Key takeaways
- Regular retinal screening helps detect diabetic retinopathy before vision is affected.
- Fundus AI analyses retinal photographs and flags those that need referral.
- In 2018 the FDA authorised the first autonomous AI system for detecting more-than-mild diabetic retinopathy.
- Image quality, referral pathways and integration decide whether AI improves a programme in practice.
The screening challenge
Diabetic retinopathy often has no symptoms in its early stages. Regular retinal photography allows it to be found and treated before sight is threatened. As the number of people living with diabetes grows, screening programmes face more images than specialist graders can review promptly.
How fundus AI works
A fundus camera takes colour photographs of the retina. AI software analyses the images for signs such as microaneurysms, haemorrhages and exudates and returns a result, typically whether the patient should be referred to an eye specialist. Some systems are designed to give a result without a specialist reviewing the image first; others assist a human grader.
What the evidence shows
In a pivotal trial published in 2018, an autonomous AI system tested in primary care clinics in the United States showed a sensitivity of 87.2% and a specificity of 90.7% for more-than-mild diabetic retinopathy against a reference standard. The same year, the FDA authorised it as the first autonomous AI diagnostic system in any field of medicine. Other products have since received approvals in various countries, each for specific indications and cameras.
Making it work in practice
- Image quality: ungradable images need a clear pathway, such as retaking the photograph or referring.
- Camera compatibility: many products are approved with specific cameras only.
- Referral pathways: positive results must reach ophthalmology quickly.
- Integration: results should reach the patient record and the diabetes care team.
- Scope: AI screening for diabetic retinopathy is not a full eye examination; other conditions may need separate assessment.
Where it fits
Fundus AI can extend screening to primary care, diabetes centres and pharmacies, where specialists are not on site. It works best as part of a complete programme: trained photographers, good cameras, fast referral and follow-up.
Related: Medical AI solutions
Talk to our team
We can help you choose the right solution for your facility.
References
- Abràmoff MD, Lavin PT, Birch M, Shah N, Folk JC. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine. 2018;1:39.
- U.S. Food and Drug Administration. FDA permits marketing of artificial intelligence-based device to detect certain diabetes-related eye problems. News release, April 2018.
This article summarises published guidance and research for general information and is not medical or engineering advice. Device selection and use should follow the manufacturer's instructions, local regulations and your facility's policies.