A Neural Network-Based Intelligent System For Diagnosis of Selected Eye Diseases

Abstract

The human eye is a vital organ of vision useful for most daily activities. However, there are many diseases, including cataracts, glaucoma, diabetic retinopathy, age-related macular degeneration, and retinoblastoma, which affect the human eye and can lead to blindness when not diagnosed early for treatment. The advent of digital health techniques has led to the provision of fast, costeffective, accurate and automated diagnosis of common human diseases. This paper discusses the development of an intelligent system for the diagnosis of glaucoma and diabetic retinopathy eye diseases. The diagnosis is done using digital image processing techniques to analyse fundus eye images of suspected patients and Artificial Neural Network to classify the images as infected or not. The system prototype was implemented as a stand-alone application using MATLAB software and experiments conducted on three publicly available fundus image databases. An average accuracy of 95% was obtained with a neural network that classifies an input eye image as healthy, glaucoma or diabetic retinopathy infected. We intend to predict the stage of eye infection and investigate the use of non-fundus images for diagnosis in our future work.