OptiGuard: Generalized, Attention-Driven & Explainable Glaucoma Classification

Glaucoma, often dubbed the silent thief of sight, encompasses a spectrum of ocular disorders causing irreversible optic nerve damage, often eluding early detection due to its asymptomatic nature. While conventional diagnostic methods like Optical Coherence Tomography (OCT) and Visual Field Tests (VFT) are effective, their limited accessibility can result in delays in early intervention. The growing patient volume and scarcity of ophthalmologists have spurred research into automated glaucoma detection systems using Retinal Fundus Images (RFI), a promising and cost-effective alternative. However, existing automated solutions face challenges including limited public datasets, poor model generalizability, and lack of explainability. This paper introduces OptiGuard, an advanced deep learning-based Computer-Aided Diagnosis (CAD) system integrated with an intuitive web interface. OptiGuard leverages the G1020 dataset for Optic Disc and Cup segmentation accomplishing state-of-the-art results and the SMDG-19 dataset for robust glaucoma classification, achieving a generalized model with strong cross-dataset performance. Notably, OptiGuard provides visual and quantitative explanations, enhancing transparency and fostering trust among healthcare professionals.