Medical-AI trial: finding early glaucoma from fundus photos with AI (GlaukomAI) — preventing blindness and cutting waits (NCT07668193)

1Glaucoma takes sight that does not return

Glaucoma damages the optic nerve at the back of the eye little by little, narrowing the visual field, and because lost vision does not return (it is irreversible), it is one of the leading causes of blindness worldwide. Early on there are few symptoms, so it is often advanced by the time it is noticed.

That is why early detection matters, yet a confirmed diagnosis requires several tests (eye pressure, visual field, fundus, OCT) and repeated specialist visits, creating waits and delays.

2GlaukomAI, reading fundus photographs

The GlaukomAI this study evaluates tries to widen that entry point, aiming to detect early glaucoma by analyzing fundus (back-of-eye) photographs with AI. Per the registry summary, the study is conducted at an eye institute (IRCCS Fondazione G. B. Bietti) in Rome, Italy, in two phases.

Phase 1 enrolls 200 participants (100 with diagnosed glaucoma and 100 healthy controls) to assess, in a case-control design, how accurately GlaukomAI can tell them apart. The primary outcome is diagnostic accuracy: sensitivity (not missing disease) and specificity (not misclassifying). The overall size is 1,200, including an assessment of referral accuracy.

3Confirmed in two stages

The trial is built in two stages: first confirming that it can tell the difference, then examining whether the referrals it prompts were appropriate.

1. 1Stage oneEnrol 200 — 100 diagnosed with glaucoma and 100 healthy controls
2. 2Measure accuracyEvaluate sensitivity and specificity in a case-control design
3. 3On to stage twoExtend to 1,200 in total and evaluate the appropriateness of referral

Glaucoma is a leading cause of blindness worldwide because lost vision does not return; early stages carry few symptoms, and confirming a diagnosis requires repeated visits for pressure, field, fundus and OCT testing. A fundus photograph can be taken as easily as at a health check or an optician, so AI reading could sort referrals sensibly into crowded specialist clinics.