Google's ARDA eye AI passes 1M patients across three countries

Google's diabetic-retinopathy screening AI has been used on more than a million patients since it moved out of the lab, and the team behind it is now writing up what the scaling actually took. A Comment in Nature Medicine dated 23 September walks through the deployment across three health systems in India, Thailand and Australia.

The named anchor sites are Aravind Eye Care System in Madurai, Rajavithi Hospital in Bangkok, and Lions Outback Vision at the Lions Eye Institute in Western Australia. It is co-written by Google researchers Richa Tiwari, Rajroshan Sawhney and Kasumi Widner with clinicians from each site: Kim Ramasamy and Ravindran Padmanabhan at Aravind, Paisan Ruamviboonsuk at Rajavithi, and Angus Turner at Lions Eye.

"Scaling a successful deep learning tool across three highly distinct settings," the authors write, "offers cross-cutting insights that may inform the expansion of healthcare artificial intelligence globally." Two researchers we track flagged the paper in our Who's Who feed the day it went live.

The distribution goal is bigger than the current headline number. Google now licenses the model — trained on more than 100,000 retinal scans — to Forus Health, AuroLab and Perceptra, and the group has said publicly it is aiming for six million AI-supported screenings across India and Thailand over the next decade. Thailand's Department of Medical Services is folding the tool into public-sector hospitals through a national innovation program.

The Comment itself is a lessons piece rather than an outcomes paper: per-country accuracy, referral rates and integration costs at each site are not broken out in the material available for this write-up.