Research On Algorithm For Medical Image Segmentation Based On Fuzzy Cluster And Level Set

Medical image segmentation is one of important research topics in image processing.And it has a wide applications and research on clinical diagnosis,medical image 3-D reconstruction and visualization,medical research,etc.Due to the particularity of imaging equipment and imaging environment,medical images have complexity and ambiguity,which makes it difficult for traditional image segmentation algorithms to obtain satisfactory results when segmenting medical images.Image segmentation algorithm based on fuzzy C-means(FCM)can describe the ambiguity of medical images well.However,the FCM algorithm has many drawbacks such as poor denoising ability,easy to lost details,low accuracy and so on.This thesis mainly studies the FCM algorithm,and combines the level set method to improve the algorithm’s segmentation ability.The main works are as follows:(1)The FCM algorithm and its improved algorithm are studied.The defects of FCM algorithm are summarized,and its improved algorithms such as En FCM algorithm,FCM＿s algorithm and FLICM algorithm are introduced.By performing segmentation experiments on medical images,the performance of FCM algorithm and several improved algorithms are analyzed.(2)A FCM algorithm based on mean spatial constrains and adaptive weighted kernel function is proposed.In order to improve the ability of the FCM algorithm to segment the complex structure of medical images and overcome the effect of noise on the segmentation quality,an improved algorithm is proposed.The improvement of the algorithm includes three parts:Firstly,the algorithm adds a Kernel function and replaces the Euclidean distance with the Kernel distance,making the algorithm can process the nonlinear region in the medical image;Secondly,the algorithm adaptively weights each cluster,making clusters with small sample size not easy to be misjudged,and improving the algorithm’s ability to describe details;Finally,the algorithm adds a mean spatial constrains to the objective function,making the denoising ability of algorithm is enhanced.The algorithm in this thesis can accurately segment medical images,and has strong accuracy and denoising ability.(3)A medical image segmentation algorithm based on FCM and level set method is proposed.The FCM algorithm has poor ability to describe the boundary.Considering that the level set method can obtain smooth boundaries through curve evolution,this thesis combines the FCM algorithm with the level set method.For the sensitivity of the CV model to images with intensity inhomogeneity,a CV model based on local area information is given.Then,an algorithm combined the improved FCM with the improved CV model is proposed.The main principle of the algorithm is: Firstly,use the FCM algorithm to segment the medical image for the first time;Secondly,select the clustering information of the target area to initialize the zero level set function,and use it as the initial contour of the CV model;Finally,use the CV model for the second image segmentation.Through curve evolution,accurate segmentation results can be obtained.The algorithm makes up for the shortcomings of the fuzzy clustering algorithm in boundary description,and also overcomes the excessive dependence of the level set method on the initial contour.The algorithm in this thesis can accurately locate the regional boundaries of medical images,and has better accuracy and efficiency.