BLOG: AI doesn’t know glaucoma but just might figure it out
Click Here to Manage Email Alerts
Artificial intelligence offers incredible opportunities in medicine, helping us cut through petabytes of data to define trends in disease and potentially discover new treatments.
In eye care alone, it has offered to make diagnosis, measurement and tracking of diseases like macular degeneration much more precise and less labor intensive. In cataract, it has already helped us develop formulas that promise greater accuracy, and AI-driven robotic systems may soon perform the surgery itself. These are impressive advances, but I believe glaucoma rather than retinal disease or cataract represents the biggest opportunity for AI.
In retina, we have a few main categories of disease that we treat: macular degeneration, diabetes and retinal detachment. In wet AMD, practically speaking and for now, we have only one approved class of medications — the anti-VEGF drugs. Yes, there are several choices in this class and more classes to come. In glaucoma, by contrast, we have seven classes of medications (see if you can name them). While much study has led to preferred trends in using these diverse drugs for glaucoma, how much better could we be with better analysis by AI?
In cataract, AI is indeed helping us optimize our power formulas, but how much better can they be than current formulas, especially when the greatest source of inaccuracy is uncorrected dry eye, a problem with which AI can’t currently help us. If the future of cataract surgery is adjustable lenses, how important is refining the preoperative formula? As for robotic cataract surgery, I’m very bullish about this technology but mainly because it addresses a shortage of our workforce, not because it’s likely to create a significantly safer or more accurate surgery than a highly skilled surgeon can accomplish today.
In glaucoma, we already have many different surgical approaches and a host of new procedures on the horizon — some with and some without implants. Which of these procedures should be employed in which patients, and in which combination, and when should we stick with drops? AI with a large data set can give us better answers than our own clinical impressions.
And while we’re at it, what exactly is glaucoma? Currently, we have some rough definitions about a group of diseases that generally stop when we lower eye pressure. Combing through anatomic and functional imaging along with treatment data from literally millions of patients, it’s likely AI can help us define and categorize this nebulous entity that impacts 60 million people worldwide.
AI becomes most valuable when it helps us predict the future, and knowing the future for glaucoma patients would be more valuable than in any other condition we treat. If we humans can be a little humble about it, we might find that AI can help our patients figure out this disease better than we can.
Follow @DrHovanesian on X, formerly known as Twitter, and Instagram.
Collapse