The Research Based On Adaptive Mechanism Of Level Set Algorithm For Medical Image Segmentation

Due to the constraints of medical imaging technology and the complexity of human organ structures,it is the complex diversity and dissimilarity of medical images that makes image segmentation more difficult.The variational level set algorithm is widely used in the field of image segmentation with its flexible topology adaptability,convenient curve evolution energy allocation scheme,simple and effective numerical solution method.However,there are still two shortcomings with the application of the variational level set algorithm in medical image segmentation field.Firstly,the lack of generalization ability has significant influence on the sensitivity to the initial contour and evolution control parameters.Secondly,the optimization degree of the driving force of the level set curve evolution.Although many variational level set methods based on edge information,region information,or integrated edge and region information have been developed,it is still a question to optimize the driving force of level set curve evolution to achieve rapid and effective convergence.Based on two classical variational level set algorithms(MS-RSF model and Local Gaussian Distribution Fitting model),this paper studies the above two problems.The specific work is as follows:(1)As for the driving force of the MS-RSF level set curve evolution,this paper puts up with a solution that adds an adaptive weight coefficient function to the energy equation,which can adaptively control the direction of the curve evolution.At the same time,an adaptive correlation function is set to control the iterations of the curve evolution adaptively.According to the experimental results,the adaptive MS-RSF model possesses accurate precision and high segmentation speed for the medical images with noise,intensity inhomogeneity and weak boundary.(2)As for the sensitivity to the initial contour and the generalization ability of the parameters in the Local Gaussian Distribution Fitting(LGDF)model,this paper adapts the Mean Shift clustering results to perform the initial contour setting of the LGDF algorithm,which makes it closer to the target contour than manual initial contour.Based on the local mean and variance information of the images,the LGDF model establishes an energy function and the evolution control parameter is changed to a function related to the number of clusters.Then,this paper takes advantage of the initial contour to perform level set evolution to realize the aim that reduces human-induced interference,increases automation degree and promotes robustness of MS-LGDF method for medical images.