MR Brain Image Registration

With the gradual development of visualization technology,the doctor will medical image as an important visual patients disease diagnosis basis,and through the rich more and more clear image information to identify the patient's condition and determine the corresponding treatment.Medical image registration is the core of the medical image fusion technology,one of the clinical diagnosis and treatment in modern medicine,and plays an important role in preoperative planning.Thus medical image registration and fusion is a hot spot in current research,is one of the difficulties of the study.Mri(Magnetic Resonance,MR)has high soft tissue resolution,image in the imaging process has the advantages of multiple parameter acquisition,any direction imaging,clinical application in recent years more and more widely and deeply.Demons algorithm,the author of this paper was unable to complete the problem of multimodal medical image registration,medical image registration based on mutual information is studied and the medical image registration algorithm based on Demons,through the analysis of the principle of their advantages and disadvantages and puts forward a kind of Demons algorithm based on joint structure information and local entropy,so as to realize the Demons algorithm for multimodal medical image registration precision,and the multiresolution framework using multilevel b-spline to algorithm is improved,and further improve the accuracy of the algorithm,through the contrast experiment analysis prove the superiority and feasibility of the proposed algorithm in this paper.The main work is as follows:First,we studied the research background and significance of medical image registration,development present situation,the basic concepts and the current existing problems;Medical image registration based on mutual information is analyzed,the relevant theory and principle of(POWELL)and the POWELL algorithm,Particle Swarm algorithm,Particle Swarm Optimization,PSO)as well as the two hybrid Optimization method to the analysis and key.Finally made in this paper,the evaluation standards for the quality of registration,by MR brain image simulation experiment to the POWELL algorithm,PSO algorithm and the hybrid optimization algorithm to do a comparative analysis of both.Secondly,based on optical flow model and one of the most famous Demons thought origin,the basic principle of the algorithm and the detailed analysis has been done for the process of the algorithm.To the local registration using b-spline features after image registration of Demons algorithm is further registration,the registration precision is improved.Use of MR brain images and compares the simulation results of several algorithms,proved the Demons algorithm is not suitable for multimodal medical image registration.Finally,because many modal and the characteristics of the image,in order to eliminate the differences,this paper proposes a fusion image structure tensor trace values of two description information and local entropy of image description method,the gray level information and structural information joint decision of the fused images;Then by Demons fusion algorithm to get the geometric structure of image registration,which Demons algorithm to solve multimodal image registration problem.Through the analysis of the insufficient of using gaussian filter for regularization,proposed under the multiresolution framework using multi-level b-spline deformation field of registration regularization processing method,MR brain images of the simulation experiments show the superiority and feasibility of the proposed algorithm in this paper.