Diabetic retinopathy is a major cause of blindness worldwide, impacting over 130 million people. Despite progress in ophthalmic imaging, most cases are diagnosed only after significant and often irreversible retinal damage has occurred. Current standard imaging techniques, such as fundus photography and angiography, often fail to detect early molecular and cellular changes in the retina associated with diabetes.
Researchers at the University of Coimbra in Portugal have introduced a new texture-based analysis method for optical coherence tomography (OCT) images to identify early retinal alterations linked to type 2 diabetes. The study was published in Eye and Vision on September 3, 2025.
The research used a rat model involving a high-fat diet and low-dose streptozotocin to simulate diabetic conditions over a period of 12 weeks. More than 80 retinal scans from both diabetic and control rats were analyzed using advanced image processing techniques. Specifically, the team applied a gray-level co-occurrence matrix (GLCM) approach to measure texture parameters across different layers of the retina.
Among twenty features assessed, eight—including autocorrelation, cluster prominence, correlation, homogeneity, information measure of correlation II (IMCII), inverse difference moment normalized (IDN), inverse difference normalized (INN), and sum average—showed notable differences in diabetic retinas. These changes were especially evident in the inner plexiform layer and photoreceptor segments. Notably, seven of these metrics had also been found altered in an earlier study focusing on type 1 diabetes models.
The researchers observed that despite minimal thinning of retinal layers and delayed oscillatory potentials—which are typical signs detected by current diagnostic methods—there was no major inflammation or vascular leakage present. This suggests that texture-based changes can be detected before traditional markers appear.
"This research paves the way for developing AI-assisted diagnostic tools that automatically screen for preclinical DR based on retinal texture signatures," according to the authors. "Integrating this analysis into routine OCT imaging could allow ophthalmologists to identify patients who show microscopic structural disruption—even when their vision appears normal." The statement continues: "Such early detection may help tailor personalized care, prevent irreversible retinal damage, and reduce the global burden of diabetic blindness."
Further clinical trials will be necessary to validate these findings in humans and improve algorithms for broader screening purposes.
