A Classification Approach Using Fuzzy Inference System for Non-Proliferative Diabetic Retinopathy

Abstract

This paper proposes a classification technique using fuzzy inference system (FIS) for non-proliferative diabetic retinopathy (NPDR). The input variables to FIS are the extracted features from the pathologies of NPDR. Abnormalities like exudates and microaneuryms are segmented for feature extraction. The pathological aspects of NPDR leads to fuzzy if-then rules that effectively handles the fuzziness present in some of the features. The role of FIS in NPDR classification replaces the training phase in learning based classification methods. The performance of the proposed fuzzy approach is analyzed for stages of NPDR and diabetic retinopathy classification on various databases. The accuracy of 98.2% is observed for NPDR classification in Messidor database.