Chandigarh PGIMER finds AI 85% effective in diagnosing diabetic retinopathy

Under a central government-funded pilot project, the ophthalmology department of PGIMER has found 85% efficacy in diagnosing the disease through AI, allowing minimal intervention of a physician

Artificial intelligence (AI)-powered detection of diabetic retinopathy at public health facilities may soon become a reality.

As part of the pilot project, being conducted only at PGIMER in the country, three AI platforms were used to diagnose 250 patients in Panchkula and Mohali. (Shutterstock)

Under a central government-funded pilot project, the ophthalmology department of the Post Graduate Institute of Medical Education and Research (PGIMER) has found 85% efficacy in diagnosing the disease through AI, allowing minimal intervention of a physician.

The disease is a diabetes complication that affects the retina in the eyes due to damage to blood vessels. Symptoms include blurred vision, colour blindness, poor vision at night and floaters, and in severe cases, blindness may also occur.

As part of the pilot project, being conducted only at PGIMER in the country, three AI platforms were used to diagnose 250 patients in Panchkula and Mohali.

“The motive was true diagnosis of the disease in comparison with eye specialists. Advanced Eye Centre of PGIMER tested three algorithms on the fundus (back of the eye) images, out of which one showed 85% accuracy, while others gave reasonably well results,” said Dr Vishali Gupta, professor, community ophthalmology.

Currently, patients are dependent on trained opthalmologists to detect the disease by looking at the image reports of the eyes.

However, retinopathy tests done through AI-based cameras can give fast medical results without the involvement of a doctor. The test will have high sensitivity and diagnostic accuracy and will be more cost-effective for the patients, according to PGIMER experts.

“At public healthcare facilities in India, especially in rural areas, people do not have access to retina specialists. Majority of diabetic retinopathy patients have no symptoms until very late stages, and delayed detection can lead to loss of eye sight. Fundus photography is useful for screening and monitoring the rear of the eye. Therefore, by using fundus camera, PGIMER is able to precisely detect clinical signs of the disease using AI,” said Dr Mona Duggal, associate professor, community ophthalmology, PGIMER.

The doctors said once approved by the central government, the AI-powered screening will be recommended to state governments to give people access to cost-effective test at public hospitals.