Computer Aided Diagnosis of Aging Macular Deterioration Via Convolutional Neural Network

Abstract

Aging Macular Deterioration (AMD) is a leading eye problem most commonly experienced by the old age people. If the problem is untreated over a prolonged time period, it results in permanent blindness. This eye problem is caused due to the damage of macula lutea which is a central region of retina needs for visualizing very fine details. However, only early detection can exhibit it from becoming severe and protect vision. This method proposes an automatic screening of all the three stages of AMD (i.e.) early (DMD), intermediate and late (WMD) using Convolutional Neural Network. A set of 400 color fundus images are taken for experimentation out of which 190 images are affected AMD images and 210 images are non-AMD images. Here, first the images are subjected to an image segmentation technique which adds-on the advantage of improving the accuracy of the system. Fuzzy c-means clustering is used as the image segmentation technique. Then the segmented images were trained and experimented using Convolutional Neural Network. This work thus obtained an overall accuracy of about 95.65%. The experimental results verify the effectiveness of this method.