Machine learning framework for Diabetes Retinopathy detection using hybrid logistic regression

Abstract

Diabetes Retinopathy(DR) occurs due to an injury to the retina, ultimately steering towards sightlessness. DR does not provide any early clues or signs and unnoticeably associates a new blood vessel towards the back portion of the eye, resulting in clots of blood in the eye, bleeding of eyes and distorted vision. The conventional methods are failed to produce the maximum classification accuracy. Therefore, this article is focused on implementation of hybrid logistic regression (HLR)-based machine learning model for classification of DR.Initially, histogram equalization is used to enhance the region of DR image. Then, segmentation of microaneurysmsis performed by using image morphological operations. Further, features extracted using gray level co-occurrence matrix (GLCM), which shows the internal relationship of DR disease. Then, selection of features is carried out using the Gaussian Mixture Model (GMM). Finally, HLR model is applied to perform the multi class classification operation. Simulation results shows that the proposed method resulted in superior performance as compared to state of art approaches