Google's ARDA eye-screening AI passes one million patients
TL;DR
- Google's diabetic retinopathy model has now been used on more than a million patients across three health systems in India, Thailand and Australia.
- Google licenses the model to three commercial operators, Forus Health, AuroLab and Perceptra, rather than running the screenings itself.
- The program targets six million AI-supported screenings across India and Thailand over ten years, at no cost to patients.
Over a million patients have been screened with Google's diabetic retinopathy detection model across three health systems in India, Thailand and Australia, according to a Comment published in Nature Medicine on 23 September 2026.
The anchor sites are the Aravind Eye Care System in Madurai, Rajavithi Hospital in Bangkok, and Lions Outback Vision at the Lions Eye Institute in Western Australia. The authors, a mix of Google staff including Yossi Matias and Sunny Virmani alongside clinicians at those sites, frame the piece as a lessons-learned write-up rather than a new results paper.
"Scaling a successful deep learning tool across three highly distinct settings…offers cross-cutting insights that may inform the expansion of healthcare artificial intelligence globally," they write.
Google does not run the screenings itself. It licenses the model, Automated Retinal Disease Assessment, to three commercial operators, Forus Health, AuroLab and Perceptra, who deploy it in-country. The stated target is six million AI-supported screenings across India and Thailand within ten years, at no cost to patients. Two of the clinicians we follow in our Who's Who index flagged the paper on the day it ran.
The Comment does not publish per-site accuracy or treatment-follow-through numbers in the version available; most of the listed authors are Alphabet employees, and the work was Alphabet-funded. Clinical co-author Paisan Ruamviboonsuk discloses consulting relationships with Roche, Biocon Biologics, Abbott, 4DMT and Bayer.
Shared on Bluesky by 2 AI experts
-
Over a million patients were screened for diabetic eye disease as Google scaled its clinical tool across India, Thailand, and Australia. One of the biggest challenges wasn't the model itself; it was sorting out camera o…
View on Bluesky →
Originally reported by nature.com
Read the original article →Original headline: Practical lessons in the global scaling of clinical AI: from one hospital to over a million patients screened - Nature Medicine