Sep 5, 2017
Could Artificial Intelligence Revolutionize the Treatment of Eye Diseases?
Artificial intelligence is in the news more frequently than ever, but it’s not typically associated with vision research.
That may not be the case for long: incredible progress is being made in the application of artificial intelligence (AI) to the detection and prognosis of eye diseases such as age-related macular degeneration (AMD) and diabetic retinopathy.
In conditions such as AMD, damaged blood vessels at the back of the eye lead to vision loss of varying degrees. Diagnosing these diseases can be complicated, involving the close analysis of images of the retina.
Today, the human retina is routinely imaged with a variety of tools, such as optical coherence tomography (OCT) and fundus photography. Adaptive optics is another example: the exciting new technology enables the visualization of single photoreceptors, the eye’s light sensitive cells that are damaged in many blinding eye diseases.
Experts study these images in order to diagnose eye diseases and study their progression and response to treatments over time. This is where AI enters the picture.
Researchers at Stanford University are developing algorithms that are intelligent enough to differentiate a healthy eye from a damaged one. It’s a form of automated detection, an approach that could revolutionize retinal imagining and enable clinicians to pinpoint and treat eye diseases at earlier stages, before the retina is severely damaged.
Using software engineered from Google, the Stanford team trained computer models to get better at interpreting fundus photographs, showing that their model can successfully pick out an image of an eye damaged by diabetic retinopathy—the work was recently published in Ophthalmology. AI specialists refer to this as a form of “deep learning,” where machines are subjected to patterns of data—in this case, images of eyes—until the act of recognizing them becomes deeply ingrained.
In Canada, automated detection is being advanced by Dr. Choudhry, co-founder of the Vitreous Retina Macula (VRM) Specialists of Toronto clinic. He and his team are at the cutting-edge of deep learning applications, and are hard at work developing their own detection software.
With projects similar to Dr. Choudhry’s and the work at Stanford progressing across the globe, we could very well be approaching a new frontier in retinal imaging, and a better future for those living with retinal diseases.
Gargeya, Rishab & Leng, Theodore (2017). “Automated Identification of Diabetic Retinopathy Using Deep Learning.” Ophthalmology, 7, 962-969.
Join the Fight!
Learn how your support is helping to bring a future without blindness into focus! Be the first to learn about the latest breakthroughs in vision research and events in your community by subscribing to our e-newsletter that lands in inboxes the beginning of each month.