In conversation with Ophthalmology Times Europe, Anat Loewenstein, MD, illuminates the outcomes of using an AI-based fluid quantification for remote retina monitoring and analysis.
At this year’s virtual Angiogenesis, Exudation, and Degeneration conference, retina specialists updated a worldwide audience on the latest advancements in the field. Anat Loewenstein, MD, Professor and Director of the Department of Ophthalmology, Tel Aviv Medical Center, Israel, focused on home monitoring with OCT. In this video, Prof Loewenstein illuminates the surprising outcomes of using an AI-based fluid quantification for remote analysis.
Editor's note: This conversation has been lightly edited for clarity.
Hattie Hayes: Hello, I'm Hattie Hayes, editor of Ophthalmology Times Europe. Today, I'm in conversation with Professor Anat Loewenstein, a speaker at this year's Angiogenesis Symposium. Her presentation, on longitudinal AI-based fluid quantification and its implications for remote patient monitoring, is our topic of discussion today. Professor Loewenstein, thank you so much for joining me. Would you please give us a brief overview of your presentation?
Anat Loewenstein, MD: Thank you very much, Hattie, it's a pleasure for me to give you a few points of interest and ideas from my presentation at Angiogenesis, which [is] based on fluid quantification on a longitudinal basis and how it can help remote patient monitoring in the management of our patients with macular degeneration. So, we know that when we look at OCTs, it allows us actually to visualize the retina very well, but we are used to looking at it as a three-dimensional parameter. Now, when we look at it over time, in a patient that is doing daily home OCTs, we have another parameter, which brings it almost to be in four dimensions.
So it gives us, really, the disease dynamics and the dynamics of response to treatment. And I've been talking about the home OCT and fluid volume classification and analysis for quite a while, from very many aspects. The topic of presentation during Angiogenesis is to actually categorize the fluid volume trajectories of patients with new vascular macular degeneration that are using the home OCT. So, my topic at the Angiogenesis meeting results from a post-hoc analysis, from a retrospective analysis, that included almost 300 eyes. These patients performed almost 10,000 self-images of OCT during the study. The patient's task, during all our studies, is to perform daily OCTs, and then the scans are automatically transmitted and secured on a cloud and analysed by the AI-based Notal OCT analyser, that we term NOA. And the analysis is submitted, [in] a report, to the physician and to the patient. It includes the raw B-scans, just the B-scans, the calculated fluid volume, the fluid volume maps and the fluid volume trajectories, and also gives an alert once a preset threshold was met. So I'm going to detail a little bit about this.
You can have exactly the fluid type. You can see if it's subretinal or interretinal. You can see it on a fluid thickness map and you can really categorize the fluid exactly where it is and how it looks. So the presentation at Angiogenesis shows almost 300 trajectories that were done in patients with neovascular macular degeneration. The follow up in this particular study is not very long, because it's basically aimed to look only at the fluid volume and what's happening to it.
We were able to identify four categories. One, that all during the period of observation, there was no fluid activity. The second, category two, was that there was recurrent fluid reactivation, but in between, there was complete resolution. In category three, there was some reactivation, and not complete resolution. In category four, there was never a good resolution. And so it is a very interesting way to look at what's happening to our patients. We were able to show the differences in visual acuity. We were able to show that there is a worse vision in eyes that had more fluid, and this is something that we know, but now we can really analyze it by the amount of fluid that exists. Also, there was a trend towards longer time of disease in eyes that had the lack of complete resolution. So just showing the importance of fluid in the in the way we manage our patients.
We can think, maybe, that according to the different categories, we can also differentiate the way we manage the patients. For example, in a patient that has no activity during all the time of observation, we will continue the monitoring. In the patient that there is lack of complete resolution, maybe sometimes you can follow it up and observe, and only when the fluid flares up, to treat. So, many such things. The presentation includes also case studies that show very nicely how these examples can happen: with the B-scans, with the trajectory and with the fluid thickness maps, the way that it responds to treatment and recurs, and when does it occur.
The aim of the presentation is actually to show how analysis of the fluid volume according to time, the temporal fluid volume dynamics, allows for categorization and maybe also decision or thoughts on how to treat the patients. In the past, we looked at classifying subtypes of CNV. Maybe now we can look at subtypes of fluid behavior. And because we have so many OCTs on each patient, this is something that can be very helpful in the follow up and management of our patients.
Hattie Hayes: That's wonderful. Thank you. Do you think there's anything in these data that your audience might find surprising?
Prof Loewenstein: I think, whenever I'm talking about home OCTs, some of the audience is surprised at how our elderly patients are able to actually perform the tasks of doing home OCTs. I know it's surprising, but more than 90% of the patients can provide gradable images, and they continue to do so even on a long term. They continue to do almost six or seven scans a week, so almost every day. So this is one thing that is surprising. Another thing that maybe was a little surprising to me, was the sensitivity of the AI system is higher than that of what exists in between retina specialists. So it's very sensitive and accurate. I know that we all know what AI is and how helpful it can be. But it's still always surprising to see how much it does better than us.
Hattie Hayes: That's great. And finally, what are the key points you hope your colleagues will take away from this research?
Prof Loewenstein: I think that the point that people might take away from this presentation is that nowadays, we are dealing with longer intervals between injections and between times that we have to see our patients. There is a way to continue to monitor them and not to be surprised by things that can happen in between the 3 months, sometimes 4 months, that we will not see the patients between injections. So, I hope that it will be clear that using home OCT with the temporal dynamics of the fluid realization will help the thoughts that the retina retina specialists might have. How can we not see our patients with AMD, for example, for such a long time between injections.
Hattie Hayes: It sounds like such promising technology for both patients and retina specialists. Thank you so much for speaking with me.
Prof Loewenstein: Thank you, Hattie.