Aravind Eye Hospital and Google: AI that screens diabetic eyes at scale

India has far too few eye specialists for its diabetic population. A Madurai hospital put an AI screening tool into clinical use to help close the gap.

By the DigitalHelp4U team · 15 October 2026

Diabetic retinopathy can cause blindness if it isn't caught early, and early detection needs a trained specialist to examine images of the eye. In 2019 India was reported to be short of more than 100,000 eye doctors, and only around 6 million of its 72 million people with diabetes were being screened.

What they did

Google and Verily developed a machine-learning model that reads retinal photographs and flags patients who need referral. It was trained over about three years on roughly 128,000 images, each graded by several ophthalmologists. Aravind Eye Hospital in Madurai, Tamil Nadu, which sees thousands of patients a day, put the system into clinical use.

What changed

According to Aravind's chief medical officer at the time, the tool let doctors spend more time with patients and increased the number of screenings the hospital could do. A 2016 study in JAMA had found the algorithm highly accurate at detecting referable cases, while noting more research was needed on long-term outcomes.

What clinics can learn

• AI supports doctors, it doesn't replace them. The value was in freeing specialists for the cases that need them.
• Start with a narrow, repetitive task. Screening one type of image is a well-defined problem, which is why it worked.
• Volume is the hidden win. More patients screened means more patients helped and more revenue for the same staff.

How we would apply it

Most clinics don't need medical imaging AI to start. The same idea works at the front desk: an AI assistant that answers timings and fee questions on WhatsApp, books appointments and sends reminders, so doctors and staff spend their time on patients.