Deep learning algorithm performs similarly to manual grading of GA on an OCT data set
Hasenin Al-Khersan, MD, of Retina Consultants of Texas, spoke with Ophthalmology 360 about an AAO 2025 study that found AI algorithms could accurately measure ophthalmic lesions from OCT scans.
Hasenin Al-Khersan, MD:
My name is Hasenin Al-Khersan. I’m from the Retina Consultants of Texas. The work I’m presenting on is related to geographic atrophy. It’s great that we have new therapies available for this disease. However, it can be very difficult to measure geographic atrophy lesions in clinic. What we looked at is designing an AI algorithm that does this in an automated fashion. Now, this is not the first time that this has been done, but what we did is we used a real-world patient cohort. These are patients taken from real clinics. That means they had not only geographic atrophy, but also wet macular degeneration, and they had images that were taken from multiple devices.
In that case, this type of study had never been done before. What we were able to show is that these AI algorithms, in ours specifically, was able to accurately measure lesions, not only in patients who had just geographic atrophy alone but also if they had neovascular macular degeneration as well. Also coming from multiple devices. We believe that this has real-world implications and that this can be a tool we can deploy in our clinics to help our patients and to guide our therapy decisions as well.
Related Content
