Comparative Analysis of different Machine Learning Classifiers for Prediction of Diabetic Retinopathy

Abstract

Diabetes is a metabolic disorder which leads to high blood glucose level. This condition leads to a number of secondary complications like Diabetic Retinopathy (DR), Neuropathy, etc. DR commonly affects the eyes. Out of various characteristics of Retinopathy, exudate formation is a major concern, as it leads to functional loss of retina and eventually leads to visual impairment. Thus, it is important to detect the presence of exudates in the eye region which can be done by processing of fundus images. In this paper, the detection of exudates in the fundus images has been done through the implementation of various blocks of CAD system using python. The designed system classifies the given input image as Healthy (class 0) or DR affected (class 1) by considering the existence of exudates in the fundus image. Also, performance of machine learning classifiers like KNN, SVM Linear, SVM Polynomial and Decision Tree have been compared through various performance metrics.