Research with the Power to Change Lives, with Dr. Ervin Sejdić, Research Chair in Artificial Intelligence for Health Outcomes, NYGH
From chatbots to wearables, AI is about to change healthcare in ways we cannot predict. Dr. Ervin Sejdić, Research Chair in Artificial Intelligence for Health Outcomes at North York General, urges healthcare organizations to step up as leaders in this space.
For this special edition of Healthcare Change Makers, HIROC has partnered with the Association of Family Health Teams of Ontario (AFHTO) to bring you the voices of health leaders advancing emergent technologies in clinical practices.
In today’s episode, we speak with Dr. Ervin Sejdić, Research Chair in Artificial Intelligence for Health Outcomes at North York General, and Faculty Member at the University of Toronto. An engineer by training, early on in his career Dr. Sejdić identified a passion for data analysis and an interest in clinical medicine. His career today brings those two passions together.
We talk with Dr. Sejdić about his team’s ongoing research in dysphagia, sensor-based prediction of falls, analysis of electrocardiograms, and the analysis of electronic health records.
Dr. Sejdić says the application of AI in medicine is still in its infancy. He sees the analysis of medical images coming as the next big splash for AI in clinical practice. We also talk with Dr. Sejdić about how we can make use of wearables, leveraging the sensors in our devices to learn about patient behaviour.
The next challenge for Dr. Sejdić is finding ways to translate his research from university and hospital labs into real-world solutions. He’s optimistic about working with start-ups and industrial partners to develop his research into products that can change lives.
At the 2021 AFHTO conference, Dr. Sejdić presented a session on “Artificial Intelligence and Robots in Primary Care” alongside Dr. Mohamed Alarakhia. After this, check out Dr. Alarakhia’s episode to learn more about how these experts are building a foundation for AI in medicine.
Mentioned in this Episode
- North York General Hospital
- University of Toronto, Electrical and Computer Engineering
- Holland Bloorview Kids Rehabilitation Hospital
- Tom Chau
- Healthcare Change Makers, Episode 29 with Daniel Pepe
- Association of Family Health Teams of Ontario (AFHTO)
Imagine you could step inside the minds of Canada's healthcare leaders, glimpse their greatest fears, strongest drivers, and what makes them tick. Welcome to Healthcare Change Makers, a podcast where we talk to leaders about the joys and challenges of driving change and working with partners to create the safest health care system.
Michelle Holden: Welcome to Healthcare Change Makers, a podcast produced by HIROC. From communications and marketing, I'm Michelle Holden joined by Philip De Souza. Today we're talking with Dr. Ervin Sejdić, research chair in artificial intelligence for health outcomes at North York General Hospital in Toronto. In living our vision HIROC partners to gather perspectives from across the continuum of healthcare. Through this podcast, we bring you the voices of these leaders. At the 2021 conference for the association of family health teams of Ontario, Ervin and Dr. Mohamed Alarakhia hosted a session on artificial intelligence and robots in primary care. Today we'll be talking with Ervin a bit about the session, about his work at NYGH and how technology and data solutions can be applied to healthcare. On another episode, we chat with Dr. Mohamed Alarakhia about AI and automation in primary care. So we hope you'll check out that episode after this. Hi Ervin, thanks for joining us on the podcast today. To start off, can you tell us a bit about yourself and what you do?
Ervin Sejdić: So I'm Ervin Sejdić, I'm a faculty member at the University of Toronto in the department of electrical and computer engineering. And I'm also a researcher in artificial intelligence for health outcomes at North York General Hospital in Toronto as well.
Michelle Holden: So that title research chair in artificial intelligence doesn't sound like a role that we would've thought existed 10 or even 15 years ago. What was your first connection with AI, if you don't mind sharing and how did you envision it being applied to healthcare?
Ervin Sejdić: So I started looking at AI and more in general, like data science, machine learning, AI during my grad school. I was really interested in data analysis in general. So I'm an engineer by training and unlike other engineers who like to build stuff, I like to analyze data, right? So I started kind of playing with data in my grad school and my entire PhD focused on theoretical machine learning and new algorithms for machine learning. And by the end of my PhD, I really kind of this sat down and just kind of asked myself, "What do I want to do with the rest of my life?" Right. And I was always interested in medicine and clinical aspects of our lives. So I'm like, "Let me look for a position that will enable me to broaden my horizons in that area." So I joined Tom Chau's group at Bloorview Kids Rehab in 2008 as a post-doc and ever since 2008, I've been playing with machine learning, AI, data science in the healthcare.
Michelle Holden: That's amazing. What did you learn in the post-doc at Bloorview that kind of got you to North York General Hospital today?
Ervin Sejdić: So interestingly enough, the first project I started working on was with a clinician from North York who unfortunately passed away a few years ago, but he was my clinical lead on one of the project. And what I really learned as a post-doc is that in theory everything works, in practice nothing works. So that was a big revelation for me because throughout my PhD, I was really looking at ways to analyze the various data, right? And typically we would assume certain data and then we would simulate this data and then analyze it. Right. But in real life, once you are dealing with patient's data, you cannot just simulate tens of millions data points. You are most of the time working with few hundred data points at best, right? So suddenly all these algorithms that you can develop and they have great potential. You can throw them away because in real life, you're dealing with much smaller data set and you need to be innovative how to apply these algorithms to such a small data sets.
Michelle Holden: And so speaking of that, how are you doing that and what are some of the projects you're working on today at North York General Hospital?
Ervin Sejdić: So there are three kind of four major thrusts that I'm involved with. One of them major thrust is the analysis of swallowing difficulties. So typically what is dysphagia? So it's the feeling when the food goes the wrong way. So we typically kind of cough in order to expel the food from a tube that's going to our lungs, right. But for people with dysphagia, they might lose that ability. So we are trying to develop a small noninvasive device that we can position on the neck of a patient. And we kind of listen, quote unquote, to swallowing sounds. So we use kind of combination of swallowing sounds and data science in order to come up with a decision, whether a swallow is safe or not and whether certain physiological events have happened.
The other kind of thrust of my work deals with prediction of falls. And can we use, again small sensors, non-invasive sensor to predict falls down the road. We don't care in predicting falls once they occur. We want to predict them ahead of time. So that's a bit more challenging and we're trying to predict falls outside the lab environment. So we want to predict falls in real life environment. The third thrust is general analysis of physiological data. So I have a partnership with few clinicians looking at the analysis of electrocardiograms and can we predict ACS occurrences, so acute cardiovascular syndrome, right? Or heart attacks based on 10 short seconds of ECG that are recorded pre-hospital at patients home. And we are able to do that quite accurately. And the last kind of thrust is the analysis of electronic health records. And we are trying to kind of use natural language processing and other modern machine learning techniques to mine through these electronic health records and to develop phenotypes for various purposes from mental health to potentially even phenotype for swallowing difficulties, right? So predicting dysphagia based on the patient history, medical history.
Michelle Holden: So I just wanted Dr. Sejdić to find out someone like you who's really busy and like many other healthcare leaders in your field who are busy. In your opinion, what are one or two things that organizations can do right now to prepare for a future with AI? Given the limited time they have.
Ervin Sejdić: I'm going to suggest something out of the box solution. I think we are past the time when patients would only see clinicians. We need to think how our organizations can involve people like myself in day to day clinical care. I strongly believe that data and what we can learn from data will be as powerful as let's say clinical assessment one on one that a clinician would do, right. And we need to position ourselves for that time because it's coming sooner than we think it is.
Michelle Holden: Yeah. That's a really good point. I think there is a big role as you mentioned for non-clinician in that clinical area. So I'm glad that you mentioned that. One of the reasons we're chatting today is because at this year's AFHTO conference, your talk with Dr. Mohamed Alarakhia was about the AI and robotics in primary care, the current and the future state. So in your experience in research, in what area specifically has AI grown and improve to assist in clinical practices?
Ervin Sejdić: The application of AI in medicine is still in its infancy. And the most realistic examples that I've seen so far are really the use of chatbots for quick assessment of patient condition, or maybe a recommendation of patients for various vaccines or follow up visits. So chatbots in general, I think are the first kind of forte of AI into the medical world, right? The real life application. The other ones like... My group also does a lot of development of novel algorithms for the analysis of medical images. And I think that's going to be the next kind of big splash that AI will make in the medical world, is that we will be able to use AI algorithms to quickly analyze some of the medical images and come off with a summary for a clinician. And that's not to replace radiologists whatsoever.
We are actually hoping to enable radiologists to see actually more patients, because currently for example, if you're ordered to have a medical imaging done, you most likely you will never see a radiologist. So what we want to see is actually enable these physicians to see patients rather than to be stuck in a basement of a hospital analyzing thousands of images that they need to analyze every day. So I think those two applications of AI will be coming to us very soon. And last but not least, I think what's really interesting is going to be the use of wearables in medicine and especially for smartwatches, smartphones, how we can use these sensors that we are really carrying with us all the time? How we can use to actually learn about the patient conditions or the patient's behavior. I think that those three things will be very important in the next few years.
Michelle Holden: I think you make some really good points there, especially because we tend to out of the box just think of AI as something that almost and robotics takes away from that face to face. But what you're saying is it's actually giving it more of a face, being able to engage more with the clinician. That was a really great example there. I'm also taking it down a notch a little bit. As on this podcast, we do try to engage leaders and talk a little bit about how they share what they learn with mentorship. So as a mentor, supervising masters and PhD degree candidates at the University of Toronto, what's one thing that you've learned from students and mentees.
Ervin Sejdić: One thing that I've learned from my students is that they're much smarter than I am. I have to admit. I'm quite amazed with what my students do, my trainees do. And I'm quite amazed by the solutions they come up with. I've been working as a faculty member for since 2011. And I'm quite amazed by recent developments that my students did. I think I would never be able to do that by myself. And I don't think I'm as smart as they are because I'm really blown away by some of the recent solutions that they proposed.
Michelle Holden: I like that it's very humble. So if you weren't an engineer applying technology and data solutions to healthcare, what would you be doing?
Ervin Sejdić: If I were, I would be probably an MD because that was one of the reasons why I ended up in medicine because I really wanted to go to med school. And for some reason, one reason or another, I did not end up going to a med school. But then I ended up doing close to what I really wanted to do that is I still see patients. Like I see their data and I analyze their data and I try to diagnose certain issues based on the data they provide. So I'm kind of doing what I wanted to do, is just I'm not an MD.
Michelle Holden: So can you tell us one thing that would surprise others to learn about you? It can be about data, but it could to just be about anything in your life.
Ervin Sejdić: The biggest surprising fact was, and that's not... People can easily find this about me if they Google my name, is when I was awarded the presidential award by president Obama in 2016, when I was in the states. There was the biggest surprise of my life is. I always knew that science plays a major role in every modern country, but it did not hit me until I was actually in the white house meeting the American president that it has that kind of a role and that it's taken seriously. So what I would encourage is the reason for saying this, I would encourage all young students to actually consider science in general. Science and research as a potential way to contribute to this society. Many excellent students end up going to med school because they really wanted to help the humanity, but you can help the humanity in other ways. And especially as a researcher, you can really help many people without even realizing it.
Michelle Holden: That's a really nice message to kind of end off on. But before we do, I do want to bring in and see if Philip you have any questions that you want to ask to Ervin.
Philip De Souza: Oh, I thought it was really good. I feel so inspired by that last thought. I'm like, "What can I enroll in? How could I help humanity?" And it actually tied in, that how you ended off just now tied into your first point about how you kind of landed where you are now. And you told yourself you wanted to broaden your horizon. That's what I wrote down. And look where it's gotten you. And I must add, when I say it own advice you, you could still go to medical school.
You can still continue to broaden your horizons. And I appreciate that. And no, all those four points you mentioned, the things you're doing right now at North York and with your team, I guess one question is how did you land on those four initiatives and what's next for those four? Do you see something where you're partnering with a startup or a for-profit or other not for-profits to further kind of get those initiatives out and scale them.
Ervin Sejdić: Yeah. So that's a really good point. At the end of the day, I am a translational researcher. I really care about the translation of whatever I do into the actual products, right? So one of the things that we are thinking about is actually for our swallowing work, we are at the stage where we have to within the next year or two decide whether we're going to have a startup and start mass producing this device, or are we going to license this technology to somebody who will do it for us? Because we are at the stage where we are ready to do it, and we need to get this type of device into hands of many, right? And you bring an excellent point. I really like the translational aspect of my work. So I'm always keen in partnering with industries.
So even for my fall assessment work, I'm already thinking how we can actually have a startup or maybe partner up with a potential industrial partner, right. To actually develop it into a product because R&D is one aspect, but developing a product is a completely different beast. And typically you cannot easily do it in a university lab or a research lab in a let's say a hospital, right? You really need to have a company to do it. So that's something I'm actually looking for. And I'm hoping that within the next year or two, being back in Toronto I'm actually able to do it and find these industrial partners that are willing to work with me. And I have resources to work with university based researchers.
Philip De Souza: That's fantastic. And I'm sure anyone listening now feel free to reach out to us and we'll connect you to Ervin and so maybe we can make that a reality. Because we at HIROC are all about our subscribers sharing knowledge across the healthcare system so that those lessons learned and these innovations can be spread right across the country so that we create that safer environment for patients, for staff, for patients' families, everybody involved, the community. I asked this morning when we interviewed Mohamed, I asked him, sometimes thinking about topics such as AI, robotics data, it could some people listen to it like, "Oh, how does that affect me? Or what's the point of that?" Some could say that.
So what would you say to people who are listening now who want to start that conversation at their organization to get the topic of thinking about AI, robotics data, all the stuff you mentioned today into people's minds that they could think about it, maybe prioritize it in as part of your strategy going forward? What advice would you give those people?
Ervin Sejdić: Well, I would give them a simple example. Think back to 1990s. And if you were wondering how cell phones, we call them back then cell phones, right? Not smart phones. How cell phones can change healthcare. We could not even predict it. Nowadays smartphones are becoming one of the tools in a day to day healthcare settings, right?
Philip De Souza: That's True, yeah.
Ervin Sejdić: It does not matter. At the end of the day, yes, we use these smartphones for actually calling people. But that's like, I don't know how you guys use your phone. I think that's the 5% of the use on my phone that I actually call people. Right. There are many more things that we do with smartphones and they're essentially changing the way we do healthcare, right? Think back even to a simple thing. Clinicians or even if you did any type of clinical work.
I still remember carrying those pagers around not long ago, right. I don't think anybody carries pages anymore. Simple as that, obsolete technology, right? So these things, think back how nobody could predict how smartphones could change healthcare in general, clinical care. Now we cannot live without it. The air is about to change the healthcare in a ways we cannot addict. And if our organization don't act on it in a timely manner, they will be playing a catch up game. So it's up to the leadership, right? And in every organization, leadership have their own ways of thinking. But simple at AI, we are either leaders or chasers. So it's up to them, right?
Philip De Souza: Yeah, you're absolutely right. I like that point you bring up. So to those people who want to start the conversation, I guess a good way of starting it is thinking back to how far we've come in the past even five, 10 years, not even 20.
Ervin Sejdić: Yeah, smart phones really now have taken us to a next level. And it's going to be the same thing with AI, not AI itself, but the solutions provided by AI right? So my team is working now on, I'm trying to recruit few students to work on few different apps for smartphones that will basically tell people, "Hey, do this simple experiment and we can predict whether you are having an asset of mild cognitive impairment or not." Or, "Hey, can you turn on this app and walk for a few minutes to, and we'll tell you whether you're prone to falling within the next few months or not." Right. So these are big things.
Philip De Souza: Yeah. And then, I recall another episode, we talked about similar to what you're speaking to now and I think it was Dr. Pepe we had interviewed with, and he mentioned that it was a better picture of the patient and their story. When you're able to identify all these elements throughout their history and their health and it only helps the clinician have a better perspective and say, "Oh, you know what, yes, you have this X, Y, Z happen and we should take a look at that further, et cetera." So, no, it's a good point. And I guess the last question I was going to ask before we let you go, is what's next? What's the next big thing you're working on? But I think you kind of told us about these apps.
Ervin Sejdić: Yeah. So I'm trying to translate some of these research findings that myself and others have come across, right. And then we try to develop some apps, some simple solutions that can be easily distributed across Canada especially to rural areas, right. Which have difficulties. If you live in a rural area, you have difficulties assessing a family doctor. But that's the issue nowadays, even if you're in a urban area, right. So hoping to kind of spur the growth of healthcare in all of ways. So that's kind of my long term goal. And again most of my work, I liked something that I heard a few years ago when one of the cancer specialists said, "You can survive almost any cancer as long as you detect it early." Right. So that kind of stuck with me and led me to believe you can prevent any major or you can control any major health or issues by detecting it early. So I'm hoping that AI and variables will provide that early detection of onsets of various diseases.
Philip De Souza: No, I hope so too. And we're very inspired by you and the work you and your team is doing. Yes so congratulations. And I think that's all from me Michelle, pass it back to you.
Michelle Holden: Yeah. I just wanted to reiterate a little bit of what you said Philip there, and that if there are any other organizations or people listening who want to reach out to Ervin, do let us know, because it sounds like connecting you guys is something that we can do and in a way we can help. So it's a really exciting project that you're working on and I can't wait to see how they grow. Thank you Ervin for being with us on this show today and we really wish you luck in the next projects.
Ervin Sejdić: Thank you. And thank you for spreading the word about my work.
Michelle Holden: You've just been listening to our interview with Dr. Ervin Sejdić, research chair in artificial intelligence for health outcomes at North York General Hospital. For more on AI and robotics in healthcare, don't forget to download our episode with Dr. Mohamed Alarakhia. And while I have you, did you know that HIROC has an AI guide? Search for it under resources on our website. For more information about HIROC and to listen to other episodes of healthcare change makers, go to hiroc.com. Thank you for listening.
Thank you for listening. You can hear more episodes of Healthcare Change Makers on our website HIROC.com and on your favourite podcasting apps. If you like what you hear, please rate us or post a review. Healthcare Change Makers is recorded by HIROC's Communications and Marketing team and produced by Podfly Productions. Follow us on Twitter at @hirocgroup or email us at email@example.com. We'd love to hear from you.