Prediction of diabetic retinopathy using machine learning techniques

Diabetic retinopathy (DR) is a complication of diabetes attributed to macular degeneration among patients with type II diabetes. The early symptoms of this disease can be predicted through annual eye checkups. This prediction can help such patients prevent vision loss before the disease progress to retinal detachment. Thus, creating awareness among diabetic patients about this disease is necessary to prevent vision loss. Thus, there is a need to develop a computer-assisted method that can effectively predict the disease. The proposed system uses adaptive histogram equalization (AHE), hop field neural networks, and Adaptive Resonance Theory (ART) for image enhancement, blood vessel segmentation, and blood vessel classification. The proposed system analyzes the disease and classifies the disease level effectively with high accuracy. The system can notify users about the stages of the disease. The proposed system can be evaluated with clinical and open fundus image datasets such as DRIVE, STARE, MESSIDOR, HRF, DRIONS, and REVIEW for DR prediction. Physicians evaluated the system and concluded that the result of the proposed system does not deviate from the quality of disease analysis and grading; the proposed technique achieves 99.99% accuracy. Evaluations conducted by ophthalmologists and witnesses confirmed the quality of the proposed system.