AI Enhances Diabetic Retinopathy Detection

A joint research team from Skoltech, the University of Sharjah, and the AIRI Institute has developed an AI-powered system to automate the analysis of retinal images for diagnosing diabetic retinopathy. This breakthrough could revolutionize early detection and improve diagnostic accuracy, potentially reducing blindness caused by diabetes-related eye disease.

AI-Powered Retinal Image Analysis: A Game-Changer

Diabetic retinopathy is a serious condition that can lead to permanent blindness if left untreated. Traditionally, analyzing retinal images to diagnose this condition requires manual segmentation, a time-consuming and error-prone process that can take between 10 to 40 minutes per patient, depending on the physician’s expertise.

However, researchers have now developed an AI-driven solution that delivers instant diagnostic results, allowing ophthalmologists to review and confirm findings more efficiently. The study, published in Pattern Recognition Letters, demonstrates how the AI model automates the segmentation of blood vessels in retinal images, eliminating human error and significantly reducing examination time.

Overcoming Challenges with Advanced Neural Networks

One of the main challenges in retinal image analysis is detecting microvessels, which are crucial in diagnosing various eye diseases, including diabetic retinopathy and atherosclerosis. Existing AI models have struggled with this level of detail, often misclassifying vessel pixels as background noise.

The research team addressed this by:

• Developing a novel neural network architecture optimized for recognizing microvessels.

• Implementing an adaptive threshold algorithm to refine vessel segmentation.

• Enhancing model sensitivity through data augmentation techniques.

The results have been remarkable, with the AI system achieving over 97% accuracy and a sensitivity rate exceeding 84% on the industry-standard DRIVE dataset.

Expert Insights on the Breakthrough

Lead author Melaku Getahun, a Ph.D. student at Skoltech, highlighted the significance of sensitivity in AI models for retinal analysis:

"Achieving 97% accuracy is not that difficult due to the nature of the data. It is the sensitivity that matters the most. It reflects the ability of the model to identify microvessels, which previous models struggled with."

Senior Research Scientist Oleg Rogov, the study’s principal investigator from Skoltech AI, emphasized the challenges of limited training data:

"Through the careful application of data augmentation and processing techniques, we managed to significantly improve the model’s performance. Our adaptive threshold algorithm played a crucial role in enhancing sensitivity and accuracy."

Future Implications: A Standard Tool for Eye Disease Screening

As the AI system continues to evolve, researchers believe it could become a standard screening tool for ophthalmologists, enhancing early disease detection and improving patient outcomes. The ability to spot unhealthy blood vessels early could lead to timely interventions, ultimately preventing vision loss in high-risk populations.

Professor Rifat Hamoudi from the University of Sharjah echoed this sentiment:

"This can help in the early diagnosis and prevention of eye diseases that are difficult to treat, such as diabetic retinopathy, which is prevalent in populations with high incidence of diabetes."

The study was conducted by the Biomedically Informed Artificial Intelligence Laboratory (BIMAI-Lab), a joint initiative between Skoltech and the University of Sharjah, headed by Assistant Professor Maxim Sharaev and Professor Rifat Hamoudi. The research also included contributions from Professor Ahmed Bouridane, an expert in AI-driven medical data analytics.

Reference:

Melaku N. Getahun et al, FS-Net: Full scale network and adaptive threshold for improving extraction of micro-retinal vessel structures, Pattern Recognition Letters (2025). DOI: 10.1016/j.patrec.2025.01.019