A breakthrough in predicting the progression of Diabetic Retinopathy (DR), a common microvascular complication of diabetes and a leading cause of preventable blindness globally, has been made by a joint team from Shanghai Jiao Tong University and Tsinghua University. This significant advancement was recently published in the esteemed journal Nature Medicine.

Led by Professor Jia Weiping, from the Department of Endocrinology and Metabolism at the Shanghai Jiao Tong University Affiliated Sixth People's Hospital and the Shanghai Key Laboratory of Diabetes, along with Professor Li Huating, the team collaborated with Tsinghua University's Vice Provost and Dean of the Medical School, Professor Huang Tianyan, and Professor Sheng Bin from Shanghai Jiao Tong University's School of Electronic Information and Electrical Engineering/AI Key Laboratory of the Ministry of Education.

This achievement follows the successful development of the “DeepDR” intelligent diagnostic system for diabetic retinopathy in 2021. The new system, named "DeepDR Plus," is a further advancement that employs deep learning on sequential image series for early warning of complications associated with diabetic retinopathy, based on retinal images. This innovative system holds the potential to guide new strategies in screening and prevention of diabetic retinopathy worldwide.

DR often presents with hidden symptoms in its initial stages, potentially leading to permanent visual impairment or even blindness when severe. Early screening and intervention are crucial in preventing and managing DR. While annual retinal photography screenings are advised for diabetic patients with none or mild DR, this practice faces significant challenges in implementation and prevalence, especially in middle and low-income countries, due to economic and healthcare resource constraints.

This research, for the first time, utilizes a large-scale medical image longitudinal cohort encompassing over 200,000 diabetic patients from multiple countries and ethnicities. It innovatively considers the progression and onset of diabetic retinopathy as random variables within the screening intervals, applying survival analysis and sequential distribution probability modeling. This approach has successfully enabled risk alerts and timing predictions for the progression of diabetic retinopathy.

The "DeepDR Plus" system, based solely on baseline retinal images, can accurately predict individualized risks and the timeline for DR progression over the next five years. This AI-driven personalized screening interval, particularly beneficial in developing countries, significantly enhances the efficiency of retinal photography screenings. This research provides new evidence for diabetic retinopathy screening, prevention, and treatment guidelines, potentially impacting future clinical practices and healthcare costs for diabetic retinopathy, contributing significantly to global diabetes management.