Performance Evaluation of Binary Classification of Diabetic Retinopathy through Deep Learning Techniques using Texture Feature

One of the main causes of loss of vision in diabetic patients is Diabetic retinopathy (DR). Automated methods are important medical applications for detecting and classifying the disease type into normal or abnormal ones. Fundus images are obtained from the retina using a retinal camera, one of a non-invasive diagnostic technique that offers a way of examining the retina in diabetes patients. We present in this paper a system for the detection and classification of DRs. Our approach is divided into two main steps: in the first step, we use local binary patterns (LBP) to extract texture features, while in the second stage, we analyze extensively the state-of-the-art deep learning techniques for the detection and classification tasks. ResNet, DenseNet, and DetNet are used as deep learning techniques. Preliminary results show that ResNet, DenseNet and DetNet can obtain 0,9635%, 0,8405% and 0,9399% of accuracy, respectively. In addition, we also evaluate the performance of each detection configuration.