So thanks for thanks Chris Rozell
it's nice to be with you and

when do is give you an overview
of this project on and

I've been doing since before many
of you were born it will feel like.

Which is about deep brain
stimulation forty pression And

just to get started it's been
supported by a number of grants

I'm going to be talking about what we call
in medicine the off label use of a device

that means I'm going to tell you
about is not approved by the F.D.A.

but we use devices that are approved
pretty much by the F.D.A.

and we are the purposing them
still have to work under

very carefully controlled
conditions with the F.D.A..

There is a patent involved that
we filed back in two thousand and

three that was issued and
then licensed to a company and

I consult to that company and
to get some fees for that licensing.

Didn't help in terms of their clinical
trials and that's another story

I want to introduce you to my team
because I want you to appreciate the.

Of people participate in this kind of
study these are people that are currently

involved in the studies at Emory there's a
clinical team that requires a surgeon and

psychiatrists and people to.

Really interact with the patients
there's a team of people involved in

biometrics and mechanisms that pretty
much covers the board of imaging

electrophysiology and
psychophysiology I'm really trying to get

metrics of what's happening in your brain
what's happening in your body what's

the behavioral readout on people with
depression with things in their brain and

then there is the modeling
which is how I know Chris and

Robby Tara we all have a graduate
student they need to revive who's a B.

any student here you may know.

And and it also involves computer science
and some meat machine learning so given

that this is a broad audience there's
probably something for everybody here.

But I want to introduce
you to is the idea of.

Diseases of the brain being
disorders of circuits you've heard

the new buzzword conic Tomic said.

Basically it's fancy words for what we've
always known is brain areas are wired

together we used to not be very smart we
always say there are lots of cell types

but we grossly thought about how big
swaths of brain areas talk to each other

we whack somebody in the had already had a
stroke in the last big area of brain while

you made inferences about the role
of that region in behavior and and

also what was the wiring diagram of how
those regions talk to each other and

you could look those up in anatomy
books going back hundreds of years and

now we have scans to
look at it directly but

concept whether or not you think about
Parkinson's disease which was really

in the modern era the first
evidence based application of brain

stimulation to a circuit disorder
started in tremor really is.

The work horse imparts instance disease.

Is how can we build a model of disorder
like depression which doesn't really

seem like the kind of thing that you
think about a circuit top of thirty but

I want to convince you that like
Parkinson's where there's problems with

slowed movement in gait there's also
slowed thinking in trouble with your

emotions Similarly it's a similar kind
of circuit in in depression where it's

predominantly a mood problem but
it also has problems with motor and

cognition and autonomic or
drive states in the hypothalamus

people don't sleep they don't eat they
have dysregulation drives and libidos and

they have trouble with stress hormones so
that think about the idea that

symptoms might map to a particular
region or set of regions and

the D.B.S. does is it impacts
specific knows or pathways or

networks that might impact one or
more behaviors and.

And I think that thinking
about wiring diagrams in.

A region a pathway may be
logically targeted for

particular symptoms that these
things all work together and

how to think about networks in that way so
what are we talking

about with deep brain stimulation of use
the term a number of times of abbreviated

it it's basically an approach
surgical approach that

involves implanting at least
a now implanting small electrode.

You can see the Quitman on the left
electrode it the active contacts there

conventionally four of them in new ones
you can have eight there are spaced a long

wire that's in case been
silicone under M.R.I. guidance.

And you can pretty much a surgeon can put
these pretty much anywhere in the brain

that they want not restricted to
the surface it's not restricted to

subcortical areas it's not restricted to
grey matter white matter it's pretty much

while a surgeon say no there's no
brain area safe from them and so

it's really just a matter of ensuring
that it's a safe trajectory and

target because there are blood vessels
there are things you don't want to hit but

on the way you should think about it is in
an operating room with the patient awake

you can put a stereotype tick
frame on their head you can

kind of see the set up here patients drape
they can talk to you from the front they

are awake they kind of lay on
a lounge chair con have set up and

you can do a small bird
hole place God can you and

then place having pretty
determined the location

the trajectory the safety avoiding vassals
and whatever else you might want to avoid

On place this lead where you want and
have one of these contacts B.

or more than one B.

where you want them to be it then
comes out of the top of the head

it's attached to a connector
cable that is under the scalp and

then you can put in a nother cable
under the skin in the neck and

hook it up to a little
battery pack the I.P.G.

implantable pulse generator that
usually goes under the collar bone and

the whole thing is internalized and
with a little magnet like you do with

charging your phone you can
sit with a little P.V.A.

and adjust the frequency
the current the pulse

with which one of the contacts one or
more and basically Dr Kurt.

To the place in the brain that you've
implanted so that's a device so

the issue is is where to put it.

And who to put it in and
what you want to have happen so to cut to

the chase because I want to cover a large
ground you might say well brain's a big

place I mean how do you decide where to go
to well that was twenty years before so

again so before your parents were
born you know you know that.

That basically we looked at
different groups of depressed people

we studied them with scanning F.

M.R.I. wasn't even around to and we map
the metabolic and blood flow signals

throughout the brain that distinguish
depressed and non depressed patients we

started to treat people we looked at how
the brain changed on various drugs how

it changed on psychotherapy how it changed
on placebo how it changed with these C.T.

we compared patients who had neurological
depression or regular depression and

we started to build a map of the brain
areas that participated resting state

not getting found C We also wanted
to get an idea about what brain

areas were most important to
negative mood because when

you see a depressed person you know people
don't sleep their appetite is disrupted.

They lose interest things don't feel
pleasurable might feel guilty there's

a number of signs and symptoms but the one
thing which is why people kill themselves

is because it hurts the psychic pain
of a depression will always make you

not feel pleasure because your brain is
busy with a very uncomfortable sensation

I've never been depressed except you know
having a bad day it's not depression but

judging by the hundreds of people
that I've seen over my career it.

If you've been depressed you know what I'm
talking about it's not having a bad day

it's impervious to a negative state and
one of the things we tried to get an idea

about was since you can't really see
me a way that kind of mental pain

you can at least have people you can at
least map what grief is like bring R.P.

even in healthy people a personal sad
memory and what we found was that

by looking at how different treatments
changed the brain we looked to see where

the over a lot of brain areas that were
provoked when we made people sad and

we started to get a pattern so
that treatments across

the board will end up impacting
this region here in C.

end of depression recovery the sub general
single it's Broadman area twenty five in

the anatomical tales tables it also
goes with having the low activity in

your frontal lobe so
when you correct low frontal activity

you also drive down single that
activity and there you know it and

so when we did experiments on
provoking even just you know

losing a loved one a pet taking
care of someone sick something

that you can personally activate is making
you stand we saw the exact opposite

activate area twenty five turn off the
frontal cortex so we started to take these

big number of regions that went with our
treatments to actually isolate in on

the relationship between the singular and
the frontal cortex and

just because of the dominance
of the singular good activity we

decided that there was a target a bull
location we knew that that region

needed to down regulate its activity or
you didn't get well with treatment and

that was really the rationale as you
can see me and I showed you the size of

that wire it's like angel hair
pasta I mean it's not a nano.

Fiber is I'm sure someone in the audience
has probably built already but

you know you've got this this.

A centimeter long active area on this lead
you've got contacts that are a millimeter

and a half cubed So the length separated
by a millimeter and a half and

we've got kind of this blob that we
can kind of stick a lead in there and

have one or two of the contacts be
right in that hot zone that red

area here and so basically we made
an approximation on an M.R.I. scan and

figured out where the approximate hot
spot was across all the experiments.

And we basically put one of the leads in
there we also knew back at that time that.

What turned off activity in movement
disorder in Parkinson's was that high

frequency would sort of jam the circuit
and we very naively said well you know

high frequency gives you the polarization
block all of that sort of like wanting

to turn down metabolism not but you know
without was our logic is that will turn

on high frequency stimulation in the area
that we'd like to down regulate and

we'll see what happens and that's what
we did we put it in turned it on and.

I'll tell you a bit about kind of what
happened acutely not on every contact and

not every current but it was the threshold
of current that when you turned it on.

The negativity just kind of went away and

if you turned it off it didn't
come back immediately and

then the patients are sort of confused if
it was there or not sort of ignored it did

nothing bad seem to be happening there was
nothing we could do to make it worse so

we left it and picked a current that kind
of went with what the Parkinson's Disease.

People have been doing for
at that point you know for

over ten years and
look to see what happened in and

here's what happened in the first
six people this here on the.

Hamilton depression rating scale score
it it's basically a severity rating so

for poor you it's been standardized
it's used in all the clinical studies

I want you to follow that because it's
supposed to be some kind of you know

ground truth again you know not but

the best we've got on and if you just
look at what happened in the hospital and

we didn't know what we were doing we kind
of gave a little stem every day we were

going to do a scan at the end of the week
and actually you can see that that score

dropped the dotted line is a fifty percent
drop in a person's score that's considered

a win fifty percent response in any
clinical trial is considered the response.

Threshold and as you can see
that everybody now only six of

the four of them that eventually
got better in the two that didn't

all kind of showed this effect
acutely within the end of the week we

implanted the battery pack and
gave them a week Reston and

actually saw that they sort of
progressed a little bit as a group but

now we just sort of wait expected it from
Parkinson's disease you turn it off it

goes away Parkinson's you you turn down
the frequency from one hundred thirty

Hertz if you turn down the current If you
turn it off you've got about twenty thirty

seconds before a terminal come back in
turn it on it it stops immediately it was

more subtle It wasn't only quite clear
in no rapid acting any depressants was

not something that was known and
so you can see the course over

the next six months of coming in making
adjustments that after about two months

you got this separation of four people got
substantially better and and two people.

Nod and we were pleased
nothing bad happened is were

patients just to share it's not like a
first episode these people were people who

had recurrent depression they'd all
actually had now in this episode had

failed electrical impulse of shock
therapy they had on average had failed

over twenty medications a number of
medications these people came in on ones

nots I mean you know who comes in on you
know ten medicines you know you kind

of keep looking for a helmet or a pass of
doing something profoundly ill people so

if you know people are depressed
these are not common OK So

to be a nice fellow I just want you
to appreciate who we would thing to

put electrodes into people with
basically have exhausted options and

it took us a year to have anybody but
for us a case but

we persevered we expanded
from six to twenty

you can look now at at the continuous
stimulation over a year

this is on average we had a sixty percent
response rate again not blinded we had no

idea what would happen you don't want to
start blinding before you know what's safe

what happens you know you might do good
for a while and then it goes away maybe it

get side effects on May be you crash I
mean you know it's totally unknown so

that was the approach just open label and
see what happens and when I moved here

we enrolled in another cohort of
patients you may know Bob gross and

neurosurgeon and
Emory is also in the B M E Department.

He's our surgeon new so the psychiatrists
picking the patients based

on their own thoughts on meeting
the inclusion criteria and

again we we look at the curve they look
very similar but now we had our first

clue that maybe it wasn't for
everybody we only had forty one per.

Respond instead of sixty and again that's
only a few people in a small study but

again you know when you start to have
variance you start to be able to figure

out actually what you're doing and and
over time I want you to appreciate that

in the Tronto group actually we just
go to the next Actually to me just

one of the important things to know was
you know we designed the experiment.

Well we had a logic but it probably
wouldn't work and we would need to know

why if in order to back justify
why had we implanted people so

we did you know what we knew we
did PET scans at the beginning and

then we did PET scans late and
actually what we found was that

these people had over active twenty five
and low activity in their frontal lobe

like their stuck sad was sort of
the pattern and we reversed it so

the first time out of the box in that
small sample we were able to at least

get some some support of data that
we might be on the right track so

what happen going forward will in Toronto
they now have patients first patient was

implanted in in March of two thousand and
three so it's a long term experiment

because you know if you're doing well
when you have an implanted device.

It's no reason if there's nothing going
wrong not to keep it going on and so

there are long term follow up at six
years was what's around sixty percent so

if you got better you stayed better
we now have ten year experience and

actually we're doing a little
better than they are in toronto

of the original group in
Toronto they've explained to.

You know they have five years on us but at
two suicides late actually in people who

have gotten better so again you know it's
not as straightforward as you only tell

yourself because you feel bad and
that's a whole other discussion

we had patients in Toronto die of cancer
had other problems you know died well but.

Of the other things that you get
when you're middle age or older and.

And people this kind of float away
loose to follow up usually because

not because they're doing well because
the batteries die have to replace them so

they have half the number still
in the old we did thirty four

actually this group is this slide is
of twenty eight we explain to three

only one for non-response on patient
after five years decide they didn't like

having a machine and take this chances
another patient actually didn't like being

well go figure as for
the analysts not for their own tests.

But what I wanted to see is if you
got better you stayed better and

we started to explore things like Will
does your brain heal after six months

what happens if you do a blinded
discontinuation well over a couple of

weeks it's not like movement disorder
where you start shaking immediately it's

a slow drift takes several weeks for
the score to go back up

kind of creeps up on patients it's
not a slam you know I feel bad

you turn it back on it you recover it and

actually what we've seen over the course
of ten years is as batteries die

when patients are home that they come in
because they start to have symptoms then

you can start to model what is the time
course and can you recapture and

this is actually the most meaningful you
know when you tell someone well you know

sometime over the next two weeks I'm going
to turn it off people start to get nervous

but you really want is just the thing
stops working in dystonia movement

disorder on that takes slow
to recover movement disorder

where where people are twisted
in have fixed postures it

takes a while to go away and
it was experience.

That batteries would die and
patients would do fine not everyone but

that you don't necessarily have
that jamming signal like you have

in Parkinson's where it's like noise
canceling headphones you interrupt and

also lation And when you turn it off it
comes right back because you didn't fix

anything question is this maybe
there are some plasticity maybe

maybe don't go back to where you were and
well it looks as though that

you know you can recapture I
have to be implant the battery.

We're not fixing people in some kind
of plastic way it's not like you have

a broken bone and you have a cast for
a while in the casket come off and

just rehab so other people have taken
this logic different kinds of logics

people have looked in the reward system
directly I mean I take the attitude of

to implant someone I mean profound
I mean depression can be modeled as

everyone has in him Donia low response to
pleasure little motivation not everybody

has mental pain and so you can model or
think about pushing positive

pushing motivation is your primary
driver to me I kind of go like this turn

negative off and it's pretty clear from
talking to the patients we operate on that

we don't give them anything they make it
very clear you took something away and

man the rest is up to me and
that is fundamentally an important thing

to think about in terms we have to model
this people have tried looking in the.

Part of the negative prediction
error pathway people have looked

directly in the media for brain bundle to
really push motivation these things work

its will harder to maintain
the state which is

an interesting fundamental
question what's happened is that.

This has rushed I would say by industry
to do clinical trials because these.

Open Label studies were so.

It's not my word I never
thought it was miraculous but

you know when people this ill get
better and you have enough of them and

they stay well everyone's as
how can we get this to market.

Do not do that before it's time because
there's nothing more agonizing to see

a ninety person trial fail not
because it doesn't work but

because your employment was wrong
you didn't consider variance and so

St Jude actually implanted did a pivotal
trial of what I just described they got to

the halfway mark with ninety people and
at their chosen him point

active didn't be changed but because there
implanted and continue to treat over.

Many people out past five years
now they looked at two years

fifty percent response rate so
failed trial.

Working in people same kind of problem
with the battery dying in needing to

replace the and so we have to kind of
think about everybody gets excited.

Then you're like your hopes are dead
then you pick yourself up and

just like all good Zen practitioner so
fall down four times get up five

and the question is then what
are the ways to get yourself out of a jam

because if you don't really just like
what's the logic to go in pick a target

you can have lots of lot you
know lots of different logic.

And then you've got to test it the same
way is when it doesn't go the way you

think then the question is
is partly what do I do and

the starting place and that is who's
fault is it and I say that as a joke but

that actually is fundamentally the point
how to why figure out what went wrong by

figuring out who's behavior can I
change what variable can I change and

the choices are Blame
it on the psychiatrist

everybody blames them because
they've got witchcraft for how they.

Choose subjects' and
they can know how to take patients care

of patients really well but they don't
have biomarkers of who they're picking and

why can blame the surgeon miss a target
bad surgeon that's actually a good

one to blame not because you want to
annoy a surgeon never know a surgeon

they are the most critical piece of
this besides who the patients are but

that their behavior can be
modified much easier with data and

then you know you can blame industry they
jump the gun don't go there you know

changing them and you have to always
consider that maybe you got lucky you had

some really spurious finding and
you know people get magically Well

if you stick stuff in their brain I mean I
better get it while it's expensive it's so

dreary you know OK I can accept that or
logic was just wrong or to all

placebo I think the on discontinuation
means it's probably not placebo so

we went with because we're emitters
I'm an immature part Marilee is go to

the where you know what is different maybe
let's start with a simple minded notion of

the surgeon you know was not as precise as
we thought they were because how would we

know well we can look in standard space
that where the where it's implanted in

Toronto actually when people saw this
data that responders and non-responders

had a wide area being stimulated if this
was definitely not a surgical problem it's

kind of a big sweet spot in Qana take the
average of that in aim at that very bad.

Bob fantastic surgeon puts it exactly
the same place every time and

does no better at explaining responders
and non-responders but we had a clue about

from the PET scanning in those first six
here was the change pattern and said

we don't in responders and non-responders
we get decrease blue is indeed.

Crease and in activity we don't
miss the primary target but

if you look at well follow the yellow
circles just for handicapping But

look at the pattern of where
there are other areas of change

red is increased blue is decrease there's
this missing stuff in the non-responders

that's not getting out into the periphery
it's not getting into the network and

well we had an idea of what the network
was we had maps of white matter we picked

it to be hitting twenty five but

to be hitting connections of twenty
five to all the areas that we've seen

with other treatments and
at that time when have tracked.

And now we have white matter we have
methods with diffusion imaging that we can

now up the connectome with imaging and so

we went to look at can we model
backwards what's good and

what's bad based on that in with
Cameron MacIntyre medical engineer and

he said the choice of his being
the degree here we look to see

what are we hitting and the logic for
the neuroscience as psychologist is look

as I've kind of made a joke people can
be non-responders for many reasons but

if this is something to do with us getting
better than they are to share something so

let's not compare responders and
non-responders and

small samples let's look with the imaging
what do we hit in everyone and

every voxel has to be shared by every
responder at six months and what we've got

was this man then we we modeled the with
tissue models that Cameron McIntyre

writes about and what we found is there
was an identifiable map shared by

everyone on their tracks based on where
they were being stimulated if you started

to look at well what was the same or
different between non-responders what got

most of it and non-responders but
obviously not.

Enough and we putting one on each side and
we found that we were missing one or

the other of the subcortical connections
to the basal ganglia in Thousand this and

more importantly the forceps minor
getting to the frontal cortex and

we could prove it by the fact that we
had people that by trial and error with

the psychiatrists you know changing which
contact they were stimulating over time

that we had converters that between six
months and two years people would change

and team said no it's about me I change
their meds by five milligrams or

I added some treatment no you
can't control that you can control

did you change the location if you only
looked at where we were stimulating in

people that converted from non-responder
still responders that over time

we recaptured the missing
regions on the track maps so

I said OK we've got we've got
a revision of how we might want to do

our procedure it's not
a spot eyeball that X.

percent and it's from the interior
comma sure to the front of the corpus

callosum and we'd sit there
the you know to surgeons and

me we'd look at coronal images in the O.R.

and say I think I see just enough of the
basal ganglia to go there I mean this is

how we did it is total total
subject to witchcraft like making.

Spaghetti sauce you know by group this was

no this is our now our new
template it's four bundles

we're at the intersection of
four white matter pathways and

we can drive around on a single
deterministic truck up on a laptop and

drive around with a little cursor
modeled about the size of the volume of

the tissue activated at the at
the current that we use and

we can find the place that gets our
template we basically did that.

In the O.R. the son wrote a program for

us to do that we would sit and
eyeball and transfer where we were to

Bob the surgeon we'd all agree
that we were in the same spot and

then over the course of six months we
left it there and never changed it and

what we got is this curve we had eight
out of eleven seventy three percent

better at six months on compared to
the forty one we had with trial and

error by the sky interest so and
we had one patient that we went back and

looked at the truck map and said there's
another contact that might be equally

good changed in that person converted so
this is now our routine.

And so what's the next thing we want to
do we've now we've now limited a huge

part of the variance without acting
asking anybody else to change any of

their behavior and
when you run a team and herd cats.

This you know you have to kind of
pick your battles because people need

a justification to change their
expert you know instincts and

so with this now this is how we do our
surgeries and now we can concentrate it's

all show you on on trying to actually
standardize other things to understand

what we're doing so let's talk a little
bit about about the what I mean what

changes with chronic stem and I told you
in those first six people we said well

this is going to take a while we write
our our ethics permission let's look you

know part way three months sounds like
a good number you know we'll wait to see

it's going to take a while figure out what
we're doing and and then six months and

you can see that whatever the switch was
going from said to well we missed it we

ended up doing two PET scans
pretty much in the same state and

that three months Ange change in the six
months changed where the saying so

so the question is is.

Why was that such about I.

And what might we do different so I think
that it kind of goes back to what did we

think we knew at the time this is
the logic in the field that you're sick

you got some number time passes you
passed some fifty percent threshold and

then you're well it's kind of this linear
process that's why you wait a period of

time let's look at the data so we've now
got three cohorts I've gone the in Atlanta

we've got the original group that we
just did the anatomical targeting

new team figuring it out we
had a blinded first month and

you can see that red line it kind of goes
down kind of slowly is a little more

linear the green line is that second Cold
War where we knew exactly where to put

it we turned it on a little bit in
the operating room again a month off but

look where they
are starting after a month.

Of nothing having gotten exposed
to good skin in the operating room

it kind of goes down and
then you have a couple weeks on and

you start to see this deck or
I'm in again it's not linear and

this third cohort is our latest
cohort where actually we do

repeated stimulation in a good spot in the
operating room because want to record so

they're trying to do it many times they
come in at that operating room and

have they don't progress and then you turn
it on again it's like they've been primed

they're getting better much faster so
we have something that the data is

telling us that we've got to think
about it a lot differently and and

so now we have to actually think about
what our models of what we're actually

doing that are we doing something
that's totally plastic or

do we actually have some sort of a local
minimum that these people are stock.

And with the right location and the right
threshold of stimulation we pop them out.

And maybe you don't pop them all the way
out of that hole they fall back in but

maybe you can pop them out and hold them
there for a while and then if you do more

stint there's a different time activity
curve so this is one model the think about

the other arm that the need has
been trying to think about is and

that you know looking things up it's more
of a homey a static plasticity model is

if you start here you turn it on in the
right place it goes in one direction and

you turn it on and it bounces around
until a con of gets into the gets

trained in the right space and you know
with Robby Tara Chris we're selling

a model these and
in different ways but but

the idea is the data is telling us
that just linear is not good and

what other evidence do we have what we
did PET scans here as well in that first

cohort and we looked at our responder
non-responder and and made the following

observation kind of I went back to this
data that had been sitting there for

five years that I couldn't
really understand before but

with this new idea in mind so
if you look at regions that are directly

connected along the white matter bundle
you actually see this pattern responders

are equal the non-responders at
the target remote from the target.

These are scans done before surgery

a month after surgery where it's been off
for a month and then at one month and

six months of active spin in
the same location same setting

it's done already in the O.R.
in those regions you reset it.

And then there are regions that
are second order connections not

directly in our blueprint like
the prefrontal cortex premotor quarter

there in the lateral Korda
sees that they don't change

from exposure to stim in the O.R.
but you turn it on you start to see that

it goes up and here's just
an example of both groups look good

at a month of stem maintained
in the responders and

last and then on our sponsors So again
something to model where should the signal

be detected you know we can we've
got ways to do that now but

this is setting the stage for how could I
follow that and the way to follow that now

is what we've been doing in our third
cohort of patients where how can

we look at where before we could
only look in the O.R. you know wire

room make measurements try to capture
local field potentials are E.G. or.

Now actually the device company has
installed has split the standard.

Pulse Generator into two parts half of
it is gone the conventional clinical

deliver the stimulation the other
half actually is God a recorder and

it can record local fuel
potentials not great not for

long but chronically you can
collect fifteen minutes of a lift

off the electrode while it stimulating so
you can divide a fifteen minutes

you can take a little epochs over the week
you can bring patients in every week and

you can follow along at what's
happening in the area twenty five so

if the story can be modeled at the target.

We can do it and it's back up
well we know that remote changes.

It's actually where a lot of the action is
that maybe the network will come back and

they'll be a biomarker off that lead that
would be great because it's implanted it's

easy you could model the information there
because the network changes it's going to

impact twenty five we'll have to look but
we're head to our bet we're going to look

at high density E.G.
Brian how in the lab is doing

very elaborate on physical models of
what we're actually stimulating so

that we can know is the template really
what we think we're starting to think

about how can we monitor
patients in in ways beyond

asking them how they're doing and
we've started and have been for

a while looking to clean the operating
room things like skin conducts and

responses and heart rate to really get a
read out of the autonomic system goes out

right in our network and then something I
want to show you is is actually we can see

that these people are different enough to
talk to him is read it on their face and

question is OK then let their
face tell us what's happening so

we started by saying let's
get an idea in the O.R.

to kind of type What's the behavior
of things people say so

you know people have read about this
it's very dramatic of depression switch

got written about which always kind of
annoyed me the beginning but it is.

And it comes in kind of two parts there's
kind of as change in the state of the body

where people feel lighter or they can
breathe a sale kinds of like really

interesting things but if you line them
all up we call that an interest of

awareness change and then and
that happens actually when you're

in a contact that hits a single IT
that's what's uncommon right or left.

And that it's seen in it's
not a specific signal so

it's a partial effect when you get.

This and then you get this second behavior
this is the good behavior this is

the dramatic one people feel connected
they feel like doing stuff they sell kinds

of funny things like one guy says
the tension is gonna I could walk my dog

One guy wanted to clean out
his garage I mean it's just

you know there are great if
you're writing your memoirs to

describe these funny things that that
people say but what they're saying is

there's stuff I want to do the breakers on
it can't do it over the break this off OK.

It's not happy it's not like a film like

going to Disneyland it's not
a he Donek seeking thing it's.

Kind of like a locked in and
now I'm not on and

so if they ever think they're thinking
about it really is a will to action and

it's things I've talked Alina about that
really this is this intersection between

the motor and the emotion and
it's part of a movement disorder and

it's part of depression and that's when
you get the singular that uniquely

get into forceps minor into
the bilateral frontal lobes so

it's got actually a readout of when you're
in that it only happens on the left so

even the good one on the right doesn't
produce this phenomena the good one

also gives this big jump in heart rate
well big jump you know four beats but a.

Demonstrably all type two effect
versus type one effect so

we're starting to get even
an autonomic readout of this report so

that that's quite interesting but
the most important thing is can we get

a readout that could actually help you to
know in the operating room that you're in

the right place and we'll go to smart to
do was supposed to talk in this lab and

electrophysiologist was to actually
try to model what is see if Act

the first time you hit the right
spot from the track Tiger Feet so

we know where the right spot is it gets
some well we looked at when that happened

when we were stimulating randomly
along the eight contacts and

what he got is if you're looking at left
recording or right recording the left side

in blue it's a big broad band change
in the local field potential.

The right side doesn't do a lot to
the left side it's not a lot of cross talk

if you're doing the right recording
the epsilon will stem also gives a bit

of fact but it's much more limited so

you've already got just like the behavior
just like the autonomic that the two

hemispheres are not responding to stim
in the same way even though the track

cover fee is identical hitting the exact
same connections on both sides so

you have something biologically
different on the two sides and this

drop in the Alpha on the left actually
predicts how fast they get well so

do I think that an alpha decreases
what happens long term no but

this effect in the operating room when
you turn it on when you turn it off and

what happened what's the net effect seems
to predict something that we want to put

as a placeholder The other thing that's
good is if we look at that good spot that

we know get some well versus the adjacent
spot that we know doesn't work again

the the main bands that are driving
the specificity are theta and alpha so

we've got two things going for
us there and now we can start to look at

what do we have in the cortex and
this is post dog Allison waters work

is let's see again when we're
in the right spot and we go into

Hertz can we actually get an idea of what
the cortical response of the brain is and

it turns out that there's a hundred
millisecond very reproducible evoked

potential response that starts at the
target moves to the front of the brain and

then sweeps over the back of
the brain consistently in a person

over months coming back every month and
across people and

if you want to look as could you use that
to know if you're in the right spot beyond

the truck to verify the truck telegraphy
Well here's being in the right place and

the wall in place into.

Jason in two people this person
the adjacent wasn't so bad but

in this person you can see just
by looking at the Peto program

that you can tell you are getting
the forty or the hundred millisecond.

Part of the sweep so starting to get
a tuner from the electrophysiology

we've even started to look to see if we
can decide on what dose to use because we

usually use about four or four and a half
we have to turn it up we started three and

a half Well here it looks like three and
a half is very different from form a half

maybe we started to love so that maybe
we can actually pick a first dose or

let the brain's response tell us what the
right dose is for a person is different

people may be different from
each other and we can then

even trying to look at maybe when you
give the therapeutic stem not to hurt one

hundred thirty Hertz not unilateral but by
lateral what's the net effect we want and

this is beneath some work that would
be in his thesis it's very preliminary

at just an Alpha this is the response if
you're in exactly the right place this is

therapeutically what that person is going
to get the first time their brain a month

after surgery gets the therapeutic
stem and this is what it looks

like if you're adjacent is sort of
backwards from what you think that

actually Alpha is a totally
noisy signal the wrong place

it's like being on your radio I'm lookin
fer the station I got static tellin

and so modeling this is going to be really
important to know not just Alpha just

on the worst person to talk about left or
physiology so house could be dumbed down

for me but the idea is this is
not just a map of location it's

Which band in what location in what model
and that's going to take some machine

learning process that I think
there's a lot of expertise here.

But what we really want to know
is have got that thing in there

in their chest we're measuring all
the time can we actually get a readout

of what's happening at
Area twenty five this or

something that could help us to tune
it maybe even have a closed loop and

a Shawn was here from University of
Maryland on a Howard Hughes for a couple

years back to medical school couple months
ago he took the brain radio data and

he did a very simple analysis
of the one over noise slope

he brought them in they came in every
week we did a recording in the lab and

he modeled what happened in the left side
in the right side so I told you about

the two sides are not the same in terms of
behavior not the same in terms of their

acute response without a sista
the left side goes on get tough.

Ladder slope starts steep and
gets flat over time and

the right side actually seems to go
steeper over time in terms of this noise

slope they're doing something
almost the opposite of each other.

This is quite significant this was noisier
but then he looked at let's just see

what's happening every week of what you
are severity score is let's just look at

the behavior and the slope and what
you get on the left is you get flatter

if you have a partial response or for
sponsors a binary effect you're sick or

well at the level of slope on
both sides of the brain so

we're starting to get a readout that we
might actually be able to use to tune but

it's not continuous It didn't really
work well until we said sick well and

I want to just kind of end with the idea
that this concept of the reset.

Is really important we're seeing that
there's kind of two basic States sure I'm

sure there is a middle state that
behaviorally we know this to this is data

on the self reports and to me See it
goes on it looks pretty stable and

then it starts to get on stable after
they're actually doing kind of well after

like two two and a half months here's
one patient they didn't really great and

then they lost it here and then they
settle back down here's a person did great

for five months and then lost it is
there a way to kind of have a clinical

readout of that because the clinicians get
to where they think they know what they're

doing again they listen to what they say
they listen to contacts they go I think

I have a you know some life stressors
where I want to know what the machine

wants to know is one way turn it up maybe
you've settled maybe it's history says

maybe you didn't start on the right
place maybe you're losing the a fact so

how might we look at that well the face is
really interesting in depression and this.

This She's an amazing cartoon.

Barely had depression and

she described that part of what was
really bad about depression is that.

She knew she looked off so people could
tell but she reached a point where she

wasn't sure if what she was feeling and
how she looked it was the same and that's

this almost emotionally Praxair that
your face doesn't reflect where you are.

That's a clue everybody you know you can
tell when someone's having a bad day but

you OK and depressed people get to where
their faces are not right they kind of

grimace when they're smiling it's odd and
there's the biology of emotional

facial expression and it's right in our
network in are all edgy you learn about

facial expression of a spontaneous model
where it's your lateral motorcoach or

sees that there's an emotional
medic system of the face and

it's actually right in the singular
where we have a dominant

abnormality in this part of our network so
we got some him.

And importantly his graduate student.

And our psychiatrist who's the one that
says well I can tell a rough patch

from needing an adjustment or
at least on thinking about it and

they went in we started filming every
weekly session when the psychiatrist

would talk to them in the first you
know thing is you know how do we go and

they have some conversation for
a while we've got it videoed and so

hard basically stripped out
the audio took the visual and

did stuff to it basically looking
at the whole face and then actually

applied three different machine
learning algorithms to those videos and

the first question is Andrea gave her
sick ROVs the transition period and

well and when it turned out was the most
important thing is everyone thought

there was this third category another
stupid categories they were sick and.

Well and for the ones that Andrea
said were transition outs in

yellow sometimes they looked like the well
and sometimes they looked like the sick.

And in fact by looking at this this is
now moving to the next stage we took and

reduce this large set of
features to forty two features

that best characterized the states and
then she set up a training and

testing set and said OK what do
I know to be true early sick and

late Well I believe let's model the system
on that and this is just not real data but

the simulation of the results what we
want to know is when you're into that

eight to ten weeks and
there's a question on if you're sick or

well what does your face say you
are based on the model of second well and

sixty two videos that are in the situation
behaviorally thirty three of them

more than half your well
they says you're well.

And only you know what third of them is
turn it up you know you're actually sick

so we're looking at this now it actually
uses the readout rather than talking to

patients to try to make inferences about
their behavior so I just want to end with

the idea that you know the last thing is
can we actually figure out who to pick and

again I trust the scans because we have
scans or at least that's something

we can model and we've already learned
from actually studies of drug and

therapy is that with PET scans we've got
bomb markers that if your brain is in

the upstate you do great on drug and
therapy will not touch you and

if you're in the down state you do great
on therapy and drug won't touch you and

we can make predictions so
can we do that by thinking about

treatment resistance we've had this
clue that treatment resistance

is something to do with this area twenty
five so we've started to look at the F.

M.R.I. How are regions connected to that
different in different kinds of resistant

patients and it turns out that the
connections to our from the frontal lobe

doesn't help us to know who the D.B.S.
patients are but twenty five

connection to that mid single at that area
that we had the hypothesis about the face.

That's only in D.B.S. patients so

there's something about that last
connection in our circuit that

seems really important to this kind of
patient that might do well with this and

I put the SIM for Constantine in he didn't
think I was going to do this I did it

the idea is now that we know that the
imaging can help us we can start to look

as he did with his graduate student as
new methods to model the network this

was done with a group of free agents
before we even knew about our template and

as he's done that is really interesting.

Two approach to modeling is here's
twenty five kind of that a nexus

between different groups and

we can start to subject our data do
these kinds of analyses as well so

I'm just going to end and I realize I
used the time but we can talk after.

What you realize that this is a real
collaboration between people taking care

of very sick patients because when we're
wrong we still have people we have to take

care of we have to do our best we have to
improvise these people have got devices in

their brain we have to help them to make
educated decisions we have to keep them

well if you listen to what they say you
actually realize they told us all along

that you know this was not an easy thing
for me and the longer they get out

the more they are sure that we didn't
actually properly consent them that we

hadn't told them how much work they would
have to do they weren't even have agency

to be able to deal with the idea when they
were in the pit where they were going but

you know we don't we're not
happiness devices this is not

change people fundamentally this allows
people to be whoever they would be if they

weren't constantly sick and
I found this so I was going through

my computer I pulled this obviously
many years ago recognized that Kerr.

If you can everybody read it this is
the funniest thing because I think I need

to take this one out this was usually
my work course about what we do or

don't do you know because
people want to be somebody else

you know you know sorry it's
kind of a buzz kill but

we return you to who you were with kind
of a long delay in life it's hard.

But it's better if you're not the press
getting through this notion that you're

out of the pit but that really you know
you just sort of want to be even keel and

life has ups and downs but

most of the time it's just kind of
not trippin and I'll end there.

I'm sorry used up the time I know
everyone has to go to class but.

I think I think you've you've nailed it so
one of the things that we want to do

in our new grant is actually
do the E R P in the O.R.

So that actually we could see if you
need high density to see the signal so

right now we know where the source
is we know what sensors give us

the major effect and the are you can't put
two hundred fifty six channels on the head

I mean if it was epilepsy patients you'd
have stereo grids but this is where would

you put it on the surface conveniently
because we need that frontal pole and

something at the vertex we can
work around surgical space so

the plan will be to do that
exactly as we did in the lab and

then can first see if we can get the
signal and then we can start to see if we

can make a guesstimate on the on
the dose so the thing we'd really

like to do is it may be individualized
I didn't have it in the side but

some patients you see a step function at
four volts not three and a half full so

it really may be that you
can set it based on the dose

response curve or it may be that because
those initial things were done later

it's already settled and there's
a reason that we go you know a month and

then have to turn it up so
did we understood at the beginning or

is no matter where you start
it's going to settle from

a home a static plasticity point of view
and you're going to have to push it up so

I think you know that could be a marker
for tracking along the device so

you know in a perfect world
we'd work harder on the signal.

On the device because then we could
actually model it in real time without

having to put another lead in someone's
brain you could always do that I mean

people do that that less hardware would
be better less stress lessens to break so

if you're nihilist like I am you
know like less things to break

as opposed to more data but
no perfectly often.

Thank.

You.