News
Article
Author(s):
OCT-A angiography can help reveal early neurodegenerative disease signs, offering a non-invasive method for identifying high-risk patients through retinal analysis.
(Image credit: AdobeStock/Pixel Matrix)
Amir H. Kashani, MD, PhD, is a professor of ophthalmology at the Wilmer Eye Institute in Johns Hopkins University in Baltimore, Maryland. In an interview with Ophthalmology Times, Kashani discussed a research paper on OCT-A angiography's potential in detecting early signs of neurodegenerative diseases like Alzheimer disease. The study aims to establish a consensus among researchers on analyzing retinal blood flow data. By examining capillary-level changes in the eye, the research offers a non-invasive method to identify high-risk patients for cognitive impairment. The work highlights the potential of AI and machine learning in processing complex medical imaging data to improve early disease detection.
Note: This transcript has been lightly edited for clarity.
Ophthalmology Times: You recently had a very interesting paper published. Can you give the name of that paper and the journal that it was published in?
Amir H. Kashani, MD, PhD: We published in Alzheimer's and Dementia, and the the topic of the publication was using OCT-A angiography data in epidemiologic studies, especially as it relates to Alzheimer's and neurodegenerative diseases. So it's a very exciting and growing field, and I think it was a timely publication.
OT: Can you talk a little bit about how OCT-A is achieving this sort of biomarker look in the eye for Alzheimer disease and other conditions?
Kashani: So OCT-A, first of all, for and I think most people are starting to become familiar with it, but for anybody who isn't, OCT-A is a really excellent modality to measure capillary level perfusion in the eye. Because of the optics of the eye, we can, we can do it non invasively, which is hard to do in the brain, because you have to penetrate the skull with whatever modality. So things like MRI, PET scans, they just can't measure capillary level blood flow. So with advent of OCT-A we can now measure blood flow in the retina, and the idea is that it's a proxy for perfusion in the CNS, in the intracranial space. This has become a very widely accepted idea that at least some aspects of pathology that occur in the brain for Alzheimer's disease, and perhaps for other diseases like vascular cognitive impairment, are reflected in the tissue of the retina. The question is, can we capture a picture of this, literally, in the retina, and hopefully tell us something more about what's happening in the intracranial disease process?
OT: In conducting this research, how was the study designed to analyze these photos and this data?
Kashani: The study that we published was really a first attempt at gaining consensus among a lot of stakeholders. So OCT-A has been around for about 10 years, and we were involved in the initial kind of FDA approval process with one of the main manufacturers and initial utilizations. So it hasn't been around that long. A lot of people have collect collected data, and one of the things we found is that it's a little bit tricky how you use those OCT-A data and how it's acquired. That's where really this consortium started to come together, is to try to get a handle of, "how do we do these analyzes," so that one site in Europe and another site in Asia and the US, we can all get consistent interpretations of our data and make sure we're using the best data possible.
That was the, really, the motivation for this paper. What we did is we had a group of 7 or 8, you know, universities which are running really large population based studies and who have collected thousands and thousands of data sets from OCT-A, and we tried to get everybody around the table and say, "what is the best way of analyzing this data so that we don't have these differences?" Because right now, there's a lot of differences between published studies published by group X and group Y. That was really the goal of this particular study, is to get some consensus in terms of study design, analytical methods, and how do you use good quality data to generate the results that you want?
OT: Taking it one step further with more of a consensus amongst different stakeholders in the area, in the field, how might this apply to future technologies, whether we're talking about AI or whether we're talking about a large language model in the space?
Kashani: I think it has huge implications for how we collect data, because, one of the things that we've learned from one of the study sites, which was the Framingham Heart Study, was that, when you collect data from both eyes of an individual, we don't necessarily see the exact same correlation in the data between 2 eyes. That could be because of the way the data is collected, or it could be because of biological differences between blood flow to each eye. We're not absolutely sure about that. So a lot of times, people will a conductive study, and they'll dilate 1 eye, because they don't want to inconvenience the patient. Or, for example, in neurologic studies, the patient has to perform another task after the eye test. So if you dilate both, it can't be functional. It turns out that that may be an important variable. So if you just randomly dilate somebody, rather than systematically doing it, you you may be introducing a bias one way or the other, or you may be missing something. So that's one thing we found at one of the sites.
I think the other things that we found is that you know how you collect this data, in terms of the size of the scans and how much of the retina you're looking at, makes a big difference in how we analyze it, because there's small vessels and there's large vessels, and there's small vascular disease and large vascular disease. Those things, I think, had been a little bit overlooked in the past, but I think that it's coming to to the forefront.
In terms of AI, the obviously, the data sets that we're collecting, they're huge data sets, and these are large volumes, and we're just starting to get an appreciation for how these capillaries in the retina connect to each other and connect to large vessels. The role of AI and machine learning in analyzing these vascular patterns, I think, is going to be very important. We've already started developing software of remotely assess the quality of data at these different sites. One of the things that came through was that, you know, site X has 10,000 images that they need to review, and it's a very labor intensive process, because you need somebody who's experienced in reading these images, and often times, it's clinicians who are doing it, but they don't have time to read 10,000 images.
So one really active application of AI right now is to develop a method of looking through 10,000 images, picking the best ones for the analysis, and then leaving the rest as a secondary analysis, or saying, well, these may not give us as reliable results. That, in and of itself, is a big step forward, because all the sites were saying, we don't have a really great way of doing this.
OT: For the average ophthalmologist, or someone who's conducting these scans, looking at these scans on an individual level, what takeaways would you have for them as they start to incorporate this kind of research down the line?
Kashani: I think this is really exciting, because I think it's putting ophthalmology perhaps in the driver's seat, or at least in a position where we can really help our colleagues in neurology and even other specialties to risk-stratify patients. So I'm not so sure we're going to diagnose Alzheimer disease. I think our neurology colleagues are very good at that, but they're not very good at saying who is going to develop Alzheimer disease in the next 10 years, and who should we send for MRI testing or cognitive testing or PET scans, because each one of those tests is hard, much harder to do than OCT-A, which takes about 10 or 15 minutes, and for which you don't even have to dilate somebody. So MRIs are a couple hours and cost is very high. Cognitive testing requires a very experienced neuropsychologist to sit down an hour and a half of discussion back and forth. Language can be a barrier if the patient doesn't speak. You're not going to get good cognitive evaluations. PET scans require radiation to the body.
So, the issue is, how are you going to identify all these people, to screen them. The other real practical implication is, when you want to do a clinical trial to try to find a treatment you want to find a group of people who are at high risk for the disease, and you can't do MRI on thousands and thousands and thousands of people. It's just not feasible. So hopefully we can provide a method for our neurology colleagues to say, hey, out of these 10,000 people, this 1000 are the highest risk group based on the scans that we've done. Let's try to focus on them and expand our really valuable resources on those groups of people.
OT: Are there any other areas you want to speak to, or anything you feel didn't get mentioned?
Kashani: It's really exciting to be able to look in the eye and assess capillary level changes in people who don't necessarily have a problem right now, but are on their way to developing a problem. We've done this now in a couple of different studies before this consortium, and I think that's where some of the interest is starting to arise. You know, when we look at people who have, for example, diabetes and hypertension, we notice that there's an increased chance that they have fewer capillaries or fewer areas of perfusion compared to people who are not diabetic and hypertension, and that may be reflecting an intracranial process as well. Similarly, we found people who have specific kinds of genetic mutations, even though they have no visual symptoms or cognitive symptoms, tend to have lower capillary density in their retina, which might mean that they're more predisposed to developing cognitive impairment in the future, whether from vascular disease or from Alzheimer. I think we're providing very useful information by looking at these subtle, but really important changes, in the eyes of people and trying to correlate them with their clinical and cognitive status from a neurology standpoint.
OT: Is there an age range with this kind of consensus looked at, or was it, you know, were we looking at a specific population of retirement age or younger?
Kashani: I think in general, cognitive impairment tends to be a problem, primarily in the elderly population. Alzheimer disease is the one we all traditionally think about, but stroke and vascular cognitive impairment, which is kind of not the typical stroke that we think of, but more microvascular, chronic stroke are all really big problems, and it is very hard to distinguish those categories among the older groups. So that's primarily where we're interested, but we don't really know if this has applications earlier on, and it would certainly be interesting to study patients earlier to see if earlier causes of cognitive impairment are associated with it. I think we're we're pretty sure that we can pick out the vascular component of a disease process that is contributing to cognitive decline. The question is, how do we implement that in our studies, reliably and repeatedly, so that everybody can get the same results?
Don’t miss out—get Ophthalmology Times updates on the latest clinical advancements and expert interviews, straight to your inbox.