Diabetic Retinopathy Classification Using Deep Learning Techniques to Enhance Findings with Color Normalization Techniques

Abstract

Diabetic retinopathy (DR) is a widespread vision problem due to diabetes that can lead to vision loss if it is  not detected and treated early. In the last few years, deep learning techniques have shown promising results in the automation of the detection and classification of diabetic retinopathy from retinal images. However, the performance of these techniques can be further improved by addressing the challenges due to the color variations in retinal images. This paper presents an implementation of a diabetic retinopathy classification system that combines color normalization techniques, namely Histogram Equalization and Contrast Stretching, with Convolutional Neural Networks (CNNs). The proposed approach aims to enhance the findings by reducing the impact of color variations on classification accuracy.