AI Just Transformed Diabetic Eye Disease Screening…

AI Just Transformed Diabetic Eye Disease Screening Published by TALMedora Editorial | Ophthalmology | Artificial Intelligence | Diabetic Care Diabetic macular edema is the leading cause of vision loss among working-age people with diabetes. It affects approximately seven percent of all diabetic patients worldwide and in India, where diabetes prevalence continues to rise at an alarming rate, the scale of this problem is immense. The standard screening pathway, fundus photography followed by specialist referral, has a fundamental and long-standing flaw. Between 71 and 86 percent of patients referred for diabetic macular edema evaluation based on fundus photographs alone do not actually have the condition. Three in four specialist eye clinic appointments for suspected diabetic macular edema are unnecessary. The consequences are significant. Ophthalmology clinics are overwhelmed. Patients face unnecessary anxiety and waiting times. Healthcare systems absorb enormous costs for appointments that produce no clinical finding. And the clinicians who need to focus their expertise on genuine cases are buried under a mountain of false alarms. A landmark randomised clinical trial just published in JAMA has provided a rigorous, evidence-based solution. Researchers conducted a stepwise evaluation across multiple care settings in Hong Kong, beginning with a prospective silent-mode validation of 603 diabetic patients, followed by a multicenter randomised clinical trial involving 276 patients with suspected diabetic macular edema referred through a territory-wide screening programme. The intervention was an AI-powered optical coherence tomography system, a deep learning model designed to analyse three-dimensional OCT scans and determine with high accuracy whether diabetic macular edema is genuinely present before a specialist referral is made. Optical coherence tomography is already the gold standard diagnostic test for diabetic macular edema. It provides three-dimensional macular imaging that fundus photographs, two-dimensional by nature, simply cannot replicate. The innovation in this study was deploying an AI system to analyse OCT scans automatically, rapidly and at scale within the screening pathway itself. The system was designed with three layers of clinical intelligence. An image quality assessment model first filtered out ungradable scans. A DME detecti