Computational Analysis of Segmenting Retinal Blood Vessels Using Fuzzy C-Means

Abstract

Segmentation is a critical phase of the image processing system as it excerpts the objects of one’s interest. The image segmentation results in vigorous and high grade of accuracy and is very much helpful in analyzing diverse image modalities. In diabetic retinopathy, the segmentation is conducted on the blood vessels of the retinal images which assists in classifying the retinal images as benign or malignant. The images have been obtained from the Digital Retinal Images for Vessel Extraction retinal dataset. The dataset comprises 20 retinal images. The research paper proposes a hybrid segmentation technique using Fuzzy C-Means and Haar feature extraction. The input retinal images are enhanced using a Gaussian filter and exposed to Fuzzy C-Means for segmentation. The Haar feature extraction is applied to the segmented retinal image to extract the blood vessels. The conducted segmentation has been analyzed by obtaining the readings for the 8 performance evaluation parameters. The proposed segmentation technique achieved an average specificity of 94.965% with an average accuracy of 85.77% and a Negative Predictive Value of 94.197%.