Alphabet's Verily Using AI To Prevent Diabetic Eye Disease

Google AI Diabetic Eye

Diabetic retinopathy and diabetic macular edema are two of the world’s leading causes of preventable blindness in adults, and now Google has outed an ongoing clinical study, handled by Verily, that seeks to use AI to help screen for those eye diseases faster and more efficiently than ever before. Specifically, Google has announced that the study is now considered a clinical trial thanks to its first official clinical venue; Madurai, India’s Avarind Eye Hospital.

The study involves an algorithm that’s been developed over the past three years or so, and is aimed at using standard footage of eyes, captured during routine eye doctor visits on perfectly normal equipment, to screen patients for the diseases in question. The way it all works is by sending captured eye footage off to Google’s servers to be run through the machine learning process of the algorithm. Screening results are then delivered back to the doctor immediately, facilitating quicker and easier screening, diagnosis, and most importantly, earlier treatment when cases are found.

The study is currently taking place globally, though in a limited capacity. The focus right now is of course India, where some 66 million diabetes patients go without diagnostic screening for associated eye diseases each year due to a shortage of qualified eye doctors. Under those conditions, this study will mean doctors spend less time waiting for results, less time setting up screenings, which sometimes involve separate appointments, and less diagnostic time and lab resources checking through the massive amount of screening data.


Verily’s system, it should be noted, is for screening, not diagnosis. It’s not so accurate as to be able to flag patients it knows for sure do have one of the listed conditions. Rather, it can tell patients who have signs of the conditions versus patients who show no signs and almost certainly don’t have one of the eye diseases it screens for. While this doesn’t eliminate the doctor’s role in the screening process entirely, it does eliminate a vast swath of would-be screening data.

Google formed Verily as a part of Google X way back in 2015, and began research into the intersection between Google’s technological wheelhouse and the medical world. One example is its acquisition of Lift Labs, a startup that created Liftware, utensils that use AI to compensate for shaky hands and avoid food spillage.

While all of this would be enough news to get excited under any circumstances, this algorithm has received a CE mark from the European Union, meaning that it’s considered to be on par with medical equipment designed for the same purpose. What this means is that Google and Verily have the vetting of the EU behind them as they spread the study and the algorithm out, making it quicker and easier to foster wide adoption.


The kicker here is that this is a machine learning AI, and it does just what its name says it does; it learns. The more eyes this machine looks at, the better it’s going to be at its job. It’s also going to be getting a better picture of what the human eye is like on a grander scale, and the implications of that in the wider fields of optometry and opthalmology are nearly endless. So long as nothing goes wrong, it’s only a matter of time before this algorithm expands its suite of use cases and sees wide adoption throughout a wide spectrum of eye-related medical use cases.