The Health Pulse

[MUSIC PLAYING] GREG HORNE: Hello, and welcome to another episode in our brand new series of The Health Pulse podcast. I am your host, Greg Horne. And in season two of our podcast series, we will be focused on health innovation, and looking to uncover where technology and new approaches will change the world of health and life sciences.

So as you know by now, we are producing the podcast in two formats. And if you've been an audio listener, then I'd like to suggest that you also check us out on the SAS software channel on YouTube. And, of course, we always welcome your questions and comments on both the YouTube channel, and at our email address, which is thehealthpulsepodcast@sas.com.

So for this episode, I'm going to be joined by my guest, Dr. Patricia Maguire. And Patricia is a professor at University College Dublin. And she is the director of the UCD Institute for Discovery. So welcome, Patricia. Thank you for joining me this morning on the podcast. Can you just tell me a little bit about yourself and your background, and what you do at UCD?

PATRICIA MAGUIRE: So thank you so much, Greg. It's a real honor to be here today. And so I am a biomedical scientist. I studied science in University College Dublin. And while I was doing my PhD in Muscle Biochemistry, my father actually took his very first heart attack. And I got very interested in platelets and blood clotting.

And also, I suppose, from the perspective of what can go wrong when platelets come together and form a clot or a thrombosis, which leads to so many heart attacks and ischemic strokes around the world. So I got really interested in platelet biology, and that's where the focus of my research lies these days.

GREG HORNE: Fantastic. And your role at the college now, so tell me a little bit about that and what you do today.

PATRICIA MAGUIRE: Yes. So I am a professor of Biomedical Science, and my research is focused on platelets. And I'm sure I'll speak more about that to you, Greg. But I'm also the director of the UCD Institute for Discovery, which is an institute that is focused on-- I suppose our mission is to drive forward interdisciplinary research in our university. And we really believe that when people come together from different disciplines, the magic happens.

GREG HORNE: Fantastic. Yeah, I'm a big believer in that myself. And Patricia, I'm assuming you're in Dublin today. Is that correct?

PATRICIA MAGUIRE: I am, absolutely. I'm in very sunny Dublin, Ireland today. And it is now currently around lunchtime.

GREG HORNE: Fantastic. So we're doing an early morning one for me. And here in Toronto it's beautifully sunny, but it is minus 2 degrees. So winter is definitely on its way.

PATRICIA MAGUIRE: Yeah. Same here. It's really cold here today as well, actually.

GREG HORNE: Oh, wow. OK. So Patricia, the other thing we like to do with everybody on the show is find out a lot about a hobby or something you do out of work when you're not biochemistrying. So tell me a bit about that.

PATRICIA MAGUIRE: Yeah, so when I was younger I actually trained as a ballet dancer. And I used to do ballet six, seven days a week when I was in school. So I still love to dance even all these years later.

And a couple of years ago I organized-- in the university, I organized a Strictly Come Dancing competition, which I organized, but also danced in as well. But we raised over $25,000 for charity as well-- has been thoroughly enjoyable event.

GREG HORNE: Fantastic. So I do like to dance, though I have two left feet and trip over constantly. But that's a brilliant thing to do. And the Strictly Come Dancing thing sounds like a lot of fun.

All right. So let's get back to the subject in hand. So you're using analytics at the very core of your research. So tell me about why did you take that approach, and how does that play out in the real world?

PATRICIA MAGUIRE: Yeah. So I have to kind of take a step back for a minute and tell you a little bit more about platelets and where this all came from. And it actually goes back to my dad as well.

So my dad's ill health journey started 25 years ago. And I saw-- I suppose living that with him I saw the need for better diagnostics in the clinic. And that's where I suppose-- so platelets are really fascinating. So they circulate around in our blood.

But the really understudied thing about platelets is that as they circulate around in our bodies, they pick up information in real time. So there are really like a sore inside as a ready source of biomarkers. And that's what my lab has focused on for the last 20 years, is getting this information out of platelets.

And we've generated a lot of data in that process. And that's made me incredibly passionate then about trying to intersect all of the data that we've come from from a biomedical science perspective, and bring that together with machine learning and analytics.

GREG HORNE: Very interesting. And I understand as well that you're using Python in your work. How's that working for you?

PATRICIA MAGUIRE: Yeah. So we're using-- we're coding in Python. We are coding in R. And we're also coding in SAS. So we actually use the SAS Viya platform to bring together all our data. And we've coded in every language. And we're using SAS. So we're using SAS Viya platform on Microsoft Azure.

GREG HORNE: Oh, brilliant, which is kind of the future for us in health care, really. Just a quick spin-off on that one then. You have, I guess, some personal data in that. Did you have any problems ethically in using a cloud to deploy your software?

PATRICIA MAGUIRE: Yeah. Actually, so everything-- so we're in Europe, so everything is GDPR-compliant. So one particular project that we're working with is we're working with the three main maternity hospitals in Dublin and Ireland. And they're really busy, busy hospitals with about 12,000 each live births every year.

So you have to go through a lot of ethical approval to do these studies. You have to be incredibly GDPR-compliant. But we worked really closely with the hospitals in terms of being GDPR-compliant and making sure that the platform we're using is. So it was just-- I suppose bringing the right people together has allowed this project to accelerate the way it has.

GREG HORNE: Yeah, that's very good to hear. Because it's something that we hear a lot about is this kind of move to cloud being very difficult. So for our listeners, let's get into our project.

So you've got this wonderful AI stack in your university. You've got all your knowledge of platelets and the like. So how are you now translating that in something real that we can kind see and feel?

PATRICIA MAGUIRE: Yeah. So a number of years ago, we started using this idea of platelets as a ready source of biomarkers. We began using that in a project on preeclampsia, and we found new diagnostic markers for preeclampsia So now I have to reverse a little bit and tell you a little bit about preeclampsia. So preeclampsia actually affects one in every 10 pregnancies. And--

GREG HORNE: Wow.

PATRICIA MAGUIRE: Yeah, It's huge. And it's something that unfortunately pregnant women sometimes hear about too late. And the issue is globally. And these are probably an underestimated figure, because they're not really categorized in lower income countries.

About 50,000 women and 500,000 babies die every year because of preeclampsia. And not only that, because the only cure right now for preeclampsia is delivery of the baby, about another 5 million babies are born worldwide prematurely every year. And that comes with huge, huge issues for that baby, in terms of especially neurological issues for that baby, especially if they're born as early as 25, 26 weeks of pregnancy.

So right now there is no diagnostic test for preeclampsia. Preeclampsia-- there's lots of rule-out tests. But preeclampsia is still diagnosed the same way as it was diagnosed in the 1960s. It's still diagnosed by reading blood pressure and looking at proteinuria. So there has been really no disruptive technology in that area at all.

And with the figures that I've said, the amount of women and babies dying, but also this affects a lot of pregnancies-- one in every 10 pregnancies. And it can affect women in so many different ways. It's not just one disease. It's like umbrella disorder.

So to actually find a way to diagnose preeclampsia, you need to use, I suppose, lots of information, to give the clinician all that information together in one way. And that was what we set about doing. So this is a project, the AI_PREMie Project.

So what we're doing is we're bringing in the diagnostics that we have from blood that we've gotten from our knowledge of platelets, and we're putting that together with all the information that is available on that woman throughout her pregnancy journey. So all her blood tests, all protein, her blood pressure measurements, everything, we're combining all of that information together along, I suppose, with clinician gut instinct. We're building that into the algorithm as well. And so we're putting that all in together.

And you see, you can't do that with standard statistics or standard analytics. You need to have machine learning involved in that to actually unbiasedly pick out what are the main markers that are driving the diagnosis of preeclampsia in that algorithm.

So because it's such a-- because preeclampsia, I suppose, as I said, is such a difficult, complicated disease or diseases, you need a difficult, complicated way of analyzing the data, and that's using, I suppose, machine learning and AI to be able to pull out the right information and the right knowledge at the right time.

And so the goal here is to be able to feed that directly to the clinician. So when the clinician needs that information, they're making the decision to diagnose preeclampsia. But what also we have built in is when they should deliver that baby.

And that's huge, because if they can-- every hour for that baby inside, in utero, counts. So if they can leave that baby in for another couple of days, a week, that's going to make a huge, huge difference to that, to basically the quality of life for that baby and its family.

So being able to deliver and categorize and bring all that information together, there's multi-modal information, if you like, to gather and deliver it to the computer screen of the clinician at the right time. It's almost this idea of augmented intelligence, that you're giving them all that information. So that's the dream. That's the goal.

GREG HORNE: Well, you mentioned augmented intelligence. That's something I talk about a lot, actually. And the idea that people shouldn't be concerned that AI is going to replace their doctor, but the doctor using AI is going to be far more efficient and far better at diagnosis than the doctor who doesn't use it.

So I kind of want to think then about the next step. So how do you collaborate with other universities, and with industries as well, to kind of take your idea and start to grow that into a real-world use case?

PATRICIA MAGUIRE: Yeah. So right now we're-- as I said, we're in the three Dublin maternity hospitals. So we're covering 50% of all the births in Ireland. So yeah, I mean next steps for us is to be able to bring that nationally and internationally. And it's working to-- really working-- we're-- right now we're funded by an Irish government funding agency. So we're in a kind of a challenge societal good funding.

But you can really see if industry, the power of industry come in behind this, to be able to deliver this product to every woman who needs this. Because really, what's our goal here? If I kind of-- and my future dream on this AI_PREMie project is so that-- and everybody who's involved in this project. It's not just me as the team lead.

Everybody that's come in on this project really wants to get this algorithm or this knowledge out into the real world, into production, that we can actually make a difference, and hopefully save some of these lives. I mean, so having industry behind that is huge, because we'll never be able to do that, I think, by ourselves.

Because if you look at how algorithms that have come out, or working AI that's come out of universities, there's like a chasm there. There's like a-- they all-- sometimes a lot of them fall down to actually get into production. So this is a huge, huge issue. And to be able to put that into production would be a dream for us. Because it would absolutely save lives.

GREG HORNE: I think that's why I like what you're doing so much. And we talk about University of Alberta having the same kind of setup as you've got. Because I've been to so many events now where a researcher will stand up and say, I've come up with this wonderful example of how we could save lives in the real world. And now I finished my PhD. I'm going to put that on a shelf, and it gets forgotten about.

And we see all the complexities of software as a medical device, and we see all the other issues coming out of it. But at the core, we need to translate things into the real world. So from your perspective, how is that process? How was it to get the hospitals to adopt this in the first place? Was it a difficult thing for them to start using it?

PATRICIA MAGUIRE: I mean, we have champions. And I think the real thing here is bringing people together. And I'll keep going back to that. It's the team. It's the multidisciplinary bringing people together around a cause. I think that's how this has worked so well.

And then just-- you've got to do it in a very GDPR way. And we had to make sure that the data was done-- we did everything properly with ethical approval. But people jumped on board on this. And hospitals jumped on board. So it's working really quite well, actually.

GREG HORNE: And when you think about things that are blockages, and things that cause barriers to successful change in health care, what do some of those barriers look like? And you've already mentioned some of the things you've done to overcome them. But in general, when other people are looking at this, what are the barriers they're going to come against?

PATRICIA MAGUIRE: Yeah. I mean, the first barrier I can think of is this chasm between all this university research, where there's incredible research going on. But as you say, it just sits on a shelf and it's never kind of-- that gold and diamonds that's in there are never extracted to make a difference in the real world. And that's something that I'm incredibly passionate about is getting these algorithms into production.

And I think you need something really, really steadfast, and really that can actually do this. So software-- and that's the reason that we chose SAS Viya. That's the reason that we chose to do this in a cloud scenario using Azure, is that these stand up in the real world. So that's the first thing is the chasm.

The second thing really is almost, and it's another passion of mine, is this idea of the democratization of AI technology to basic researchers in a university. So I know personally, I come from a biomedical science background. And we did very basic statistics, very basic analytics. And to access the knowledge, it was really quite difficult.

So I'm incredibly passionate about democratizing AI and sharing that knowledge and sharing-- being able to share a platform access into something like SAS Viya is incredibly powerful for somebody that they can code in R, or they can code in Python, and they were able to use this incredible software to completely navigate the cycle and get that algorithm into production.

I mean, that's key. I think-- so those two things-- the chasm and then the democratization of AI will lead to hopefully a lot more translation of a lot of the incredible work that's going on in universities globally, and get it into the real world, that'll make an absolute difference to people's lives.

GREG HORNE: Absolutely. I couldn't agree with you more. I think that's a really important thing for us to start looking at. How do we cross that chasm and get people to do this in the real world?

So let's think about now, as we kind of come towards the end, the future. So what is your future aspiration, both on a personal level for your projects and the things that you're working on, but also in a broader sense? What do you think is the future for AI? And on the timescale, like, is it six months, a year-- what-- play out the timescale for that as well.

PATRICIA MAGUIRE: OK. So I'll start with my personal view. So I mean-- and I suppose I've spoken a lot about our preeclampsia project. And personally, and I think it's personal for me, but also for everybody that's involved in this multidisciplinary team, is we really want to get our AI_PREMie test to every woman who needs us all across the world. And that would be the dream. And if we can do that, then we will save lives. So that's I suppose my personal dream.

But in terms of the AI in health care and where this could go, I mean, if we can get this, if we can get AI-PREMie up and running in every hospital environment globally, there's a real potential there then to completely disrupt diagnostics in health care globally.

We're using diagnostics that have come from our research. But we could use any diagnostic. If you have the formula, and you've got this working, and it works and it stands up in a real-world environment, then we can use different diagnostics, and maybe slightly different patient characteristics. And then translate this into any disease.

And that's, I suppose, the incredible thing here is that the potential, if you can-- because I mean, I looked actually, and there is-- I cannot find an example of algorithms that have come from universities that are actually working in the real world. I mean, wearable devices come the closest. And there is some kind of AI in robotics and surgery, where they-- but it's still surgeon-controlled AI.

So imagine if we can actually have-- we could completely disrupt diagnostics globally by actually having a solution that brings in all the information that's available. Because when you think about it, Greg, all this information is available. But it's a matter of amalgamating it and bringing together in a really useful way. And it's-- again, it brings me back to that augmented intelligence.

And timelines-- I think if we got-- if you can do it for one project, then why not do it for thousands? And that's-- so the timelines on that are really just, I suppose, based on funding and imagination.

GREG HORNE: But it seems to me that, particularly your preeclampsia work-- and sorry, this just came to me as you were as you were talking there-- particularly your preeclampsia work, it seems to be a very low cost against a very high reward.

And so I would be very hopeful that folks would see it that way, and you would start to see a rapid uptake. Does that ring true? Is it low cost to high reward? It seems to be what you've described.

PATRICIA MAGUIRE: Yeah, I think so. I think so. I mean, because we're cloud-based, but you could even do it locally and just have the machines speaking to the cloud intermittently. I mean, this technology exists in the clinic. So yeah, I think it could be easily achieved, yeah.

GREG HORNE: Fantastic. Well, thank you, Patricia, for all those insights and the discussion today. I think it's been really interesting. And we're going to throw it over to the listeners and the viewers now and our audience. So please email your questions to thehealthpulsepodcast@sas.com. Please add comments on the YouTube channel as well.

I'm really interested on this one to hear your views on this democratization of AI, and how do we get people who are not researchers, not statisticians, to understand the value of augmented AI? So let's get those comments in as well. Because this is the sixth and final episode in the current season. And we're going to be looking to put together a new season for next year early 2022.

So those comments and the likes do help us pick guests and to formulate where we go next with the podcast. So please keep them coming in. Please remember also to subscribe and share the podcast, either through your podcast aggregator, or through the YouTube channel.

And just to wrap up, I've been your host, Greg Horne. So this has been the sixth episode, final episode, in season two. And I hope you've enjoyed listening and watching as much as kind of I have in making it. It's been a really good, fun thing to do. I want to thank you for joining us today, and look forward to bringing you another season hopefully early in the new year. So thank you very much. Bye.

[MUSIC PLAYING]