Research Of Medical Image Segmentation Based On Cluster Analysis

Medical image segmentation is an important application of image segmentation technology based on the field of medical image processing.It is a method to extract the information of the region of interest in medical images,and is a crucial step in the process of medical image processing from image acquisition to image recognition.Medical image segmentation method has been paid great attention by researchers since it was proposed at the beginning.Segmentation method based on a lot of image information and mathematical theory has been put forward.Among all kinds of segmentation methods,the image segmentation algorithm based on clustering analysis is an appropriate application of clustering theory in the field of image segmentation.For the problem of image segmentation,the essence of clustering method is to divide pixels according to the similarity of features between pixels so as to achieve the purpose of image segmentation.However,for the present clustering algorithm itself,there are many problems,which seriously affect the computational efficiency and segmentation effect of image segmentation.In recent years,with the development of machine learning algorithm,the research of medical image segmentation based on cluster analysis has become a hot topic in the field of medical image.This paper mainly studies two kinds of clustering analysis methods commonly used in medical image segmentation.Firstly,the application of GMM in medical image segmentation is studied,and the initialization method and parameter optimization algorithm commonly used in Gaussian mixture model are improved.Secondly,a genetic fuzzy clustering segmentation method based on spatial information is proposed based on the traditional FCM model.Finally,based on the platform of MATLAB2018 a,the new model in this paper is combined with medical images for segmentation experiment.From the experimental results,the improved model segmentation effect in this paper is better than other methods.(1)Through theoretical analysis,programming,implementation and experiment comparison,the study such as threshold method,the area method,the edge method,neural network segmentation algorithm,graph theory,algorithm,clustering analysis of medical image segmentation performance,comprehensive analysis and according to the subjective and objective evaluation,finally choose the comprehensive analysis of the performance of the optimal clustering analysis as a segmentation algorithm of further improvement.(2)Because of the difficulty in solving parameters and Numbers in image segmentation and the difficulty in getting into the optimal solution of game and part,the process ofparameter optimization is slow.To Gaussian mixture model segmentation rapid convergence,this paper use the K-all get together to complete the initial division,and according to the classification of pixel values after the initial value of iterative EM algorithm are given to speed up the iteration algorithm to the optimal solution,thus greatly reduce the number of iterations algorithm and effectively solve the EM algorithm to solve the parameters of the random initial value point that GMM into local optimal solution of the problem,and then make the segmentation region is complete,at the same time,before the segmentation,using anisotropic filtering for image preprocessing,denoising smoothing and enhance the image of detail.Finally,an improved image segmentation algorithm based on Gaussian mixture model is proposed.(3)In the process of image segmentation,the traditional fuzzy c-means clustering algorithm ignores the effect of spatial information on image segmentation performance,resulting in poor image segmentation accuracy and easy to be affected by noise.The method USES the improved genetic algorithm to initialize clustering center,improve the efficiency of the division of the algorithm,and combination of spatial information improved FCM membership function,improve the anti-noise performance of the traditional fuzzy clustering segmentation model,at the same time,compared the improved algorithm in the add noise data clustering,add noise of synthetic images,medical image segmentation,the experiment results,this method was verified under different experiment obtained the better segmentation effect.