Automated Classification of Glaucoma Using Retinal Fundus Images

Abstract

Glaucoma is an irreversible chronic eye illness that prompts vision loss. It progresses slowly without easily noticeable symptoms. Computer-aided diagnosis (CAD) of glaucoma in the early stage is needed, which is fast and more accurate. In this work, the empirical wavelet transform (EWT) and correntropy (CE) feature-based novel method has been proposed for the classification of glaucoma stages. In the proposed method (PM), the preprocessed fundus images are decomposed into various frequency components using EWT decomposition technique. Further, correntropy based features are calculated from decomposed EWT components. Afore, student's t-test algorithm has been applied for the selection of significant features and features with higher t value are ranked first. Finally, random forest (RF) classifier is used for classification of glaucoma stages. The obtained classification accuracy using tenfold cross-validation is 91.48% and 94% for two-class and three-class classification, respectively. The proposed method is ready to assist the ophthalmologist to diagnose glaucoma.