Multi-Class Peripapillary Atrophy for Detecting Glaucoma in Retinal Fundus Image

Abstract

The characteristics of glaucoma that can be observed through a fundus image is a peripapillary atrophy (PPA). Hence, identifying glaucoma based on fundus images can be carried out by observing the PPA occurrence. Research conducted by detecting the presence of PPA has been done a lot but still uses two classes of PPA namely no-PPA and PPA. It cannot distinguish between mild-PPA and severe-PPA which causes treatment equalization. Support Vector Machine (SVM) has had success in classifying PPA, so it will be used to classify PPA from the retinal fundus image dataset into multi classes, i) no-PPA, ii) mild-PPA and iii) severe-PPA. Multiclass PPA classification can detect glaucoma and also know the severity which then determines the treatment and treatment to be carried out. Testing on two datasets containing 110 images of retinal fundus (85 as the training-set and 25 as the testing-set), the proposed method gives high accuracies of 95% and 94% respectively.