Australian Study Demonstrates High Accuracy of AI for Detecting Diabetic Eye Disease

A groundbreaking Australian study has shown that an AI-powered portable retinal camera can reliably identify diabetic eye disease with over 93% accuracy in real-world clinical settings, including general practitioner and endocrinology clinics, as well as Aboriginal health services. Conducted over two years between August 2021 and June 2023, the trial involved more than 860 participants with diabetes, who took retinal images using an AI-enhanced device while awaiting their appointments.

The AI system was trained on a vast dataset of over 200,000 retinal images graded by a panel of 21 ophthalmologists. Participants received a scan result via a QR code, linking them to their results, and those with signs of diabetic eye disease were referred for further specialist assessment. The results from AI assessments were compared with traditional human grading, serving as the gold standard.

Findings indicated a high accuracy rate of 93.3%, with 86% participant satisfaction and 85% clinician approval. Despite these promising results, the study also identified areas for improvement, such as enhancing image quality, reducing false negatives, increasing follow-up adherence, and developing strategies for diverse communities. Dr. Zhuoting Zhu from the Centre for Eye Research Australia emphasized the potential of AI to extend eye screening access, particularly in remote and underserved areas, thereby facilitating early diagnosis and treatment.

The study underscores the potential of AI technology to revolutionize diabetic eye screening by making it more accessible, cost-effective, and integrated into routine healthcare, ultimately aiming to prevent blindness in millions worldwide. More information can be found in the published paper in the British Journal of Ophthalmology.

Source: https://medicalxpress.com/news/2025-09-ai-scans-accurately-diabetic-eye.html