Hemorrhage Segmentation on Retinal Images for Early Detection of Diabetic Retinopathy

Abstract

Diabetes mellitus is a chronic disorder that can lead to serious complications, including diabetic retinopathy, which affects the eyes and can potentially lead to blindness. Rapid identification of diabetic retinopathy is crucial to facilitate quicker and more efficient treatment for patients. This study aims to segment hemorrhages in retinal images using the Laplacian of Gaussian (LoG) approach in conjunction with threshold-based segmentation and analysis of region properties, including eccentricity. The LoG approach is utilized for its ability to detect edges, features, and abrupt variations in image intensity, thereby optimally highlighting the bleeding lesion area. With accurate segmentation, it is hoped that early detection and monitoring of diabetic retinopathy can be improved. This research uses the IDRiD, DR_2000, and DIARETDB1 datasets, recommending the use of IDRiD and DIARETDB1 for optimal results. Through this methodology, it is expected to make a significant contribution to reducing the risk of blindness in diabetes patients.