Severity Assessment of Diabetic Retinopathy Through Automated Segmentation

Diabetic Retinopathy (DR) is a microvascular complication of diabetes mellitus that impacts the retinal vasculature, leading to progressive vision impairment and potential blindness if left untreated. This study investigates the efficacy of automated segmentation techniques in the evaluation of DR severity. The objective is to establish a precise and efficient methodology for the quantification of critical retinal biomarkers, including blood vessels, microaneurysms, and optic discs. This study used advanced image processing algorithms such as Hough Circle Transformation, Canny Edge Detection and Contrast Limited Adaptive Histogram Equalization to extract these features from fundus images. Refinement to the images was achieved through contour detection and elliptical element filtering. The performance of each method on the features were assessed by pixel-wise evaluation and the methods yielded high accuracy. The automated segmentation of fundus images allows for the assessment of DR severity, which enables ophthalmologists to perform timely therapeutic interventions. This method enhances diagnostic accuracy and minimizes human error in DR diagnosis, optimizing patient management strategies. This approach found in the study not only aids ophthalmologists but also provides quality care to patients through early detection and continuous monitoring of the retinal features, particularly for underserved diabetic populations at risk for vision threatening retinopathy.