Study On Medical Image Non-rigid Registration Based On LBM Model

Medical image registration consists in finding a transformation that realigns two images taken by different sensors or at different times to establish the correspondence relation in spatial location between the template image and the study image. When the image registration process finished, the anatomical point contained in images correspond with one another, which is beneficial for providing the complementary useful information of differnent images to improve the aid of images in clinical diagnosis and treatment. Medical image registration becomes very important basic research topic of bio-medical engineering domain. With the existence of nonlinear deformation and noise, medical image registration can be divided into the rigid and the non-rigid registration. Compared with the rigid registration, the non-rigid registration is not ripe. No good method can satisfy the ideal requirements in the complex deformation simulation, the registration accuracy, the registration efficiency and the automation degree at same time. For the above mentioned problems, the main work in this paper is as follows:This paper introduces the concept and step of the medical image registration, analyzes the features of the current non-rigid medical image registration methods, and summarizes the urgent problems of the non rigid registraiton methods. Especially, a non-rigid viscous fluid registration method of medical images has been studied. How to use the idea of hydromechanics to slove the medical image registration problem and the physics significance of this method have been deep understanding. What is metioned above establishes the base theory of the whole innovative research work. Simulated by computer, this method uses the directly discrete combining with ROS method which has high complexity and time consuming process to solve viscous fluid partial differential equations. With the increase of image resolution ratio, the consuming time is longer and the registration efficiency is lower so that the practicability of this method is poor.In order to improve the overall performance of medical image non-rigid registration, this paper presents a new medical image non-rigid registration method based on LBM, which transforms medical image registration problem into viscous fluid physical problem to take advantage of the fact that fluid particle movement can simulate any free complicated deformation. In the registration process of this paper, template image and image pixel is considered as viscous fluid and fluid particle respectively. The external force from the gray level difference between the template image and the study image drive registration. Simulated by use of LBM, The flow of template image pixels under external force which make template image generate displacement field and deformation is obtained. By using the Jacobianto to judge the displacement field, the image topological stucture is maintained. When the gray scale value of the study image and the template image is near consensus, the image registration has finished at which time the external force disappears and the fluid flow stops. The LBM equation of compressible viscous fluid with external force is set up and deducted completely, in additional, the boundary conditions, initial conditions and modulus of speed are defined in this paper. Simulation results show that the non-rigid medical image registration performance has been improved comprehensively; especially the efficiency of image registration. This method is very suitable for large deformation registration occasions. For the medical image with a certain noise polution, this method can also achieve precise registration and strong robustness.At the end of paper, the paper summarizes the research work and expects the next work.