Early and Efficient Detection of Glaucoma Using Image Processing and Deep Learning

Abstract

A Chronic eye disorder called glaucoma leading to irreversible blindness by damaging the optic nerve of the eye. It is provoked due to exalted intraocular pressure inside the eye. Detecting glaucoma is the most challenging process in case of open angle glaucoma (OAG) due to lack of initial symptoms. Detecting glaucoma in the early stage is required to facilitate appropriate monitoring, treatment, and to diminish the likelihood of vision loss. In this paper, we propose a method to analyse and categorize the fundus image as glaucomatous or healthy image by considering cup to disc ratio using image processing techniques and feature extracted through Deep learning. The assessment of CDR is the foundation to detect glaucoma, the CDR value will increase from 0.6 – 0.9 when affected by this disease. In order to consider other medical parameters for glaucoma detection and to automate the detection process Deep Learning-Convolution neural network model is implemented. Overfitting is avoided by adopting data augmentation technique. To make the system user friendly and interactive Graphical user interface (GUI) application is developed. The system is trained and the results demonstrate that the technique had a good accuracy in classifying the fundus images as healthy or glaucoma.