Research And Application Of Image Registration Technology In Sequences Images Of Prostate DWI

Medical image registration is an important direction of modern medical image processing technology applications. Its main task is to correct images by removing the parallel movements, rotations and deformations of the images. Prostate DWI is a sequence of images which are taken under different magnetic field strengths during a certain interval. Because the drift and deformation of the images will affect the diagnosis of doctors dramatically, the registrations are needed for prostate DWI. How to improve the accuracy of the registration among the image registration processes is an important task for clinical diagnosis. The main contents of this paper include:First, the background, significance and current achievement of the image registration are depicted in this paper, and then follow the detailed explanation of the image registration process, the basic models and the classification method of the image registration.Second, in order to correct the drift appearing in the prostate DWI sequence images, a rigid registering method based on mutual information and optimized ant colony algorithm is applied in this paper. Mutual information reflects the degree of alignment between two images and it is the key issue of the image registration. This paper systematically discussed the definition, nature and calculations of the mutual information, and also discussed the ant colony algorithm. Since the ant colony algorithm falling into local optimum easily, the ant colony algorithm has been optimized by introducing a new factor to the state transition probabilities and applying local pheromone update and a global pheromone update. Then the prostate DWI sequence images have been registered by using the optimized ant colony algorithms. Experimental results show the efficiency of the improved ant colony algorithm for image registration.Third, in order to solve the deformation of the prostate DWI sequence, nonlinear registration(FFD) model based on B-spline deformation method freedom has been applied. The functions and curves of the B-spline has been introduced in this article briefly, and then the uniform grid FFD model are applied to the images after rigid registration for nonlinear registration. Experimental results show that the deformation of the prostate DWI image has been dramatically reduced by using the method proposed in this paper.