Segmentation of Optic Cup and Disc for Diagnosis of Glaucoma on Retinal Fundus Images

Abstract

Glaucoma has been attributed to be the leading cause of blindness in the world second only to diabetic retinopathy. About 66.8 million people in the world have glaucoma and about 6.7 million are suffering from blindness as a result of glaucoma. A cause of glaucoma is the enlargement of the optic cup such that it occupies the optic disc area. Hence, the estimation of optic Cup to Disc ratio (CDR) is a valuable tool in diagnosing glaucoma. The CDR can be obtained by segmenting the optic cup and optic disc from the fundus image. In this work, an improved U-net Convolutional Neural Network (CNN) architecture was used to segment the optic disc and the optic cup from the fundus image. The dataset used was obtained from the DRISHTI-GS database and the RIM-ONE v.3. The proposed pipeline and architecture outperforms existing techniques on Optic Disc (OD) and Optic Cup (OC) segmentation on the Dice-score metric and prediction time.