AI Detects Diabetic Eye Disease With 93% Accuracy

A new Australian clinical study has demonstrated that AI-driven retinal imaging can accurately detect diabetic eye disease with more than 93% accuracy, even in non-eye care settings. The findings suggest that artificial intelligence could play a key role in making diabetic eye screening more accessible and efficient.

Study Overview and Clinical Setting

The two-year trial was conducted between August 2021 and June 2023, involving over 860 participants with diabetes. Screenings were performed in the waiting rooms of general practice and endocrinology clinics in Melbourne, as well as at an Aboriginal Health Service in Western Australia.

The study was led by Associate Professor Lisa Zhuoting Zhu and Sanil Joseph from the Centre for Eye Research Australia and University of Melbourne, in collaboration with Professor Mingguang He from the Hong Kong Polytechnic University.

The findings were published in the British Journal of Ophthalmology.

Technology and Methodology

The trial employed an automated portable retinal camera powered by an AI algorithm trained on over 200,000 retinal images, graded by 21 ophthalmologists. Participants self-administered their scans while waiting for their medical appointments.

Each participant received a printout with a QR code linking to their scan results. If signs of diabetic eye disease were detected, they were referred to an eyecare specialist for further evaluation. To validate the AI’s performance, all scans were compared with human grading assessments, considered the clinical gold standard.

In addition, both patients and healthcare providers participated in a satisfaction survey to assess the usability and acceptability of the AI screening tool.

Key Findings

• The AI system achieved an accuracy rate of 93.3% compared to expert human grading.

• 86% of participants reported being satisfied with the AI technology.

• 85% of clinicians rated the technology highly.

The study demonstrates the feasibility of incorporating AI screening tools into routine clinical care for people with diabetes, even in non-specialist settings.

Areas for Improvement

While results were promising, the study identified several areas where further development is needed:

• Enhancing image quality to reduce ungradable scans

• Refining the AI algorithm to lower the rate of false negatives

• Improving follow-up processes to ensure referred patients act on recommendations

• Developing targeted strategies to meet the needs of diverse and underserved communities

Clinical and Systemic Benefits

According to Dr. Zhu, AI-driven scans hold strong potential for rural and remote regions, where access to trained eyecare professionals may be limited.

"AI scans could be a great benefit in rural and remote areas where there is a shortage of trained eye care specialists," she stated. "It is also a cost saving for the health system, as it enables early screening to occur without the need for an eye care specialist for every patient."

Sanil Joseph emphasized the patient convenience factor:

"People with diabetes often have many medical appointments and prioritize appointments with other specialists over eye care. The AI scan enables them to combine their eye test with other medical visits."

Reference:

Sanil Joseph et al, Effectiveness of artificial intelligence-based diabetic retinopathy screening in primary care and endocrinology settings in Australia: a pragmatic trial, British Journal of Ophthalmology (2025). DOI: 10.1136/bjo-2025-327447