Detection of Sub-Retinal and Intra-Retinal Fluid Accumulation in Retinal Optical Coherence Tomography Images of Diabetic Patients

Diabetic maculopathy, a complication of untreated diabetic retinopathy, impairs vision by affecting the macula through pathologies such as microaneurysms, exudates, hemorrhages, and fluid accumulation. This study focuses on detecting intraretinal and subretinal fluid in optical coherence tomography (OCT) images. This study proposes a novel image processing pipeline involving custom denoising, automated region of interest (ROI) extraction using edge detection, and active contour-based segmentation to delineate fluid boundaries. The method was evaluated on 500 diabetic macular edema (DME) images from a public dataset, achieving an average accuracy of 98.96%, sensitivity of 99.25%, specificity of 98.83%, and Dice coefficient of 0.95. These results were validated against ground truth annotations. Compared to state-of-the-art deep learning methods, the proposed approach offers interpretability and lower computational complexity, making it suitable for clinical deployment.