AI Based Automated Detection & Classification of Diabetic Retinopathy

Diabetic eye disease poses a significant threat to vision, affecting those with diabetes and potentially leading to blindness if left undetected. To tackle this issue, recent progress in deep learning and artificial intelligence have revolutionized Diabetic Retinopathy Detection. Convolutional Neural Networks (CNNs) have emerged as a powerful tool, displaying impressive accuracy in automatically diagnosing and classifying DR severity levels from retinal fundus images. With access to large annotated datasets and the use of transfer learning, these models have shown robust performance, even with limited labeled data. The method involves several essential steps, including Data Collection, Preprocessing, Augmentation, and modeling, culminating in fundus image categorization into severity levels such as No DR, Mild, Moderate, Severe, and Proliferative DR. Through the utilization of GPUs during model training, computation time is significantly reduced, resulting an exceptional performance, achieving an approximate 81% accuracy rate on a dataset of 35,126 publicly released retinal images from eyePACS on Kaggle. By expediting the detection and decision-making process, this automated DR screening method holds the potential to effectively control or manage the progression of diabetic retinopathy, ultimately leading to improved patient outcomes and timely interventions.