CNN Based Computer Aided Ophthalmic Diagnosis Systems for Classifying Eye Conditions

Accurate classification of ocular conditions is crucial for the development of computer aided ophthalmic diagnosis system. Traditional methods relying on expert analysis can be subjective and time-consuming. This study presents a machine learning-based approach to classify eye images into six categories: normal, glaucoma, strabismus, bulging, uveitis, and cataract. We employed convolutional neural networks (CNNs) using dataset from Kaggle that consists of 2893 images, validated for quality. Our study focused on optimizing the training process using two popular algorithms: Adaptive Moment Estimation (Adam) and Stochastic Gradient Descent (SGD). Adam, known for its adaptive learning rate and momentum, typically accelerates convergence, while SGD, with its simplicity and efficiency, often provides better generalization. Results show that the Adam-optimized CNN achieved a higher accuracy of 94% and faster convergence compared to the SGD-optimized model's 48% accuracy. These findings show Adam's effectiveness in handling diverse ocular image data and enhancing diagnostic precision. In conclusion, the proposed method demonstrates significant potential for automating the diagnosis of ocular conditions, offering a scalable and efficient solution for clinical applications.