Google eye AI scales from one hospital to a million patients

A single deep learning eye-screening tool went from one hospital to over a million patients across India, Thailand and Australia. That is the scope a Google-led team of Richa Tiwari and co-authors set out in Nature Medicine, reviewing what did and did not travel when the same model was pushed into three very different health systems.

The abstract describes the effort as scaling "a successful deep learning tool across three highly distinct settings" and pitches the resulting insights as generalisable to the wider expansion of healthcare AI. That is roughly all the public-facing abstract commits to; the concrete numbers, workflow adjustments and country-level failure modes sit behind the paywall.

The three anchor sites on the author list are the Aravind Eye Care System in Madurai, Rajavithi Hospital in Bangkok, and Lions Outback Vision at the Lions Eye Institute in Western Australia. Google has separately said the same diabetic retinopathy model has supported more than 600,000 screenings in clinics around the world and is licensed to Forus Health, AuroLab and Perceptra, with a stated goal of six million AI-supported screenings across India and Thailand over the next ten years.

"Seven years later, we're grateful to bring this technology to Thai patients with diabetes but also Thailand's public health system as a whole," said Rajavithi Hospital's Dr. Paisan Ruamviboonsuk in Google's October 2024 announcement. The abstract publishes no per-country accuracy figures.