Retinal vessel segmentation based on adaptive fuzzy local information C-means clustering

Retinal vessel morphological changes correlate closely with ocular and cardiovascular diseases, aiding in their evaluation, screening, and diagnosis. In order to extracted the information from retinal vessels, a retinal vessel segmentation method based on adaptive fuzzy local information C-means clustering is proposed. After using contrast-limited adaptive histogram equalization for image enhancement, the features of retinal vessels are extracted by B-COSFIRE filters. Finally, the segmentation of fundus vessels is achieved by adaptive fuzzy local information C-means clustering algorithm. On DRIVE and STARE datasets, the method proposed achieve the average sensitivity of 0.6841 and 0.7585, the average accuracy of 0.9458 and 0.9463, respectively. The proposed method provides good segmentation performance, and its segmented vascular network has good integrity and continuity. Compared with the feature space FCM method, our method has significantly improved the sensitivity of detecting retinal blood vessels.