How AI is making eyesight-saving care more accessible in resource-constrained settings
Nearly a decade ago, teams from Google Research and our foundational research partners including Aravind Eye Hospital in India and Rajavithi Hospital in Thailand began exploring how AI could help reduce preventable blindness caused by eye diseases such as diabetic retinopathy. Our goal was to help doctors reach more patients who lack access to screening, so we set out to investigate whether AI could identify diabetic retinopathy, a treatable but leading and growing cause of preventable blindness. Many people with the condition don’t get the regular screenings and prompt care they need, so we developed an AI model to help detect this complication. To date, the model has supported more than 600,000 screenings in clinics around the world.
We’ve spent the last several years exploring how to increase AI adoption into real-world clinical settings. And now, we’re licensing our diabetic retinopathy model to partners in order to bring this tool to communities across India and Thailand — two countries with a significant shortage of eye specialists. Our hope is that this will help improve outcomes for millions more diabetic patients who are most at-risk.
Globally, at least 537 million adults are estimated to be living with diabetes. In the Asia Pacific region alone, there are nearly 227 million people living with the condition. Nearly half of people with diabetes will develop diabetic retinopathy, which can happen when diabetes impairs the blood vessels that nourish the eye’s retina. Over time, this damage may lead to blindness. Early detection and timely medical intervention can help preserve their eyesight, so regular screening is key. However, due to limited healthcare specialists in many parts of the world, many people go undiagnosed and untreated, ultimately jeopardizing their vision.
Developing an AI model to help clinicians detect diabetic retinopathy is one step toward a potential solution. But for population-scale impact, we need to identify ways to expand diabetic retinopathy screening on a large scale. That’s why we’re taking this next step of partnering with healthcare providers and health-tech partners Forus Health, AuroLab and Perceptra. These partners will work toward securing local regulatory approvals to bring the model into clinical care systems across India and Thailand. Over the next 10 years, these partners will aim to deliver a combined 6 million AI-supported screenings to resource-constrained communities at no cost to patients.
We’ve also been working with the Thai Ministry of Public Health’s Department of Medical Services (DMS), which is responsible for the country’s diabetic retinopathy screening program, on implementation research and cost-effectiveness analysis. This collaboration is bringing our AI into Thailand’s National Innovation program and will pave the way for a partnership between Perceptra and DMS to apply the diabetic retinopathy AI model in public sector hospitals and help deliver impact at population scale.
We continue to see AI’s transformational potential in healthcare, especially in early disease prevention, detection and diagnosis. And by sharing our models, we aim to support others in building valuable medical AI tools that advance care and improve health outcomes for communities around the world.
