Modular Neural Network for Detection of Diabetic Retinopathy in Retinal Images

Abstract

Modular feedforward network method is introduced to detect diabetic retinopathy in retinal images. In this paper, the authors present classification method; the Modular Feedforward Neural Network (MNN) to classify retinal images as normal and abnormal. Publically available database such as DIARETDB0 including high-quality normal and abnormal retinal images is taken for detection of diabetic retinopathy. Modular Feedforward Neural Network is designed based on the extracted features of retinal images and the train N times method. The classification accuracy by MNN classifier was 100% for normal retinal images and 86.67% for abnormal retinal images. In this paper, the authors have explored such a method using MNN classifier which can detect diabetic retinopathy by classifying retinal images as normal and abnormal.