Yari Has Questions ...
May 24, 2024
10 min watch
Save
Yari Mitchell has questions for Jay S. Pepose, MD, PhD: Expanding eye care’s scope with AI
Transcript
Editor’s note: This is an automatically generated transcript, which has been slightly edited for clarity. Please notify editor@healio.com if there are concerns regarding accuracy of the transcription.
Hi, I'm Yari and I have a lot of questions, and this is Dr. Jay Pepose. He has a lot of answers. Let's see if they match! Hi Jay, thanks for joining me today.
Pleasure to be here, Yari.
So, I'm wondering there's so many burning questions I have for you and the top of mind is AI. AI is such a buzz word and I'm wondering just in general: are you using AI in your daily life at all?
Well, I think AI is just really penetrating ophthalmology now and it's fantastic really, because I think it helps us in so many ways. It helps us in diagnosis. I think it's going to help us in screening patients. And I think it's going to help us in really being better doctors if we embrace it. You know, I look at it like years ago, before we had Google searches, right? Google searches and Pub Med enhanced our ability to practice medicine. I think the same thing will happen here. We're already seeing it in many areas, right? It's really good in imaging. Anything that has imaging just lends itself to AI. So, you think there's eight million patients with non-proliferative diabetic retinopathy. How are we going to screen all those patients, right? How are we going to take care of them, and how are we going to know the patients who are more likely to progress and need treatment? So, you know, there's a perfect example where we have AI. Now we have three approved AI devices that do screening, automated screening. And you see it evolve, now that not only are they starting to screen but you're seeing articles coming out about deep learning for prognosis. So it's going to be a paradigm shift.
That's amazing. So how do you think care will be administered if you're thinking ahead, five, 10 years as AI continues to evolve into our space and deliver more information for you and enhancing your decision making. What's your vision for what the practice of the future looks like with AI integrated?
Well, I think we're going to see this evolve more and more. I think first we're going to see AI in the doctor's offices in optometry and ophthalmology. Then we're going to see it in general practice because they're going to want to know who to refer. And then I think it's going to take one step beyond that, where you'll have a patient who'll take their iPhone out and they'll have an app and it'll do a fundus photo and it'll send it to the doctor for analysis. And we're already seeing that for example in retina, like with Notal Vision, we're starting to see patients do home screening, first to see when they go from dry to wet. Now they have home OCT. It's just amazing, and all that information is being sent to the doctor. And there's a good example; So, let's say you see a patient and they get an injection for AMD, and they're not seen again for a month or two, and they're wet again. If you have home screening, you could see: did they get better and then get worse, and if the drug really worked, or maybe the drug never worked, and they need to be on a different agent. So, you could see how that improves the quality of care and becomes more personalized medicine.
Yeah, powerful. It's going to change the way care is delivered, bringing it to the patient, and then helping the treatments to be more effective sooner. You're able to intervene sooner. Very cool.
Oh absolutely. Anything that requires multimodal imaging or lots of information. So, here's another example: Keratoconus. You know, we have topography, we have tomography, we're starting to get biomechanical data, right? Genetics, environmental things. Do you rub your eye? Do you sleep on one side? Do you have sleep apnea? I mean, how do you put all of that together? AI puts it together for you. So, you can enter all that information and AI gathers that. And even in the office as you're talking to a patient, it could be doing your EMR, right? It could be doing your coding. It could be in the background giving you a differential diagnosis. It can just enhance the way we care for patients in so many ways. And I think ultimately it will make medicine more efficient, because right now, all of us are feeling this crunch. You know, we can't see that many patients. We can't do it efficiently. Here's an example, it's a great example. The last meeting, I went and saw a device — it's still in prototype — and I put my head in this visor and the device started talking to me, asking me questions like “Look to the right, Look to the left,” and it asked me “How many spots do you see?” I realized in that maybe 40 seconds, it did a visual field by asking me how many dots I saw.
Very cool.
And by the time I was done with this whole process, and as I spoke, it said, “You can speak softer, you don't have to speak...” It started talking to me back! And I realized, I asked them why, and they said “Well, we may have several of these in a room. We don't want you to speak so loudly that it's a HIPPA violation of the person next to you.” And so, a device like this within six minutes did 100 tests.
Amazing.
And it marked which ones were not normal. It did OCT, and also, the device, I was talking to the inventor and he said it could be programmed in a way so that it knows how many patients are in your waiting room at any time. And so if you came in, let's say you're on Plaquenil, and you need certain tests to monitor. I'd say out of the six tests, three are mandatory. If the waiting room is filled, it would do just those three tests and the next time you came back, it would do all six. So, it improves your flow.
Yeah, that's so many ways that it can help. But on the flip side, do you have any concerns? What kind of keeps you up about AI at night?
Well, I certainly have some concerns about you know the way data... You know, AI is only as good as the data that it's trained on, right, and the analysis of that data. We've seen hallucinations with ChatGPT, right? I put in my own CV and it came out with schools I never went to, and-
Oh, congratulations!
Yeah, so you know you have to be sure that the data is right. So it doesn't... And that's where the doctor comes in, right? It's not going to replace the doctor, and I think it'll evolve and get better, but that is of some concern: You have to validate the data.
For sure. Garbage in, garbage out. There you go. So, Jay, I'm also curious: AI is coming, it's going to be in the future. What are you interested in today? What have you integrated recently that you are very excited about?
Well I mean, in the sense we're already using AI. For example, some IOL formulas are based on AI, right? Intraoperative aberrometry has some component of AI. You know, so we already have AI in the background of our current practices, and we start to see publications now of machine learning, keratoconus, comparing the two eyes, right? Things like that that are hard to do by hand. You know, you need something in the background. And I think we're going to start to cross over into other fields with AI. So for example, I think we're going to see someone come in for OCT, there will be analysis in the ganglion cell layer, and it'll predict if you have a cognitive problem through AI. So, it's going to cross over into neurology, I mean, when you think about it, we're a specialty where we're visualizing the brain, right? An extension of the brain. And so a lot of pathology that occurs, you know, with Alzheimer's, and other kinds of disorders like that, probably will have a retinal clue to it: A retinal trigger. And so I think we're going to start to see ophthalmology reengage with specialists that we didn't interact with as much in the past.
I like it. Interesting. So, Jay, I have two totally different questions for you, and then we can wrap up. I very much appreciate your time. So, one is, what is one piece of advice that you had early on in your career that has really stuck with you that you've carried with you as you've progressed?
Well, I think one bit of advice that I've gotten, and that I would pass on is to be inquisitive. I mean, I think that if you're an inquisitive person, and you're interested in learning all the time, first of all, things become fun, and also, it makes it easier to sort of play a leadership role because you're sort of thinking out of the box. I think being inquisitive is really important. Reading about things that are not exactly in your area, and I think most of the paradigm shifts that we've had have been because we've taken a technology that didn't come out of ophthalmology and brought it in. You think about Wavefront, you know, we're talking to astronomers, right? So, being inquisitive I think is really important.
I like it. So, I'm curious, if you were to be an ice cream flavor, what would you be and why?
Probably strawberry 'cause it tastes good.
I love it. I love it. Well, thank you, Jay. I really appreciate you answering my questions. I think your answers matched, and you know have a great day.
Thank you so much. Pleasure to be here.
Hi, I'm Yari and I have a lot of questions, and this is Dr. Jay Pepose. He has a lot of answers. Let's see if they match! Hi Jay, thanks for joining me today.
Pleasure to be here, Yari.
So, I'm wondering there's so many burning questions I have for you and the top of mind is AI. AI is such a buzz word and I'm wondering just in general: are you using AI in your daily life at all?
Well, I think AI is just really penetrating ophthalmology now and it's fantastic really, because I think it helps us in so many ways. It helps us in diagnosis. I think it's going to help us in screening patients. And I think it's going to help us in really being better doctors if we embrace it. You know, I look at it like years ago, before we had Google searches, right? Google searches and Pub Med enhanced our ability to practice medicine. I think the same thing will happen here. We're already seeing it in many areas, right? It's really good in imaging. Anything that has imaging just lends itself to AI. So, you think there's eight million patients with non-proliferative diabetic retinopathy. How are we going to screen all those patients, right? How are we going to take care of them, and how are we going to know the patients who are more likely to progress and need treatment? So, you know, there's a perfect example where we have AI. Now we have three approved AI devices that do screening, automated screening. And you see it evolve, now that not only are they starting to screen but you're seeing articles coming out about deep learning for prognosis. So it's going to be a paradigm shift.
That's amazing. So how do you think care will be administered if you're thinking ahead, five, 10 years as AI continues to evolve into our space and deliver more information for you and enhancing your decision making. What's your vision for what the practice of the future looks like with AI integrated?
Well, I think we're going to see this evolve more and more. I think first we're going to see AI in the doctor's offices in optometry and ophthalmology. Then we're going to see it in general practice because they're going to want to know who to refer. And then I think it's going to take one step beyond that, where you'll have a patient who'll take their iPhone out and they'll have an app and it'll do a fundus photo and it'll send it to the doctor for analysis. And we're already seeing that for example in retina, like with Notal Vision, we're starting to see patients do home screening, first to see when they go from dry to wet. Now they have home OCT. It's just amazing, and all that information is being sent to the doctor. And there's a good example; So, let's say you see a patient and they get an injection for AMD, and they're not seen again for a month or two, and they're wet again. If you have home screening, you could see: did they get better and then get worse, and if the drug really worked, or maybe the drug never worked, and they need to be on a different agent. So, you could see how that improves the quality of care and becomes more personalized medicine.
Yeah, powerful. It's going to change the way care is delivered, bringing it to the patient, and then helping the treatments to be more effective sooner. You're able to intervene sooner. Very cool.
Oh absolutely. Anything that requires multimodal imaging or lots of information. So, here's another example: Keratoconus. You know, we have topography, we have tomography, we're starting to get biomechanical data, right? Genetics, environmental things. Do you rub your eye? Do you sleep on one side? Do you have sleep apnea? I mean, how do you put all of that together? AI puts it together for you. So, you can enter all that information and AI gathers that. And even in the office as you're talking to a patient, it could be doing your EMR, right? It could be doing your coding. It could be in the background giving you a differential diagnosis. It can just enhance the way we care for patients in so many ways. And I think ultimately it will make medicine more efficient, because right now, all of us are feeling this crunch. You know, we can't see that many patients. We can't do it efficiently. Here's an example, it's a great example. The last meeting, I went and saw a device — it's still in prototype — and I put my head in this visor and the device started talking to me, asking me questions like “Look to the right, Look to the left,” and it asked me “How many spots do you see?” I realized in that maybe 40 seconds, it did a visual field by asking me how many dots I saw.
Very cool.
And by the time I was done with this whole process, and as I spoke, it said, “You can speak softer, you don't have to speak...” It started talking to me back! And I realized, I asked them why, and they said “Well, we may have several of these in a room. We don't want you to speak so loudly that it's a HIPPA violation of the person next to you.” And so, a device like this within six minutes did 100 tests.
Amazing.
And it marked which ones were not normal. It did OCT, and also, the device, I was talking to the inventor and he said it could be programmed in a way so that it knows how many patients are in your waiting room at any time. And so if you came in, let's say you're on Plaquenil, and you need certain tests to monitor. I'd say out of the six tests, three are mandatory. If the waiting room is filled, it would do just those three tests and the next time you came back, it would do all six. So, it improves your flow.
Yeah, that's so many ways that it can help. But on the flip side, do you have any concerns? What kind of keeps you up about AI at night?
Well, I certainly have some concerns about you know the way data... You know, AI is only as good as the data that it's trained on, right, and the analysis of that data. We've seen hallucinations with ChatGPT, right? I put in my own CV and it came out with schools I never went to, and-
Oh, congratulations!
Yeah, so you know you have to be sure that the data is right. So it doesn't... And that's where the doctor comes in, right? It's not going to replace the doctor, and I think it'll evolve and get better, but that is of some concern: You have to validate the data.
For sure. Garbage in, garbage out. There you go. So, Jay, I'm also curious: AI is coming, it's going to be in the future. What are you interested in today? What have you integrated recently that you are very excited about?
Well I mean, in the sense we're already using AI. For example, some IOL formulas are based on AI, right? Intraoperative aberrometry has some component of AI. You know, so we already have AI in the background of our current practices, and we start to see publications now of machine learning, keratoconus, comparing the two eyes, right? Things like that that are hard to do by hand. You know, you need something in the background. And I think we're going to start to cross over into other fields with AI. So for example, I think we're going to see someone come in for OCT, there will be analysis in the ganglion cell layer, and it'll predict if you have a cognitive problem through AI. So, it's going to cross over into neurology, I mean, when you think about it, we're a specialty where we're visualizing the brain, right? An extension of the brain. And so a lot of pathology that occurs, you know, with Alzheimer's, and other kinds of disorders like that, probably will have a retinal clue to it: A retinal trigger. And so I think we're going to start to see ophthalmology reengage with specialists that we didn't interact with as much in the past.
I like it. Interesting. So, Jay, I have two totally different questions for you, and then we can wrap up. I very much appreciate your time. So, one is, what is one piece of advice that you had early on in your career that has really stuck with you that you've carried with you as you've progressed?
Well, I think one bit of advice that I've gotten, and that I would pass on is to be inquisitive. I mean, I think that if you're an inquisitive person, and you're interested in learning all the time, first of all, things become fun, and also, it makes it easier to sort of play a leadership role because you're sort of thinking out of the box. I think being inquisitive is really important. Reading about things that are not exactly in your area, and I think most of the paradigm shifts that we've had have been because we've taken a technology that didn't come out of ophthalmology and brought it in. You think about Wavefront, you know, we're talking to astronomers, right? So, being inquisitive I think is really important.
I like it. So, I'm curious, if you were to be an ice cream flavor, what would you be and why?
Probably strawberry 'cause it tastes good.
I love it. I love it. Well, thank you, Jay. I really appreciate you answering my questions. I think your answers matched, and you know have a great day.
Thank you so much. Pleasure to be here.
