Deep Dive Into Diabetic Retinopathy Identification: A Deep Learning Approach with Blood Vessel Segmentation and Lesion Detection
Kamal Upreti, Anmol Kapoor, Sheela Hundekari, Shitiz Upreti, Kajal Kaul, Shreya Kapoor, Akhilesh Tiwari
Abstract
In the landscape of diabetes-related ocular complications, diabetic retinopathy stands as a formidable challenge, reigning as the leading cause of vision impairment worldwide. Despite extensive research, the quest for effective treatments remains an ongoing pursuit. This study explores the burgeoning domain of AI-driven approaches in ocular research, particularly focusing on diabetic retinopathy detection. It delves into various diagnostic methodologies, encompassing the detection of microaneurysms, identification of hemorrhages, and segmentation of blood vessels, primarily utilizing retinal fundus photographs. Our findings juxtapose conventional machine learning techniques against deep neural networks, showcasing the remarkable efficacy of Convolutional neural network (CNN) and Random Forest (RF) in segmenting blood vessels and the robustness of deep learning in lesion identification. As we navigate the quest for clearer vision, artificial intelligence takes center stage, promising a transformative leap forward in the realm of vision care.