Multi-modality Medical Image Registration Research And Implementation

Medical image registration is a classical research in the field of medical image analysis and processing. Modern medical technology can not be seperated with the support of a variety of medical images, however, it requires more for image quality and information. For images which were got in different circumstance with movement because of imaging technology, issue displacement and deformation. Medical image registration is to make their position return to entire corresponding, which is signality for clinical disease prevention, observation and monitoring. In order to obtain comprehensive patient information multi-modality image information fusion is needed. Medical image registration is the basis and precondition for kinds of medical image fusion. That establishing medical imaging application software platform with information technology for process and results visualization of medical image processing and calculating the quantitative analysis and information feedback will significantly improve the diagnosis and treatment. On this background, multi-modalilty medical image registration is studied in this article, perfomance and effciency in mutual information method is analyzed and improved. A finite element method driven by mutual information is also provided. In the end a medical processing visualization platform is designed and implemented on the basis of algorithms. Main research completed is as follows:First, there is a summary and review for the status of image registration techniques here. Main research was around mutual information algorithm suitable for multimodality medical image registration. Caculational method and process of mutual information are analyzed. The distertation focused on the robustness and efficiency of this approach. That maximum mutual information method is too dependent on the existence of overlapping regional also the mismatch problem were discussed.Using normalized mutual information with high-dimension instead of maximum mutual information will extend the current pixel to the neighborhood. In order to solve large computation and time-consuming of registration, this paper adopts multi-resolution analysis application for the image registration optimization. The acceleration algorithm is achieved by image multi-resolution decomposition with wavelet pyramid and a more rational multi-resolution strategy is confirmed through experiments.Then, in the soft tissue non-rigid image registration area, Bspline-based free form deformation model method is analyzed and experimented. Based on the idea of physical model combined with image intensity and the finite element analysis to solve the stress problem, a mutual information-driven finite element medical image registration is provided in this paper.In this method, discrete finite element as a physical model, balance equation is constructed by the principle of minimum potential energy, the algorithm with mutual information energy-driven function can accurately simulate the deformation zone and fully meet the rigid displacement.Finally, visualization of the image registration algorithm is implemented. According to characteristics and commonness of various registration methods, heavy ion radiation treatment planning software platform is as the context. Using and recombining ITK for the algorithm framework, VTK as a visualization tool, FLTK for interface design, in the light of existing medical image processing and visualization Software, a medical imgae registration visualization platform is designed and implemented for 3 dimension medical image display, pre-processing and registration.