Abstract
The increasing prevalence of ocular diseases such as diabetic retinopathy, glaucoma, and cataract highlights the urgent need for accurate and scalable automated screening systems to support early diagnosis and prevent vision loss. In the United States recent advances in medical image analysis and deep learning have enabled the development of intelligent frameworks for retinal disease detection using fundus images. In this study, an automated ocular disease screening framework is proposed using deep learning architectures, including a conventional Convolutional Neural Network (CNN), VGG16, MobileNetV2, and ResNet-15 V2. The models are trained and evaluated on a curated dataset of 4,217 retinal fundus images categorized into four classes: Normal, Cataract, Diabetic Retinopathy, and Glaucoma. Experimental results demonstrate that ResNet-15 V2 achieves the highest classification accuracy of 90.52%, while MobileNetV2 provides competitive performance with lower computational complexity, making it suitable for resource-constrained environments. The findings confirm that deep learning-based screening systems can effectively enhance early ocular disease detection and support practical deployment in modern ophthalmic care systems.