A CDR - Based Image Processing Approach for Glaucoma Detection: Validation on DRISHTI-GS and ACRIMA Datasets

Abstract

Glaucoma is among the most frequent causes of vision loss in patients who suffer from retinal diseases. This disease is highly dangerous since in its initial phase it manifests itself asymptotically and leads to irreversible vision loss. Thus, early detection of glaucoma is an essential step that will help detect and cure the disease. In this work, an algorithmic method based on computer-aided technology has been proposed for glaucoma diagnosis using retinal fundus images. The main idea is to design a simple glaucoma screening method. Firstly, pre-processing is applied to the retinal image in order to extract regions of interest, remove noises, enhance the contrast and smooth the image using Gaussian filter. Image processing techniques are applied for segmentation of optic disc and optic cup. The Cup-to-Disc Ratio (CDR) is then calculated, which is an important clinical parameter in diagnosing glaucoma. The fundus image is classified as normal, borderline, or glaucoma on the basis of calculated CDR value. The outcome clearly indicates that the proposed method is accurate, efficient, and useful in detecting glaucoma using computer-aided approach.