Research On Medical Image Registration And 3D Visualization

The medical image registration is a research hotspot at present, a research field with the cross of multiple disciplines, and an important application on engineering of computer graphics and image processing. The medical image registration not only applies on medical diagnosis but also on establishment of the operation plan and the radiotherapy plan, the track of pathological change, and the evaluation of treatment effect and so on. The visualization technology of medical image is a powerful assistant method, can make up the shortage of imaging equipments and provide perfect 3D image for users. Doctors can view and analyze from multi-angle and multi-hiberarchy, therefore doctors can analyze and process the medical data effectively. The visualization plays an important pole on assistance diagnosis, surgery simulation, guided treatment and so on. Hence, researches on the medical registration and visualization play an important part in academy and have wide applications.Mutual information(MI) is a kind of theory that is used more frequently in latest years, whose many advantages make it often used in the image registration field. This thesis emphasizes Mi-based medical image registration, whose objective is to raise a registration algorithm that has better performance in all the speed, precision and disregistrated rate. Firstly, based on contour images, a registration algorithm which has perfect speed and better precision is raised, to be applied on some fields which require perfect speed but common precision, such as surgical navigation. Secondly, based on the following researches: research on all the dimensions of image transformation, such as translation and rotation which have different rising curve respectively, research on many characters of images, such as edges, centroid, and research on similarities, such as MI, fuzzy gradient similarity, a triple level registration strategy is raised, which firstly match images roughly according to different dimension respectively and then match images accurately on all the latitudes. The proposed triple level registration strategy is a registration strategy that can harmonize the precision, speed and disregistrated rate effectively.Likewise, this thesis studies medical image visualization and introduces two kinds of visualization methods, surface rendering and volume rendering. Because of the requirement of three-dimensional cutting function in medical image processing system, this thesis emphasizes to implement the Shear-Warp based volume rendering algorithm. This thesis introduces the rationale of the Shear-Warp based volume rendering algorithm, modeling language UML and modeling tool Visio, and analyzes the reason and advantages of UML modeling of algorithms. The UML models of the Shear-Warp based volume rendering algorithm is emphasized, according to which the Shear-Warp based volume rendering algorithm is implemented in Visual C++ 6.0. The algorithm will play an important pole in medical image processing system analysis.Finally, the medical image toolkit, MITK, is adopted, which is developed by Medical Image Processing Group, Chinese Academy of Sciences, to analyze and implement a medical image processing system. Image segmentation, image visualization, image registration and data display are implemented in one system. The user interface of this system is friendly and operated easily.