Evaluation of Cup to Disc Ratio for Glaucoma detection through Optic Cup and Disc Segmentation using K-means Clustering with Morphological Operations and comparing with Adaptive Thresholding method

The aim of the study is to develop an automated method to detect Glaucoma from fundus images by using K-means clustering and comparing with adaptive thresholding. Materials and Methods: This study focuses on detection of glaucoma from fundus images by segmenting optic cup and optic disc. K-means clustering and adaptive thresholding is used to locate the optic cup and disc in fundus images. The total sample size is 40 and pretest power for the sample size is 80 %. Results: The accuracy obtained for K-means clustering with morphological operations is 91.1177 % and 83.6288 % for adaptive thresholding. Significance value of accuracy is obtained as 0.000 (p<0.05, 2 tailed). Conclusion: It is observed that the accuracy of the proposed K-means clustering with morphological operations is significantly better than adaptive thresholding.