An Improved 3D Medical Image Segmentation Method

Brain diseases have gradually become the worst killer endangering human life and health,and as an effective tool for disease diagnosis,to realize the automatic segmentation of threedimensional images has become the current research hotspot.Based on the fuzzy c-means clustering(FCM),this paper discusses the existing problems,studies the existing improvement methods,based on its shortcomings proposes the 3D neighborhood weighted Image segmentation algorithm for 3D medical image segmentation.First of all,by studying the principle of FCM algorithm,this paper analyzes and expounds the shortcomings of FCM algorithm,such as its sensitivity to noise due to the lack of neighborhood information,its results are easy to converge to local optimum and its parameters are not adaptive.By studying the existing neighborhood weighting algorithm,this paper points out that it has the disadvantages of insufficient use of neighborhood information and not giving the neighborhood interval a separate membership degree,proposes an improved objective function,and gives a new clustering center calculation method based on the improved objective function,constructs a complete neighborhood weighting algorithm,and increases the noise offset through the use of neighborhood information Resistance ability.Secondly,extend the algorithm to 3D,and a new 3D neighborhood weighted image segmentation method is proposed.Because the initial center of clustering will greatly affect the clustering effect of the algorithm,which will reduce the accuracy of segmentation to a certain extent.By studying the current selection algorithm of initial clustering center and discussing the advantages and disadvantages of various algorithms,the current selection method of initial center based on density function is improved,so as to ensure that the selected initial center is the optimal solution.Combined with multi-modal MRI fusion method,a three-dimensional neighborhood weighting algorithm is proposed.Finally,in order to solve the self-adaptive problem and local optimal problem of the proposed algorithm,the particle swarm optimization algorithm with comprehensive learning is used to improve the existing problems,so that the single point fuzzy weighted index and the regional fuzzy weighted index of the 3D neighborhood weighted algorithm can be adjusted adaptively,and the global optimization of the comprehensive learning algorithm is matched The ability of optimization further improves the accuracy of the algorithm.Finally,based on the above research,proposed an adaptive 3D neighborhood weighted image segmentation method eventually.