DETECTION OF GLAUCOMA FROM RETINAL FUNDUS IMAGES USING STRUCTURAL FEATURES

Abstract

Glaucoma and Diabetic retinopathy are the two major eye diseases that affect human eye. Irreversible damage and partial loss of vision is caused by these diseases if not treated at the earliest. Glaucoma is a painless neurological disease in which fluid pressure in the eye increases constantly, which damages the optic nerve and thereby affects the sight of the patient. It is considered as the one of important reason for blindness. Conventional screening methods enable us to identify these diseases only after it has caused partial damage (25% or more) to the eye. Detection of glaucoma at the earliest stages using computational decision support systems can help overcome this problem. Highly specific quantitative information of the Optic Nerve Head (ONH) or optic disc (OD) and other retinal structures can be obtained using fundoscopy. This paper proposes the detection of glaucoma using structural features extracted from retinal fundus images. The optic cup to disc ratio, rim to disc ratio and cup shape analysis are used as parameters for identifying the symptoms of glaucoma from the fundus image. The extracted features are fed to Support Vector Machine classifier, with which an accuracy of 80% is obtained.