Automatic Features Extraction from the Optic Cup and Disc Segmentation for Glaucoma Classification

Glaucoma is a disease that progressively affects the optic nerve, the leading cause of blindness worldwide. One of the most assertive strategies to make the diagnosis is Optical Coherence Tomography (OCT) which identifies anomalies in the anatomy of the optic nerve. OCT is a high-cost exam, so some works in the literature have been using computationally expensive deep neural networks to analyze images on retinal fundus images to diagnose glaucoma. As an alternative to these approaches, in this work, we propose a low-cost computational method for extracting characteristics of the optic nerve anatomy (i.e., optic cup and disc segmentation) through the processing of retinal fundus images, which is used in conjunction with lower computational cost classification algorithms (i.e., support vector machine (SVM)), is capable of performing accurate diagnoses. The most dominant attributes were identified using shapely adaptive explanations (SHAP) and local interpretable model-agnostic explanations (LIME) analysis. More specifically, the more precise the extraction of features, the greater the accuracy of the classifier.