Hemorrhage segmentation by fuzzy c-mean with Modified Level Set on CT imaging

Abstract

Intracranial hemorrhage is the primary cause of death for patients in early age. Lives of these patients depend on proper or accurate detection of Intracranial Hemorrhage (ICH). This paper presents a hybrid method of fuzzy c mean (FCM) clustering and modified version of distance regularize level set evolution (MDRLSE). Membership values and cluster center are used in FCM clustering for dividing hemorrhagic images into different clusters. One cluster is selected from the FCM clustered images which is used for initialization of MDRLSE. Level set function used in MDRLSE does not require re-initialization process. So convergence speed of active contour propagation becomes faster. Numbers of iterations are reduced in level set method because large time step is used by the MDRLSE method. Proposed hybrid method is evaluated on CT mages of 20 patients suffering from ICH. Result of proposed method provides highest accuracy as compared with FCM method and thresholding method.