RETINAL VESSEL SEGMENTATION AND CLASSIFICATION BASED ON MODIFIED FUZZY C MEANS ALGORITHM AND CONVOLUTIONAL NEURAL NETWORK

Currently, retinal vessels play an essential role in the diagnostic process of retinopathy. Though, precise segmentation of retinal vessels is vital for pathological examination. In this research, a new system was developed for precise blood vessel segmentation and categorization. Firstly, the retinal images are collected from Digital Retinal Images for Vessel Extraction (DRIVE) and STructured Analysis of the Retina (STARE) datasets. Then, modified Fuzzy C Means (FCM) clustering algorithm was utilized to segment the blood vessels from the collected images. Besides, feature extraction was accomplished by using Enhanced Local Binary Pattern (ELBP) and Inverse Difference Moment Normalized (IDMN) to extort the features from the segmented images. Lastly, the attained feature values are forwarded as the input for Convolutional Neural Network (CNN) classifier to identify the normal and abnormal blood vessels and it's abnormality. An experimental conclusion of the proposed system also shows that the improvement of classification accuracy up to 1.6% to 1.8% as related to the existing system.