 I'm Dr. Timothy Patrick from Biomedical and Health Informatics. Dr. Campbell is winging her way down to Buenos Aires today, so she asked me to introduce the speaker for the seminar. And our speaker is Dr. Olga Emus from Milwaukee School of Engineering. I think we're going to have a very interesting talk. Dr. Emus said that she would welcome questions during the talk, and at the end we'll have maybe 10 minutes, a few more minutes, for questions at the end of the talk, okay? So, without further ado. Thank you for having me here. As Dr. Patrick said, I'm on faculty at the Milwaukee School of Engineering. I also have my own consulting company called iBiotech. And so actually the work that I'm going to talk about today is more on behalf of what I've done as part of my consulting work along with some of my students at MSOE. So I know this is a joint UWM at Marquette and MCW seminar. I'm actually a graduate of Marquette MCW joint functional imaging program back in a day. So this is a good opportunity to kind of come back to my roots and speak with the new graduate students. So what I'm going to talk about today, the title of my talk is Assessing the Utility of EEG or Electroencephalographic Coherence in Evaluation of Treatment of Red Syndrome with a very specific drug that is currently being evaluated for FDA approval in the United States. It's really data analysis project for me here in the context of a large multi-center clinical trial that is actually continuing at the same time. So if you look at the logos here, these are all the contributing organizations that are involved in this trial or in the subset of the trial. So the three logos that I have here, one is my own. Neuron Pharmaceuticals is the company. It's a pharmaceutical company that is the sponsor of this clinical trial, this new drug that they've developed and now they're clinically assessing and hoping to get the FDA approval for sales in the United States. They're a New Zealand registered company, headquartered in Australia. They do have offices in the United States. They do have offices in Europe as well. They're relatively small compared to the large pharma companies out there that they're really, really small guys. They have this one drug that they're currently trying to get past the FDA approval for three conditions. One of them being the traumatic brain injury, non-acute traumatic brain injury, something that I've done working as well. Fragile X syndrome, which is believed to be one of the underlying causes of autism, and RET syndrome, which is a very, very complex syndrome that I'm going to talk about today. So Neuron Pharmaceuticals actually contracted Jordan Neuroscience Company and Jordan Neuroscience Company contracted me. And JNS and I have a really long-standing history. We've worked together for a number of years on some of the technology that JNS develops and they developed these small portable EEG recording and analysis systems that can be used wirelessly actually in clinical settings as well as in military settings. So they got quite a bit of funding from DOD and they have done work with their systems so that they can be actually used out in the field with the soldiers. So Neuron Pharmaceuticals came to JNS, JNS came to me. And in fact, in this EEG analysis, JNS and I worked together and helped Neuron with the EEG analysis of their data. So these are really all the other universities and organizations that are involved. I had MSOE students helping me with the data analysis. They've done the statistical analysis of the EEG data. Baylor College of Medicine, the Texas Medical Center. They're really the world-renowned center for RET syndrome. So they have this Baylor RET syndrome center and it's one of the top places for this condition worldwide. So they're the PI on this particular study in terms of the clinical assessment. And of course, we had some other clinical sites as well. Vital Systems, Inc., they're the clinical research company. If you know anything about clinical trials, you usually, you know, you have a sponsor that would be the pharmaceutical company and then you have all these clinical sites and a bunch of other consultants helping out. Well, Vital Systems, they actually manage the clinical trials. They make sure that everything is per FDA regulations. They're compliant with everything. They can also provide additional services. So they've done statistical analysis for a subset of data. We did the other subset. So it's a phase two randomized, double-blinded. I know it's a lot of words. I'm going to kind of explain some of that. Placebo-controlled dose escalation clinical trial of NNZ2566. That's the name that Neurin gave this experimental drug. Obviously, it's not the name that if they get the approval that would be marketed. Traffinotide is the name that the World Health Organization assigned to us. It could be later on a generic name if they get their approval in red syndrome. So phase two, I'm going to explain that in just a second. Double-blinded is really important. So this is, it's one of the requirements of the FDA. If you're running clinical trials, if you know anything about drug approval process, it's got to be double-blinded. In other words, the subject or the evaluator neither knows who receives the placebo or who receives the drug. So it's completely closed off. The patient doesn't know. In this case, it's children and adults involved in the study. Nobody, their parents or caregivers, nobody knows, including the physicians. So it's a double-blinded study. In fact, we completed the entire analysis without knowing whether we're looking at placebo or treatment data, whether we're looking at low dose or high dose, until the very last moment when the study was fully closed, they let us actually, they opened up the data set so that we could assign statistical results to what we've concluded. So it's a phase two clinical trial, and these are the primary outcomes, primary and secondary outcomes. The primary outcome is safety. So in the phase two clinical trial, we're looking at whether or not the drug is safe. Can it be tolerated at what dose? How well is it going to produce any sort of adverse effects? Things like that. The secondary outcome is efficacy. In other words, does it work? What does it do? Does it help to treat this condition? Does it help to alleviate the symptoms? But it's really secondary. Usually pharma companies try to do more within this secondary outcome because the clinical trials are incredibly expensive. So if they can show efficacy in phase two, it's a huge plus. But again, it's a secondary measure. The primary concern is to demonstrate safety. So as I said, Neuron is a sponsor, and part of their funding actually comes from the International Red Syndrome Foundation. Please stop me if you have any questions. So if you're not familiar with the drug development process, at least in the United States, although it's really quite similar in Europe as well and Asia, in the United States anyway, the FDA really requires that you have three phases, three official phases that will take you to the FDA approval of the drug. And these take years and millions and millions and millions of dollars. They're tremendously important. Everybody says that the drugs are so expensive and they are, right? We all know that. But the process costs an enormous amount of money. So phase one, in this study we're here, we're in phase two. But what happens in phase one? Well, when you get to phase one, you've already evaluated your drug, your medication on animals. You've done all your bench testing. You know it should be well tolerated in humans. You know what the effects will be. You've done your homework. And so now you actually recruit a relatively small population of healthy subjects. So they do not have the condition which you're planning to treat. You have a healthy population, about 50, sometimes less patients. And you simply assess the safety aspect. Do they tolerate the drug? What kind of side effects would they develop? You also look at dosage. You already have an idea obviously from your previous studies. What is the safe dose? Well, here you're looking at how well the patients will tolerate it. Side effects, right? So this is, that's the big thing, to demonstrate that it is safe. After that, you move into the phase two. And so this is where we are with the Neuron clinical trials. In phase two, you can see you've got a much larger population. So now you're looking at 500 or so patients, something around there. And now these are the patients with the condition. So you're done assessing healthy population. You're now looking only at the patients affected with this particular disease. Your primary outcome is still safety. So that you're still trying because, believe it or not, for the FDA, safety comes first and efficacy second. And you can see logic in that, right? First, you've got to show that it's safe. And then you have to show that it actually works. And those two things have to come together. Because if they're, you know, working for certain conditions, but creating some serious damage and effects on something else, that is not acceptable. So safety first, then efficacy. So here, again, primary outcome is safety. But now we're also looking at whether or not it's effective. So here we're also giving the patients placebo. So you're comparing the effects of this drug relative to placebo. So some safety, safety is a, I'm sorry, safety is the primary, efficacy is the secondary outcome. Past phase three, you're moving on to a really large study. Now you're looking at thousands of patients. They all have the condition. It's multicenter. It can be international. And in fact, phase two is often international. But it's not, it's really not necessary. If you're looking for approval in the United States, you probably, you try to recruit the U.S. sites. If it's closer, the FDA, you have a chance for the FDA to visit and it's a little bit easier. But plenty of bigger pharmaceutical companies will have sites worldwide. In phase three, you do try to recruit sites around the world, or at least in markets where you want to actually place your medication. So now you've got thousands of patients and you're looking at long-term effects. This is, you know, it's still not licensed. You're still considered an experimental treatment. Everything is under informed consent and it's highly, highly regulated. So hopefully with all these three phases, you've shown that the drug is A, safe, B, effective. And now the FDA will give you the clearance and you're ready to license your drug. After that, it goes into open circulation. It's used. And we enter what's called a phase four. It's really not an official form for approval anymore, phase for approval anymore because you have the approval. It's really the surveillance. So now it's the drug is in use and FDA will be monitoring how effective it is and how safe it is. So that's kind of, and it takes, you know, many, many years. This is, you know, 10, 15 years later down the line. This process can take a really, really long time. Can I ask a quick question? Just about this. Would the use of simulated or in silico trials help speed this up or reduce some of the cost? Well, you know, it's going to be supplemental data in most cases. You really need to have human subjects. There's no other way. And they're lucky if they're not going to stop the trial in the middle for some sort of unforeseen adverse effects which happens very often. So, yeah, no, it's really these three phases would involve human subjects as the main, as the main goal. And efficacy only on phase three or phase one? You're also looking in phase two for efficacy. But it's not the key outcome of the study. The primary outcome in phase two is still safety. If you can show efficacy, it's great. Now, if you showed only safety but no efficacy in phase two, I have a feeling FDA would not like that. So then they would say, look, no, you've got to do more studies. We're not going to let you go into phase three. This is not acceptable. So while theoretically you're still primarily concerned with safety, at this point you need to be showing that it's working. In phase one, you don't have to show any efficacy whatsoever. Although companies try. If you can show efficacy through the whole three phases, it's a big plus essentially for you. But here you don't typically have actually disease population. So how would you? We have one more question. So this kind of goes along with the last, well, prior question about the simulation and modeling. And do you also do, is that more of like, I guess, a preliminary step before these clinical trials instead of with these? And also, do you guys do animal testing and whatnot? Yes. Prior to this? That's a good question. Yes. The simulations in animal testing come in preclinical work. OK. So you've got all your lab work. You've got your actual preclinical trials, which you also have to complete before you can enter into this one. And so the precursor to all that we've done here were a whole bunch of animal studies where they did show that it's effective and in this case they had knockout mice with red syndrome mutation. And so they've shown that it does a certain set of things that I'm going to talk about. And only then, and yeah, they do simulation studies and all that stuff. But here, you need to treat the patients and see what happens. OK. Any other questions? Yeah, keep them coming. It's a lot easier to present when there's a discussion going on. So in this presentation, I will focus on what we did. So myself and J&S. Now, not because I'm so self-centered and I just want to talk about my work, but because most of it, other than this, is really not my work. I will talk about the results of what was found. But I can competently talk about what I've done. That being said, it's a big challenge. Most of you guys, I'm assuming, are graduate students, post-docs, and so on. And you're working on projects that are federally funded. You get NIH grants and SF grants, whatever it may be, American Heart Association. And so you're used to presenting your work at a conference. You come in, you give a presentation. You're happy to talk about your results. And that's the way it is with federally funded research. You are obligated to give back to the public and tell them what you found. That is not how it works with corporate-sponsored research. They tell me what I can or cannot say. I have two confidentiality agreements. One with J&S, one with Neuron Pharmaceuticals. And so I am able to talk about EEG coherence, but I'm not able to talk about the final results that we've come up with, unfortunately. Not because they want to hide it, but because we're still working through it. And so it's going to be challenging. I'll try to present as much as I can. And I hope you can understand. But that's the nature of this type of work, unfortunately. So that being said, I'm going to focus on the EEG analysis using something, a technique called magnitude-squared coherence. It's an FFT or Fourier transform-based approach. It's very, very common used in a whole variety of biomedical signals. And it's really, very useful with EEG. It's been shown to work for decades. Excellent technique. And so what we're trying to look at in EEG specifically, we're looking at the connectivity or the functional relationships between brain regions. And we see how it's going to get affected in patients with red syndrome. And what can this drug, the NNZ2566 do? Is it going to help if there are some changes associated with the way the brain communicates? If we see those changes in red syndrome patients, can we fix it to some extent? So Jordan Neuroscience, where do they come in? Dr. Jordan, who is the CEO of this company, also is a board-certified neurologist with a specific expertise in EEG and neurophysiology. So in neurology, you can have a number of specialties and his is very specific to neurophysiology and EEG. He, so he in this case is acting as a clinical consultant on the EEG analysis aspect of this idea. And I'm actually acting as a data analysis specialist. So we worked hand in hand when it came to the EEG data. So I've been talking about red syndrome, but what is it? In fact, I have to claim my own ignorance. Before I started on the study, I never heard of red syndrome. And then I learned about this. And it's incredibly scary and very concerning and sad. So in 1966, that was the first time when red syndrome was described. And by Dr. Rhett, who was saying hands to name red syndrome, it's a single gene neurological disorder that occurs almost exclusively in girls. So 96, 97% of the time it occurs in girls. If boys are affected, they're affected much more severely, although it is really hard to imagine how much more severely one could be affected with something like this. So what is happening there is in the first six to 18 months of age, these girls are developing normally without any sort of science that are a warning science, I guess I would say, to pediatricians or parents. Although in retrospect, parents would say, yeah, my baby was just on the commerce side, slapped a little too much. But nothing really that would trigger pediatricians to say they're having some sort of developmental disabilities. Now at about a year or 18 months of age, they're almost right away, becomes this period of regression that continues throughout their life. If the kids were able to speak by then, they lose their ability to communicate verbally and that progresses. To the point that later in life, some of them are not able to have a coherent speech at all. They are no longer able to use their hands in a purposeful manner, so they no longer are able to grab for a cup if they want to. They develop these very stereotypical hand movements and they're kind of rhythmic. So they're very specific to this condition. They have gait abnormalities. At some point, many of them are not able to move at all. They, in addition to that, develop a lot of neurological conditions such as seizures. So many of them, about 60, 70%, would experience ongoing seizures, recurrent seizures in some cases, 150 or so per month. Almost always, they would have abnormal EEG and by abnormal EEG it would not necessarily be seizure EEG, but it would include spiking and slowing activity, so patterns that we find abnormal in general cases. The characteristic trait of this condition are autistic-like behaviors. In fact, the symptoms when it comes to mental state are so similar between autism and red syndrome that it's actually believed that in girls, red syndrome is the primary cause of autism. At least that's based on the data that we have to date. So there are a lot of genetic similarities as well between autistic children and children with red syndrome. It is, fortunately, very rare. It occurs, well, here's the prevalence, 1 in 10,000 to 15,000 females age 2 to 18 years. So it's a pretty rare condition. So what do we know about this? Why does it happen? Well, we know it's genetic and we know it's a single gene issue. So the gene is MECP-2, 2, MECP-2, and it's an X chromosome gene. That's what we know. And there's a mutation in that gene that causes red syndrome. There are different mutation types and they can account for some differences in the condition that we see among the patients. These patients have a relatively long life expectancy. So they would live to their upper mid-40s. However, it is a very dependent lifestyle. So they are not able to take care of themselves. They do need a full-time caregiver. As I said, many of them are not able to communicate or walk or move around. In addition to all the symptoms that I've listed in the previous slide, they also develop dementia, anxiety, anorexia, depression. They have epilepsy. They have a slew of cardiovascular problems. Apnea, asthma, those are just most common manifestations of red syndrome condition that you can have. Hence, it's the syndrome. It really compiles a lot of different diseases and conditions in one. The cost on society is enormous on the family in particular. So if you just look at the direct medical and supportive care services, it becomes higher than $20,000 a month just to care for a child with this condition. This does not include physical therapy, special education and all that. That comes on top. So it's a really debilitating and financially draining condition. And what's really scary about this is there really isn't anything out there that can treat red syndrome specifically. These patients are treated, of course, but they're treated for specific problems that they're experiencing. So if they have asthma, they're going to get asthma medication. If they have arrhythmia, they're going to get cardiovascular medication. And usually they have this cocktail of a whole bunch of different things. So it really is something kind of very scary and something that requires a solution. So what do we have? Well, Neuron Pharmaceutical is working on this drug. They've developed it. They've tested it on animals. And they think it has a chance. So how does this work? Again, Trefinitide or NNZ2566, I'll use those two interchangeably. So what does it do? It essentially replicates the action of the IGF-1, the last three peptides, actually, of the IGF-1 growth factor. This IGF-1 in central nervous system is really produced primarily by neurons and glial cells. So it's in the brain and it's really highly critical for normal brain development in children. It really regulates a lot of different things. One of the important things that it regulates is the synaptic development and the maturation of dendrites, so how healthy your neurons are going to be, how the glial cells are going to develop. And later on in adult, mature brain, it actually regulates how your brain responds to injury and disease. In fact, a neuron is also testing, currently, this drug for non-acute traumatic brain injury. It's been shown to really help with the response to injury. So that's another good application. They're quite far along in their face to clinical trial. They've already looked at over 250 patients on that one and it's very promising there. So what did they show in their pre-clinical work? So in their animal work, they actually showed that this Trefinitide is going to reduce the inflammation in the brain. It's going to help with this over-activated microglia, the supporting cells in the brain. And it also helps to regulate the dysfunction of synapses. And overall, it corrects the levels of this growth factor, IGF-1. Why are these things so important? Well, because it just so happens if you take all these different neurological and neurodegenerative disorders, which are different in nature, they tend to all have these three things. So if you, let's say, you take Alzheimer's or Parkinson's, for instance, and then you take Red Syndrome, very different conditions, most likely very different mechanisms that are producing it. At the physiological brain level, they will all exhibit inflammation. They will all show over-activation of microglia and they will all show some sort of dysfunction of synaptic activity. So they've shown that with this drug, it can actually help, in some cases, completely reverse some of these problems, like in non-acute CBI. But in cases of Red Syndrome, it's actually been shown to normalize it and slow down the progression of those symptoms. So in their preclinical studies, they show that if taken orally, which was one of their goals, there are really good brain levels in animals with a pretty good half-life. So that drug is there long enough to actually introduce some positive change and also has a good safety profile in adult healthy volunteers. I am not a biochemist. I'm a biomedical engineer. So I don't know as much as I probably should know about the physiological and cellular mechanisms of this, but if anybody is interested, they got a list of publications. Those are just some. You can visit their website. They got a longer list there. But I'm going to try and kind of move along and tell you a little bit about the patients, the inclusion criteria. So it is a multicenter clinical trial. Unfortunately, I can't tell you how many centers are there, but several. Let's leave it at that. The patients that we evaluated in the subset of study is 56. That doesn't mean that that's done. There are more studies going on. They just completed another study with 60-plus, more patients in a different age group. So they keep going. So this study included 56 patients and the age was 15 to 44. The study that they just completed looked at patients between four and 15. So they're trying to cover the entire population of these patients in their clinical trials. To include the patients in the study, they had to meet a diagnostic criteria for a typical RET syndrome. And there are questionnaires that I'll talk about that are filled out by parents and doctors, of course, to determine whether or not they meet the diagnostic criteria. Part of that criteria is to show that the patients actually do, in fact, have this MECP-2 mutation. Clinical global impressions scale, CGIS score. So this is one of the questionnaires that clinicians fill out that assesses in general how sick these patients are, what are the symptoms, how they behave in the environment, and so on. And it's scored between zero and seven. So anybody who gets a score of greater than or equal to four is considered to be moderately ill. So it's used also in general with all mental disorders as well. So those are the inclusion criteria in the study. This is how the study was actually set up. So there are the dosing cohorts, are the groups of patients based on what they received in terms of dose and the length of treatment. So they had three cohorts in this study. Cohort zero was two-to-one randomized. In other words, they had, it's a two-to-one ratio treatment to placebo. So they typically would have more treated patients that they would, patients receiving placebo randomized. So again, you don't know, it's blinded and randomized. Nobody knows what you're getting. This was a low-dose study. So 35 milligrams per kilo, BID actually stands for means twice daily. So that's the dose they received, or they received a placebo. They call it a strawberry solution. Here, and the treatment was 14 days. So the way it worked is they would titrate the dose to the maximum, which would be 35 here, in the first three days. Then for the next so many days, they'll give them a full dose, and then they will titrate down in the last two days. So it's escalating. I said at the beginning that the dose was escalating. That means it goes up, stays at the maximum, and then goes back down. Cohort one and two. So the second cohort, same dose, but the treatment is 28 days. So longer, and cohort two is 70 milligrams per kilo. So higher dose, twice as much, for 28 days. And you can see that, of course, they're not balanced because it all depends on the recruitment of patients and qualifications and so on. The EEG was actually recorded only in cohort one and cohort two. So we did not look at cohort zero. And actually, it became quickly apparent that 14 days is probably not enough. Uh-huh. Mm-hmm. See if I can answer it. That part, I may not be able to. Do you have a specific question or? I mean, yeah, I was just... No, usually what happens, that I can't answer. Usually what happens is, so the Baylor Red Syndrome Center, they have the most patients there would be. And so you normally, a company like this would go to a specific center, and they have patients that they see on a regular basis. And usually it's up to a clinician to actually talk to the family and say, would you like to participate in this? Your child may be eligible to participate in the study. Sometimes it's word of mouth. You know, if you know some family somewhere, they may have a particular child or somebody with this problem. You can, if you know of a child to be conducted, you can actually contact the clinic on your own and ask to participate. They will conduct all the physical tests that are required and see if you're qualified. That's usually how it's done. They typically do not have a problem recruiting for these types of trials. I think the biggest problem is, would you accept receiving a placebo and not knowing that you're receiving a placebo at the end of the day? I think that's the, to me, that's the probably the most kind of, that's the hardest thing to deal with as a parent especially, that you're hoping for treatment, but... Hang on, hang on just a second. Let me get the question. So I'm taking you a little bit off topic, but in the situation where there is a treatment, which may not be perfect, so we're trying to find something better, how do they justify giving the patient nothing as a placebo or is the placebo then the standard of care drug? That's a good question. So they would, it's a study design. Okay. Actually, in a study like this, they would compare the treatments. They wouldn't necessarily do it against the placebo. So they wouldn't do it in the phase two trial necessarily, but they would compare treatments in phase three. I don't know. It's a good question. I don't know. Yeah. You know, I don't know. I, there are plenty of people who would agree to do it with, if you are claiming that this medication is going to work better than something else, they will, you know, they'll do it. If there is also enough data to support that you can be on placebo and it's not going to be, you know, the progression of the disease is not going to speed up, you know, or it's, you know, there's lots of ethical issues, right? That's, it's really, it's a difficult question. I thought I saw another hand somewhere. Maybe not. Okay. So that's, that's the protocol. So let me tell you what they actually looked at in the study. Again, the safety was their number one concern. That was actually curious. How do you look at safety? And, you know, FDA puts a lot of emphasis on safety. And when it comes to medical devices, which is normally where I am, it's actually a lot more well-defined. I thought that, you know, medical devices and the FDA regulation of medical devices came after drug process was defined. So I thought that, you know, in the drug world it should be all color coded, but it's not. It's just, you know, this. So did you have any adverse events? How many of them were severe? What did the parents report? Did the patients feel better or not? ECG, you know, in this case, because they do experience some cardiovascular problems, what happened to their heart function? Physical exams, lab values, you don't even necessarily have to be all that specific. It's a battery of common physical tests that patients with this condition receive. Now, secondary outcome, efficacy. So the neuron really wanted to show efficacy. And so they had what they call core efficacy measures. And the one, the some of them are listed here. I don't want to go into details on this, but what I actually want to point out is these are all, for us as engineers, most of us anyway, these are really non-quantitative. They're really, really subjective. And so this is why a neuron wanted to do something new. So all these things, all those are questionnaires. They're filled out either A, so one through two, these three are filled out by clinicians. And they typically assess how severe the symptoms are. So for instance, this motor behavior assessment, MBA, deals with the core symptoms of red syndrome. So can they move? How much impairment do they have with their hand movements? Things like that. You know, the clinical severity scales deals with how severe is their disease. Again, based on really kind of subjective assessment. So it doesn't really take data into consideration. Yes, they take measurements obviously. They take your ECG and so on. But it's really still, it's a questionnaire. Same with the global impression scale. This essentially combines not only your physical symptoms but also your ability to communicate with the world. Top three concerns. That's actually given by a caregiver. So a parent would identify three primary concerns that they may have with their child at home. So what are some things that their daughter is not able to do at all, it can't cope with? What would, if you wanted an improvement and you could pick only three areas, what would they be? Again, highly subjective and very different from family to family, right? This ABC checklist, this is a very common questionnaire that's used with patients with mental illness, has to do with the behavior in the home, at school, kindergarten, whatever it may be. So again, it deals with behavior. Modified apnea index, this is actually a tool used to assess the severity of apnea condition. So this is more physiological. So if you look at these, these are really common in these clinical trials for drugs. That's what they use for assessments. So physicians, caregivers, fill these out. Yes, obviously they think about the data that they put in. It's all based on clinical assessment. But it's not a very engineering approach. So Nuren wanted to do something brand new, actually at least in these types of studies. And they said, we would like to have something that we can measure from the body. Like we measure ECG, but we can't really analyze it here because it's not as relevant. But we want to measure EEG. We want to look at coherence. They also looked at, you know, they looked for seizures and spikes and slowing as well. But they wanted to look at EEG coherence and see if there is something in there, if it's correlated to what we see in red syndrome patients. And if so, how would it get affected by the drug itself? Now, this was an exploratory analysis. In other words, they didn't really anticipate to publish it. They just actually wanted to know, can we see something? There's a lot of work with EEG and specifically EEG coherence in autistic children. Tons of literature. Nothing that looked at the treatment option, just simply to characterize it. So because there are similarities between autism and red syndrome, we thought this would be a good link. Did you have a question? But could you just give a brief definition of coherence? I'm going to do it in just a couple of slides. I have a definition of what it is. So again, not published. I'll say what I can. Unfortunately, won't be able to give you a conclusive picture. So here is how they recorded the EEG. And they did it, as I said, in cohort one and cohort two. So they're looking at low dose versus high dose. And they're looking at 28 days, at least. So this is hospital sites. So they use traditional international 1020 EEG electrode placement systems. They're standard EEG machines. If you look at it, this is the placement of electrodes that you get on the sculpt. That's the 1020 system. Now, each site had some specific modifications to this topology. In other words, for whatever reasons that they felt was necessary, they would turn off some of these channels. And we had no way of controlling this. This was kind of a major downside in the study. Each site did their recordings the way they normally would with their patients. And they have a certain amount of flexibility with that. So there was very little standardization as to how the data were collected. They used video EEG monitoring, which is a common technique. It's used in sleep studies, especially. But here, the patients were awake. In other words, they collect the EEG, but they also have, they record the patients to see what happens. In sleep studies, they need to kind of see how you sleep. Not only monitor your EEG activity, they actually need to see what's going on when you're sleeping. So here, the reason they use EEG video monitoring is so that they could actually pick out sections of the EEG when these patients are in fidgeting. Otherwise, they're always moving randomly. They can't sit still at all. They're constantly fidgeting. They're constantly blinking, chewing and swallowing. So there's a lot of motion going on. So we went into it already knowing that motion artifacts will be terrible. We just had no idea how terrible they were going to be. That was one of the findings. So they recorded for four hours. And with the video EEG monitoring and looking at the EEG samples, they would identify for us 30 to 60 minutes in each study. And that was done on site. So every neurologist or EEG specialist or most of the time EEG technologist, so not even a clinician, would look at the EEG records and send this 30 to 60 minutes of what they thought was OK. And if it's an EEG technologist, typically they're very good at running the study. And they certainly know their EEG patterns, but they're not necessarily going to be able to identify complex artifacts and say, oh, yeah, that's an artifact. That's not a healthy EEG. So that's part of the challenge. So Dr. Jordan, once we've received the data, actually went through every single file. And we set on long Skype calls doing that, where he can do EEG in his sleep. He looked at every single data segment, and he identified that that's motion, that's eye movement, that's something else. So we've actually reduced data from this 30 to 60 minutes to, in some cases, 10 minutes of what was usable. Uh-huh. Yes. Oh, the artifact rejection? No, I'm going to show you. I'm just going to show you why it's not something that can easily be done in the automated fashion. And that's one of my personal findings in this study. We're looking at all bands. So in clinical EEG, you generally try to look at, you know, delta through beta, at least. And now there's much more interest in gamma, but we couldn't actually look. We wanted to look at gamma, which is my postdoc and PhD work is all in gamma EEG. Couldn't do it because of the filter settings that the sites used and because of the motion artifact that came in exactly at 50 hertz and, you know, and this. So it was, I'm going to show you some. So what is EEG coherence? We use what's known as a Welch periodogram magnitude squared coherence. It's an approach that's been around since the 60s. It's an FFT or 48-transformed-based measure, and it looks at, essentially, in engineering terms, it looks at linear coupling. As a function of frequency. So you can think of it as correlation, but in the frequency domain, but with one significant difference. You are looking at components being in phase over time at the same frequency. Where you're looking at correlation, they don't have to be in phase, right? They can be out of phase. Here, actually, no. In fact, they have to be identical, right? If they're out of phase, they're not going to get much. With the coherence, you're looking at the phase relationship between two frequencies staying the same over time. So what I have here is what we call an unbiased MIC estimate, where here you got the cross-power spectrum. So that's the cross-spectrum computed between the two signals, and this is the product of the individual auto-spectra. You got an average here and an average in these two. So average is very important. This is where that phase continuity comes in. This is where we're going to pick up the phase that stays the same over time. That's going to give us a high value of coherence, one. That's the best case scenario. And if they don't stay in phase, we're going to come close to zero. So coherence is great. It gives you a value of zero to one for all your frequencies in the FFT spectrum, right? Zero to your sampling frequency divided by two. So that's what in a nutshell coherence does. Now in this study, right, we're looking at linear coupling between pairs of brain region, or at least that's what we're attempting to do. Very limited, nevertheless very useful, right? Linear coupling, we know brain is not a linear system. It would be kind of just silly, right? It's not a linear system. But we can at least look at that. There's a lot of utility in EEG coherence. In fact, spectral analysis has been used for decades with EEG signals. And it's been used in autistic subjects. And that was a big motivation for trying to do coherence here. Unfortunately, in autistic subjects, when I looked at the literature, there's really very little consistency in what they found. So when I did my consciousness work with gamma, you could find a gazillion publications where people were really zooming in on something specific. They said, yep, gamma power goes up when you're thinking. It goes down when you're not, yada, yada. Here, it's all, it's wide open. So there's not much consistency. So we thought, hey, why don't we look at it? Maybe we'll pick something up. So that's our coherence. What did we do? I'm going to kind of spare you the details. If spectral analysis is not something that you do, some of it may be just information you don't necessarily want to hear. But I'll make it, it's here for those of you guys who do spectral analysis. But we looked at this coherence between left and right sections of the brain, so between the two hemispheres. And we looked at frontal, we looked at temporal, parietal, occipital locations. And we also looked at, that would be this. And we also looked at within the hemisphere regions. So what happens to the connectivity within the hemispheres? We thought, hey, we're going to look at all these frequencies because the literature and autistic children said that we had something that we saw here, others saw it here, yet others saw it there. Everybody saw something somewhere. So we thought, hey, we're going to keep it open. Plus neurologists always look at all these frequencies. Clinically they look at delta through beta. And gamma is now becoming more interesting but more in the research side. So we thought we'll keep that open too. Again, that was our ideal plan until we saw the data. And until I looked at the data acquisition specifications. I got data from one site. I looked at their sampling frequency was 200 hertz. Low pass filtered at 75. I said, okay, that I can do. I can still look at some gamma. Great. I got data from another site. They sampled at 500 hertz, but they filtered at 35. So that ruled out gamma altogether. So there were all these differences just from the data acquisition standpoint. But at the end of the day, once I normalized all that and took all this into consideration, we looked at a frequency resolution of about 0.04 hertz which is really, really high and unnecessary in this analysis because EEG data with this frequency resolution would look pretty noisy. So I used some spectral smoothing. So some moving average in a frequency domain with about 20 data points. So moving average window of about 20 samples. So that of course brings my frequency resolution to 0.04 times 20 which makes it a little more reasonable. I used the Hammond window for spectral leakage suppression. There was plenty of leakage. So there was a lot to suppress. This is just a table that would give you some information as to which pairs of regions that we'll look at. So when we were looking within the hemisphere, we wanted to look at frontal-frontal connection. We wanted to look at frontal temporal occipital, frontal occipital, and then central occipital. So more kind of like towards parietal. How did we pick, and these are the channels in the EEG, how did we pick these connections and why? Well, plenty of literature supports that you got to communicate within those regions to be conscious, to pay attention, to work in anesthesia monitoring. That's kind of where I come from originally. It's kind of all supports this stuff. And also we know that there are physical anatomical tracks in the white matter that connect it. So we looked at that. I'm going to try and kind of go through this a little quicker. I know we're running out of time. Left and right inter-hemispheric pairs. So now we looked at between hemispheres as well in the front temporal regions and occipital as well. So what did we see? General observations. I'm going to kind of summarize it. Artifacts. That's my big takeaway message. In these patients, artifacts are nothing like I've ever seen before. And I've seen a lot of different EEG, including traumatic brain injury patients with a lot of seizure activity and so on. But this is something else because they're awake and they're constantly moving their eyes. They have this repetitive chewing and swallowing. They've got ongoing muscle artifact and something else that I'm going to show you which took me a little bit by surprise. What's challenging with this artifact is that we found that they really were anywhere in these frequency ranges. So whatever I wanted to look at, I couldn't because this artifact was just eating it all up. So gamma frequency was gone. We couldn't look at it anymore. A, because some sites didn't even give us that data. And B, because this motion artifact jumped right in the middle of my gamma frequencies. So we spent most of our time, in fact, on this analysis looking at every single data file to identify segments of EEG that were free of artifacts that we could use. The reason we couldn't easily use artifact rejection algorithms is because it's got all these artifacts put together. It's such a complex pattern that you can't easily automatically detect it. And I suppose a deep learning algorithm would be something that one could recommend but not in a clinical trial study. Here's an example of what a clean EEG would look like for just a few seconds. So these are your channels. It's in here. But there you got your motion artifacts on either side. This is just a few seconds of data. This is what it would look like if I've actually compressed it. So here you got this motion artifact throughout quite a few seconds in the data across all channels. And this is the same set, but expanded a little bit. Now, if you look at it, so here, yeah, it's kind of high amplitude. But in here, if you have some sort of artifact rejection algorithm, it would miss it. We've actually written some software to do pattern detection in EEG. It would miss it. It would identify it as something else in many of the cases. So muscle artifact was huge, and that's because they fidget all the time. When they're awake and they can't sit still for four hours with the EEG. Chewing and swallowing artifacts, that's what they look like. This is unfiltered EEG. This is filtered up to 30 hertz. Still comes through. So they're in the clinical EEG range. That's another thing that you can't avoid. Eye movement, that's pretty typical anyway for all of us, right? Frontal EEG would be affected with eye motion because it's just close to your eyes. But in this kids, it's continuous eye movement. It's not just our regular blinking. It's sometimes very rhythmic blinking that you observe. And then I saw something like this. So this is my coherence plot. Here I got my frequency and my coherence on the y-axis from zero to one. And I got this four traces. So different regions or different pairs of regions. I plotted this for every single subject. This is before smoothing. You can see how noisy it actually is at this point, zero, four frequency resolution. I saw this peak. Beautiful peak at 30 hertz. Just perfect coherence that goes all the way to one. And this is interesting. This is all in the right, well, this is all in the right, this is all in the right hemisphere. So right and left actually and across hemispheres. I have in here. What's interesting is what happens in the right actually more so than the left. So the next call we had with the clinician from this side, I said, you know, I'm seeing this. I don't know what it is. And he was actually kind of taking it back too. He's like, well, I don't know. And then he did some searching through the patient data and found that, well, this patient had recurrent seizures, very, very severe, very many of them. And so they implanted the vagus nerve stimulation like a pacemaker for the brain. It goes on the left side. You can see it here. It gets implanted under the skin and the leads go to the vagus nerve. So typically you would expect to see something like this in the left brain, but it went all over. So you saw this beautiful perfect peak. Hey, coherence works. That's definitely one observation you can take from there. But it's exactly, again, in my gamma frequency. So that, and that is quite common. So many of them would have something like this implanted as well. I promise I'm almost done. This is actually something I can't talk to you guys about. This is just to show you what this coherence data looked like for different pairs of regions. And as I say, I can't even tell you whether this subject received treatment or placebo. That's not neurons request. This is actually FDA. If you're showing single subject data, you cannot tell what type of treatment or plan they were on. But this is just to show you that, yeah, we saw patterns and we were able to actually group all the data. And we did get results. But because of all these artifacts and how much work went into it, you really have to think about how to interpret it and how many more studies you need to do. And I must say that neuron has been very responsible with the data that we've provided. And part of the reason why they're holding back on publishing the data is because they want to do more studies and see how can we do this better. So that's example for within the hemisphere coherence. But we found, yep, it's a good tool. We want to use it. But it's really highly dependent on the quality of the EEG data. And you need some kind of robust artifact rejection methodology. Is it feasible in the context of a clinical trial? Sure, but it would require a lot of hours to do it. If you're interested, I don't want to go through all of it. This is not my stuff. This is what they found with their efficacy measures. So first, they found that the drug is safe and well tolerated. No adverse effects, actually, adverse events. Everything was within really normal toleration level. So that was good. And again, without going into details on what's shown here, these are the scores and these different assessment categories and what they found that for the higher dose, when you look at all these core symptomatic tests as well as behavioral tests, the 70 milligram per kilo dose did way better than placebo. All of these are significant. I don't know why they're not marking their significance, but it has gone through a lot of significance testing. So all of these are highly significant. And so they believe that it actually is a really good clinical trial that it does show some initial evidence of effectiveness. So they're continuing on their work. Just want to acknowledge my students who helped me with this a lot with coherence analysis and statistics. Any last minute questions? I don't want to hold anybody. So first one is, were you able to compare the data that you said, the four hours data that you collected? Was it during the administration of the drug? So it's during the treatment and after. What about before? Were you able to establish the baseline? So there's baseline. So yes, everything is with respect to baseline. So they come in with no drug, then they're given the drug and then you monitor over 28 days. You take at least two EEG readings in between. So day 14, day 28, then they're done with the treatment and then we do another EEG at day 40. So were you able to do the coherence with the before and were you able to see the exact same thing because you were still... That's what I can tell you. All I can tell you is that there's a lot of variability. I saw tons and tons of variability. But at the end of the day, the standard errors didn't look actually quite as bad as I thought they would be. There are patterns. I don't know if they're very conclusive. That's my takeaway. That's my, I'm sorry. I wish I could show you the plots, but I can't. But yeah, we were able to do it and also baseline, day 14 of the treatment, day 28 of the treatment and day 40 following the treatment. And we did see that coherence just like their, you know, their questionnaires and other parameters showed that a higher dose definitely shows change, whereas the lower dose, it was inconclusive, completely inconclusive. That's probably as much as I can disclose. So I think you partly answered the second one. So I'll move on to the third question is, okay, have you tried like separating the components out for the analysis with the artifacts? Are you talking about like independent component analysis types in? No, not here. We didn't. That's actually something that I proposed to them. So when I went into this, the way these contracts are set up, they want me to tell you how many hours it's going to take me to do it, right? And I proposed, I thought I was like, oh yeah, I'm going to do it. Well, I spent twice as many hours except you don't get paid at some point anymore. So yeah, I was willing to do something else, but that wasn't in the contract. So it's something that we're actually thinking maybe in phase three, we're going to do another exploratory study. So independent component analysis, but honestly, I think it's just so complex with the artifacts that it's not going to pull it out. And how about adding any imaging to it in comparison? Yeah, that's a great idea. They're not, I don't think they're doing any routine imaging studies, although you know what? I don't know. It's actually a really good question. I don't know if they have any imaging data on these patients. I'm sure they do because they show that their brain size and head size, it just remains small at certain points, stops growing. So they clearly monitor that somehow, I'm sure, with imaging. That's a really good thought. Yeah, thank you. Mm-hmm, yeah. For your phase three for additional testing, would it be worth the effort to try to get all the institutions to standardize how we're recording EEG? Yeah, that's one of the recommendations we made from this is that people who are going to analyze the data, they need to contribute to the clinical trial protocol. We came in when the protocol was already approved by the FDA. And so the sites were already recruiting patients and collecting data. So we didn't get a chance. When we finally got all the sites on the same cold, they were all very willing. But at that point, they was already too late. Otherwise, yeah, it's just a matter of switching the sampling rate to the same value and setting up the filters to work. Still, there are different systems. They're not using an identical system, but that's not a big deal. Yeah, it's some, but not, you know. Do you think like epilepsy, you will be able to see a lot of activity in the higher frequency range, higher than gamma? I don't know. You know, traditionally, epilepsy didn't even look that high because the seizures would be lower frequency. So if you ask a neurologist, they'll tell you, we don't really care for gamma all that much. Especially for seizure detection, the work that I've done in seizure and just generally EEG pattern detection, they don't want the muscle artifact. And then that comes in and gamma. Yeah, yeah, yeah, yeah. I don't know. I don't think that clinically they're looking at it. Maybe research-wise, I suppose they would. But clinically... Yeah, no, for research, absolutely. But I think clinically neurologists are still traditionally cutting at beta and that's... It's easier because EEGs are read visually. They don't necessarily... Neurologists don't have time for anything else. They just read it. And they look at it as a... In fact, when I worked with Dr. Jordan, when I studied DSP, when I learned spectral analysis, nobody cared about the amplitude values. You know, it's all relative. So when he asked me for some FFT plots, I put it up and he said, wait, Olga, I don't understand. It has like 2800 something. I said, well, who cares? It's relative. It's all relative. He said, no, no, I don't do relative. I look at my EEG signal and it shows me microvolts. I need to see correspondence. And so since then, and I teach my DSP class, I told my students, you've got to normalize it. You've got to make sure it matches your actual signal because clinicians look at those measures. It's very different for EEs and CEs. So, yeah. Yeah, it's been an interesting experience, actually doing this kind of research, doing actually consulting work with clinicians. It's been a really eye-opening experience for the last few years for me, for sure. Be happy to take the follow-up questions. Absolutely. After, but at the moment, let's thank Dr. Anna. Thank you for being a very interactive audience. It makes presentations actually very good for me.