A Hybrid Clustering-based Approach for Segmentation of Optic Disc and Optic Cup

Glaucoma is one of the leading causes of vision loss. It is caused due to an increased Intra Ocular Pressure (IOP), which in turn damages the Optic Nerve Head (ONH) and progresses to blindness, if left untreated. Hence, early diagnosis of Glaucoma is extremely important. The Optic Disc (OD) and Optic Cup (OC), which are integral parts of the ONH, are the major Regions of Interest (ROIs), from which clinical indicators of Glaucoma such as Cup-to-Disc Ratio (CDR) are assessed. Manual diagnosis of Glaucoma is a highly challenging and involved task. Automated segmentation of OD and OC plays a vital role in Computer Aided Diagnosis (CAD) of Glaucoma, and can reduce the time and burden involved. In this paper, a hybrid segmentation algorithm that integrates K-Means and Mean Shift clustering algorithms, is proposed to alleviate the drawbacks in existing OD and OC detection techniques. The proposed approach has been validated on RIM-ONE r3 dataset, achieving accuracies of 97.75% and 98.64% for OD and OC segmentations respectively, and a Mean CDR Error of 1.55%.