So today's actually the finale of the GTneuro seminar series for this, this year. You guys, you guys know that this seminar series kind of encompasses celebration of neuroscience, neuro technology. And I actually think that the speaker do better, encompasses this celebration on a global scale. Michelle Johnson and as our speaker today of her career has spanned, I think four different continents. So she got her training in her BS in Mechanical Engineering and Applied Math from the University of Pennsylvania. Her MS in Mechanical Engineering from UC Irvine, and her PhD in mechanical engineering from Stanford where she focused on mechatronics, robotics, and design. I mean, she did a brief post-doc at the advanced robotics technology and systems laboratory at it. I think it's less you can't AOT a Santa Ana in Italy through and has a fellowship from NSF and nato. In her independent career, she has, she's established a research program that focuses on the use of robotic technology in order to eight, rehabilitation and improved quality of life for individuals living with neural injury. She started, and I believe the Medical College of Wisconsin and Marquette University in Milwaukee. And since 2013 has been at the University of Pennsylvania, where she directs the rehabilitation robotic research and design laboratory. Though her lab is made on a number of really important advances and a number of different fields including upper limb rehabilitation for individuals falling stroke or brain injury, using robot media therapies. And I actually heard of Dr. Johnson's work when I was a grad student and feeling really pulled between kinda this high-tech or prosthetics or and also just simple interventions for practical implementation in medically underserved communities. And for those trainees who are out there kinda feeling the same way. Um, I think that Dr. Johnson's research has really evidence that there's a lot of synergy between, between both of those both of those things. So she develops intelligent robotic systems for neuro rehabilitation with an emphasis on affordability and implementation in low resource communities. And these technologies have improved the quality of living for patients as well as yielded a new insights about plasticity during rehabilitation from cognitive and motor impairment. She's implemented these technologies and communities across Latin America and Africa. And actually right now she's a Fulbright scholar in Botswana, but it has made a nice detour to come visit us here to give a talk that I'm very happy to introduce Michelle Johnson. Thank you so much. Pleasure to to come and visit Georgia Tech. I've been here right before the pandemic in 2009. Feed shins did like a long-time. Feels like decades instead of like a couple of years anyway. So my talk today, so just actually came back from Botswana. I was there for two months and on the last leg of my my Fulbright scholarship. And so I decided, okay, I'm going to bundle that and kind of talk a little bit about my journey a bit. And then kind of what is some of our findings that were just getting out of our Botswana work? Let me start this. Did it move public create financial disclosures? I have to see this because there's a spinout company out of a pen call recuperate robotics that's kind of spinning awesome with the affordable robots work that we've been doing. To the area of kind of thinking about therapy and assistive robots you guys are probably already familiar with. So I've been working mainly waste kinda thinking about therapy robots, which I define as robots that are specifically trying to assist recovery after neurological function in my in my case, and assistive in a sense of robots it's trying to replace or help support the, the acquisition of functional activities. So one of the things that my lab does is we work really closely with MDs inclinations and we tried to study, I know what's actually happening in the clinical area. And there's a work that we're actually currently about to publish and I JSR, where we're observe that kind of the therapist and the patient is kind of in this kind of dyadic space. And where we see that the robots are functioning as demonstrators while the patient's observing that exactly three groupings that are happening, therapists observing while the patient is doing it, then the therapist helping while the patients performing. So even in this kind of setting of stroke rehabilitation setting, we see three general uses for our robotic systems. Though. If we're thinking about as I think about robots as therapists or robots as assisting therapist, I'm thinking of them as helping robots. So that's kind of our classic therapy robots, demonstrating and observing. And these are our classic assistive robots. And we work, we actually don't have a lot of robots that do it all right, that do both the fluid transition from holding onto someone and helping them and then observing them, monitoring as the human does. To the beginning of my work, I kind of spent a lot of time in thinking about motor recovery in the real world, in how we can use robots to support the upper extremity for real-world tasks like drinking, feeding. And this is one of our first symptoms that we, we had put together called the Adler system. And then we kinda more recently I've done a lot more with what I call assistive robots, thinking about making demonstrated observer real life. This is a system that is built by my current PC, is actually a week from defending his thesis. And the idea of this system kind of being a demonstrator system. It's non-contact and being able to help remotely adults and patients. And then we also kind of work on the classic humanoid that might be more demonstrator observer again and a cooperative, collaborative player with the patient. This is my colleague, Dr. Cook and Becker. I was Co advising a student of hers, kind of thinking about humanoid robots that play. Of course, this has been now kind of a pretty, I remember when I started in the date myself, but the nineties or so kind of thinking through these issues. And now there's like a lot of data that's kinda telling us these robots are actually doing something. So we know from, uh, from, uh, the body impairment, body structure, we know it's increasing strength, increasing motor controls for reducing spasticity. All these really good things that we can actually quantify now from an activity space, we see that, yes, we're, we're when they're task-specific, we see return of reaching, return of grasping. We also see improvement in activities, daily living, but that's always remained inconsistent. This is, this is, it has been a challenge. And this actually, and then we also see more in participation with exoskeletons kinda moving us into community a place. But a lot of times they're also not relevant to low-resource settings. With the challenge of cost and of those types of issues, make participation for the ride variety of the world and very difficult, right? So this, you know, brings us the kind of thinking about, well, there's some really great things that we're doing with robotics. And then there's some things that we need to consider and how do we challenge them. So the list you guys probably know, it's repeatable, it's adaptable, can help us with semi-autonomous training. We can get consistent force application. There's a really good, we can get some great reps out. However, they tend to have challenge with transfer of, of, of, of recovery to real-world type task. High costs for the most part, for most of the communities that are kind of out there that's struggling with neurological impairments. They can be super mechanically complex, which is not a bad thing all the time. But it might also, there might be a relationship between mechanical complexity and cloth. And then they are often quite huge. And they're often very targeted in one way. And so many of the real-world settings that are out there can't kind of really like deal with these types of systems in, so they're not quite accessible then to community-based settings, to skilled nursing facilities, to places in the global spaces that I've also been working in. So when we take a look and we step back and we say where these robots actually globally see that there in primarily in high incomes faces. So a lot of the data that we've gotten on the, the quick, the goodness of these systems and their effectiveness in terms of neuroplasticity driving that enabling recovery for many people in with neurological impairments, a lot of the data is coming from high-income spaces. And so my push has been, how do we take what we've learned from these complicated systems and how do we map that into maybe lower degree of freedom systems, but also mapping it into low resource status within the United States in community settings, well as in the global space and global settings. I like this map because it was 2014. I put this together when I was starting to say no, the difference between high-income species in global and goals at my standing in your high-income phases. At low and middle income settings, when we think about the prevalence of the diseases that we care about, cardiovascular cancer, diabetes, communicable diseases. We see that the, the, the higher income spaces we see that they're going to be more likely have higher percentage of cardiovascular issues. Lower issues with infectious diseases was 2014. Covid has kind of changed as I go. So I always presenting data now pre COVID, because COVID is it thoroughly through everybody off right there. And then what we see in places, for example, like what fauna we do see this area is now growing a lot, lot more communicable diseases happening. A lot more non-communicable diseases like diabetes, hypertension, and then the infectious disease rate is going down, but it's still pretty high. So these communities are not dealing with kind of this, this dual issue, right? And, and so that becomes quite challenging. But the challenge is that when they needed these robot, just like, do you know what are the challenges though when we push that? When we push more in low-resource settings, and I will say low resource settings because it's not just about developing countries. It's about the urban site, the skilled nursing facility that's down the street from Penn, right? So it's about those sites as well, is oftentimes the, so, so in just thinking about developing countries, The issue is also disease focus. We're more focused on the disease and less on how do you live after diseases. And because of low resources, the country structure might be challenging for implementing the fact that robot systems. But rehab axis is low. So that actually creates an opportunity for the technology and that the skilled therapists in physiatrist are often not available in Botswana. There are 0 for scientists with IHS or rehab medicine docs, duper important for that rehab quality of life. There are 0 for scientists and backcountry, low doctor-patient real low patient therapist ratios. The rehab technology, if it's there, it doesn't kind of exist in the quantities. And we've actually recently did a survey. One of my goals was to survey what's going on with rehab in Botswana. One of the things we found is that, for example, the World Health Organization, I don't know if you guys notice came up with the 50 list of the most important assistive technology that should be in a country. And Botswana is one of the countries that signed up on that saying we're going to dedicate to make sure that people with disabilities have access to these technologies. Most of them are like, for example, role later as walkers, canes simple things. When we, when we query and people in Botswana, like most of these things don't exist. And if they exist, because there's no local manufacturing for rehab, everything is important, important it and thou, we're now quadrupling the cost. A simple key now cost a whole lot more money than they can afford a little and a wheelchair once you import it, it breaks. No one's fix, fixes it, and so therefore, it's more expensive. So there's these really weird dynamic. So now let's fast-forward to robotics. And people are like wait, there's an oxymoron here, robotics and developing countries maybe that not so smart. But I push back and ICU. But these are the plate, the species that might need us to really think about this even more. Because one, they're fascinating that the type of interplay of disease that's happening there. But then also the, the issues of access is very startling. Here's just a quick statistics. You can kind of high-income countries, the ratio of doctors, medical doctors per million population. Here we are in, in that, in some of the areas that we consider low and middle-income, they're struggling with down to very close to the zeros. Here when we think about the density of physiotherapists, which is often, if you say were the therapists, most of the time you encountered therapists there, physiotherapists, they're not as many specialized therapists like speech and language occupational therapists. They're broadly for the most part, mostly physiotherapist. And the, the luxury of having therapists that fully specialize is not there because there are not enough of them. So the therapists have to do a lot of different things. You can see this, the disparity between the number of therapists available and what's happening example in Africa region. So this, I always have this nice slide that I always think of the, the gap. The gap is huge in the developing world, but the gap exists also in the US. And that's what I, I don't forget that because I see this gap also in the community as well. So we assume that technology can help bridge this gap. But one of the challenges with technology is how can we make them more inclusive and how can we make them more affordable? I use this definition, affordability being the extent to which something is affordable as measured by its costs relative to the amount the purchaser can't buy, right? So when we think about what is affordable, I use the World Health Organization guideline, which is really actually kind of cool because they say they actually came up with something that's economically can be appreciated. And it's based on the gross domestic product per capita. What they argue is that if you can make something and you can implement a technology that is within one GDP, then you're going to be highly cost effective. And it's talking about cost and benefit ratios. If you can focus that, okay, it'll be between one, it's less than one GDP per capita. Highly cost effective if it can be between 1 GDP and three times the GDP per capita, it's cost-effective. And then something that's beyond, that is unreasonable. So I took this and I said, Well, could we loosely use that as kind of a cue for how we implement technology and if our robots are our systems or technology systems and fit these criteria, maybe then they begin to be affordable, right? This begin I say that facetiously because with the, with the, with the, with the dollar and the differences in economics and it becomes hard to even say that, for example, the US GDP per capita is said to be 53 thousand. Of course, we know there are people that makes a lot much more money than that. When people make a lot less money than that. So just, just bear with me that this is certainly the the average. And then the three times that is 159. So when we bind rehab robots that cost 250 thousand or a 150 thousand or a 100 thousand. Most hospital systems are not blinking. They're like, Okay, that's okay. But but the community centers are like, We can't afford that. We're more like operating on the low middle-income space. When we think of a gonna, gonna is separate, this was I think their increased right now to maybe about 10. But there's still a significant difference in what is considered cost-effective. So when we were developing systems, we wanted that the entire system would be under that number or between these two numbers, right? So those are the numbers that we're working with. I find that this is actually effective in kind of making me put cost within my infrastructure of design. Some of my colleagues will say no, we want to make the best robot system possible, no matter the cause, we want to deal with the science first and then we deal with the cost. I agree on some level, but I just realized that designing and putting classes as constraint on my design and the effectiveness of my intervention forces a sensitivity within me that actually pays off in the end. I know not all my colleagues will agree, but I think that if we actually design and push with cost in mind, we might even innovate even more as a result. But this is my opinion, right? It brought me here to give you my opinion, giving you that. So we actually had a colleague in Italy working on kind of doing a review on affordable robots out there. What's going on? There are people working in this space and trying to push that envelope. They said, This is our system and we're one of the few people that have actually actually trying to put our money where our mouth it and do this type of work. But I'm happy to see colleagues like starting to push in that direction as well. The potential strategies that you see out there that when we looked at the literature using low cost robotics mechatronics systems not being committed to the sixth degree or, or 70 degree of freedom systems if possible. But if they are going to have multiple degree of freedom, thinking about cheaper material, soft materials, 3D printing, those types of things. We see that as a way people are doing that. Also using a system of low, low costs are low degree of freedom and using those in creative ways, I then one degree of freedom system, could it be modular? Could it built to a two degree of freedom system? Or can a one degree of freedom system itself be just oriented manually or a rearranged. Or you, you, you challenge it by putting 21 degree of freedom systems together. So these are kinds of strategies that I see you people doing in order to try to push this envelope of an affordable system. So one of the big things that I realize is this idea of using local manufacturing and resources. So in a sense, this is actually me coming to this conclusion after spending time in Botswana realizing that that I can price something at $5 thousand US dollars. That is why it's 11 ampulla, which is the dollar in Botswana, one US dollar. So all of a sudden my 5000 system, because 50 thousand fool in their life, we can't afford that. And then I have to go, Okay, this is interesting. So I learned very quickly that collaboration is key because using local manufacturing and resources can drive that cost to now where it's affordable. So we built an, a memorandum of understanding with the University of Botswana so that I could share the tech in my lab with the university. There. Then what we were able to do was build a lot in house that then was able to drive down the cost of the system that we're using there. So just an interesting strategy. This challenge though, economically I'm not an economics, but some people might say that sustainable or not. But maybe I need to find some folks who I can collaborate with. The practice of inclusivity, practice of providing equal access and opportunities and resources for people who might otherwise be excluded or marginalized. And those that have physical or mental disability or belonging to other minority compute community. Sometimes people go, why are you using that term inclusive entity, inclusivity. And I think I said, I guess it's the thinking that my passion for being in the space of working with people with disability was always about making sure that no one felt excluded, right. Making sure that if you had an impairment, that we had the hope of getting back. Right. I know it's I know it's not always the case for everyone, but they have no hope of getting back to your yourself. And it began with my grandmother. And if anyone have ever seen my TED talk, grandmother is dry. I'm here because when she had a stroke and I saw probably took her from this dynamic woman, she's top line was wheelchair bound. I became committed to figuring out how we can change this dynamic, right? Just say we have a question from the online audio shorter. This is from Steve. Well, he says also that this is a question about the cost part. Cost includes the time for the clinician family interface. What is the evidence between cost of robotics and their efficacy or effectiveness in securing a given functional outcome. And is that interface superior to one-on-one treatment or interfaces? Okay, so there is some evidence. So, so a lot of the evidence around robotics have been mixed in the sense that there is a clear sense and it may not be better than what's happening with the, with the physiotherapist said if you dial in intensity and you dial up intensity and you dial up the the and you match dosage that the, the, what the therapist can deliver is, is equal to you. Oftentimes see what the robot can deliver. But I don't view that as a bad thing because in these spaces there are no therapists. And therefore that's the, that's why I push that opportunity for robotics is higher in these spaces where there is no therapists. We can have that argument in America where the number of therapists and so we argue robots shouldn't be replacing therapists. That's not about what I do guys, bringing intelligent robotic systems to work alongside therapists. And that in that evidence, we maximize the therapist's ability to do more. Where I drag the argument that robotics become a superior possibility is when there is no therapists and that's the only way that people are even going to get asset. So I push back on that and say, The evidence is the fact that it can be pari gives us the opportunity BD now explore whether it can be a vehicle through which we can manage these access issues. So, but that I think is research that people like myself are trying to do within the space is hopefully the answer to the question. So one of the things of inclusivity that we were thinking about was oftentimes in rehab, even a technique. As engineers often or when we start to do experiments, we, we try to control the problem and we look at disease as having one effects. So most of the time in robotics when we began, we started just thinking about motor issues. People with motor function, stroke survivors have primarily motor function, but the reality is stroke survivors have a plethora of issues. And that oftentimes, cognition is a major part of that. And the fact that most often there is actually some really good research that shows 77 percent might have upper limb impairment. 43.9% are dealing with cognitive impairment, mild and sometimes even severe. So the question then becomes, how can we move and be able to use robot one with a larger community that my clinical colleagues push that to me, say, Hey, these robot systems are often not really inclusive. Included these other population, these stroke survivors that I have cognitive issues. How can we change that, right? The second thing was thinking that because cognitive function is not commonly use, when we looked at the literature, we looked at all the robotic systems over there, most of them, but most of the studies have done with people with mild to no cognitive issues. So there's an assumption that the people doing and using our robotic systems don't have any any cognitive issues. And therefore, the, the recovery that we're seeing is possibly challenged by that naive assumption. Because when someone gets recovery and they go out into their regular environment, the cognitive demand of doing motor tasks, as well as your, the cognitive demand of living and doing motor function begins to challenge their ability to transfer what they learn to the real world. This was my theory. And we began to say, How can we look at that and examine whether that has any truth. We also kind of said, well, if we took a simplistic approach and we said the x-axis is our is our motor impairment. High impairment here means low function. Low impairment means higher function. And then the y-axis is our cog axes. And high impairment mean low carb function and low impairment means high com function. So I say that because I may use the term interchangeably. What we see when we look at the Robotics, let's say we were, when you look at like this quadrant, these were kind of people who were already kind of mild stroke, almost fully recovered. Most of what we do in robotics is this group, low COP, the impairment, high, moderate to high motor impairment. Here, lack of effective mesh. It measures most people exclude this population. And then this population often have only **** rehab and not any motor rehab. So we realized that there was an opportunity here to really think through how we can include more within the concept of, of, of our robotic assisted therapy strategies. This is why we started looking at motor and caught together. So that plus our typical design goals kinda led us to start really rethinking. I'm going to talk about too quick to case studies really quickly. One is the work that we did in Mexico that kind of gave rise what we call our rehab cares system that we've deployed in Botswana. And then to, I'll talk a little bit about the work that my PhD student in terms of this issue of looking at motor and Cobb together and how it influences kind of the how of my influence recovery after motor impairment. How we might be able to possibly use this to determine how robotic systems could possibly function in the future. But even how we can use it to understand what might be going on after HIV and joke. So this working Mexico was kind of done little bit towards the my time in Wisconsin. I was working with this colleague who was in Chihuahua. She was a bioengineer professor there. And we collaborated together to look at this idea of just my preliminary work on getting affordable mechatronics systems out there. Though, we worked with specific stroke. Most of you are probably already familiar though. A couple of different ways that you can get ischemic stroke. Hemorrhagic. Ischemic stroke is by far the most popular, especially in developed world. Hemorrhagic stroke tend to be higher in, in developing world. Eight hundredths the strokes per year in the US. Awful lot of money. So there's, there is reason why we're, we're all thinking about stroke recovery. And stroke survivors have, depending on the quality of rehab, even in the United States, may not be getting enough therapy in order to make a difference in terms of their recovery. So this is the typical kind of characterization of the severe a person with severe stroke, having the upper extremity and then having mobility issues. This talk will be more focused on the upper extremity. And so Huth, already six to 61% of stroke survivors will have cognitive impairment, which is why we feel like this is a direction that we really need to think as a feel 65 present motor. They'll also have psych issues as well as impairments in their living, an activity of daily living. This is why even considering this space here says, chokes the individuality of the stroke survivor and why it's important to figure out where they are on the spectrum. In our case, we're looking cog and motor and then determine how do we then treat this person given that spectrum. Initially in Mexico, we, I started working with, can we just not use robotics and just use simple like force feedback have to existence as my first question. And then, and so in Mexico we develop kind of we were using force feed back wheels. And so we were putting gaming together. We were one of probably the first lousy start thinking about is can we just have people play games and those have, you know, helpful consequence. One of the things we got a grant, my colleague got a grant from Mexico and then I had some funding to put together kind of what we call a robot gym. And it's basically a computerized mechatronics. Jim For, with excuse me, six stations. And then we publish this work and we, we basically randomize people to going to regular therapy and going through the gym for upper and lower limb therapy because we want it to be comprehensive if we were going to prevent people from doing the standard of care. We also kind of looked at 20 stroke survivors and we examine that they had a variety of different strokes, various levels of function, control group, robot, what we discover in, and I don't know if I kept it in the slides. But we discovered that they were able to recover both on the upper limb and the lower limb. The lower limb did better than the upper limb. And then that we also saw that the robot shape was actually had two really great unintended consequences. Work it out in a group was fun. And people were like, Hey, can I can come back and we're like, okay, this is the unexpected plot. And then to the, the, the time and the energy to both play games at work out. Really, people felt like, hey, I am getting my exercise in and I am getting better. I'm also doing it in a way that was was enjoyable. Not to say that they didn't enjoy working with their therapist. They actually did enjoy those one-on-one interactions. But this became kind of all a really interesting thing. What are the things we found out though, was the force, the force feedback we'll for inadequate. They were too they were not powerful enough for very severe stroke survivors. We needed to be able to assist them when they couldn't move. So the robot needs to be able to be strong enough to applied torque if it's going to be a one degree of freedom to help move the system. And it couldn't do that for the very severe functioning patients. So we went back to his joy, morgue and cytokine. Can we develop a one degree of freedom system fill thinking about low caught that would help us to work with both the severe, the moderate and the miles functioning stroke survivors. So hence, that led to kind of this simple robot system here. There is four sensors in the handle. There are position encoder, very simple system. And we basically can do adaptive strategies to assist when needed and to resist if you're doing really well in terms of performance, we could then do like an anti spring or challenge you in some interesting way. We also realize on the gym idea that maybe we shouldn't be thinking about just one isolated system, often the quarter. But can we, in the idea of working with the upper limb, the lower limb, having the, the arm working different ways because these were one degree of freedom systems. We can populate them around the structure that would allow us to maybe now have multiple people working together in a gym. Why didn't we create this structure if you? So this was our first prototype of this platform. And then we decided then, okay, this platform, if we build it, we're the center part was a a gate platform in places where where they have a gate platform but no space, we would replace the gate platform with our robot system. They'd still have a gate platform, but now they would have a couple of robots in the system. We've also talked about how do we, do I have a PhD student right now looking at haptic dyads. So basically, if a stroke survivors working with a one degree of freedom robot and another is working with another. How do we have them interact? Where now we, and how does that interaction increase the motivation, increase, maybe the challenge, maybe increase and drive to plasticity that she's at the early phase that was looking at that. It turns out there is though there's a lot of work for people who work with healthy subjects doing that, but not a lot of work or people work with impaired subjects in that paradigm. So the work in Botswana then, my PhD student, we started saying, asking the question, how does cognitive issues affect motor performance? That was kind of fundamental question that we wanted to answer. And why he didn't HIV. Most people go HIV. What does that have to do its job. But at the, there, there is a long story as to why I'm in Botswana, but the bottom line is, HIV is a risk factor for stroke. I had a colleague working in Botswana who was telling me that a third of the stroke survivors had HIV. And that oftentimes the first time someone comes to the hospital with a stroke, they don't realize they have HIV and that could have been the reason for the early stroke. And that later on there is that there's knowledge that, that the HIV leads to a type of stroke that's different from the average stroke that you get when you have co-morbidities, etc. So it was actually a really fascinating place at the time. Though. Now with antiretroviral, yes. Somebody also reads mostly atrophy which will then affect muscle system. Yes. Yes. And, and what's interesting too is restarted. You're working with Botswana because it was one of the first countries were antiretroviral therapy was in place. But so many of the people in Botswana have our audio ART's. But we were still noticing neurological disorders, Worst, still noticing what they call HIV associated neurological disorder or hand. And that there wasn't a significant percentage of the population that might be actually having these cognition and motor issues. So the other issue is the HIV population. They're living, they're surviving. Now, they're becoming chronic over the age of 50. And now joke is this than the regular stroke is starting to catch up with them with the co-morbidities, hypertension, etc. So how do we deal with the complications of stroke was what was, what was put to me and whether our robotic system could help one assessed and possibly tree. So it was like a good way to think about how do we translate our stroke work into HIV and how do we deal with those with HIV and stroke? So that's kind of where my students spent a lot of his time kind of really thinking about ways we can do this. The long-term impairment, 40% of HIV will end up with some type of neurocognitive disorder. Turns out 70% of them have some type of motor dysfunction. But what's tragic is most people with HIV do not get any rehab because people assume there. Okay. All right. And so this is a this is this is actually one of the novel areas that we're pushing it and, and trying to support the need for rehab and whether robotics can support that population or not. They also then will have difficulty with instrumental daily living. And so it is a population that might benefit from one assessment and to therapy. So one of the key things that BY had my students do was think about, well, what, how did we do this? How do we try to figure out how motor performance is affected by cognition? So HIV became our example group and stroke as well. So these are two populations that we decided to study. So he targets for clinical, clinical assessment. There was a lot of work that showed that HIV predominantly affects the prefrontal cortex. And that might be more related to executive functioning, information processing, things like that. So we decided then to put together a clinical tool that can kind of look at these particular cognitive areas as well. Generalized clinical tools. I'm going to go pretty fast. So our goal was to develop a robot base metric that could quantify motor and cognitive. Put someone on a map for us to see where they are, and then we could stratify them. The next step would be to say how do we then define an appropriate rehab or robotic strategy that would then help that patient move and, and recover. So this was the study, this was first done in the USA, had 15 HIV thick stroke and then the combined was about 21. The key thing here he had, he had the moca is the Montreal Cognitive Assessment, the International HIV dementia scale. And then we also use gross box and block scale. But this was just the comparison. So people with HIV and stroke, and I'll, I'll put time, I'm going to get this and move on. So he I had him put together motor assessment from our consultation with clinician, everyone did. Only a stroke. Survivors did the Fugger mire, but everyone did the box and what Gibbs grip strength and the groove peg board. And then we had Cognitive Assessment, the module COG, the HIV dementia scale, color channels 1 and 2 and digital symbol Cody. These are for information processing. Digital poetry I was 12 are both visual, spatial and executive function. International HIV dementia scale was essentially said to classify whether you had severe dementia or not. And the Montreal Cognitive Assessment looks at five, excuse me, six domains of cognition. Though we kind of like double thing. So if you've never seen the box and block, this is like a really simple measures, like a minute on each side and you're, you're, you're trying to figure out, oh, yeah, your triangle. We're one minute to get as many boxes across. This is supposed to be for reach, gross, gross reaching and grasping motion. Let me atom. So this, There's a little bit of delay, but that's okay. This is the group that worked for those who are not familiar, essentially involves both motor and cog and it's more fine motor function. You have to put a mini pigs in and in as little time as possible. And so the faster you do it, the better your, your cognitive and motor function, right? The Montreal Cognitive Assessment, you have to be trained to administer it. But basically it looks at visual-spatial executive function naming memory, attention, language, abstraction, and delayed recall. And this was actually made famous by Donald Trump when he said, it was hard. This is the test data. So this college real test is actually like a trail making its classic cognitive test that's used. Essentially if you can count up to 12, 25, and then essentially there is, there is a process for making in the connection box. This is really a great simple test, but really good about looking at executive functioning. So if you need a quick way of looking at executive function, I recommend some of these simple access to then the robot based measures. We decided then to develop a robot based legend that was mostly motor, then one that was mostly cog with the working memory, and then one that required both to the mostly motor one. This does take some spatial, has some hog in. It's, it's, it's hard to really pull separately the COG out, but this is mostly motor. They're tracking a task and they have to keep the, the green that they're there. The robot in the green box, the robot is their icon. And essentially these are the metrics that me pull out. Mean performance error, normalized travel distance, things like that. Then this is what you would get depending on what level of function, how well were they able to track that task? Then the cognitive task with the working memory task, euro back one back to back. The back is like no memory involve every time you see a to press the button, one back. Every time you see a number repeated, press the button, and then it gets a little harder. Now you have to remember it every time you see the number before to back before press the button. So this wasn't actually, I call it facetiously a robot task, but it was really just a button press. And we look that correct responses and reaction time. So then this was the task that my student put together from the Sternberg task, which is actually a very common cognitive task that's done. The idea here is the patient is shown in a sequence that lights up. This is the COD part. They have to remember the sequence and then they have to move the robot you in the exact order to acquire the box, light it up. So now they have to remember, then move in the order of the sequence. And that was what determined it was interactive. So they were first given three box to light up. If they got it right, yea, they get for box. If they got it wrong, went back to three, got it wrong, both to any went up and down until we figured out what was the great, the best combination. We did this both with the dominant and the non-dominant limb. Because the fact that we were trying to look at motor, how does motor performance of the affected by hog as well as vice versa. And if the if the impaired arm. So in charge, we knew that there's going to be a component of the performance that's going to be biased by the impaired arm. So we had them also do a width there. Nonetheless, impaired arm or they're not impaired are important to see kind of can we tease these two things now? This is kind of what you might, the result might look like people with no impairment. And to be able to remember longer sequences and be able to move wealth of the sequence. People with motor only impairment will tend to, remember they only a little bit more over 2. They'll have less sequence of the histogram it shifted. Turns out the COG only was overlapping the motor the motor only. And the people with motor and cognitive impairment are the lowest functioning folks. Though, when you, you see it, there were 15 trials. See, everyone has kind of been learning. But if you're not impaired, you see, you see like there's a little bit of a fatigue effect because this is not easy. And if you are not left impaired, you see that you are you're able to get up to high mean I the trial number and your sequence length gets pretty high. When you're, when you have both, your, you're actually really struggling and you don't quite learn as fast about what's happening. And then if you look at the murder Omi and the column only group, they're hanging on top of each other, right? And we're like, okay, can we tease that out some more? It's implying that the COG is is affecting your motor performance. Because if you have high motor function, you're the cob, the issues that you have actually morphs your ability to perform. And intuitively that makes sense. It's funny. There's a couple of neurologists on my PhD students team and they were like, We know that well my bot robot. And it's like, you know, when we're trying to apply these things in the real world, these like I said, but, you know, you guys know that, but you never really quantify it in this way. That then allows us to be able to take that information and make use of it. So this was our really big finding it. The other thing was, could we predict the clinical measures of cognitive from, from the, the robot tasks? And this is something that was super important for us because we couldn't predict with a reasonable strength than this means that VB on the fly, we could adapt things, right if we knew and we can personalize therapy by understanding where you were on this on this made this map, this map, and then tried to drive accordingly. So here we were able to see that for colored shells, one of visual-spatial tracking, there was a strong relationship or an information coding. This is digital symbol. Coding, was that there were two strong relationships that showing that this facial span, that task that had both was able to predict, your information processing speed was pretty good, reasonable accuracy, and then also your color channels. One performance, which is your visual, visual spatial performance. We also saw that that facial fat task was also able to predict your color trails to with pretty decent, not as strong as the digital symbol coding, but the R-squared value. And the reason why we put, i've, I've put this up here, this I'm borrowing size of my PhD students work with. This is what we see in the literature where most people were reporting and using predictive values that were R-squareds in the 0.4.3 and we were able to get value that exceeded that. And so these moderate these were more moderate type results. And then in terms of motor function, that box and block the attack, we were back up to straw where what predicted boxes bond with both the facial span task as well as that tracking tasks, your normalized distance. So arguably, what we're hoping then is, now could the question was, can we use, so we can do both things. We can put people on the, classify them using the clinical scores. The clinical, this with the Bach, bach was translated to a z-score. 0 being what is normal. And minus two. Minus two is two standard deviations below the mean. That is the beginning of the moderates, the moderate function. And if you're up here, here mile the right year, mild them up. Alright. I'll I'll, I'll move quickly. You're going to be, this is the high, high group, right? Remember, and so we're, this is using those clinicals or is, but now, if we can predict these clinical scores, we can use the robotic measures to map people. And then if you're up here, this is the people we normally treat. But now if we know you're here, how can we treat you? What do we do? And it turns out there are a lot more people here than we realized. We now have the potential to on-the-fly strategy so that my PhD students are working now with happen one is that on the fly strategy, what do we need to do now to deal with people in these different groups? How do we personalize that? Tell me. So I'll just quickly that assumes you. Now we we did all this in the US. We we refined her stuff and then them full. Then we're like, okay, we're going to put it in someplace different like Botswana. And kind of test if some of the things that we're thinking about, one holds up in a new place where it's, cultural norms are different. Is it feasible, is it usable for this affordable robot to be used in this setting? Though? Most of you may know that Botswana, Southern, the southern half of the African continent, just above with South Africa. We were working with a variety of of, of kind of entities. They're University of Botswana, faculty of medicine and engineering. And we, this was funded to NIH and a couple of and the Fulbright, the US State Department to do this work. Botswana for his highest prevalence of HIV to occupational therapist, that means dealing with the upper limb only 2533 bed hospital. That means nobody was getting therapy. A third of this stroke survivors had HIV. So this is we toured lots of the, the, the, the, the centers that were there during therapy. People were being sent home right after they discharge with they came back every month for therapy. We know that that's not sufficient. So again, it was the I feel pleased to say, CAN me put our system here and what can we can, is it feasible, will the therapists, except that we'll do patients that accepted. The first goal was to develop the therapy system. They're in Botswana. This was my student. First, we brought the robots with us. We brought the robots with us. And then we commissioned our faculty of engineering colleagues to build a platform there which was locked lists lot cheaper. And then we basically did mounted the robots on the platform. And we try to be kind of sensitive to what kind of materials we could get there and things like that. Anecdote is that we thought LTE printing, that's the way to go to that. They have a 3D printer. They, to get something 3D printed took forever. And then they were very careful with the schools because, because the schools were imported, they were four times the cost of the schools in America. So dealing with those types of issues, and so ironically, they switch to wood for the coffee because it was just cheaper to get the wood. And so it made me think about 3D printing and the assumptions that we make. But anyway, so we now have two systems. One that setup, we ended up because of COVID having to robots as opposed to three, because of getting people away from each other. So there are two robot systems here. We have one in the hub, Moroni, and then wanting in a hospital, we were able to put one in the occupational therapy department in the hospital. And so right now we're running studies where we're using the, the, the robots to one assessed. And now we're actually in the phase where we're trying to do therapy in the feasibility study just quickly. The first part was we collected patients HIV, stroke, HIV and stroke. And then the second part was to do therapy with them. This was the Botswana population. Age wise, a leaner be about the same as our US population but a little bit younger in HIV and stroke. Remember I said that the HIV a stroke population appear to be younger than the wanting in us. Lot less education in the HIV in terms of over the age of 12. More moderate on average in terms of their functioning. And definitely more. They were, hadn't more motor function than the US group as well. We expanded the clinical assessments to include some mobility ones as well as some verbal fluency type testing. We did the three motor task, the same setup that we, and this was bent now done in Botswana, we did the NMAC as well as the spatial span task. Then we actually they're asked usability surveys. So we had three standard what they usually robotic the the nasa TOX, if you guys are familiar with that for task demand, how this would be Usability Scale and then the Self-Assessment Manikin talking about how excited you are, our joy and we'll just quickly, this is kind of when we plot the box and block, the way I did was I just plotted for you quickly the box to look for when the z-scores and the MLC and score. This line is our moderate functioning line. This line is our severe functioning line. And this is mild to normal, severe to severe to moderate, known as how many people are falling below that line. And this is where we normally treat right in. And so it was, it's always surprising for me to see how many people are below where we normally tree, right? And so in the HIV, only HIV and stroke and stroke population. So it suggests that we're able to stratify just even with clinical scores. But if you don't have clinical scores, can you do the same here? So just this is actually new data. So we're still in the process of analyzing. But there was a colleague who had went to Botswana before and worked with the HIV population. And these were n equals 30 was the number of people she had looked at in terms of. Was impaired and she did similar tests. The cohort in the US, when looked at that, they had left him pyramid. The cohort in Botswana. And you see that the impairments or higher. And this is actually, for me, interesting because now this is the cohort that we were just not going to get into the robotic stuff. And how much insight we can gain from that, from that population wasn't feasible. I think in general, lower is better. In this particular engagement, the self-assess mannequin, most people enjoyed the encounter. They were relatively excited by it and they felt in control. At the end of the day, there was some difference across the HIV stroke in people with HIV and stroke population. The lower numbers really fits him. For the most part. The people with HIV and joining the people with stroke had mixed reviews. I actually thought it was going to be the opposite. But what I really think was happening is that people with HIV are normally ignored. And so in some sense that they were exciting than just that people listening to them. I'm going to go quickly. Mental Demand, physical people were in terms of the different population and the therapist people were feeling that it was it was reasonable. The it was somewhat demanding, but it was reasonable. And so for most, most part, people were really frustrated with it. You see that the stroke survivors were a little bit more frustrated because they were a little bit lower in function. And so it was harder. But this is kind of how overall task demand. We see that it was kind of in the 50, 60 and the, the, the stroke survivors felt that it was much more challenging. System Usability Scale. Most people thought it was usable. 68 is kinda like the cutoff. They said, we again have this I think we need to explore is why the stroke survivors or less so with our system. But we see that the system was quite usable and that people were feeling like this is actually a reasonable thing for us to bring to Botswana sites that actually thought this was good results. I'm just going to quickly wrap up here and just save. We haven't dug as deeply as we did with us cohort to all the nuances in the murder data. And that's actually our next phase. We just finished collecting a lot of this data. And in part two, we're starting to do now the, the therapy. Though. Bottom line is we're doing the therapy part and we're starting to deploy the gains. And we're goal is to collect data on the, the, the therapy. I know I'm running out attacks and like the challenge as an opportunity and we have complex. So it is important to look at motor and COG in, in looking at kind of how do we recover function and sustain it. I think the key thing is weird. The idea of being able to creep up systems that can be effective as assessment and therapy tools. I think we're, we're beginning to provide some evidence of that. And I think in the global space, I think robots can work in that space If we get them to be manageable to, I'm not saying that these are very expensive systems are not in, in low and middle-income countries they are, but they're in very specialty places and the average person does not have access to them. So that's kind of like me trying to push this and it's the same thing in the US. You know, some, the average person may not have access to the Robotics. So some takeaways in terms of affordable therapy. If you're going to be in Botswana and places like that, I would say, we really need to think about partnering with the, with the, with the local engineers as well as clinicians there in order for these things to work really. And I, I think my biggest lesson is how the difference in the dollar and how that makes such a huge difference. And all the things I'm like, Oh, it's affordable and they're like, no, it's not. And so that was actually just an eye-opening that it's not as simple as even I was making it out to be, right. And that there needs to be just stronger collaboration if we're going to be this happen. Some is not there. And my guidelines, if anyone want to work in this space, portable, multipurpose, they still need to be effective, needs to be community-based and appropriate. And we have to build capacity without the rehab medicine doc. What I'm doing doesn't work without engineers that know are associated with rehab. Most of the engineers there are not even thinking about medicine. And if they are, they're considered technicians. They just gained the hospital and they're fixing things, but they're not innovating. They're not thinking through how to build new things. And a part of what I've been doing with the Fulbright was to educate, to try to train the next generation of engineers to be able to apply their skills within this field. So we have capacity-building is important. All right. These are all the folks that are funded us and I thank you and sorry for going over I went over quite a long answer to that. But thank you for bearing with me. Take care. Thank you for a very inspiring talk. And so I'm going to ask a question from online audience first, but anyone here who has a question titled faster and of my back on my math, let's say. And so this question's from Eileen says, this is with reference to the motor cognitive portion of the talk. As far as cognitive function or tests, is your team primarily looking at memory function? Are there other cognitions besides memory tested for the good question? So if you saw the diversity of clinical tools that we were looking at, We were digital. Symbol. Coding is information processing. We, we did the general mocha, the Montreal Cognitive, that looked at a variety of different domains whatnot. It didn't delve deeply into any one, any domain, but gave us insight. In, in Botswana, we added language, we added verbal fluency as a part of the domain. The color channels 1 and 2 are looking at visual spatial functioning and then executive functioning. So we are actually looking at what the different domains are and as well as working memory and trying to figure out how each of those domains relate to our our task. What we found out quickly was in the robotic tasks that we weren't doing. It seemed like executive function and working memory and information processing. We're most sensitive to what we, what we are doing. So those dead became kind of like the domains that I think we're kind of going to delve in much deeper. Let's see. I have a question, but I actually am sensitive to the time, so maybe I'll ask you during marijuana one-on-one. If anyone does have questions, please feel free to just come to the podium afterwards. I think that will just thank our speaker now.