Research On Methods Of Fuzzy Clustering For Segmentation Of Brain Tissue Using MR Images

With the rapid development of computer technology,magnetic resonance imaging(MRI)technology has been widely used in various stages of medical research and clinical application.Among them,medical image segmentation technology is the basis of other medical image analysis technologies,and has a very important position in the MR image analysis process.The quality of image segmentation will directly affect the quality of other auxiliary diagnostic results.This topic studies the brain tissue segmentation technology of MR images,and aims to segment brain tissue into three categories of white matter,gray matter,and cerebrospinal fluid through a clustering algorithm.In brain MR images,due to noise,bias field effects,partial volume effects,and differences in the physical structure of the human brain,the edges of image tissue will become blurred,making it difficult to accurately segment brain tissue.This thesis focuses on the problem that the fuzzy C-means clustering algorithm(FCM)will be affected by noise and bias field effects when segmenting brain MR images.The main research results obtained are as follows:(1)Aiming at the problem of low accuracy of FCM algorithm for the segmentation of noisy images,an adaptive nonlinear weighted local information fuzzy C-means algorithm is proposed.The algorithm fully considers the local grayscale information and local spatial information of the image,constructs a new similarity measure,uses the similarity measure to construct a local fuzzy factor,and applies the fuzzy factor to the fuzzy local information C-means algorithm(FLICM).And because a new distance measure is used to describe the distance between the pixel and the cluster center,the segmentation accuracy of the algorithm is further improved.From the experimental results show that the algorithm can obtain good brain tissue segmentation results and has strong noise suppression ability.(2)Aiming at the effects of brain MR image bias field,this thesis introduces local bias field information on the basis of bias field correction FCM algorithm(BCFCM),a bias field correction FCM algorithm is proposed.This algorithm adds a constraint item containing local bias field information to the objective function of BCFCM,so not only the advantage of BCFCM algorithm in the bias field image is maintained,but also because the addition of local bias field information constraint item makes the estimation of the bias field is smoother,so that the segmentation of the image edge area is more accurate.From the experimental results show that the improved algorithm has better segmentation result than BCFCM algorithm.This topic studies the brain tissue segmentation technology of MR images.Aiming at the two problems of noise and bias field,two improved algorithms are proposed to improve the accuracy of the algorithm for segmentation of noise images and bias field images.