Advancing Optic Nerve Degeneration Diagnosis through CNN: Towards Better Eye Health Outcomes

Abstract

Glaucoma is a common and progressive eye condition that represents a major global health challenge, frequently resulting in irreversible vision loss due to late diagnosis. To tackle this issue, we introduce an innovative web application developed using Flask, which harnesses Convolutional Neural Networks (CNNs) combined with transfer learning. This application categorizes the severity of glaucoma into four distinct levels: pre-glaucoma, mild, moderate, and severe, providing users with a straightforward interface for uploading eye images for analysis. In addition to classification, the system delivers tailored medical recommendations based on the predicted severity, advising patients on potential treatments, lifestyle adjustments, and the necessity of regular eye examinations. Our method not only promotes early detection of glaucoma but also empowers individuals to take charge of their eye health, potentially decelerating the progression of the disease. Existing systems typically depend on image processing techniques and CNNs for detecting glaucoma, often classifying patients simply as "Glaucoma Detected" or "No Glaucoma Detected." However, these systems face challenges such as limited interpretability, high implementation costs, dependence on human expertise, and a focus on specific imaging techniques, which can diminish their effectiveness. To overcome these challenges, our proposed solution introduces advancements in data collection, preprocessing of diverse datasets, selection of a custom CNN architecture, and thorough training, validation, and testing of the model. Our system goes beyond binary classification by integrating glaucoma severity categorization, offering personalized health recommendations, and featuring a user-friendly web interface. This multi-class classification strategy improves diagnostic accuracy and ensures the system's relevance across various clinical settings. By incorporating a multi-class classification approach, our system enhances diagnostic accuracy and ensures applicability across various clinical scenarios, ultimately improving patient outcomes in glaucoma management.