Khan – Advanced Noninvasive Neuromonitoring in ECMO

Dr. Imad Khan is an assistant professor in the Division of NeuroCritical Care at the University of Rochester. His research focuses on monitoring cerebral perfusion and pathophysiology in patients with […]

Episode Transcript

(00:00):
Everyone, I have the distinct pleasure to introduce
today's speaker, doctor Maad Khan, who is the
assistant professor of neurology,
neurosurgery,
and medicine and director of research in the
division of neurocritical care, department of neurology at
the University of Rochester School of Medicine.
Ahmad completed his neurocritical care fellowship here with
us at the University of Maryland 2015 to

(00:22):
17.
After completing his neurology residency at Stony Brook,
he
distinguished himself,
for his meticulous clinical work and work ethic.
We really put him to work. He was
a terrific fellow to work with, enthusiastic and
friendly and hardworking.
He offered several first author publications from his
time as a fellow with us.

(00:42):
1 of pilot study of NEARS and ECMO
patients and the other investigating brain histopathology
after ECMO.
He is an accomplished clinician scientist, now an
educator. His collaborative approach has allowed him to
be very successful enrolling patients into
challenging clinical trials like ICECAP and boost 3
and RAISE and TRACK TBI.

(01:03):
Throughout his 7 years at Rochester, he's
focused his clinical efforts in improving care for
comatose patients who've suffered cardiac arrest
and has led a number of QI initiatives
in that regard like,
the cardiac arrest survivorship clinic.
He's an ex he's a burgeoning expert in
the complex field of monitoring cerebral perfusion

(01:25):
pathophysiology
in patients with anoxic brain injury
from cardiac arrest or cardiogenic shock.
And you'll hear some of that today with
the collaboration of an,
optical engineer, epileptologist,
cardiac intensivist,
cardiac surgeon.
He developed
a paradigm for multimodal monitoring in patients undergoing

(01:45):
VA ECMO.
And most recently, and maybe you can update
us on this, is his lab group's NIH
r o one proposal
was scored in single digits.
So with that, I'd like to introduce doctor
Khan to give his top titled advanced
noninvasive neuromonitoring
in ECMO.
Thanks so incredibly much, Kyung Jin. It's deeply,

(02:07):
deeply humbling.
It's it means the absolute world to me
to be invited back to talk
in front of you guys.
Maryland is the place I call home. I
mean, it's a place that made me,
and, you know, the neuro ICU team is
the team that kind of forged my interest
and forged my career path essentially. So I

(02:28):
couldn't be here without you guys, and, it's
it's just incredibly rewarding to be talking to
you guys today.
You know, I you know, so as as
Gunjan said, I am at Rochester. I've been
in Rochester since, since I graduated,
since I graduated in Maryland.
And, you know, I'm gonna talk about,
my
project path that has,

(02:49):
revolved around non evasive monitoring and ACL patients.
Real quick, my disclosures,
that r o one that scored did get
funded. So January,
the project started. So this is my that's
why, it's an MPI with a biomedical engineer,
collaborator
as well as, we've got some funding from
Boost and Icecap and some some studies from

(03:10):
that. So,
kinda take you through things.
We'll we'll quickly go through definitions because this
group is a group that taught me ECMO.
So,
we'll define
real quick what I mean by when I'm
when I say, hypoxic skin brain injury and
coma.
And then, we'll go through the use of
noninvasive neuromoders,
to measure these,

(03:30):
factors and then,
the steps that we're taking next.
You know,
I think the this group,
is well versed probably in what extracorporeal membrane
oxygenation is.
The centrifugal pump provides the blood flow. The
gas exchange unit provides oxygen and carbon dioxide
gas exchange,
all,
artificially.

(03:51):
It uses for cardiogenic shock, cardiac arrest, and
acute respiratory distress syndrome.
But the pump speed, not the blood flow
is set by the clinicians. So we're trying
to try to approximate
the, perfusion to the body
by,
by setting a speed, essentially, and the actual
blood flow to the brain and the other
organs
depends on a myriad of factors, systemic vascular

(04:13):
resistance,
intravascular volume, and any obstructions in the circuit,
any kinking of the tubes, things like that.
So, and then gas exchange, of course, is
submitted clinician as well. So you're really taking
up,
the physiology of the brain,
and the organs completely in your hands when
you're when you're on ECMO.

(04:33):
There's no clinical set of monitoring of how
much oxygen or blood flow or perfusion the
organs are, are getting, and the organ that
we care about in particular is the brain.
So as you all probably have seen,
particular to VA ECMO and and in particular,
it provides continuous blood flow by the centrifugal
pump. You typically gonna see someone who has

(04:53):
a very
small pulse width,
when they're fully dependent on ECMO. And this
is a variable. It changes with time when
the patient's heart is fully dependent or the
patient's circulation is fully dependent on ECMO. You
know, you have very,
small pulse width, even you and as that
heart,
gets stronger, as recovery happens, or if you
place another form of mechanical circuit or support

(05:15):
like an Impella,
you know, that pulse width might increase.
So pulse utility might change,
and you're providing that blood flow retrograde of
the aorta typically in peripheral
cannulation.
And,
when you're when you're providing retrograde up the
aorta, it's gonna compete with native heart flow.
So you know? And you have no cardiac
function. You are kind of here's your femoral

(05:37):
artery cannula coming in,
and your pump your your, ECMO blood is
going to essentially
go all throughout the entire aorta and supply
the entire both hemispheres of the brain equally
with, well oxygenated well carbon dioxide removed,
ECMO blood. But as the cardiac function starts
improving or increasing,

(05:59):
that mixing zone such as the the where
the native flow the native,
blood is going to mix with the ECMO
blood, and that mixing zone might change or
might move
depending on how your cardiac function is improving
or changing, depending on how,
what the speed at which your your,

(06:20):
ECMO pumps,
might be,
revolving at. Basically, the
the the power with which you're providing blood
into the body might challenge against the power
with which your heart is natively ejecting, and
that might eventually lead to asymmetric perfusion,
or a different perfusion of the brain compared
to the rest of the organs.

(06:40):
ECMO can cause injury.
As well described, hemorrhages. It can cause,
interventricular hemorrhages, subarachnoid hemorrhages, ICH, thrombosis,
leg ischemia,
all part and parcel of
or all due to our inflammation.
This is schematic
of kind of imagine your mind's eye kind
of a line going down the middle,

(07:00):
a schematic of blood going through the vessels.
This is the artificial this half of the
screen, the left half of the screen, the
artificial,
PCFE cannula,
that blood is flowing through. And on the
right half of the screen is your native
blood vessels, your aorta, your cardiac arteries, your
your cerebral vascular.
As blood goes through the,

(07:22):
the,
artificial cannula, it's going to activate complement. It's
gonna activate neutrophils. It's gonna,
lead to
an up an upregulation of thrombin and platelet
activation, essentially. The coagulation cascade is also activated
through contact contact activation through the,
the biomaterial.
And then as it enters into your body,

(07:43):
I mean, it's going to activate anaphylotoxins
on the complement system. The neutrophils, which are
activated earlier, They extravasate into the, into the
body's tissues.
The endothelial
the endothelium of the native, vasculature is also
upregulated.
So there's a myriad of inflammatory,
factors and complement factors that are that are

(08:03):
upregulated in patients undergoing ECMO,
it's not simply the physics
of the blood going through the the tubing
of their body that we have to take
into account.
So
in ECMO, ECMO is usually started urgently versus
emergently. You never have the luxury of having
any neuro exam beforehand oftentimes and particularly in

(08:26):
cardiac arrest,
where you have cardiac arrest without ROSC and
you're kind of crashing onto ECMO.
To your traditional definition of, ECPR, extranial portal,
cardioclonal resuscitation,
you don't really have the opportunity for a
premorbid neuro exam. You can't really be, selective,
when it comes to understanding how much brain
injury this patient might have already incurred from

(08:47):
cardiogenic shock or from cardiac arrest.
In ARDS, for example, your patient might already
be paralyzed, prone,
with a very limited opportunity for neuro exam
before you choose to go on ECMO.
And they bring that hypoxic ischemic brain injury
pathophysiology
to ECMO with them. So
not only can the ECMO actually cause injury,

(09:08):
that we described from mechanisms
poorly described mechanisms, mind you, in the previous
slide, but the patient's bringing with them this
path of mechanism,
from their underlying injury that or the underlying
illness that necessitate that come in the first
place.
And so, I mean, what's happening in the
brain, in the healthy brain, it's kind of
a half a schematic here.

(09:28):
There's the
I like to kind of divide the brain
into that's basic very basic unit, the neurovascular
unit, where the neurons are communicating with the
endothelium, the endothelium,
the endothelium
is encased by the astrocyte's foot processes. Right?
And so these astrocyte foot processes are creating
the blood brain barrier, the the, the sheath
that is that is kind of allowing for

(09:49):
selective
movement uptake of of proteins and molecules.
And that's all the neuron communicates with this
endothelium via the astrocytes, and so there can
be kind of a a upregulation of perfusion
when the neuron needs it. And it can
communicate with the cap the capillary to dilate
or constrict or that arterial to dilate and

(10:10):
constrict by communicating with by by the astrocyte
when it's all working.
And when you have hypoxicamony brain injury, you
know, you
have a kind of a mishmash of of
pathomechanisms
that are happening here because, overall, you have
this massive
metabolic crisis from the lack of perfusion, from
the sudden lack of perfusion,

(10:31):
the sudden lack of of of
provision of glucose, of oxygen.
The brain is extremely oxygen hungry as you
know.
The the neuronal membrane starts breaking down because
of the lack of ATP.
The,
aerobic metabolism
has ceased. So,
you know, if you can provide as best
CPR as you can, at least give some

(10:53):
forward flow, but it's still suboptimal to the
native to the native circulation that you just
had before you before you rested.
So there's basically a a metabolic crisis happening
in the neuron as well as the astrocyte,
and astrocyte gets injured. The foot processes
shrink and get injured. There goes your blood
brain barrier. The blood brain barrier is now
broken, and the neutrophils

(11:15):
that are that are, upregulated
can be, more prone to diapodizing into the
brain tissue. And And then you also have
these because of the,
activation
of platelets and and and clotting factors.
You can have microthrombi
kind of
create be created or move into capillaries,
creating kind of a

(11:36):
patchy,
pattern of blockages in the, microvasculature in the
brain,
leading to a no flow or it's kind
of a in the no reflow phenomenon when
you have,
CPR and native circulation start again, or in
our case, ECMO start again. So even when
you have
flow restored,
you still have this ongoing path of mechanism

(11:58):
of the lack of reflow in certain parts
of the brain because mic because microthrombi
and,
microvasculature
are blocking that, that flow. So you get
an uneven,
kind of
metabolic crisis or an abnormal metabolism ongoing into
parts of the brain and,
oftentimes, the more metabolically hungry parts of the

(12:19):
brain.
And that is a process that's ongoing, of
course, post ROSC. So the the party only
begins at ROSC. So
so we looked at,
you know, what is how do we characterize
a brain injury in ECMO patients?
And we tried to get as close to

(12:40):
the truth as possible. So we started trying
to look at patients who had died, and
and, looked at their autopsies, their brain autopsies.
So when looking under the microscope, we saw
a new, a kind of a a vary
a variety of histopathologies.
This is work from,
slides taken or, autopsies taken at Maryland as
well as Rochester and Michigan.

(13:01):
We had,
injuries, like, such as microhem microinfarctions,
spartoratory glands, the the pons,
microinfarctions,
from kind of these small capillary,
bursts.
We had,
very, ventricular vacuilization,
some signs of cerebral edema.

(13:22):
And, you know, all of these
were,
not only cortical, but subcortical as well. So
you have injury in various different parts of
the brain,
and a lot of these were thought to
be in the same kind of realm or
or or, spectrum as hypoxic ischemic brain injury.
But we could be so certain that that
this injury wasn't also caused by

(13:43):
the pathomechanisms that have been put forth from
the ECMO machine itself that we reviewed in
the previous slides. So you're kind of complicating
the pathomechanisms
a lot more
by adding ECMO on onto this patient who
already has
significant,
hypoxia to the brain.
So
how do we monitor for these situations? How

(14:05):
do we monitor these patients? You know, it's
a very complicated situation. You have to monitor
how you understand the degree of injury that
the patient came to you with and then
how much you'd be able to understand
the degree of injury that the ECMO circuit
itself is causing. How can you tease them
apart?
How can you try to, under get understanding
then what's actionable? What could be changed? So

(14:26):
we needed to a way to detect
the degree of neurologic injury from ECMO, any
new neurologic injuries from ECMO, and also a
way to kind of predict, what whether the
patient will be able to wake up or
not because that decision or that prognosis is
going to play a key role in whether
this patient, for example,
is a a good candidate for a VAD,
for example. What's the what's the destination therapy

(14:48):
after ECMO, remembering that ECMO is merely a
bridge. ECMO itself, to be honest, is not
going to
save the life. It's not gonna reverse everything.
It's only going to kind of temporarily pause
everything so the dust can settle so you
can figure out what to do next. Is
the heart transplant? Is it the heart going
to naturally
resolve or improve or heal? Lung naturally gonna

(15:09):
resolve or improve, or heal, or is it
a bad or a more durable,
mechanical circulatory support?
So you need a monitor that is noninvasive
for at least you know,
we
I mean, in my at least in my
experience,
I would be kind of squirmish to put
a invasive monitor in patients who

(15:29):
might be prone to,
hypocoagulability
due to the,
the consumption of clotting factors and complement cascades
and stuff like that, that we outlined in
the previous slides.
You know, we need a monitor. We definitely
need a monitor that's at the bedside because
these patients are extremely challenging to transport.
Your scans might give you a nice

(15:52):
spatial resolution image that can tell you,
the degree of injury throughout the entire brain,
but it's merely a snapshot. It's kind of
tough to,
act immediately to try to, stave off secondary
brain injury from just merely a scan. We
need something that is continuous because of that
as well.
So as I mentioned, like, most temporary. Right?

(16:12):
You're gonna either recover,
organ transplants, bad, or you die with withdrawal
of life sustaining therapy. And so it's a
high stakes game when you're talking about prognostication
on this.
Like, what was clearly expensive as we all
know, you know, from a couple of studies
that we found, during COVID, you know, COVID
without,
pay COVID patients without VV ECMO. We're talking

(16:33):
about 16 days on average, hospital stay and
$11,000
of their, was their cost. When you add
ECMO into the mix, you know, we're talking
about 30 days of stay and over $200,000
on average. The patients who had ECPR, I
found a paper that talked about their average
cost being something like $81,000
and, post cardiotomy

(16:54):
ECMOs,
were on a $100,000,
their cost of stays.
So, clearly, we need something that this is
a high stakes game, essentially, as we said.
So
what modalities are we dealing with then that
are noninvasive,
continuous at the bedside?
Well, the one that is
widely fairly, I think, by now, fairly widely

(17:16):
described,
at least well described in multiple retro observational
studies as retro retrofitters studies, but not in
a clinical trial to my knowledge,
but, is NEARS. This is,
commercially available, multiple
FDA approved devices, including
devices that we had tried in Maryland when
I was a fellow.

(17:36):
This is how I started off. Dan Hurd
put
told me to start putting, a NEARS device
on people's hands on in, cardiac ICU.
And, you know, we took a look to
see if that could spot,
any onset of neurologic or unilateral and neurologic
injuries.
You know, to remind ourselves, NEARS, near infrared

(17:57):
spectroscopy, it measures
oxygen, tissue oxygen saturation. It's noninvasive. Right? It's
cost effective.
The disposables are probably
on the order of, like, $100,
200,
and they're easy to place in your forehead.
They're somewhat unreliable.
The, you know, the reliability has improved with
time, but there's still there's still plenty of
papers describing their unreliability.

(18:19):
They're,
they can be easily,
falsely
falsely elevated rather in patients with hyper with
hyperpigmentation
with patients who have more melanin in their
skin.
So that can lead to to to significant
issues that we saw in COVID, for example,
pulse oximetry as well,
with disparities actually, with this racial disparities,

(18:41):
so they can be fooled.
And STO 2, the thing they're measuring, the
saturation tissue oxygenation,
it's only a surrogate a cerebral blood flow.
It's it surger blood flows are I'm sorry.
S t o 2 is also dependent on
the other parts of the content of arterial
oxygen equation. The amount of hemoglobin you have,

(19:02):
the the
the saturation of your peripheral oxygenation or peripheral
blood oxygenation,
and also to a smaller extent, the dissolved
amount of oxygenation, the partial pressure in your
your arterial,
in your arterial system. So STO so you
could have, for example, hyperoxygenated
blood, theoretically, hyperoxygenated blood from that small circuit,

(19:22):
and your CDFs
might still be suboptimal
at a time, but your STO 2 might
still be tricked that way if you're if
all your if your hemoglobin and your STO
2 and your PM 2 are
normal or should be therapeutic
in theory. So,
that's you know, it's not the perfect model,
but it's still widely used because it is

(19:44):
fairly cheap and and and easy to put
on people's heads. Stromine ourselves is a light
source, near infrared light source.
Oftentimes, they have a detector that has 2,
2 different detectors rather. 1 a little bit
closer, 1 a little bit farther. The idea
is that the light source,
shines a light into the brain's,
cortex. It bounces off the the or I'm
sorry. It's differentially absorbed by the red blood

(20:06):
cells in the cortex,
and it's,
the light that is reflected
back up, it allows you to understand
how much oxygen in hemoglobin, how much deoxy
hemoglobin there is,
and that can give you essentially a saturation
of the tissue.
So because of the caveats that have been
previously described,

(20:26):
we,
started looking at
diffuse Coriolis spectroscopy. This is this is, Regine
Chay and,
Erfan Dhar.
Erfan is a PhD student who just recently
defended it off of all his work, and
Regine is my engineer colleague,
who
specializes in this, type of instrumentation. This is
not FDA approved. This is a lab built

(20:48):
instrument. And to my understanding, there's only 15
labs that can do this,
and it's mostly
the majority of them are Regine's classmates.
So the very small
field of people who are using diffuse correlative
or spectroscopy.
And it diffuse corrosive DCS, it measures perfusion
directly. It measures CBF directly.

(21:09):
In NIERS, you're you're using near infrared lights
that bounce off of RBCs,
and the RBCs,
depending on their oxygenation state, absorb light differently.
Okay? And the near detector is going to
tell you what,
wavelengths of light have been absorbed. Okay?
In DCS,
the light is shine. It's there's still a

(21:29):
near infrared light. It's, produced by a laser,
and then it's bouncing off of
moving red blood cells, and the detector is
detecting basically the speckle of that of that
light. So,
essentially,
it turns into this thing called the autocorrelation
curve. And so when you have fast flow,

(21:50):
autocorrelation curve looks like this. When you have
slow flow, the autocorrelation curve shifts.
And so it's delay it delays it in
time. So you can essentially
have a a more accurate surrogate of
particularly cerebral blood flow. So,
we call this relative blood flow. That's been
well described in in,

(22:10):
in the literature behind DCS.
You know, we and also, I'll tell you
I'll show you a little bit of of,
information about validation
studies
in a second. So we are our design
was,
crafted as such. So there's 2 probes that
are,
custom 3 d printed in our lab.

(22:30):
They're slim. They can fit on your forehead.
There's a light source,
this fiber optic source
kind of coming,
directly into this pad that little prism inside
this pad that bends that light 90 degrees
so that this light going this direction then
gets bent. And then when you look at
this pad, there's like a laser, essentially, that

(22:51):
will go into your scalp. It'll go about
1 centimeter deep into your into your cortex.
This these fiber optic cables attached to a
laser off, in a in a on a
cart off the screen.
This is the,
the light source wavelength.
Specifically,
this this wavelength specifically reflects off of RBCs.
The source detector separation, that's the source detector

(23:13):
separation, which is really important. This this duration
this, distance is what determines how deep you
can detect. But the the the the further
you,
you separate the source and detector, the deeper
into the brain you can theoretically probe. However,
the the higher your signal to noise ratio
I'm sorry. Your lower your signal to noise
ratio will be, the more noisier your signal

(23:33):
is. You could overcome that by increasing the
power of the laser, the light source,
but then you can burn the patient. So
there's a there's a little kind of a
tight, tight rope to,
to walk there. So, you know, pretty much,
standard that's been well described, not only in
DCS, but also in NERS, is 1 centimeter
source

(23:54):
vector separation for your superficial to kind of
measure your superficial
perfusion in the scalp, the superficial tissues, and
then 2.5 centimeters to measure
your deeper brain tissue like your cortex.
Subtract the 1 centimeter data from the 2.5
centimeter data, and you can and you can
get just a quart just a cortical,

(24:14):
perfusion
alone. And this this concept has also been
described in in previous,
NERS machines as well that are FDA approved.
Our machine gets one data point every 2
seconds. We have, this was a our first
for this machine was our first foray that
was kind of before we got funded. So
it was already built,
for a different study altogether. So we kinda

(24:36):
just started applying this on the head using
pilot funds.
We had a right and left hemisphere,
probe, and it would alternate every 4 seconds.
So right, and then 4 seconds later would
be another right one, but it's right, and
then 2 seconds later, left, then right, then
left to get a data point alternating.
I reach the seconds.
And this is us putting our first this
is a regime putting on on our head,

(24:57):
and this is what it attaches to this
laser light source right there. It also has
a,
photo avalanche detector that can
detect the photons coming in.
Alright. So DCS has been validated for a
number of years before we started our foray
against,
a a chiaroscopy membrane MRI, against xenon CT,

(25:18):
TCD, PET, all,
quote, unquote, gold standard methodologies for measuring,
CBF in vivo. Right?
This displayed study was, a rodent model that
underwent,
progressive hypercapnia
with the ASLR neuron light and the on
the y axis and the DCS,
CBF on the x axis.

(25:38):
So kind of just showing the correlations of
validation studies. So we this has been so
we had come into this with this knowing
this validation data already.
So then we validated our own instruments.
This is me doing a breath hold test.
So we first,
you know, this is seconds on the x
axis. We first put on our heads, and

(25:59):
we kind of breathe normally.
And then I take a nice deep breath,
and then I hold it for 30 seconds.
It's the pink shaded time period right here.
So we can see that it's my nice
deep breath.
And then we can see CBF
totally rise
until I start breathing again. That's 30 seconds.
And we had TCD data as well.

(26:20):
It,
which I'm not showing in the screen right
here, but it correlated with the TCD data
pretty well.
So
that was
that was our first instrument, and I'll show
you data in a second what that showed.
But,
that was DCS.
We did not only want to stop with
just measuring CBF.
We want to also measure the activity of

(26:43):
the cell that we really care about in
the brain or at least,
that we thought we only care about the
brain, which is the neuron.
There are now the a lot more emphasis
being placed in the astrocytes as well in
the, the microchialia.
But in the, to measure the brain's actual
activity, we measure the EEG. Obviously, excellent temporal
and spatial resolution as well as studying in

(27:03):
coma, and no no surprise there.
Quantitative analyses from Persist have really, advanced our
understanding of how to classify coma in these
big populations,
particularly in Hibb patients.
And so that's what we we added to
this entire mix.
We also performed auditory brainstem response, auditory evoked
potentials, essentially.

(27:24):
We measure the latency to wave 5,
which is right there. So
from a series of clicks, we are,
recording
the electrical stimulus at various,
at, at the auditory
cortex essentially.
And, we're measuring wave 5 particularly because,
we we theorized that that could give us

(27:45):
an idea of any injury subcortically
kind of in the in the, brain stem
as well as subcortical nuclei.
And that was kind of informed by,
our autopsy data that we had seen a
lot of, not only cortical injury, but some
cortical kind of injuries that we had seen,
on early in this talk.
So we wanted to at least have some

(28:06):
sort of measure, for subcortical injury.
But more lately, I've been,
programming a mismatch negativity at Veradigm for this
because,
we what we do is we play a
series of clicks in the 3 earbuds in
these patients who are comatose,
and measure this. And what's been described before
is you as if you play a particular

(28:26):
paradigm of clicks and include an oddball sound
and oddball tone that's different than the rest
of them, If the patient has patients who
can recognize that oddball tone
have a bit more,
it's been shown to be more sensitive for
being able to awaken or specific for being
able to come out of coma. So if
you, I love most of our tests clinically

(28:47):
used are specific for poor outcome. Mismenselectivity
might be specific for
good outcome for recovery from coma.
So,
we do that using the attributable potentials.
So
combining them all together,
we created our 1st pilot study, which we
enrolled for a number of years called non
invasive neuromonarchin in ECMO.

(29:08):
It was an observational cohort. It was a
convenience sample.
And the first thing we wanna do is
just establish feasibility and combining all these modalities
together. And then,
we wanted to see what we can detect
as far as signatures of brain injury,
using DCS, EEG, and the ABR.
We included VAs or VVs to start with.

(29:29):
We didn't wanna be, you know, we wanna
be,
beggars, not choosers here.
To start off with,
eventually, we narrowed it down to VA commos
only,
starting in 2022. We started off, like, in
2020.
COVID did significantly delay our enrollment,
but we got enough,
to move on to for for the funding.
These are patients who were cannulated in the

(29:50):
past 24 hours, and they were adults only.
We'd exclude you if we couldn't put a
pad on your on your forehead,
facial trauma. Any ECMO that was expected to
last less less than one day or preexisting
neurologic disease,
would exclude you from the study. So what
we did was we'd consent you, and then
we'd monitor you for about 2 to 4
hours each day.

(30:11):
I couldn't do longer than that only because
we needed to have a grad student sitting
next to our t our DCS machine at
all times because, again, it was, you know,
our very first,
our very first foray into making this machine
in for the ICU setting. It was not
designed for that initially, so it's not the
most robust And, indeed, it did break twice,

(30:32):
so I'll I'll show you, later.
We would also I didn't mention this. We
also added TCDs as well. We do we
do TCD spot checks ourselves,
to measure, CBF against DCS. We'd measure the
auditory bone brainstem response. We'd get their vital
signs, by plugging into the Philips monitors we
have
and downloading all that data.
And we do continuous EEG,

(30:54):
for a number of days. We wouldn't we
just leave it on just for the tech's
sake,
and we would, get their exam every day
as well. So we do this every day
while the patient was on ECMO.
When the patient would start weaning,
if the patient start weaning, not everybody did,
of course, we would try to extend that
session by sitting there,

(31:14):
for up to to 6 hours for 12
as long as 8 hours, I believe we
had. So we try to capture a prolonged
period when they were weaning
so as to kind of see changes in
CBF
as we step down on the ECMO speed
and also if we were changing the, the,
sweet gas flow.

(31:34):
We try to capture a post accumulation
day. I think if they survive decalation, we're
gonna capture one measurement post accumulation
to kind of get an internal control on
native circulation,
or on circulation does not echo like that.
And then,
on discharge, we get a CPC score. And
then at 3, 6, and 12 months, me
and a couple of med students would call

(31:55):
the patients to get a Glasgow outcome score
and then a tick score,
and telephone,
survey based on cognitive,
injury.
So,
basically, we had 22 patients from that duration.
We had a long break for COVID.
First thing we found was it's feasible. We
had no adverse events, namely, we didn't burn

(32:15):
anybody. We had to decrease the the power
of the, the laser,
per, OSHA standards.
But we didn't have any rashes, no disruption
in patient care.
We didn't piss off any nurses. It's the
most important thing.
The probe broke twice,
and, that was something that informed,
our future probe design.

(32:37):
And we had about 10 out of 90
days, 11% of the days that we did
DCS data 4 were discarded due to low
signal noise ratio.
That's about and that happened in about 3
patients.
So,
I think because of probably because of poor
contact to the forehead.
One thing that was important was the auditory
brainstem response was mostly unusable because of poor

(32:57):
signal noise ratio.
We found that the
impellers,
about half of our patients had impellers,
and that created a
noise, a electrical noise that would completely obviate
our our auditory brainstem response. But not only
was it the impeller, it was also other
things like the like the rest like the
ventilator as well. So,

(33:20):
you know, in later patients, more more recently,
we have a couple grounding cables, and we're
working on that right now. But the our
initial, you know, our initial foray into using
monotonic voltentials has been pretty challenging.
We so that was our first thing we
found. Okay. So analyzing these patients,
we
one of the other things the first thing

(33:41):
we that kinda popped out of us just
by looking usually looking at our data was
that patients who are comatose
had this
asymmetry in their CBF. Their left and right
hemispheres
seem to be more asymmetric compared to the
ones who are awake. Remember, these were,
all comers and ECMOs. Some of them were
just sitting there intubated
talking to you while writing. They were completely

(34:03):
awake, and some of them were completely comatose.
So when we analyzed our our first 13
patients, the comatose patients were never defined as
never being able to follow commands,
versus the ones who were,
you know, able to follow commands at least
once and and would majority of time follow
commands. So, you know, kind of a mismatch
there as far as numerically concerned. But what

(34:24):
we did was we measured asymmetry by taking
their this is their RBF tracing, the the
blood flow tracing from the DCS. This is
their map data
as time goes.
This is taking their relative blood flow and
and plotting it against all different map values,
so map values in x.
So we get and we do that for
each hemisphere. Right? We get the average

(34:46):
relative blood flow at each map point. Okay?
We do the same thing for left hemisphere,
plot it out. K? And then we'd subtract
each other to get asymmetry.
Alright? So this we term this asymmetry the
asym RBF
is the
sum of all this, the sum, under the
under the curve, essentially,
giving you percentage. Okay?

(35:07):
This on this y axis is is just
the mere the absolute value of the delta,
the difference between the right and left hemispheres
at each, for relative blood flow at each
map point. That's why it's delta RBF map,
and it's described as a percentage. Alright? But
when we summed all those percentages up under
the under the curve, we get a total
asymmetry RBF.

(35:28):
Alright? So it's
pretty cooked, this data.
So we do that for the the non
comatose or the comatose patients. We do this
for
the day that was that was most asymmetric.
The patients had at least
the the overall I'm sorry. Not the almost
symmetric, though. The sum of, essentially,
the patients who were comatose

(35:49):
had an overall
higher sum asymmetry compared to the non comatose
patients.
So that was something that was some somewhat
interesting. We're trying to trying to figure that
out.
We
when you when you saw the 4 patients
who are comatose
and you, plot their delta RBF map, their
their,
difference out versus versus the map versus the

(36:12):
range of maps on the x axis,
something, first of all, that was that that
is good to know is that look how
short or small this range of maps is.
It's, you know, kind of owing to the
patients being pretty dependent pump dependent on ECMO,
when we measure them. If we had measured
them for 24 hour, epics, for example, maybe

(36:32):
we would have seen
more broad
map, ranges. You know?
But
I think that this a small range of
maps will be observed
probably because we only measured them for 2
hours, and they were more often than not
pump dependent. So pretty stable on on, map
wise on ECMO.
The majority of this is this is, when

(36:54):
right was greater and blue is when the
left was greater. So you can see that
I mean, 3 for 3 out of 4
of them first of all, for all of
them, there was some range of maps or
some points in in map where that asymmetry
was minimized.
And then beyond outside that range of map,
that asymmetry kind of grew. So outside this

(37:14):
optimal map, for example,
patient became more asymmetric, more towards left to
side being higher than right. Same thing here.
At some point below, the right became a
little a little,
higher, and beyond that, the left became a
little higher, essentially, a little bit of noise
in between.
So
it seems like we had more points in
the left than the right,

(37:35):
for the higher maps,
and we wondered about this. We we plotted
the asymmetry RBF percentage
against
p c o two and pH. Right? We
were wondering about whether this has something to
do with auto regulation.
And so when we plotted it,
the asymmetry RBF versus p c o two
and PH, and this is from the post
auctioneer blood, which we've theorized was the blood

(37:56):
the brain was seeing more than anything else,
the the the correlation for comatose patients was
different than the correlation for the wake patients.
Now the if you were comatose, your ASMRBF
was more likely to be correlated
to p c o 2. You were more
likely to be higher asymmetry
when with the more p c o 2
you had.
So,
same thing, you know, similarly

(38:18):
inverse of that, the pH or the the
the higher p the lower pH you had,
the higher asymmetry you had.
So we wonder whether this something something to
do with autoregulation impairment,
in patients who who are comatose.
And
more importantly,
I'm gonna skip over this for a second.
We've all seen the regulation curve kind of,

(38:39):
you know,
the the brain's blood, the vasculature is supposed
to be able to kind of autoregulate
the flow going through it by either dilating
constricting for a
wide range of maps.
And
when you're heavy, you know, you sometimes become
passive
pressure passive rather or you become right shifted.
So for either you either have a much

(39:01):
narrower range of maps that you stabilize your
CBF for or you're you require a higher
CBPR map to maintain the same CBF.
And so
I'm sorry. So what I mean to say
is that
one thing we wondered whether is this patient
with Hiby already was prone to having dysregulation,
and then we are taking blood and shooting

(39:23):
it retrograde up the aorta. And is it
possible that the patients who are comatose were
not able not able to kind of
regulate their their their,
perfusion as well as the ones who are
awake,
and then they were able to see theoretically
more their left hemispheres are able to see
a higher kind of blood flow compared to
the right because

(39:44):
it's going up the left more directly. These
are all all our patients were all our
patients who were, comatose were VA ECMOs, and
they were we never had any central cannulations.
We were all peripheral cannulations.
So one thing that we were wondering about
was whether this could be
some sort of a marker of,

(40:04):
is,
a I'm sorry. Some sort of it could
be due to some sort of thing,
caused by the retrograde perfusion of the aorta.
So
a very shaky kind of finding,
and we're gonna we weave that into our
future study, that I'll describe in a second.
Anyway, so you all have, heard about the
auto regulation,
Lawson curves.

(40:24):
We started
probing, autoregulation with our data. So, you know
that we've heard of PRX, respiratory activity index,
best described method of measuring cardiac I'm sorry.
Best described mission method of measuring cerebral autoregulation,
has been described in TBI, cardiac arrest, stroke,
ICHs.
The PRX is the slope of the lesson
curve that we just talked about,

(40:45):
and you you basically take a moving correlation
of ICP versus MAP to get PRX.
That requires a invasive ICP monitor and requires
variation in MAPS and ICPs.
And if you were to plot the PRX
remember, loss of curve is y axis is
CBF. Okay?
In the PRX in this curve, the y

(41:05):
axis is now PRX,
and the x axis is still CPP or
max or map. If you plot all the
different,
slopes
out at every map point,
theoretically, if you are able to get a
wide enough,
picture of the patient's map range, you will
get a optimal map where that that slope
is minimized. And that's the optimal map where

(41:25):
the patient is, you know, the sweet spot
where the patient is not going to have
any perturbations of CBF.
And so
challenges in our case were that in VA
ECMO, the delta map, as you saw, could
be minimal.
Maybe because we were measuring too short or
maybe because the patient's map changes are minimal
because they're very dependent on pump.

(41:46):
So to solve this, we correlated DCS derived
BFI
with the the blood flow index I've described
before, the RCBF measure. We correlated with map
2 different ways. The traditional,
you know, moving correlation with map,
in instead of PRX, this has been described
as DCSX, and this has been described before
in previous papers.

(42:06):
But we also correlated it with map using
a wavelet coherence,
and I'll show you why in a second.
So
our this led us to find our our
third finding, ECMO,
ECMO patients with hypoxic sickle brain injury had
a higher degrees of cirrhotic circulation than those
without hep b.
So
in this in in this analysis, we did

(42:26):
5 patients with Hibb, 5 patients without Hibb.
We defined Hibb using
CT criteria, CT scans that had loss of
gray white differentiation
and cell phone definition. These patients also were
all comatose.
One of them who had ARDS and VV
ECMO. The rest of them were cardiac arrest,
so VAs.
The normals were 5 patients who were did
not have any CT scan injuries.

(42:47):
They were all VA ECMOs.
We set this limit for quote, unquote intact
auto regulation,
meaning that lesson curve slope as less than
0.4. That's been based off of previous literature.
So using your traditional method where you take
your RBF,
you know,
and,
your map as well, so we kind of
correlate that against map, and you get this,

(43:08):
this autorerelation curve.
Sorry. You get this autorerelation plot rather, DCSX
versus time, and we
write this line at 0.4.
If you're above 0.4, that badness, and we
sum that up and give you an area
under the curve
for each patient.
That is one way of measuring lot of

(43:28):
relation.
However,
a lot of our maps were, as you
as I saw, just
a small amount of delta, a small amount
to change your maps. So
we would only be able to calculate
this we won't be able to calculate the
DCSX
if your delta map was more than 5
millimeters mercury. It would be,

(43:50):
it wouldn't we couldn't really tell if there
was if you had less than 5,000,000 of
mercury change in your map over a period
of time, then we wouldn't be able to
tell if there was a difference or a
correlation
with RBF. It could've just been noise.
So a lot of our a lot of
our data we had to throw out because
a lot of it wouldn't was not,
a delta of, greater than 5 millimeters mercury.

(44:12):
So we did, so the next thing we
did was a coherence calculation. We took our
RBF
data, and then we had the map data.
The red line is map data. K? And
we wanted to see how well does by
breaking down the the wavelets or the wave
the the the, wave of the data, how
well does it correlate essentially with the between
the 2 different, types of data, essentially?

(44:34):
So we break the map and relative blood
flow into frequencies, apply it over time. Okay?
This is the RBF,
relative blood flow, frequency on the x axis,
time in the y. I'm sorry. The way
around. So if you can see on the
y, time in the x, and map,
same thing.
So we run it through a wavelet coherent
equation on MATLAB,
and we get essentially a coherence. How coherent

(44:55):
was the different frequencies,
like, did, breaking them down to co components,
the different frequencies,
how co coherent was it, and the over
the range of frequencies we saw.
We use the we we kind of paid
particular attention to the this range of frequencies,
0.05
to to 0.1.
As that has been previously described,

(45:16):
frequency is a correlation that,
I'm sorry. Collation that when your frequencies before
has been described as being associated with the
endothelial function.
So we've broken it in 20 minute segments,
and amplitude is this color here.
So we wanted to make sure that this
coherence was not just noise. So we also
do some really in-depth math that

(45:37):
our my my grad student did,
create random,
data as well
to compare it against. So we kind of
ran this through,
a 1000 simulations,
essentially,
of different shift. And we just shift the
the, the the, data
stream,
a 1000 different ways randomly, And we created
a 1,000 different wavelet coherences

(45:58):
and kind of put them on a histogram,
and we create this randomized,
just kind of normalized,
curve as well. So whenever there was a
significant difference
between our data that we calculated and this,
this randomized sample, that way we could trust
that as being more significant coherence than compared
to noise, if that makes sense. So the

(46:19):
upshot here is is that for DCSX,
we we, as I mentioned, we only wanna
use a delta map of more than 5,000,000
mercury. The rest, we discarded.
So for each of our patients, here's how
much we used how much data we used.
Okay?
And then we had to discard the rest
of this. Right? So up to 70% we're
discarding.
Right? For when we did wavelet coherence analysis,

(46:39):
WCA,
we we were able to use upwards 99%
of all our data.
And so when we plotted out, you know,
by looking at the uninjured versus injured,
looking at our old gold standard, the DCSX,
it was the patients who were neuroinjured,
had both on both hemisphere well, almost both
hemispheres
had a higher DCSX, higher correlation

(47:02):
compared to the uninjured groups. Okay? So it
made correlation. It made significance for the the
right hemispheres.
Just missed it for the I'm sorry. Just
missed it for the right hemisphere. It made
it for the left hemispheres. When we did
it for the wave of coherence analysis, it
was statistically significant on both hemispheres.
And,
so we took this to be,
a real measure our measure of,

(47:24):
cerebellum dysregulation.
Okay. So what we also use the we
combined BCS and EEG to probe the neurovascular
unit.
So,
a real quick review of,
neurovascular unit function. Neural activity, as I mentioned,
stimulates arterial or capillary perfusion.
Postsynaptic, you know, glutamate release leads to vasodilation,

(47:44):
and it's all mediated by the astrocytes as
well. Neuronal metabolism leads to an increase in,
or a drop rather in the partial tissue
oxygen duration.
It uses your oxygen in your brain, and
then that can lead to a sustained vasodilation.
So initially, synaptic activity
activates the capillary
and forward tell tells the blood to come
to the brain. And then as you're using

(48:06):
your brain as the oxygen drops,
it tells the the the vasculature to keep
going, keep keep refusing.
So So it rely this all relies upon
a functioning neurovascular unit. When you have HIBI,
what happens when you don't have the function
as neurovascular unit?
So we probed that. We had a full
montage of c g. We used persist to
quantify,
a power band analysis over 2 minute windows.

(48:28):
Fast versus slow ratio, which been described before
as an alpha delta ratio in some of
the hemorrhages, for example. We added beta into
the mix. So alpha plus beta as the
fast frequency versus the,
delta as a slow frequency.
And we could betas because we were measuring
our BF at the frontal lobes, and beta
is often up correlated to front more prominent
frontal lobes.
We correlated the ABDR, this fast versus slow

(48:50):
ratio with BFI in the wake versus lepatic
patients, defined as the patients who were not
able to follow commands,
on the majority of the neuro checks.
And so the 4th finding we had was
with these these gray ones are the encephalopathic
patients. These the ones with the white backgrounds
are the ones who were awake. Alright? So
we see a better correlation between the blood

(49:12):
flow,
DCS blood flow, and the alpha beta delta
ratio
on the patients who are awake, meaning their
neurovascular units were essentially more likely to be
intact. We see a much more messy scatterplot
with, let's say, less
significant correlation in the patients who were comatose.
Sedation was used equally in both groups. Sedation

(49:33):
can be confounded, but sedation used equally in
both groups. And to be honest with you,
it wasn't much at all. No one used
Versed.
Some of them used Presidex,
and a very small amount of propofol, but
equal no significant difference in both groups.
4 out of 5 of these psychopathic patients
had Hiby on their CT scans.
One of them did not have a has
did not have a CT scan at all,
and it was comatose without any sedation given.

(49:57):
So when we put them on a,
a PARP plot, essentially,
we can see that the ABDR BFI ratio,
the core correlation rather, was significantly higher for
the neuro in for the group one was
a normal significantly higher from the normals compared
to group 2, which is the injured group.
We correlated against our quote unquote previously described

(50:17):
gold standard ADRs as well. We saw almost
a similar thing. We did not reach difficult
significance for the left hemisphere, But in general,
we saw the same kind of, pattern emerge.
But our our,
metric made significant correlation.
So
we used that to evaluate or we we
we took that to mean that we could
we could determine whether patients

(50:39):
who were in telepathic had decreased neurovascular unit
function.
So this is a pilot study. It was
feasibility. Right? Small, heterogeneous, awake, not awake, BBVA's,
various definition of groups. You know, we kind
of analyze it as time went. Initial probe
design would break if you looked at it
wrong way,
and it was frail, obviously, to get delays
in enrollment.
So what do we do next? Right? We

(51:00):
we tried and tried since 2020
to get an r, and we finally
succeeded
a number of years later
on the 4th attempt.
In between all this, I was writing a
parallel k. The k scored, but then the
k,
did not score any better. So I and
the reviews for the r was just make
me a co, a co investigator,

(51:21):
co PI MPI rather for the r. So
I've abandoned the k and went to the
r, and we became,
co PIs for the r and succeeded. We
want to homogenize population, aim for only VA
ECMOs, peripheral cannulation only, carotid shock, plus cardiac
arrest,
only.
The majority will be will have some sort

(51:41):
of cardiac arrest in the either during cannulation
or previously or just prior to it. Only
comatose patients. No more awake ones. GCS less
than 8 within 6 hours.
Homogenize the monitoring. So CT everybody can get
a CT head, part of standard care. No
more will some patients will not get it
at all. Everyone's gonna have it. Everyone's gonna
get the same continuous EEG, DCS, and and

(52:02):
we're gonna add NEARS to the mix. We're
actually buying a NEARS machine as well. We're
gonna add that DCS
so that we can when you get CDF
and tissue oxygenation, you can get oxygen extraction
freight or fraction, and you can calculate
metabolic rate of oxygen usage.
Add,
volatile potentials. We're gonna try coma. We're gonna
try basically mismatch negativity,
volatile potentials,

(52:23):
loop grounding cables.
This is our new probe design. So not
only have DCS, we have a nearest source
as well. It's separate light source and detectors
as well.
The first aim is conduct combine the parameters
we just talked about to predict a coma
outcome,
put into a machine learning prediction model,
which we have a collaborating lab with.

(52:43):
And the goal is to pay 68 patients
over 4 years, 48 to our training set,
20 for validation.
Ultimately, try to see who will actually wake
up so we can evaluate their ability for
bridge therapies.
2nd aim is to compare the asymmetry RBF
and the cardiac arrest patients that are noniclone.
This is my excuse to pay put this
entire thing on non echo cardiac arrest patients,
which is really my passion now.

(53:04):
The opportunity to to to do that, some
non echo cardiac arrests. Only VA echo,
patients,
this is, like, as a comparison,
only VA echo patients without, with out of
hospital VTVF cardiac arrest and then and then
comparably,
only comatose VTVF
out of hospital cardiac arrest without ECMO.
We predict that the higher ASMRB it is

(53:24):
our our grant prediction was that we wanna
use this as an aim to kind of
suss out whether this ASM RBF thing is
real or not, whether it's specific to ECMO
patients. This is specific sign just for ECMO
patients.
Theoretically, you shouldn't have that sign if you're
a patient with Hiby, but not on ECMO.
And so maybe this if that's true, maybe
there's future trials I can optimize ASMRBF by

(53:45):
changing
ECMO speeds.
Then there are 3rd most exciting I think
is to develop a slightly different version of,
CBF monitoring
that's way cheaper. This whole setup that I
showed you before,
it's about $60,
probably 70 if you add in nearest.
SCOS would make it $8,000 on average.

(54:06):
And if we can do this, it's we
can do this spatially as well, put into
a 10, 20, kind of electrode,
placement as well. And,
you know, actually, it can be as cheap
as $1,000. But in all in all honesty,
it's probably be, like, $8,000,000,
and this would pave the way to making
it a multicenter trial. So,
we would have to have this successful before

(54:27):
we can go for, like, a multicenter,
a collaborative agreement.
So and we've already started building this, and
this is kind of like our, exploratory kind
of analysis on this.
Alright.
So, you know, I couldn't do this without
the entire army of people behind me.
Regine Che is my medical engineer collaborator. Yifan

(54:48):
just,
defended on all this data and and and
successfully got his PhD. Olga is my epileptologist.
Mark is my cardiac intensivist.
We've got med students, undergraduates,
you know, p h PhD students all over
the place.
And then we've got collaborating labs that, you
know, David Bush actually is a person who
first induced the regime in 1st place. I
met him at a LSO conference. He was

(55:09):
presenting
DCS data on on neonatal,
children
going like well.
And then, you know, we couldn't get anywhere
without our cardiac surgeons and support from our
cardiac from our cardiac ICU.
And then, of course, I couldn't get here
without Benjen, without Niraj, Dan Hur pushing me
to put this on people's heads, you know,
without Benjen opening my my my mind and

(55:30):
all the dreams that, he, you know, kind
of inspired me to have. And I couldn't
get this couldn't get here without Shocktrauma. I
really couldn't get here without you guys. You
know? I couldn't get here without,
you know, everything and every single,
grueling nights that, you guys put me through.
It was the best. So
with that,
that's it.