Baptist Health Miami Neuroscience Institute invites Dr. Aviva Abosch to discuss innovation, leadership, and discovery in the Marie Curie Women in Neuroscience Lectureship.
OK. You're good to go. Good, Christine. All right, gonna um Small group in physical presence, but it'll increase um I think over the time of the lecture ship, but there's a number of people on Zoom, so it's part of our effort to bring uh leaders. From around the country to lecture, uh, faculty. Uh, students and uh medical assistants about important things in neuroscience, and today, Avivo Bosch, who's the chair of the department of neurosurgery at the University of Nebraska in Omaha, is our speaker. She's the Nancy Keegan and Donald Volta chair. And the Division became a department when she started as as per one of her requests. And Aviva was a resident when I was a junior attending, so I started in 19 not to give away dates or anything, but I started in 1992 as an assistant professor after moving from University of British Columbia back down to San Francisco. After having been in San Francisco as a fellow 88 to 90, back to Vancouver in 1992, came back and started residency. In 1992, and graduated in 1999. Uh, she went on to do fellowships in functional neurosurgery. I hope I get the sequence correct, Toronto. Yeah, and the and I, yeah, so she went to Montreal, the original Montreal Neurologic Institute where Wilder Penfield mapped out the human homunculus and published on that, a very famous theater, just like Harvey Cushing had at Brigham Women's Hospital with snorkels to suck away the evil humors from the human nose, um, and then went to Toronto and uh first stop was Atlanta, Emory University. And uh moved once or twice and finally landed in um Nebraska. So we're um I'm delighted she came. It's a nice circle to see one of your former residents, and there are a number of them from UC I've become department chairpeople around the country, which speaks to the uh quality of the residents we selected and the quality of the education and the mindset for becoming academic neurosurgeons in the future. So, Charlie Wilson, who is the chairman of neurosurgery for 30 years, Avive is a big fan of his. Um, when I started, I had an introductory interview with Charlie for about 3 minutes. Literally came in the office, sat down, and said, all right, Michael, welcome. Nice to have you back. It's my job to make you famous, and if every Neurosurgery faculty person at UCSF became a department chairman, I think I've done my job well. So, Doctor Wilson would be very proud of Aviva, Viva, please. Thank you. So first of all, I just want to set the record straight. I did not overlap with Wilder Penfield at the MNI, so, um, and secondly, um, uh, those of us who trained under Mike McDermott have just an enormous debt to repay him. He was kind of the conscience of our department, um. And we looked to him for what was right about patient care and about how to practice neurosurgery. So it's a pleasure to be here, Mike, and uh it's absolutely extraordinary to see what you've built and what's what the potential is. So I've had a lot of fun today. Thank you. Uh, I am a stereotactic and functional neurosurgeon in Omaha, Nebraska, uh, as Mike said, a chair of, um, the Department of Neurosurgery at University of Nebraska Medical Center, and, and for those of you who, who stick to the coasts, we're the part of the country that you fly over between New York and LA, um, in fact, if you bisect the country north, south, east, west, that's Omaha. We are literally the center of the country. And what I'm gonna talk about today is really in a sense, the future of stereotactic and functional of my, my area, uh, because everybody knows that we do DBS or lesioning, and I'm not gonna, we're not gonna have that, uh, debate today, but, um, we do neuromodulation essentially for the treatment of things like Parkinson's disease, essential tremor, dystonia, etc. But right now what we engage in is called open loop stimulation, if we're doing um DBS. So open loop stimulation means that I implant a stimulator into a precise target in the brain, uh a basal ganglia structure, the thalamus, uh STNGPI, and then I use electrical current stimulation to alter patterning, firing patterns coming out of that structure. Um, to treat the patient's symptoms, movement symptoms. But the stimulator itself has no idea what's happening in the patient's brain, has no idea what the symptoms are, has no idea whether the patient is awake, asleep between medication doses, etc. So when I say this is the future of, of neuromodulation, what I'm talking about today is ultimately the, the search for closed loop stimulation. So relevant closed loop stimulation, meaning that we have biomarkers in the brain that a device can sense and respond specifically to that, to those biomarkers. So. Um, I, uh, have a few disclosures. That's where my funding comes from. I have funding from, um, the National Institute of Nervous Disorders and Stroke, um, from, uh, the National Institute of Drug Addiction, um, from, uh, industry, uh, and, um, uh, through collaborative, um, University of Nebraska, uh, funding. And um here's my outline. I'm gonna talk a bit about sleep and sleep deprivation. Those are some of our residents, uh, a bit about Parkinson's disease and its treatments, and then a research protocol that we have um labored long and hard over to uh look at the treatment of sleep disorders. Uh, so, um, This is, um, a rower on Carter Lake. In fact, my daughter rowing on Carter Lake in Omaha, and there are a whole bunch of very, very, um, precisely coordinated movements, um, neurotransmitter systems, um, neuromodulatory circuits in the brain that have to come into play to be able to, she rows very well. I'm, I'm not, um, not just that I'm biased, but, but, but systems that have to come into play to allow her to do exactly what she's doing. Sleep is the same thing, only more so. And we know far less about sleep than we know about motor systems in the brain and their dysfunction. So, sleep by definition is a reversible physiological state with reduced mobility and responsiveness to sensory stimuli. That's the classic definition. And, um, when you look across the animal kingdom represented there, everything that we look at sleeps from fruit flies to mammals, OK? Everybody sleeps. And if sleep is that highly conserved across the animal kingdom, you have to assume that it's important. So, so what is it actually for? I mean, it's worth worth asking that question. Well, the evolutionary biologists will tell us that sleep is really crucial for conserving energy and then timing our behavior for conditions that are optimal for both acquiring food and for escaping predators, which allows the species to to to survive. And um it's probably worth pointing out that in um mammals and, and increasingly in these other animals uh across the animal kingdom, uh sleep can be divided into wakefulness, REM sleep, rapid eye movement sleep, where we as as mammals dream, and non-REM sleep, OK? And uh as we go further and further into the down the, uh, the chain of command of the animal kingdom, it turns out that these, these three states are present, um, pretty pervasively. So I guess one question we should ask is, do you need a brain to sleep? Show of hands, who, who thinks that you need a brain to sleep? OK, I mean, that's a reasonable thought and um Uh A group of very enterprising graduate students at Caltech, um, we're studying an organism called Cassiopeia. It's an inverted comb jelly that you see over there. It basically sits around and pulsates with its tentacles and captures prey and uh goes about its business. And what they did was to sneak into the lab late at night and poke these things. And the things would startle they that means they would go from a quiescent state to to kind of a looking dazed stay dazed and confused, and then they would start to to function the way they do um during a normal uh wake time. And they were able to show through physiological monitors that the number of pulsations during the day was different than the number of pulsations at night, that through physiological probes, they could distinguish between two states, and, and, and yes, these things sleep. Now, what you have to know about Cassiopeia is that it doesn't have a brain. It has a distributed neural network. So now we know that anything with neurons seems to sleep, but you don't actually need an organized brain as we think of it to to sleep. So again, sleep is that important if it's that highly conserved. So, How is that, how, how do we function? How do we, um, go from waking to sleeping? So, so it's, it's good to know about, um, these two processes, uh, represented in the top figure. You see in the red, um, line, that's sleep pressure process S. in the blue line. Uh, that's our circadian arousal drive. And what you see, um, during the, um, transition from day into night is that when your sleep pressure reaches a peak and your circadian arousal drive reaches a nadir, uh, uh, uh, you know, the bottom. Um, that's when we fall asleep. And so the gray area represents sleep. And then those two, those two curves reverse their, um, their status, and over the course of the night, circadian arousal rises and homeostatic sleep pressure is relieved because we've been sleeping, and then we wake up, OK? So it's so in order to maintain that that form of opposing. Uh, processes, there's an incredibly well orchestrated, uh, system of neurotransmitter, uh, release, uh, in different parts of the brain that has been well mapped out, not in humans so much as in other animal systems. So very, very finely orchestrated. And, um, that has led us to appreciate the fact that there's um an anatomy and a neurochemistry to sleep, um, that allows for these two key processes to control our waking and our sleeping. And when we disrupt those, uh, two key processes, we just and we disrupt the, uh, neurotransmitter release that's responsible for allowing us to be awake or conversely to be asleep, um, we cause big problems, uh, for the nervous system, for health, etc. So, um, Functional uh imaging fMRI studies have come into the fore and allowed us to to pursue this kind of work in uh human subjects. And it turns out again that there are a variety of different nodes within the brain and neural. Networks within the brain that work in opposition to one another, uh, to allow us to transition from uh waking to sleeping, and when we're awake and not sleep deprived, to maintain our focus on a task. When we're sleep deprived. That all goes to hell in a handbasket, if you will, um, because the, the, the, um, role of the thalamus to, to sort out the control between um uh uh various networks like the default mode network shown in uh blue and the frontal parietal attention uh network uh shown in pink, that goes away and um we start to uh have difficulty. And sleep deprivation, uh, results in bidirectional changes in brain activity and connectivity that then secondarily result in widespread changes both in cognition and in affect. Now, those changes result in decrease our ability, um, a decrease in our ability to encode uh activity in the hippocampus, and that leads in um in uh uh um in time to our inability to process emotional expression. We don't express emotions the way we should when we're sleep deprived, and we don't sort out the emotions of the people around us. So we fail to register other people's emotions. That leads to problems with interpersonal functioning. It can lead to hyperactivity and attention, pure disagreement. You know, when I think back to being sleep deprived as a resident and having to go to the emergency department, that summarizes every interaction I had with the emergency room physicians. Um, and it also leads to something that is referred to as non-optimal, uh, incentive-based decision making. And that's a kind of a lengthy term, but when you think about it, that's why this always seems like a better choice, um, for dinner in the middle of the night than this does. That's, that's what we mean there. So, chronic sleep deprivation, what is it? Um, it turns out that the definition of, uh, chronic sleep deprivation is anything less than 7 hours of sleep a night. I'm going to repeat that. Anything less than 7 hours of sleep a night is considered sleep deprivation, and if you do it repeatedly, that's chronic sleep deprivation. And so it this turn it turns out is epidemic and industrialized nations such as our own, and the CDC was so alarmed, the Center for Disease Control in Atlanta was so alarmed by the epidemic in the United States that they, um, in a, in a big way, uh uh actually investigated in 2014. And began to generate maps like this. This is a map, an interactive map of the United States. It's been updated in 2020. And um it shows you um age adjusted prevalence of sleep deprivation for the United States. And so Reno, um, uh, Nevada in general, um, the mid-Atlantic states, uh, and Northeast, um, here's Florida down here, and before you start to think that you're, uh, doing well, that's a Florida map of sleep deprivation by count. the little stars where we, where we are. Um, but, you know, OK, so only say 35% of us are sleep deprived in, in the state of Florida or in, um, Dade County. But in fact, when you think about medicine and the military, we're the worst of the worst. We're the ones who are supposed to be the most vigilant, but we're the ones who, um, are suffering from, uh, sleep deprivation. And so consequences for this have been unknown, you know, long term um uh consequence of this unknown, and, and by the way, that's from your local press. Um, so, um, what the, the CDC started to look at correlations with chronic sleep deprivation is, this is what they found. There's a group of disorders that correlate with sleep deprivation. These conditions include MIs, coronary artery disease, stroke, asthma, COPD cancer, arthritis, depression, chronic kidney disease, diabetes. So there, there are things that travel with chronic uh disorder, uh, uh, chronic sleep deprivation. And in addition to that, Sleep deprivation is incredibly common in the neurodegenerative disorders and in the neuropsychiatric disorders. So in depression, schizophrenia, Parkinson's disease, etc. OK, so that leads to the question of how we study it. Well, um, the gold standard for studying sleep deprivation is actually, um, uh, uh, called a PSG or a polysomnograph. And that's a combination of EEG and uh extraocular, um, uh, uh, monitors, movement monitors, uh, and capnograms, um, all, uh, in conjunction, uh, allow, providing the data that allows a sleep expert, um, To, uh, make a determination about whether somebody's awake or in non-REM sleep or in REM sleep, or if they're in non-REM sleep, what stage of sleep. And why this is important is because there are differences, critical differences to the various stages of sleep. And so, for instance, um, any limited capacity for the adult mammalian central nervous system to repair itself happens in two phases of sleep in REM sleep and in slow wave sleep. And so it's really important to track how somebody is doing in terms of um uh these various stages of sleep relative to age-matched uh normal uh controls. Now, what, what is not a gold standard is this, OK? There are all sorts of wearables that are commercially available that allege that they can monitor your sleep well. And I can tell you that through my research I've looked not only at these, the, the Fitbits, the Apple watches, etc. but also at $3000 research wearables, you know, wrist mounted, um, uh, bands that you wear across your forehead and compared them to the polysomnograph, and they don't do nearly as well. So just a, just a cautionary tale out there. Um, this is from the Journal of Clinical Sleep Medicine. It turns out that the sleep experts reviewing the PSG, the combination of EEG and all of those other heart rate monitors, etc. um, are about 83%, uh, uh, in agreement with what stage of sleep anyone is in. So why is a neurosurgeon interested in sleep? Well, um, now I will transition and talk a little bit about Parkinson's disease to explain why this is fascinating to me. So you've all seen patients with Parkinson's disease. It's a combination of, um, the, the classic motor triad, it is tremor, bradykinesia, um, and rigidity, and then we start to treat patients with not with um dopaminergic type medications and patients respond in, um, with these stereotype movements, these dyskinesias, because we're giving back dopamine in a way that's not normally regulated by the, by the body. Uh, by the brain, um, and so, uh, there is a Really solid level one evidence now that talks about the efficacy of deep brain stimulation for the treatment of the motor symptoms of Parkinson's disease. Obviously, we're not curing Parkinson's disease, but we've gotten pretty good at treating the motor symptoms. And, um, you know, here is a classic example of a patient. A young woman with Parkinson's disease trying to move. Here she is, um, following deep brain stimulation surgery, hates her haircut. She's wearing a base or a, a cowboy hat because of that, but she's now able to go back to her job as a large animal veterinarian, um, because she can move again. Her, she's not fighting her body to move. So everybody's seen those videos, they're powerful. Um, that's what we do. We, we use dopaminergic medications, and when those start to, to, when the on, uh, periods, you know, which is to say the, the, I'm able to move well on my medication when those periods start to, to dwindle, then we talk about, uh, deep brain stimulation surgery or, uh, lesioning for, um, for the motor fluctuations. And we feel good about that, right? Neurologists feel good about it, neurosurgeons feel good about it. The patients feel good about it. But we're leaving a lot of stuff on the table in Parkinson's disease. There's a whole bunch of non-motor symptoms of Parkinson's disease, separate from the tremor, the rigidity, and the bradykinesia that we don't touch. And those non-motor symptoms affect every organ system you can think of in the body. And so patients lose the ability to smell. GI motility goes down, and, um, patients wind up with chronic constipation, which can be so bad, it creates its own problems, etc. Uh, and interestingly, even though we think of the classic triad in that netter, um, drawing there of tremor, rigidity and bradykinesia when we think about Parkinson's disease, that's when patients get diagnosed, um, with the onset of tremor or the motor symptoms. But the constipation, the sleep disorders, the, the lack of smell, the depression, that predates the onset of the motor symptoms by at least 2 decades, a decade, 2 decades, something like that. In fact, when you look at the non-motor symptoms down here in the aggregate, they're far more disabling for patients. As I said, they precede the onset of the motor symptoms by a decade or more. Um, they're more insidious in onset and therefore harder to diagnose, and they're less apparent to clinicians, uh, and, and less effectively treated. So I decided to focus on the sleep disorders of Parkinson's disease because I was tired of the experience of having a patient in my office to talk to me about awake brain surgery and having, having them fall asleep. That's pretty profound, right? You can't talk about something more intimidating to a patient than, or to any of us than awake brain surgery, yet patients would nod off during that, um, discussion. So it turns out that um Parkinson's disease is, is notorious for sleep disorders. It's when you actually start to study your patient with PSG, I would, I would put to you that it's a 100% prevalence, um. The sleep disorders include insomnia, sleep fragmentation, where they may go to sleep fine, but they're constantly waking up during the course of the night. Excessive daytime sleepiness, which is what I was seeing in my clinic, sleep attacks, sleep apnea, restless leg syndrome, periodic leg movements of sleep, REM behavior. Disorder and hypnagogic jerks. So, so it doesn't seem like a big deal, so you can't sleep. But here's a patient, um, this is an old, uh, file uh from one of my movement disorder neurology colleagues of a patient who's fast asleep, proven by a PSG, you know, the gold standard for measuring sleep. So he's fast asleep here. Um And I should say also that when, when the spouse sees this, the comment we get is, wow, he doesn't move like that when he's awake. Doesn't sound like that with that clarity of voice when he's awake. So what'll happen at the end of the video if I let it play to the end is he, he, he launches himself out of bed. So, so what's going on there? Um, well, when I go to sleep, if I'm sleeping normally, my thalamus shuts down activity. So I actually can't move. During REM sleep. If, if your thalamus doesn't do that for you, then during RAM you actually start to act out your dreams. And that's a pretty dangerous position to be in when you're supposed to be sleeping. So what do we do? We talk to our patients about sleep hygiene, and that's the, the, you know, have the same ritual before you go to bed every night. Be consistent, get exercise on a daily basis, avoid caffeine and alcohol after 4 p.m., but if you're drinking before 4 p.m., it's probably a problem. Uh, avoid late evening meals, no napping, no scotch at midnight, Doctor McDermott. Um, establish bedtime routine, and if you're unable to get to sleep, to fall asleep within 15 minutes, get up, do something else. Make the bedroom comfortable. OK. So, That's as good as it gets. Um, there's, there's also a variety of sedative hypnotics, which are used for treating things like REM behavioral disorder and the other sleep disorders. And what they do is they mask the problem. They don't give you back natural sleep. And so you don't wake up feeling rested. So, about the whole time, the time that I was going through all this and learning, educating myself about sleep, a variety of different, um, case reports, case series started to come pop up in the literature about DBS and sleep. So what happens when you put in sublanic nucleus DP DBS in patients? They come back and they say, yeah, my tremors better, but oh my God, my sleep is better. So kind of interesting, um, level 3 evidence uh that STNDBS improves sleep, but there was no validation for it, no possible mechanism, and we didn't know how to leverage it. So, uh, together with colleagues at the University of Colorado, where I was working at the time, we decided we were gonna, we're gonna sort this out. We're naive, but uh it's good to have a goal. And we set about trying to figure out how we could measure sleep better, and we decided that we needed something in the home first to figure out who these patients were because it was too expensive to put people in the hospital for long-term PSG monitoring. So we actually invested in these wrist, uh, mounted, uh, actigraphy watches, which have heart rate and 3 heart rate monitors, 3 dimensional accelerometry, etc. and can get at when somebody is sleeping and when they're awake, even though that they fail when it comes to, um, sorting out the various stages of sleep. And, um, we were able to show that these were effective in, um, uh, uh, telling us when, uh, patients were asleep versus awake. And what you're looking at here are these patchwork quilts represent, um, green asleep, uh, when the patient climbs into bed at night and, uh, sleep, sleep has started. Red is when the patient awakens in the morning and the sleep cycle is complete. Black is sleep, and anything else other than black is awake. The patient is in some stage of arousal. And so what you see in 33, patients with Parkinson's disease from my population in Colorado is that they're not sleeping. This is terrible sleep fragmentation, and these are the people who are falling asleep in my clinic. So that was great from a kind of a superficial, you know, um uh uh 30,000 ft level of what was going on in our patient population. But how do we, how do we get a better handle on it? Well, this is worked by Andy Schwartz from the University of Pittsburgh, um, who is a brain computer interface, uh, guy. Um, uh, it's a schematic that tells you about, um, how we can monitor electrophysiology in, in somebody's brain. So there's scalp recordings that we can do when we get a kind of a 3 centimeter distribution of what's happening, but we're kind of far from the source of the activity. There's electrocorticography, which can give us 0.5 centimeter in terms of spatial resolution, um, but you have to do surgery to put it on the surface of the brain. Then there's local field. Potentials which you put an electrode into the brain and you get a sense of what the entire population of neurons surrounding the electrode is doing. Or you can do single unit activity, um, where you look at individual neurons or, or a handful of neurons. Um, but the problem with the single unit activity is that it's hard in a human to get long-term single unit activity. Um, but we've been talking there about things like the Utah array, which has been used. Power, uh, informed brain computer interfaces. So we decided to look at local field potentials, which, as I said, is the, the aggregate activity, the summed electrical fields of neurons surrounding the tip of the DBS electrode, for instance, and that can give us long term um information about what's happening uh in the brain while a patient is uh being um recorded. So, a couple of things to know about the local field potential spectrum. It's divided into various uh power bands, and you see them represented there and um alpha, beta, um, gamma, um Theta, etc. and each one of those frequency bands tells us something different about what's going on. So what we'd have to do though is process the, the signal, um, uh, and through uh means such as courier transformation, uh, we can then uh begin to look at these specific bands and correlate them with disease states and the symptoms in patients. Now I should say that I did not come up with this by myself. This is worked by Peter Brown at University College London, where he had externalized uh DBS leads in a patient who was off of his Parkinson's medication. Um, and what you see when a patient is off and Bradykinetic and fighting their body to move is this spike in beta activity. When you give the patient back their, um, uh, uh, Parkinson's medication, beta drops and you see a surge in gamma. And so right there, beta versus gamma uh activity in the local field potential spectrum can tell you what, if you knew what was going on in the brain and you could sense that, you could tell if the patient was between their medication dose and needed more stimulation, for instance, um, or was, you know, had just taken their medication, was able to move, and therefore would become diskinetic if you ramped up the stimulator. So it gives you the possibility to work towards a closed loop device. So we set up an experimental paradigm when I was actually, this was at the University of Minnesota, we had a group of patients in whom we recorded 24 hours over 3 days of local field potentials from externalized DBS leads, um, and we were looking at motor control then and uh there was a lot of work that good work, good publications that came out of that work, um, to cut to the chase, we were able to show that. The local field potential activity that we recorded um could be used to um figure out whether or not we were actually in the STN or outside of the SDN so it's useful for navigating in the operating room because beta activity increases when you enter into the SDN and then drops when you come out of the SDN. So that's useful. Um, and that actually led to that work led to a software program that now is run on a variety of different, um, microelectro reporting uh racks to help people who don't have electrophysiologists in the operating room figure out where they are. Um, the, um, it turns out as well that when you're trying to program a patient, the peak beta activity is the sweet spot for the contact that that you should be using to to program the patient, uh, for their DBS. So that's really helpful. You can begin to automate that process if you can record, um, where peak beta activity shows up in a. Or polar electrode and then um when I change a battery at 8 years, a decade after initial implantation of the DBS lead, you can still record the local fuel potential activity, um, from the tip of the, of the extension cable, which means that it's useful for the long term, um, if you want to use it, if you want to use that as a biomarker to power, um, uh, a closed loop device. So all useful information and um. Uh, around this time, I, uh, very, um, uh, engaged and intelligent, um, uh, MD PhD student came into sort of wandered into my office and wanted to have a project. And so there were 9 hour overnight recordings that we haven't looked at at all. So I gave them access to this data set, uh, and asked the question, uh, of whether or not local field potentials told us anything about sleep state. Um, this is a, um, hypnogram up here, and this is this, the stair step, uh, montage is the collected interpretation of the PSG below, uh, and the various sleep stages. You can see a patient transitioning from awake to REM sleep to stage 12, etc. and then cycling during the course of the night. And there may be a couple of awakenings, but for the most part, you're cycling between those sleep stages during normal sleep. Um, we thought maybe if we could use local field potentials to um figure out what the sleep stage was instead of having somebody in a formal sleep lab with the, um, all these electrodes attached to them, we could, from a practical standpoint, save on battery life because patients don't need stimulation if they're sleeping, right? Because they're not trying to move. So that was from a practical standpoint, but then we figured there'd be all sorts of other things we could learn about human sleep and disrupted sleep and Parkinson's disease. So, um, Andy Techrewal, the MD PhD student, uh, started to look at the local field potentials that were recorded during sleep, and what he found, you can see very readily in this figure from one of our papers, was that beta activity, um, recorded in the um uh from these implanted electrodes actually mapped very nicely uh to REM sleep. And you could superimpose the hypnogram on the totality of the local field potential spectrum that we reporting, we were reporting from these patients. So that was really curious. We also found that we had dramatically diminished RAM and and slow wave sleep in our Parkinson's patients, as other people had reported, but this is the reason that the patients wake wake up so, um, so unrested after night. After a night's sleep. The other thing we found were these conserved, uh, spectral, uh, patterns, um, within these different, uh, bands. So awake doesn't look anything like non-REM sleep, doesn't look like REM sleep. They're different patterns. Um, these are our patients are represented over here. And we saw this clustering of these various, um, uh, uh, local field potential energy bands that were representative of the stages of sleep. This led us to think, well, maybe we could use machine learning to um take this local field potential data and determine and spit out to a device what stage of sleep somebody was in. And so we used a support vector machine, a kind of a rudimentary effort at machine learning to do exactly that. And we were, it turns out we were successful in predicting what stage of sleep was, uh, what stage of sleep a subject was in. Relative to the PSG that we had from experts, so we built these uh patient specific models that accurately predicted um what stage of sleep a patient was in. The problem was that they failed to generalize from one patient to another. They were really useful within subjects. So we figured we could probably do it better by uh roping in a um computational neuroscientist. And um we, um, because I thought, I thought it was the cat's meowity use the support vector machine, which got us only so far. There's a guy named Joel Zeibelberg, who might be the last remaining computational neuroscientist not working for Google or uh Tesla, and, um, his graduate student, uh, uh, or MD PhD student, Elijah Christensen was able to take this what we had done and, um, completely upended using an artificial neural network. On the, on the data, um, Uh, sets that we had recorded and come up with with a much more uh beautiful um algorithm, um, that blew the model performance that we had come up with out of the water. And in fact, um, using, um, Elijah and Joel's, uh, artificial neural network, we were able to prove our accuracy to 91%, um, uh, and it generalized between subjects. Uh, it didn't do so well where we had, um, smaller data sets. So for instance, because REM is so diminished in these patients, because slow wave sleep is so diminished in these patients, it was less accurate, uh, for those, but we had an overall, um, predictive value that was much better. That led ultimately to this, uh, NIH um uh study funded by the National Institute of Neurologic Disorders and Stroke, um, to figure out if we could do this even better, uh, in our patient populations. So what we set out to do is to figure out whether dysfunctional subthalamic nucleus activity in Parkinson's patients actually correlates with the sleep fragmentation that we see in these patients, and if we could leverage that activity to develop what we call adaptive stimulation algorithms. So in other words, we know that sleep gets better, um, anecdotally in our patients who have SDN stimulation. Can we drive that stimuli, that, that observation? And specifically stimulate in certain phases of sleep to push them into the more restorative sleep. So that was really the goal of this work, and um we um the the holy grail of this field now, as I said initially is figuring out What biomarkers that are relevant to various symptoms and disease states we can use to inform a closed loop device and actually have patient responsive uh stimulation, whether it's for obsessive compulsive disorder, Parkinson's disease, motor symptoms, or sleep, craving an addiction, um, in, uh, and drug addiction, etc. And so we took a group of patients, um, uh, at the University of Nebraska Medical Center and at our partner site, Stanford. Um, and through these externalized recordings have now recorded uh 20 subjects, um, and done replicated the work that we did in the University of Minnesota population, um, in more careful, um, uh fashion, um. Partnered, as I said, with Medtronic Research Division, um, because they're the ones who produce a um A research enabled. Pulse generator that allows us both to stimulate um and record. So it's fine to do this in a hospital with externalized electrodes, but ultimately we have to be able to take it in the home into the home and see if it's gonna work. So what does this look like um when we're actually doing it? Well, you see a schematic here of a patient who's connected to a recording equipment with an implanted DBS lead and an externalized extension cable, which is just, we don't actually externalize the DBS lead. We run a percutaneous extension, an extension cable through the scalp, uh, connect it to the uh recording equipment, and that allows us to record local fuel potentials while the patient is um connected to a PSG and um. Correlate um the PSG with the local field potentials. So there's, I'll just play a little video to show you what this looks like when we're um doing the recording. So here's a patient um about to get, he's, he's at uh in bed. We have an externalized DBS lead coming out of his head connected to the recording equipment and um what you're seeing here is the raw local fuel potential data. That's been preprocessed and here is the hypnogram based on the PSG which we will then be correlating to the local fuel potential data. So it's a very powerful way to get at these this multi multimodal data set to begin to answer some questions about sleep in um in humans. Um, we, uh, had, we, we, um, uh, have retained the expertise of some sleep experts from around the country, including, uh, Cleet Kushida, who's at Stanford, um, and is one of the reviewers for the sleep boards, uh, uh, for the American Society of Sleep Medicine. And we came up with, with Clet's uh guidance, we came up with a with a process that actually increases the percent consensus of the sleep experts when they're um uh focused on the sleep recordings, which improves their accuracy by more than 83%. It improves it to to uh. Uh, almost into the um uh 92% range or 96% range, and um this is actually my first publication in this in sleep medicine, so um it was kind of uh interesting um to, to, to live in this field. Um. Additionally, what you're seeing here is um a couple of subjects um you're looking at or one subject you're looking at the hypnogram here that was generated from the um consensus um review of the sleep experts, and here you're looking at color coded at um the various stages of the local field potential um. Mapped to the hypnogram. And what we noticed in our patients, um, without stimulation at night, was that roughly 3 minutes before they wake up during those repeated awakenings overnight, there's a spike in, um, in, uh, beta activity, uh, in their, um, uh, uh, local field potential spectrum. So, in other words, that spike in beta which precedes waking up, is potentially a target for stimulation. So we started to stimulate um first in a random pattern and then specifically uh relative to the um what we were seeing in the beta activity, and it turns out that we can maintain people in longer sleep stages through targeted stimulation, adaptive stimulation, so stimulation that's specifically firing when it, when, when we detect a certain biomarker. So, um, beta power also decreases at the onset of um of a sleep bout, uh, it increases before each waking period after sleep onset and then it's highest during wake after sleep onset. So, um, these sorts of clues allow us to begin to uh leverage this therapeutic finding for uh the treatment of patients. We've gone back to the artificial neural network that that Joel Zeberberg and his uh graduate student came up with and refined it further, and we can now do a better job predicting sleep stage from an an implanted STNDBS electrode than the sleep experts can do on the PSG, which is fascinating. Um, just to comment about the wearables, um, as I said earlier, even the best of the wearables, the research, um, uh, focus wearables are only a fraction as good as the PSG in terms of, um, determining what stage of sleep somebody is in, uh, accurately. They, they tend to overrepresent RAM and N1 and underrepresent N2. So we're now at the final phase of this study where we take it back into the home, and we have this uh RC+S device which is is rechargeable so that we don't kill the battery by streaming uh nighttime sleep data. Um, and it allows us to both stimulate and record what's going on at the tip of the electrode it's sitting in the STN, but it also allows us to to embed into the IPG into the um battery, the smart battery, if you will, um, a uh uh adaptive stimulation algorithm that's is. Specific to the localeal potential changes we're, we're seeing. So patients are um off the home with these implanted. We have a randomized, uh, we've randomized patients to a week of no stimulation, a week of regular clinical stimulation, and a week of adaptive stimulation to see what sorts out their sleep the best. So pretty exciting, um, that we're standing on the, on the edge of, uh, beginning this, uh, in-home trial. And I hope that, um. This has given you a better understanding of why sleep is important, despite the demands of work and um our uh personal lives, um, because sleep is, is fundamental to what we do as organisms. Uh, sleep is also highly abnormal in various disease states, in particular in Parkinson's disease with as yet no effective treatments. And by studying the role of basal ganglia in sleep maintenance and its disruption in Parkinson's disease, we hope that we'll be able to come up with mechanisms of sleep in humans and sleep disruption and be able to leverage these findings for better therapeutics for our patient populations. So uh shout out to my collaborators at the various institutions who've been involved with this work. Um, it's really been a, a lot of fun to uh work with the, the combined expertise. I'm not a sleep medicine, uh, person, but Cli Kushida has been spectacular. Uh, Joel Zeibelberg, I mentioned that he's at York University in Toronto. He's a computational neuroscientist. Um, uh, Casey Halpern, uh, and Andrew Sitaroff at University of Pennsylvania, um, and John Thompson at the University of Colorado, in addition to my, my team, uh, at University of Nebraska Medical Center. It's been a lot of fun. So thank you for your attention. Uh, if there are questions, I'd be.
Related Presenters
