There's the algorithmic level,
15:22which is the the individual steps that
15:25you have to take to solve this problem.
15:26And anyone who is in a cognitive
15:28psychology class in the late 90s,
15:30there's lots and lots of
15:31boxes and arrows for lots,
15:32you know these these super complicated
15:35models that did chess and cooking and,
15:37you know flight path organization and
15:39said this is how the brain works.
15:42And then there's the implementation
15:43level and that's what what's actually
15:45doing the computing in a computer,
15:47it's silicon chips and a brain, it's.
15:49Green.
15:53So I feel like I think about these things,
15:55and I said, can we align them in kind
15:57of a meta theory because they're
15:59not necessarily mutually exclusive?
16:00We can think about social motivation as sort
16:02of addressing this computational problem,
16:04right? We want to make friends.
16:05We want to learn about other people.
16:07What are the steps we go through
16:09to do that predictive coding is?
16:13You know, really fits in sort of
16:15this algorithmic level and it's
16:17aligned with these sort of low level
16:19information processing demands where
16:21we deploy our eye movements and seems
16:23how we learn from different kinds of
16:26statistical regularities in the world.
16:28And then from the implementation level,
16:30we have excitation and EI and
16:32we have just regulated arousal.
16:34Cells firing you know too much or body
16:38state and brain state up or down.
16:41Right.
16:42So these theories generate a lot
16:45of hypotheses,
16:46but what we don't know and so I
16:48was briefly went through this but
16:50you know hundreds of papers on
16:52on many of these topics.
16:53There's a lot of research in autism
16:55as I'm sure you are all aware,
16:56but we we need to kind of know which
16:58ones work before we put it on our wrist.
17:01And even then it's not perfect.
17:02It's almost called the police.
17:04So I'm going to talk about biomarker,
17:05not the police,
17:06just just 911 in general.
17:07And ASD, I'm going to talk to you
17:09about the tools that we use here.
17:11All right.
17:12So EG what's EG EEG is let
17:15me know if I stray too far.
17:17It's the sum of the ongoing cortical
17:20activity recorded of the scalp
17:22reflects the excitatory and inhibitory
17:24postsynaptic activity in the cortex.
17:26Way we measure it is we take
17:28this soft spongy cap with little
17:30electrodes in the wires.
17:31You put it on your head and we use
17:33those little electrodes with a very
17:35expensive amplifier to amplify the
17:36very faint electrical signals generated
17:38by your brain into a recording.
17:40And the recording looks like this.
17:42So each one of these wiggly
17:44lines is the voltage.
17:45Measure your scalp, and this is amplitude.
17:48So how loud, how big the voltage is.
17:50And this is time,
17:51and this is in milliseconds.
17:52This is really, really,
17:53really, really, really fast.
17:54A lot of lines means we're
17:57using lots of electrodes.
17:59And we can learn a lot of things from
18:00this ongoing EG even when people are just,
18:02when they're not doing anything.
18:04One way is we can look at the relative
18:07contribution of the different frequencies.
18:09So the fast wiggles and the slow wiggles.
18:12And what's really interesting is
18:13that the excitatory and inhibitory
18:15neurotransmitters in the brain
18:17have different time constants,
18:18which means that the really lines
18:20wiggle at different speeds for
18:22excitation and they do for inhibition.
18:24So, and I apologize,
18:25I think the language of wiggly lines.
18:27So there's people here who are like serious.
18:29EG computational researchers
18:30and they're like,
18:32you know,
18:33throwing their coffee
18:33on the floor angrily and
18:35leading. But this is still we need to agree,
18:37this is still just a lot of wiggly lines.
18:39Lots of them. So.
18:42Sitting and doing nothing.
18:43It's like a great experiment because we
18:46can use it to measure the I balance,
18:47and this is what's called
18:49the power spectral density.
18:50So what it's measuring on the left
18:53are the volume of the amplitude.
18:56I'm using my hand again of the slow wiggles,
18:59and over here the fast wiggles and
19:01this relative activity across the
19:04frequencies indexes this balance, right?
19:06Like how much excitation,
19:07how much inhibition you have.
19:09Shallower slopes, yes, shallower.
19:12Groups are more excitation,
19:14steeper slopes are more
19:16inhibition and this also changes.
19:17So this is a pretty reliable effect
19:19if you give people different kinds
19:22of pharmacological interventions.
19:23So if you give people Ambien,
19:25well actually goes the other way,
19:27so it gets much steeper.
19:28So if people are giving Ambien,
19:30given benzodiazepine,
19:30you see these inhibitions really,
19:32really ramp up. Also this happens in sleep.
19:35So that's one way.
19:36We can look at the EEG 2nd way and this is
19:41again the computational neuroscientists.
19:43You think this is too basic our
19:47event related potentials so.
19:48This is where we average the EEG activity
19:50around repeated presentations of an event.
19:52Like we show you pictures of face,
19:53we show you pictures of a house,
19:55and then we average the activity.
19:57And so these little chunks of
20:00those activity of these wiggles.
20:02And what that does is it gets rid of
20:03the wiggles that don't have anything
20:05to do with what you're interested in.
20:06And it accentuates,
20:07it amplifies the wiggles that you
20:08do care about and you get what's
20:10called an event related potential,
20:12which has relatively few wiggles.
20:14But those wiggles end up having
20:16names and they're important for
20:17learning about brain activity.
20:18And down here time this is in milliseconds.
20:22So this is really really fast.
20:24So I want to highlight this is like
20:26that's a feature about EEG and ERP's
20:29that is not captured in many other. Umm.
20:33Somebody put something in the chat.
20:34I'm going to try not to be distracted by it.
20:38And we're going to talk to you about
20:40now probably the most important ERP
20:42for this talk, which is the N 170.
20:46So the N 170 and it's in this
20:49red circle right here.
20:51When we show people faces compared to almost
20:54any other visual object in the world.
20:57Almost.
20:57You get this much more negative.
21:00Even the negative ones we call peaks much
21:03earlier peak to faces than anything else.
21:06We called the N 170.
21:07This is 170 milliseconds.
21:09This is really fast,
21:11right?
21:11We other kinds of measures that are slower.
21:17Are reflecting some different kind
21:18of thing like if you hit a button
21:19when you see a face,
21:20that's like what you saw the face.
21:21But then decision process has
21:22worked in this is at the level of
21:24like action right now.
21:25This is what is probably,
21:27we think subserving how we interact
21:28with people in the world, right.
21:30When you look around,
21:31every time I see a face,
21:32that part of brain is sort of firing
21:34up and saying up face something bigger
21:36right there and it's it's meaningful.
21:40Umm. So it's a selective to faces.
21:44It's really very early, which I'll
21:46keep saying it's sensitive to context.
21:48So if I tell you that like I'm
21:50going to show you a bunch of faces
21:51of people who are judging you,
21:52judging how you dress.
21:53In fact was a recent study,
21:55you're 170 is different,
21:57it's sensitive to gaze too.
21:59So if gays changes,
21:59when you look at a person's face
22:01and they look away, you get in 170.
22:03So movements of eyes seem to be really,
22:06really important for understanding
22:07this really early brain activity which
22:10is meaningful and social interaction.
22:11Guys are important.
22:13Can they move fast?
22:15So Umm,
22:17now just briefly this is 1 slide
22:19and it's not nearly enough on
22:21EEG findings and ASD that are not
22:23all of them and but are relevant
22:25to this talk which is one.
22:27I guess I'll start with the 4th point,
22:30which is in general patterns of
22:33findings are heterogeneous and
22:34that's the case in in many fields.
22:37But the most consistent among
22:39inconsistent findings is is delayed
22:40in 170 that people with autism it's
22:43less efficient phase processing or less.
22:45Less fluent space processing.
22:47There's also evidence that that
22:49the profile that spectral profile
22:51is atypical in that it's.
22:54It's the shape seems to be different.
22:56Specific features of how it's
22:59different can vary among studies.
23:01All right, so there's EG now eye tracking.
23:04So eye tracking,
23:05and I think this is this is
23:07actually really intimidating to
23:09even eye track we're talking here.
23:11Ever take a video of your eye and use
23:12it to figure out where you're looking?
23:16And you end up with a really long
23:18spreadsheet, just of coordinates
23:19of where someone is looking.
23:20But it's really, really useful
23:22because with that information you can
23:24figure out where someone's looking
23:25at a social scene the faces are not.
23:27You can also look to see how that attention
23:30unfolds over time. And as a bonus,
23:32you can look at pupil diameter,
23:34which is a nice measure of arousal.
23:37Eye tracking ASD again.
23:40In a nutshell. Wow. God, I spoiled it.
23:46General attenuated looking
23:47to social information.
23:49This is replicated across many, many,
23:51many studies with different stimuli.
23:53Again, with any one of these papers,
23:55I guarantee you can find another
23:56paper showing an opposite pattern
23:58or a failure to replicate,
23:59but these are fairly consistent and also
24:02attenuated pupil constriction to light,
24:03and this goes back to the 60s.
24:05The idea is that arousal is too high,
24:07and so when you flash the
24:08light in someone's eyes,
24:09there's more norepinephrine kind
24:10of going through the brain,
24:12and so it just won't constrict a lot.
24:16So briefly. Want to point out that these.
24:22Findings target different levels of
24:24abstraction but they reflect pretty
24:26modest sample sizes and it's in
24:28heterogeneous patterns of finding.
24:29So we got to find out which ones hold up
24:32and then I introduce you to the Autism
24:35Biomarkers Consortium for clinical trials.
24:37Jamie Mcpartland here is the principal
24:39investigator of this and the goal
24:40of this project is to test these
24:42well evidenced biomarkers and A5
24:44slight clinical trial model using
24:46egg and I tracking because they're
24:48inexpensive and practical like my watch.
24:50Targeting social community performance,
24:52but not only that,
24:53so we have some other measures in there and.
24:56The characteristics of the kids are there.
24:59I'm going to talk to you about
25:01some of the results,
25:02primarily in the context of
25:04the theories I talked about.
25:05So social motivation disregulated arousal and
25:08tell you about what the experiments were.
25:10So. I've talked a lot about the inland 70s.
25:12The first thing we had to do was
25:14just show kids a bunch of faces
25:16and just look at their address.
25:1870, but not like few kids.
25:19We, as you saw two hundred 399 kids.
25:23Umm,
25:24there's evidence that's delayed,
25:26but we really want to know and
25:27escale like what the characteristics
25:29of this measurement are.
25:31And we think about this as sort of
25:33measurement for indexing social innovation.
25:34And this is really,
25:35this is the experiment that the kids see.
25:38So just so we can all sort of
25:40have some audience participation.
25:42Then with eye tracking,
25:43we looked at what's called the
25:44composite Ocular Motor Index.
25:46So again evidence that people thought
25:48some look less to social information.
25:50So we showed people different kinds
25:52of dynamic and static scenes and we
25:54just measured how much people look at
25:56faces and then average that across that.
25:58So you can see.
26:00So these.
26:02Overlays here indicate sort of the
26:04regions of interest for our analysis.
26:09Again, social interest.
26:10Then we look at arousal,
26:12pupillary light response.
26:13Very simple experiment.
26:15We have this really intriguing little
26:17circle thing in the middle that sort
26:19of loops around the noise, nice noise.
26:21And then for 66 milliseconds
26:22of white flash on the screen.
26:24All we're doing is we're
26:26measuring how fast your pupil can
26:27starts to constrict after that.
26:29And there it is.
26:31And finally,
26:32we love this experiment because
26:34no one has to do anything.
26:36We just show these little screen
26:38savers and for for two minutes and
26:41we just measure your resting EEG.
26:42So we can look at the slope
26:45of the resting EEG as a marker
26:47as an index of EI balance.
26:49So the bridge results on these.
26:52On these experiments SO1 participants
26:56with ASDF slower and 170 so on the left.
26:59This is our waveform ASE is in green.
27:02TD's in blue.
27:04You can see that the group of kids
27:08with autism is significantly later.
27:10All the results are I'm showing
27:11you will hold up even when we
27:14include age and cognitive ability
27:15as a covariance on the right.
27:17This is a sort of a stacked histogram
27:19of anyone 70 so you can see a SD
27:21on the left and TD on the right.
27:23We see this sort of longer term not
27:26relative a decent chunk of variety
27:28of kids with autism will reach
27:31show this slower and with 70s so.
27:35The Ocular motor index.
27:37There's a very large effect size that
27:39kids with autism look less two faces.
27:42So this is again,
27:43controlling for age and cognitive ability.
27:46And in line with what we know
27:48about social motivation is that
27:49the less you look at people,
27:51the less you learn about people.
27:52And so here's evidence for just
27:54less interest in looking at people.
27:56For the PLR, slower PLR construction.
27:59So this is a latency, I'm sorry,
28:01it's a little delayed here,
28:02the latency of the PLR and autism
28:04and in typical development and we
28:07see that kids with autism have a
28:09slower constriction to pillar and we
28:11think this is an index of increased
28:13sympathetic noradrenergic activity.
28:16And finally, the shallower EEG slope
28:18that we find in kids with autism,
28:21it's a little hard to see,
28:23but that green line is significantly
28:25different than the blue line,
28:28which shows us that we have more excitation
28:30relative to inhibition in this group.
28:36One nice interesting thing we found out.
28:39And we hope to find out was that this
28:41also this the slope and the shape of
28:42the EEG's are tremendously reliable.
28:45So these shapes, this is just one
28:49person's power spectrum plot.
28:51The blue are on one day and
28:52the red are on another day.
28:53And these shapes are idiosyncratic.
28:56You can pick them apart and like put
28:58them together like it's a matching game.
29:00And so that's very just cool to see
29:02but also lets us know that this is a
29:05this is like a functional of reliable
29:08index of functional. Activity.
29:09So it's not like we're measuring.
29:11We're going up to the kids and we're like up.
29:13Their head still has the weird
29:14bump on it two days later,
29:15like when they go to sleep.
29:16This whole, EG power spectrum,
29:17like shifts over. They're asleep, right?
29:19When they wake up, it's different.
29:21But when they're kind of at their daily idol,
29:24everybody has the same sort of
29:26functional pattern of activity.
29:28Alright,
29:29So what did we see there is that there's
29:31evidence for reduced sodium motion by
29:33social motivation from the 170 and OI,
29:36increased arousal,
29:37increased excitation.
29:38The different biomarkers are hitting
29:40different levels of abstraction,
29:42so these aren't really mutually exclusive,
29:45but there's evidence for multiple
29:46mechanisms to work here.
29:49OK. So what are our next steps with that?
29:52Well, right now we're working on
29:53replication and larger sample looking at
29:55long-term stability of these biomarkers.
29:57So we've got kids back over,
29:59you know, past a year at this point.
30:02And a feasibility study in a younger age
30:04group and feasibility is really important
30:06for any kind of measure because you
30:08want to make sure that your biomarker
30:10can work in people who need it, right.
30:13So if you want to do early intervention,
30:14then you want to be able to
30:16want to make sure that these,
30:17you can use these biomarkers
30:19and a kid and a kid who's,
30:20you know, three or two years old.
30:23So in our process of putting
30:26these analysis together,
30:28we really dug into things like data quality,
30:31which I'm sure is riveting and I'm about
30:33to go into it more, but it's important.
30:35And I'm going to talk to you about
30:37data loss and biomarker measurement.
30:39So we had 399 kids, nine,
30:42399 kids came in for three visits.
30:45Each visit was two days each.
30:46That's a lot of people coming in.
30:48It's a tremendous amount of work.
30:50I'm looking at the clinician
30:51right now who's seen probably.
30:53Hundred of those kids nodding at me
30:55saying yes, I did all of that and
30:57it was a tremendous amount of work.
30:59And our failure on the eye tracking
31:01end is that even though we got 399 kids in,
31:05we didn't get usable EEG data on or
31:08eye tracking data on those 399 kids.
31:12So, and it turns out that this,
31:15this data aren't really missing by random,
31:17right?
31:17The kids were the most impaired are the
31:19kids who were missing the most data on.
31:21So we decided that we needed
31:23to quantify this a little bit.
31:25And I'm going to,
31:26I'll skip through most of this.
31:28This is just sort of you know
31:29how you get like how much,
31:31how you add up your data.
31:32But here's the the.
31:34Important thing is this thing
31:36called valid data right?
31:38You need to have enough valid data,
31:40which is usable data.
31:42Data you can analyze to in
31:44order to make a measurement.
31:46And if we compare groups on valid data,
31:50this is you don't have to look at,
31:52just look at this column.
31:54It's the P values for comparing
31:56the groups on amount of valid
31:58data and it's always significant.
31:59And kids with autism always
32:01have less valid data.
32:02And this continues even
32:04when you control for age,
32:07cognitive ability and data collection site,
32:10which is a ton of.
32:11It's like now we're just almost
32:13throwing things in the model to
32:15try to make that P value go away,
32:17but we weren't.
32:17These are being valid measures
32:19to control for,
32:20but it's still a big effect.
32:23But I think you know many of
32:25you would have predicted that.
32:26So this maybe you would have
32:28predicted this too though which is
32:30that in this sea of significant
32:32correlations it shows that the
32:34amount of valid data you have is
32:38reflects clinical severity across
32:40all of the clinical measures we did.
32:43The take home point.
32:43We lose the most data from the kids.
32:45We're the most impaired.
32:48And it's the rule, not the exception,
32:50so it's significantly correlated
32:52with their individual differences.
32:55And then this one is, is even more fun,
32:57which is that we're losing most
32:58data from those kids who are
33:00looking at least at the faces.
33:01So that's the thing that we think is
33:04the most important marker of clinical
33:06characterization and it turns out.
33:08So the kids who are looking the
33:10least at the faces are the ones
33:13we're losing the most data from.
33:14So the question is,
33:15how do you know that they're
33:17looking the least at the faces if,
33:18and that's kind of you take
33:20this out far enough,
33:21you don't your your measurement precision.
33:26Is the worst in the people
33:28you want it to be the best in.
33:31Um, it's like a this is kind of like
33:33an ECG system that stops working
33:34when your heart rate goes up.
33:36Right. It's really, really,
33:37really great when you're kind
33:38of like a sleep or if you're,
33:40you know, really,
33:41really healthy and you're
33:41running a triathlon.
33:42But if you go in for a stress test,
33:45it just starts to fall apart.
33:46You think that's really when we want it,
33:48though.
33:49Alright.
33:50But also I want to I've just talked
33:51a lot about the missing data.
33:53This is not specific to the ABC TV.
33:55When I first started doing
33:56eye tracking research here,
33:57a very senior person came up to me and said,
33:59oh man, you lose so much data.
34:01Nobody ever tells you about it.
34:02And they were just there sort of crestfallen,
34:04like they work so hard and so the
34:06kids are just like looking at
34:07something in the corner of the room.
34:09But we have so much data that
34:11we can really quantify it.
34:12Now usually this is a byline
34:14like kids who looked less than
34:1650% of the screen were excluded
34:18from our analysis and it's.
34:19Maybe three or five kids.
34:20So you can't really quantify
34:22it in the same way,
34:24so.
34:27Umm. This is data loss from
34:31kids in the ACT 6 to 11.
34:33Their IQ's are from 60 to 140
34:36and you know it's. Still really,
34:39really, really good data quality.
34:42But we've thought about this a lot.
34:44And so a few years ago, well,
34:45kind of working on this for more
34:48than a few years, you know,
34:49Jamie and I came up with the idea of like,
34:51how can we include these kids?
34:53Thank you. 60 to 140,
34:56that's kind of not higher IQ's,
34:58but there's a lot of kids
35:00who still don't hit those,
35:01those targets, they're more impaired.
35:03And so how can we make these
35:06experiments like work?
35:07And how can this system operate
35:09for kids who don't have an IQ of
35:11like 120 and are really motivated
35:13by $100 in a Lego set coming in?
35:16And that we don't give Lego sets.
35:18And I want to make that clear
35:19because I'm being recorded.
35:20It was sort of there's an
35:21illustrative moment,
35:22but I do not want to that to
35:23sort of be false advertising.
35:25So.
35:25So I got to think about a group of
35:27kids that's in significant need of study,
35:29kids with autism and intellectual
35:33disability and.
35:34So approximately 30% of kids with autism
35:37have significant intellectual disability.
35:38They're very,
35:39very underrepresented in neuroscience
35:41research and I would argue the
35:43reason why is that it's hard to get.
35:45Usable brain data, and I'm going to
35:47give you some numbers to come into.
35:49It's a real issue.
35:51So I did some pub Med searching.
35:54From 2020 to now,
35:55molecular Autism is a great journal.
35:57Published 214 articles.
35:59Autism, which is another journal,
36:01published 193.
36:03Autism Research published 140.
36:05In that same time span,
36:07if you go through all the papers and
36:10search for kids with IQ of less than
36:1260 who were in a study with an EEG,
36:15all the papers.
36:15You know, this is not just those journals.
36:17This is JCP and you just add up all the kids.
36:20That's 66 kids.
36:22That's 66 papers.
36:23There's not a paper with 66 kids, 66 kids.
36:27There's like that's crazy.
36:30This is this is 3 journals,
36:34548 seven or eight 45147 articles and
36:36those are like the other review articles.
36:38Those are opinions,
36:40but.
36:41It's it's still a tremendous amount
36:43of work to even just put those
36:46articles out there. Only 66 kids.
36:49So, so what's happening?
36:51Well, it's really hard to get usable data.
36:54So people know this,
36:56it's hard and that's why
36:57there's probably only 60s kids.
36:59So for characterization you need,
37:01you know, specialized staff.
37:02And Christine is a BC BA works
37:05with us and a psychologist and
37:07Julie are experts on this in
37:08the world and they're flexible
37:10with complex characterization
37:12situations and behavioral demands.
37:14The experiments need to be able to
37:16accommodate the participants needs.
37:18So a lot of experiments will say like.
37:20It's still and press a button
37:21when you see a dog.
37:22If a kid doesn't have useful language,
37:24doesn't understand what you're saying,
37:25can't read that on the screen,
37:27that it's not going to work.
37:31And then the data are usually messier.
37:33And so just the analysis.
37:34And this is like the, you know,
37:36the people who are in the back room with,
37:38you know, 50 computers.
37:38And like, we need to come up with a
37:40new kind of experimental pipeline in
37:42order to accommodate this data from
37:44these kids because they move so much.
37:46You typically need more people
37:47and your staff. But that's like.
37:50That's manageable.
37:52And another way,
37:54it's really hard for the families,
37:56like for any of you who have
37:58come to New Haven and tried to
38:00park somewhere comfortably.
38:01Now imagine that you're doing that,
38:02and you have a child who's in
38:04an unfamiliar place who has
38:06difficulty difficulty navigating
38:07these unfamiliar situations.
38:08You take you're missing a day of school.
38:11You're trying to find a place for lunch.
38:13Then it's really hard for the kids
38:15because they're going to some,
38:17because then they sit in front of
38:19this experiment that says press
38:20a button if you see a dog and
38:22you know what's going on.
38:23So. When I say usable data,
38:27that's going to be clear.
38:28These are nice wiggly lines up here.
38:29That's what we want to see.
38:31This is what happens when
38:32someone's moving their head.
38:33And that's like, they're just like,
38:35you know,
38:36adjusting their neck and it's unusable.
38:37It's muscle activity.
38:39This is most of.
38:41This is what happens when
38:42you're not you don't know.
38:44You have to sit still and
38:45you're not sitting still,
38:46and then you can't use this for anything.
38:50The task demands of these experiments,
38:52following verbal or written instructions.
38:54Again,
38:54I'm sitting still is like it's not trivial,
38:56but I keep repeating it,
38:57sustaining your attention and
38:59tolerating an unfamiliar set.
39:01And so our approach to all of this was
39:03instead of just saying it's really hard,
39:06you know,
39:06tough it out,
39:06which I think would probably
39:08not be effective,
39:09it was just to try to make it easy.
39:10So we developed what we call
39:13Pelican is the the probabilistic
39:15and active learning infrastructure
39:16for characterization neuro typing.
39:18So it's an experimental system that
39:20reacts to participants movements,
39:21their eye movements, their attention.
39:23It adaptively teaches participants to
39:25attend to the experiment and monitors
39:27the data quality so it can adapt
39:29what they're seeing in real time.
39:30There's no explicit.
39:32Instructions.
39:32And it's personalized reinforcers.
39:35And I'm going to show you just how it works.
39:36So you come in, well, first of all,
39:39someone calls you on the phone, the parent,
39:41the child, the parent on the phone.
39:42And they say, you know,
39:43what movies does your son or daughter like,
39:45right.
39:46And they say, oh, Bob the Builder.
39:48So maybe not anymore.
39:49But let's just say it's like,
39:51really, really like Bob the Builder.
39:52They've all PBS VHS tapes.
39:54So they come in, we sit them down,
39:55we put this cap on their head.
39:56And that red,
39:57that red mist doesn't mean that they're
39:59hot or smells.
40:00It means they're moving, right.
40:01So they're they're moving and it's plan.
40:04Because they start to move too much
40:05or they look away from the screen and
40:07we're measuring this with cameras.
40:09We're measuring this with a chair that
40:10kind of like monitors acceleration.
40:13We're measuring it with head movement.
40:15So you start moving too much,
40:16Bob stops, all right?
40:18So you start moving a little bit less,
40:20Bob starts playing again.
40:21So the idea is that whenever you attend and
40:25sit Stiller than you were before, right?
40:28Because having someone go from moving around,
40:30sitting completely still is a tall order.
40:32You get reinforced by this.
40:34Personalized reinforcer.
40:35Bob the Builder.
40:37So what this works out to be is
40:38we get attention in this word that
40:40I never know how to pronounce.
40:42Quiescence,
40:42stillness without verbal instructions.
40:46Umm.
40:48Always positive reinforcement and
40:50then the personalized reinforcers.
40:52Actually, this is much more effective
40:54than we thought it would be.
40:56So it's a very unfamiliar place.
40:58But you come in and something very,
40:59very familiar to you is happening.
41:01And it turns out that there's no
41:03way you can have one-size-fits-all.
41:05I mean, we have kids who love Moana.
41:07We have kids who love 80s action
41:09sort of sitcom dramas.
41:11These are all real.
41:12You know, these are not participants,
41:14but these are choices of participants.
41:15And then we have the Chicago bus system.
41:18And we'd never know.
41:19We would have never picked
41:21these up on our own.
41:22But it really helps to navigate
41:25the uncertainty of the room.
41:26And then the last thing is
41:28that adaptive trial delivery.
41:29So we're watching how you're watching
41:31the computer because if we're watching it,
41:33we're too slow,
41:34we're too inattentive.
41:36We're monitoring if you're moving around
41:37when a face pops up on the screen,
41:39right, look for showing your face.
41:40So if you are,
41:41we know that we're never going to be
41:43able to measure effect brain activity
41:44effectively from that trial from that,
41:46you know,
41:4750 milliseconds,
41:4820 milliseconds.
41:51But the way European experiments typically
41:53work is we just show you like 100 faces
41:56or 200 faces and figure we're going
41:58to some of those are going to be OK.
42:00So those of you who have been in an ERP
42:02experiment in college to earn 20 or $30,
42:05you probably fell asleep in it because
42:07they're long and they're boring.
42:08But imagine if we knew. That you watched.
42:12You were sitting still and looking when
42:14twenty faces popped up on the screen.
42:15Well, then we'd be done with that right now.
42:17I'll show you some houses.
42:18Now we'll show you something else.
42:19So we can actually make the
42:22experiment shorter for these kids
42:24than what the standard is.
42:26Would they end up being much longer
42:29than they would normally be for,
42:31like a compliant kid,
42:32you know, with a higher IQ?
42:35So this really reduces the burden.
42:38And so I'm going to talk to you about
42:39this is our twelve kids we've seen.
42:41Unfortunately this was funded during
42:42there's a little bit of pandemic happening.
42:45So we weren't in the lab as
42:48much as we'd hoped to be.
42:50But as you can see from these numbers,
42:52this is a fairly impaired group of kids.
42:55These are the averages are 100 for these
42:59numbers and these are not near there.
43:02The experiments we used we adopted 2
43:05from the ACT, so the faces task and
43:08170 faces and the static social scene.
43:10So we're just showing faces,
43:11we're showing scenes.
43:14And then the results, what do we get?
43:16So run it works.
43:17Oh God, I left it up there.
43:19So yes, but it's exciting.
43:20What these are,
43:21are these are little trajectories
43:23through the experiment.
43:24Each one of these,
43:25this is time along the X axis here.
43:27And this is 1 kid on the top and a
43:29different kids trajectory on the
43:31bottom because the experiment adapts.
43:33So everybody sees different things in
43:34different orders at different times
43:36and you want some way to look back
43:37at that and see how things went.
43:39These blue lines right here indicate
43:41when you're moving too much or
43:43you're not looking at the screen,
43:45and the lollipops indicate when
43:47the system was determined you were
43:49sitting still enough that we could
43:51move on with the experiment.
43:53So.
43:55On top kid,
43:56you can see sort of kid on the
43:58represented by the top line.
44:01Definitely had some periods of
44:02time where he's still learning the
44:04contingencies of the experiment,
44:06but towards the end. Figured it out.
44:09And we have a stable presentation
44:12of the stimuli.
44:13This kid at the bottom learned things really,
44:15really, really quickly.
44:16And what's cool is that so these
44:19different colored lollipops are
44:21different kinds of experimental trials.
44:23We're all done with the lime,
44:24felt like the lime ones.
44:26They've seen enough, good enough,
44:27good data so we didn't have to keep
44:29showing that again and we could focus
44:31on just the ERP face and house tasks.
44:35And we can get measurements.
44:37So this is a grand average ERP and it
44:41doesn't look as clean as 1 from 299 kids,
44:44but from 12 kids we're getting sort
44:46of expected negative deflections that
44:48are earlier for faces, for houses.
44:50And then if we compare these kids,
44:53in the end, 170 latency
44:55against age matched age match,
44:57kids with autism and controls from the ACT,
45:00we see this same pattern of extended latency,
45:03which is really, really exciting.
45:04The the, I guess the the
45:06really important part, it's,
45:08it's important that the 170 is later,
45:10but I think it's more important
45:11that we can measure it all at all.
45:13Because now we can know it's later,
45:15but who knows,
45:15like there's all kinds of other things.
45:17If we don't know that we can
45:19now know we have this tool that
45:21is like wildly applicable.
45:23And so in our next steps,
45:25we've just submitted this for an hour one
45:28ready to deploy it in a larger sample,
45:31tighten up some parts that are
45:32still a little rough on the edges,
45:33eye tracking,
45:34calibration and then incorporate in
45:37other biomarker experiments for that.
45:38OK, this is going to be very fast,
45:41but because I'm.
45:43Sorry, slow enemies like a heart rate.
45:46So the heart rate rate changes
45:47in response to me.
45:48When you go up steps it gets faster.
45:52When you go downstairs,
45:53it gets slower so and you can learn a
45:55lot about somebody's cardiac health
45:56by putting them on a treadmill and
45:57then measuring their heart rate.
45:58That's perhaps some of you have
46:00been in those situations.
46:02So I'd argue that anyone 70 is also.
46:06Responsive to changes in the environment.
46:08And we can learn a lot about somebody's
46:09brain from how it changes in response
46:11to different things in the environment,
46:12like such that I guess,
46:15well,
46:15what's what are the stairs for the Inman
46:1870 and I think it's social interactions.
46:21Social behavior is interactive.
46:23We do know that.
46:25You know,
46:25I gave you the list of the laundry list
46:27of all of the of all the challenges
46:29and symptoms of autism before,
46:31one of them wasn't sitting in a room
46:32alone and watching faces on a screen, right?
46:34That's TV.
46:35And a lot of the kids who come
46:36in are really good at it.
46:38A lot of us in this room are
46:39also really good at that.
46:40But in social interactions are
46:42where we have these challenges.
46:44So if social interactions
46:45are the stairs for the 170,
46:47it would be great to throw EEG caps on
46:49all of you in here at a cost like 170,000.
46:52That person.
46:54But it would be great if we could have
46:55like a version of the stress test in our lab,
46:57like a Stairmaster.
46:58What?
46:58A Stairmaster for the end 170.
47:00So how would we do that?
47:02Well,
47:02let's go back to predictive coding.
47:04So again,
47:05evidence that prediction is different in ASD.
47:10But also remember the 170
47:12sensitive to changes in gays.
47:13And if I tell you that this face on
47:15the screen is like your friend or is
47:17judging your clothes, it changes.
47:19So the animal 70s also sensitive
47:22to expectation and and context.
47:24So how do we build this?
47:26And 170 Stairmaster?
47:27We can do that by trying to simulate the
47:30relevant parts of the social interaction.
47:33So how do we do that?
47:34We have these very simple experiments where
47:36we use eye tracking and EEG simultaneously.
47:39Such that when you look at a face,
47:42there's a that's a fun animation.
47:44Right now, that's your eye movement.
47:47When you look at the eyes of the face,
47:49the face looks back at you.
47:51So this is a really,
47:52really subtle visual change.
47:53It's a really,
47:53really meaningful visual change.
47:54So if you're sitting on a bus and you look
47:57at somebody's face and this happens, this is.
48:00Meaningful.
48:01I don't know how.
48:01Depending on the bus context,
48:03it's going to mean who knows
48:04what kind of things is,
48:05but you shouldn't ignore it.
48:08So in our experiments we can control,
48:10control, predictability,
48:11to make eye contact either
48:13predictable or unpredictable.
48:14We're going to show you the
48:16designs of two experiments.
48:17So in the first one we made
48:19eye contact unpredictable.
48:20These experiments are all a little bit.
48:22Quirky because of the way they operate,
48:24but briefly, we showed people an
48:26arrow on a screen followed by a face,
48:29and if the arrow is pointing up or down,
48:31that queued them to look to the
48:32eyes or the mouth of the face.
48:33And then one of two things could happen.
48:35The thing they look at could open.
48:36So they could look at the eyes and
48:38the eyes open and make eye contact,
48:39or they could look at the eyes
48:40and the mouth opens,
48:41or they could look at the mouth in the mouth.
48:42So four things.
48:444 outcomes there.
48:45What's important is you can't
48:46predict what's going to happen.
48:47You might get eye contact, you might not.
48:50It's 50% of the time.
48:52This is our video of it.
48:55So on the right you see gaze and
48:57on the left you see the sort of
48:59washed out video of what the
49:01participant is seeing.
49:04And I apologize for the iPhone quality,
49:06the 2012 iPhone quality, but the.
49:13So that's our first one.
49:13So we're looking at eye
49:15contact to unpredictable.
49:16Second, now we're making it predictable.
49:19So crossair peers.
49:20You look to the eyes of the face.
49:23After you look,
49:23for about 500 milliseconds,
49:24the gaze will shift.
49:25Now it's looking away from you,
49:26it's going to look at you,
49:27it's looking at you,
49:28it's going to look away from you.
49:29Totally predictable.
49:33And.
49:39I look at the mouth. Nothing happens.
49:41Yeah, it's visually really subtle,
49:44but when you get looked at, you feel it.
49:49OK, we did experiment in two samples.
49:53Finally, able adult with autism
49:55and theoretical controls.
49:57And we also measured continuous
49:58measures of symptom severity,
50:00so we have the the cast.
50:03The SSS calibrated severity score.
50:07Back anxiety inventory. The gas flow.
50:09Glasgow Sensory inventory,
50:11sort of a measure of sensory
50:13hydrogen sensitivity.
50:15Here's our results.
50:16So when I I contact is unpredictable.
50:19You actually get a much
50:20bigger and 170 in autism.
50:22No latency differences and across groups.
50:26The difference between mouth
50:28movement and eye contact was
50:30associated with sensory sensitivity,
50:33autism severity and anxiety.
50:36This is an unpredictable situations.
50:39So and this is the N 170 here in Gray,
50:42this darker colors ASD and the caramel STD.
50:47When it's predictable eye contact.
50:49We now see this later and
50:51170 that we see in the ACT,
50:53although the morphology is a bit different,
50:54there's no amplitude differences
50:56between the groups and there's
50:57also no correlation with the
50:59clinical characteristics.
51:00Those are our clinical characteristics
51:01and we're going to jump over those.
51:03But in summary,
51:04so when I contact unpredictable,
51:06we get these bigger responses associated
51:09with anxiety and sensory symptoms,
51:11but when it's predictable,
51:12we don't see any differences.
51:14And we think this means that.
51:16You know real social interactions
51:18are aren't perfectly predictable or
51:20unpredictable but how you integrate
51:21that information is different
51:23in autism and it's happening
51:24at less than 200 milliseconds.
51:27It might not be faced specific
51:29but the way if you think when two
51:31people are interacting if one
51:32of them's jitter like is slower,
51:34think about when you're talking
51:35to someone on zoom is actually
51:37a great example and there are
51:38like 200 milliseconds behind you.
51:40Just most of the time just give up
51:41just like look you put in the chat.
51:43We'll talk later because that temporal.
51:46Asynchrony is just tanks the
51:49social interaction,
51:51so that might be sort of the kind
51:53of things happening in here.
51:55So.
51:57Our next steps were we're exploring
51:59how shifts in Gaze might impact
52:01impact brain response and looking at
52:03emotional faces and brain response.
52:05Basically, different stairmasters,
52:07treadmills, tilt tables for the M70.
52:10In summary,
52:11we've got different theories hitting
52:13different levels of abstraction,
52:15evidence for multiple systems being impacted,
52:19and.
52:20Promising biomarkers for more
52:22for many of them as well as,
52:25I think pretty good success in
52:27measure in in measuring and designing
52:29experiments for an underserved group.
52:30Umm.
52:31And I think that I'll briefly
52:34just say that the research and
52:37progress that I'm going.
52:39Jump over really fast is kind of
52:41an integrating different levels
52:43of that analysis.
52:46But I just want to thank this is
52:48like some of the people in the ACT.
52:51It's a lot. I mean they're all
52:53everybody's really really really important
52:54in it somehow only these really,
52:56really really important people got on
52:59this this slide and this is Jamie's lab
53:01and this is a I don't know if this is
53:04everybody's on here but it's a little
53:06I think it's if you haven't you're
53:08not on this list I apologize and then
53:10I just acknowledge that participants
53:12in the families and our support.
53:13So thank you.
53:25All right. Should I turn
53:27off the chair? Good. Thank
53:31you so much, Adam.
53:32And I'm actually the distractor
53:33in the middle of your talk
53:34was from Linda Drozdowicz,
53:35who was just complimenting you on
53:37how accessible you were making EG.
53:38So you're getting compliments
53:40in the chat questions for Adam.
53:50Adam, that was really great.
53:51I really learned a lot from you today.
53:53As always, I appreciate that.
53:55Just a quick question about eye contact.
53:57So I understand the difference of it
53:59being predictable or not predictable
54:00what the response is going to be,
54:02but what about desirable or not desirable?
54:05In some people they like eye contact, right?
54:08But I have some patience and some
54:10friends who really dislike eye contact.
54:13And So what does that do to to make
54:16your data even more complicated?
54:19I'm, I was surprised at
54:20how complicated it is.
54:21But more complicated, you know, so,
54:24but that desirability back there, you know,
54:27some people like it and some don't.
54:29Thank you.
54:31Should I talk into this? Is that OK so.
54:36I guess so on one level like.
54:40Things like desirability might
54:41be the kind of context I'm
54:44talking about right today. They.
54:49Those are the kinds of factors that
54:52are probably, I think probably.
54:54So I think they're probably.
54:58Sort of charging up that piece
55:00of brain where the end 170 is
55:02being generated from really,
55:03really early or not.
55:05So I don't know that.
55:10Like.
55:13You know, the M70, it's this,
55:15it's this peak that can be larger or smaller,
55:18earlier or later. So it's not.
55:20There's only so many places it can go.
55:23So I. I guess like that's as I think
55:26about this task, like the Stairmaster,
55:29that might be like the tilt Table, right?
55:32There might be other ways you can sort of.
55:35Poken product. Thank you. Now
55:39you're making me get my steps in,
55:40so my heart rate is going to be going up.
55:44I wonder about has there been any
55:47experiment or data comparing,
55:50I mean kids with autism in terms of
55:52their eye tracking data when they
55:54are or they are doing the experiment
55:56on screen versus in the real life,
55:59I mean as an in person experiment.
56:02Yeah, we did one here years and years ago.
56:08So. Here's the here's the big
56:12caveat to all of those experiments.
56:14So when I play you a movie on the screen,
56:16it's the same movie.
56:17Every kid who comes in sees the same movie.
56:20When I put an eye tracker on your head,
56:21it looks like just, you know,
56:23now they can just look like glasses.
56:25But if I put an eye tracker on
56:27your head and I send you out?
56:30I who knows where, where you're going to go?
56:32So there's this.
56:36It you you,
56:37you can find that people might be less
56:39prone to engage in social interactions
56:41and then they look less at faces.
56:43But probably what you want to know
56:45is for those people who weren't
56:47going to social interactions,
56:49what would they have done if they were
56:51in those socially charged situations
56:52that the other people went into?
56:55So it's sort of a.
56:58Long way of also saying I don't
57:02know and it's complicated.
57:05Methadol just also just in
57:07measurement and methodology,
57:08measurement and methodology.
57:10It's really hard to do that.
57:13And get really accurate and
57:15meaningful data for a single person.
57:17When you have a 50 kids or 20 kids and
57:19then you average it across all of them,
57:21then you know, smooth it,
57:22they smooth it again.
57:24Then there's things come out of it.
57:26But then at the individual level like this,
57:30this issue of data loss,
57:31again these are they start to creep in,
57:33people make their own experiments and
57:35usually they're not the experiments
57:36that you wanted them to be in.
57:41Mike.
57:46Adam loved your talk,
57:47great out of the park with all
57:50the highly innovative approaches.
57:52I've several questions. I'll
57:54just ask the first one, which
57:56is, have you thought about with
57:57your your new shaping paradigm,
58:01do you think that it changes the
58:03brains of these kids in any way where
58:05you might see that their
58:06inhibitory controls improved?
58:08Do you see any applications of
58:09this beyond your experiment?
58:14So. Yes, I don't think in the
58:17sessions we're doing that.
58:18There's like measurements of that we're
58:20going to be measure like inhibitory
58:22control changing because the amount
58:24of change that happens within the
58:26experiment is going to be driven by
58:28how far away they were from sitting
58:29still at the beginning, right.
58:31So you're sort of.
58:32It's like you're that baseline
58:34is always going to be a predictor
58:37of your outcome measure.
58:39And but yeah, and I think it's,
58:41I think it's a.
58:42I do think it's a different
58:45kind of experiment.
58:46And I don't know there's like better or
58:49worse than kind of the traditional ones,
58:53but it's different.
58:54And I don't know that we know
58:57exactly every way in which those
59:00differences might be important.
59:02Sometimes,
59:03and I've abandoned this,
59:04we would say to participants,
59:05try not to blink.
59:06And you're layering your layering
59:08on whatever your instructions are.
59:10You know, someone is has those on
59:11board while they're doing your task.
59:12You really don't know how
59:13much they're doing it.
59:14So it's some ways similar
59:15to what you're doing.
59:16You're just doing it
59:16implicitly. If, well,
59:17I'll say it having been my own
59:20participant hundreds of times,
59:22it feels really different, right?
59:24This is responding faster than your N 170 is.
59:27So it feels, I mean it feels like,
59:29you know, we've tied strings on to you
59:32and they're hooked into the computer.
59:33And that level of like.
59:36That level of latency,
59:39or lack of latency and perfect contingency.
59:44It's it's you have it's
59:46very implicit learning.
59:47You don't think like oh I have to
59:49sit still like you just sort of
59:52it's a little bit for me I sort of
59:55just freeze not freeze up that just
59:57stopped moving and I don't think
59:59about it explicitly so you think
01:00:01do you feel it. The reason why I'm
01:00:04asking is because we've done some
01:00:06stuff with biofeedback and we'd relax
01:00:09to keep a ball from when you relax,
01:00:11a ball levitates. And with me,
01:00:12every time it would levitate,
01:00:13I get excited, then it would drop and so.
01:00:16But I was thinking that, you know,
01:00:19you could use this as a as an intervention in
01:00:21some ways your approach.
01:00:23Yeah, sounds like that.
01:00:25Yeah, yeah, definitely have kids.
01:00:27When, like, their favorite movie comes on,
01:00:29they start getting excited.
01:00:30And then you're like, OK, this is.
01:00:32And you feel bad and it's like,
01:00:33oh man, what have you done?
01:00:38Wonderful. We're just about at time,
01:00:40so we'll have a captive audience here.
01:00:41I just like to remind you that we'll
01:00:43be back here in the Cohen for grand
01:00:44rounds next week for Doctor Lisa Gallia.
01:00:46And previously at UBC,
01:00:47now recently appointed by Camp H in Toronto,
01:00:50we'll be talking about the impact
01:00:52of biological sex on brain health.
01:00:53So please join us for that.
01:00:55But just to thank Doctor Naples for a
01:00:57wonderfully accessible deep dive into
01:00:59biomarkers in a really important area.
01:01:01So thank you so much.
01:01:03Thank you, guys.
01:01:10Do I do something here?