Feature Selection for Simple Color Histogram Filter based on Retinal Fundus Images for Diabetic Retinopathy Recognition

Abstract

Applications of learning models for text-based datasets as well as image pixels-based datasets grow rapidly for prediction purposes. Pre-processing becomes challenging in carrying out image filtering and classifying. Retinal Fundus images plays important role in Diabetic Retinopathy (DR) diagnosis and treatment planning in various stages. Diabetic Retinopathy is diagnosed by observing the variation in retinal blood vessel, exudates, micro aneurysm, hemorrhages, and the new blood vessel growth inside the retina. The objective of this study is to enrich the diagnosis for the Diabetic Retinopathy from the retinal fundus images by applying machine learning algorithms. The proposed work implements normalization, parameter tuning, and optimal feature selection method to improve the classification accuracy offered by selected algorithms like decision tree algorithm and K-nearest neighborhood classifiers. The highest accuracy of 81.99%, Weighted Average of Receiver Operating Characteristics (ROC) 0.907 are obtained by k-Nearest Neighbor (KNN) classifier due to its best performance.