Analysis and Detection of Glaucoma from Fundus Eye Image by Cup to Disc Ratio by Unsupervised Machine Learning

Abstract

The cup nerve head, eyecup, neural rim shape, and eye disc ratio are useful in identifying glaucoma early in medical practice. The most important medical sign of glaucoma is the eyecup to eye disc ratio, which is presently measured manually by restricting the bulk screening. The following approaches for automatically determining the eye to disc ratio are proposed in this work. The subcapsular image of the eye disc area is studied in the first section. K is utilized robotically to identify the eyedisc in decluttering, whereas K value is continuously determined using a hill-climbing method. Two approaches, morphological and elliptical fitting, have been used to smooth the segmented shape of the eyecup. The eye to disc ratio of glaucoma patients was computed for 60 ordinary images and 60 subcapsular images. The same set of photographs has been utilized throughout this work, and the eye to disc ratio values given by an ophthalmologist has been used as the golden standard value. Throughout the study, the mistake is computed by comparing the eye to disc ratios to this global standard number. For the morphological fitting and elliptical fitting, the average error of the K-means clustering approach is 4.5 percent and 4.1 percent, respectively. The inaccuracy can be further minimized by taking into account the inter-pixel relationship. Another approach is the method used to achieve the aim is Spatially Weighted Fuzzy C-means Clustering (SWFCM). Fuzzy C-mean clustering was chosen because to the huge error, and the method's mean error for morphological and elliptical fitting is 3.52 and 3.83 percent, respectively. SWFCM Clustering has clustered and segmented the eye disc and eyecup. The SWFCM mean error clustering method yields 3.06 percent and 1.67 percent, respectively, for the morphological and elliptical fitting. Sub capsular pictures were obtained from a famous eye hospital in pondy named Aravinda Eye Hospital for this purpose.