Machine Learning Based Approach for Disease Diagnosis of Human Retina

Abstract

The DR is a rapidly growing interrogation around the world which can be annotated by abortive metabolism of glucose that causes long-term infection in human retina. This is one of the preliminary reason of visual impairment and blindness of adults. Blood vessels alteration and retinal lesions are major disorders in retina that has been occurred due to diabetes. Most of the current systems don't have the ability of classifying severity of DR and also vessel tortuosity measurement of untwisted vessels. In this research, we are mainly focused on resurrecting a novel approach for severity classi cation of DR patient. Retinal fundus images and corresponding diagnostic stages are used as input to the system. These retinal images and diagnostic stages of DR patients are acquired from the National Eye Hospital, Vision Care (Pvt) Ltd and publicly available databases. The preprocessed color retinal images are segmented utilizing adaptive K-means clustering algorithm. A set of features which based on texture and color on the surface of the fundus image are extracted and fed it in to the classi er to classify the severity level of patient. The classi cation of severity is done utilizing ANN approach. In addition to that we also proposed a novel methodology for vessel tortuosity measurement of untwisted vessels to assessment of vessel anomalies in e ortlessly. In order to evaluation process severity classi cation method obtained better results according to the precision, recall, F-measure and accuracy in all formats of cross validation. In ROC curves also visualized the higher AUC percentage. In user level evaluation of severity capturing can be obtained higher accuracy result and fairly better values for each evaluation measurements. Untwisted vessel detection for tortuosity measurement also carried out the good results with respect to the sensitivity, speci city and accuracy.