Diabetic Retinopathy Detection using Deep Learning: A Systematic Review

Abstract

Diabetic Retinopathy (DR) is one of the leading causes of vision loss and blindness among diabetic patients worldwide. Early detection and timely treatment of diabetic retinopathy are essential to prevent severe retinal damage and improve patient outcomes. Traditional diagnosis methods rely on manual examination of retinal fundus images by ophthalmologists, which can be time-consuming, expensive, and prone to human error. In recent years, deep learning techniques have shown remarkable success in medical image analysis and automated disease diagnosis. This systematic review presents a comprehensive analysis of deep learning approaches used for diabetic retinopathy detection using retinal fundus images. Various deep learning models such as Convolutional Neural Networks (CNN), Transfer Learning, Recurrent Neural Networks (RNN), and hybrid architectures are reviewed based on their classification performance, dataset utilization, preprocessing techniques, and evaluation metrics. The study also discusses publicly available datasets, image enhancement methods, segmentation techniques, and performance measures including accuracy, sensitivity, specificity, precision, recall, and F1-score. Comparative analysis indicates that deep learning-based models significantly improve diabetic retinopathy detection accuracy and reduce manual diagnostic effort compared to conventional machine learning methods. The review concludes that deep learning techniques provide reliable and efficient solutions for automated diabetic retinopathy screening and can greatly assist healthcare professionals in early diagnosis and clinical decision-making.