Clinical Insights through Xception: A Multiclass Classification of Ocular Pathologies
Main Article Content
Abstract
In this study, the Xception convolutional neural network is employed to categorize four distinct eye conditions: cataract, diabetic retinopathy, glaucoma, and normal eye status. The model achieves an impressive accuracy rate of 92.87% when assessed on a dataset comprising 4,127 images. This success highlights the potential of deep learning techniques in the field of ophthalmology, providing a dependable method for early diagnosis and disease management. Based on these promising outcomes, the research outlines several prospective directions, including model refinement, transfer learning, multiclass classification, data augmentation, ensemble learning, interpretability, clinical integration, ethical considerations, and diversification of the dataset. These paths hold the potential to further enhance the capabilities of AI-based diagnostic systems in ophthalmology, contributing to improved patient care and early intervention, all while addressing ethical and practical concerns within the healthcare domain.