Detection of Multi-Class Glaucoma Using Active Contour Snakes and Support Vector Machine

Abstract

There are several ways to detect glaucoma, one of the most accurate is the presence of peripapillary atrophy (PPA). PPA is located outside the optic disc around the optic nerve head (ONH) and sometimes looks vague which can cause misclassification, so other parameters that can detect glaucoma are needed. The calculation of the optic cup to disc ratio (CDR) is mostly done for glaucoma detection so that CDR can be considered in addition to the presence of PPA to improve classification results. In this paper, a multi-class glaucoma detection is developed using an active contour snake to get the value of the optic cup and optic disc to measure CDR and a support vector machine (SVM) for classification. Glaucoma is categorized into three classes: non-glaucoma, mild-glaucoma, and severe-glaucoma. Hence, the model can detect its severity which determines further treatment. Evaluation using two datasets of 210 retinal fundus images (165 train and 45 test) informs that the model reaches high accuracies of 95%.