🎤 Listen to the full podcast segment on Mon Carnet by Bruno Guglielminetti
Émilie Delvoye, Director of Communications at Prompt, welcomes Dr. Ismail Ben Ayed fromÉTS and Dr. Hadi Chakor from Diagnos, who have designed a solution to diagnose diseases with AI.
Discover the podcast, in full text
Émilie: Today, I’m joined by Dr. Ismail Ben Ayed, full professor at ÉTS in the Systems Engineering Department and holder of the Chair in Artificial Intelligence in Medical Imaging, and Dr. Hadi Chakor, researcher in retinal microcirculation and Medical Director at Diagnos, a Quebec-based company that develops artificial intelligence tools to help healthcare professionals detect critical problems. Today, you’re going to tell us about your collaborative project to develop deep learning algorithms for retinal image analysis, enabling diagnoses of eye diseases and cardiovascular disease. Would you please tell me what triggered your collaboration and the development of this innovation?
Hadi: Thank you for hosting us today. We met in 2018. We met Professor Ismail during a conference on innovation. We’ve been collaborating ever since. I remember at that time in Diagnos, it was at the very beginning of deep learning algorithms, and our performance was really very low. We saw after this meeting and after working together, after less than a year, that we had increased our performance by 30%. I’ll let Ismaël explain this further.
Ismail: Those were the days, when deep learning was starting to show incredible performance in everything to do with automatic image analysis by machines. And we were approached. At the time, Diagnos was using conventional detection algorithms, and wanted to take advantage of our technical team’s expertise in developing algorithms based on neural networks and deep learning. So indeed, for the detection of diabetic retinopathy severity, we increased performance by perhaps 30% at the time with deep learning algorithms. Obviously, this is very important for the company.
Hadi: For the company and also for clinicians, because clinicians when they also start to see the impact of such algorithms on patient triage, especially triage by severity. The impact is great and they’re really going to adopt the technology.
Émilie: Of course. And we don’t have enough doctors, do we, in relation to the needs?
Hadi: Absolutely. We really see it, especially for systemic chronic diseases in general. For a diabetic patient, generally speaking, he has to see an ophthalmologist for follow-up. So, if he’s not at a really severe stage, it’s very difficult, it takes more than a year to get an appointment. With algorithms based on artificial intelligence, we’ll be able to do this triage much more easily, in an optimized way. We’ll be able to do the follow-up. And we only send real patients who need to be seen by an ophthalmologist.
Émilie: What other impacts does your solution have?
Ismail: Perhaps what also needs to be explained is that algorithms or machines are capable of extracting precise information that cannot be extracted with the naked eye. For example, microvascular structures can be precisely extracted, along with their geometric measurements. This is very complete information that can help clinicians and increase their productivity. This is a very, very, very important aspect. It has an impact on practice. And of course, algorithms that perform well, and algorithms that give clinicians more information, facilitate real-life adoption, and facilitate not only the commercialization of Diagnostic tools, but also acceptance by health regulatory agencies, and so on, and the large-scale adaptation of these algorithms. That’s an impact on health. Obviously, this project also has an academic and scientific impact. It has also strengthened Quebec’s and Canada’s position in the promising field of AI for health, and has attracted and trained talent, students, post-docs and researchers, and so on. Also, to publish scientific publications in prestigious journals. There’s that aspect too. But for us, as computer scientists, beyond the computer aspect and beyond the virtual aspect, we also like to do work that has an impact on other fields. In this case, we’re talking about the strategic field of healthcare, so it’s real life, not the virtual computer world.
Émilie: Can we say that we’re saving lives through the prevention you’re able to do, and then taking on more patients?
Hadi: Absolutely. It makes it easier to treat more patients in an optimized way, of course, and also to avoid complications and blindness. And also save lives, because we’re currently working on another project, which is hypertensive retinopathy, for example. We know that the impact of hypertension is seen directly on the microcirculation, and with tools like these we can assess details that we’re unable to assess with the naked eye. Objectively, it’s unthinkable. And to have this expertise right beside us, because we didn’t think it was there at ÉTS. Every day, we work on this close relationship between engineering and medicine. This synergy is what’s great.
Émilie: What I found interesting was that, during the preparation, you also mentioned that the eye was a place where you could detect an enormous number of problems in a non-invasive way. Isn’t that so?
Hadi: Absolutely. We ourselves have published on this, and other teams have published on the fact that the eye is an open window to the microcirculation and to other diseases. We now know, and this has even been confirmed by studies, that the eye is really a tool for early detection of all cardiovascular events, so to speak.
Émilie: Hence the importance of your work.
Ismail: There are complex biomarkers such as vascular structures that can be detected by machines and not by the naked eye.
Émilie: Excellent! Thank you so much for your beautiful work and for sharing it with us.
Hadi: Thank you for inviting us.
Émilie: If you too have great projects or innovative ideas that you’d like to develop with partners, don’t hesitate to contact Prompt, either to be put in touch with the right partner, or to find out about the support instruments available for your project.
