Research On The CT Image Segmentation Of Brain Tumor

Because of the complexity of brain structure,individual differences of patients and the variability of tumors,the segmentation of CT images has been a difficult problem.Based on the research status of CT image segmentation and the deficiency of the variational level set method,the adaptive level set of the variational method is presented in this paper of combining theoretical analysis and numerical experiment.Firstly,something important,which contains the research background,the significance,the imaging principle,characteristics of CT image,CT value distribution,the classification of brain tumor,classical image segmentation method,the research status and development trend of CT image segmentation,has been introduced systematically for the future research work.Secondly,the basic theory of the variational level set method,which includes curve evolution theory,level set theory and variation theory,is introduced.The mathematical model of the traditional variational level set method is presented and its advantages and disadvantages are analyzed.For traditional variational level set method can’t adaptively adjust the size evolution direction and speed,an adaptive level set method is presented.This improvement is mainly shown in three aspects: the external energy variational parameters from constant to functions related image gradient information;the Dirac function domain has a wider domain;the edge stopping function goes to zero faster.Finally,for CT images of brain tumors of four patients,threshold method,edge operator method,k-means clustering method and watershed method were first used to verify segmentation.The characteristics of these classic segmentation methods on brain tumor CT images were summarized.On this basis,the contrast experiment results both the variational and adaptive variational level set method are presented.The results show that the adaptive variational level set method has fast iterative speed and numerical stability,does not depend on the location of the initial curve,can automatically according to the real-time image information to determine the size of the curve evolution method and evolution speed,can detect multilayer contours of objects,narrow deep sag area and weak edge profile.