Diabetic Retinopathy Detection Using Deep Learning

Abstract

This study presents a novel approach to detecting neovascularization, a critical indicator of Proliferative Diabetic Retinopathy (PDR), in fundus images using deep learning techniques, specifically transfer learning. Neovascularization poses a significant risk to individuals with diabetes, potentially leading to blindness if not detected and treated promptly. Traditional image processing methods have struggled to effectively identify neovascularization due to its random growth patterns and small size. In response, this paper explores the efficacy of transfer learning, leveraging pre-trained models such as Inception ResNetV2, DenseNet, ResNet50, ResNet18, and AlexNet, renowned for their automatic feature extraction capabilities on complex objects. By harnessing the power of deep learning, our proposed method aims to enhance the accuracy and efficiency of neovascularization detection, offering promising advancements in early diagnosis and intervention for diabetic retinopathy.