Research Of Surface Reconstruction From Medical Imaging

This dissertation researches the problem of surface reconstruction from medical imaging. Topics include the extraction of contour lines from medical images, surface reconstruction based on skeleton, surface reconstruction using distance fields and the speeding up techniques for extraction of iso-surface. This presentation is divided into six chapters.In chapter 1, a survey is given about medical visualization. In its long research history, several types of techniques for surface reconstruction from contours have been addressed:?triangular surfaces;?particle systems;?superquadnics;?implicit surfaces;?others, such as B-Splines surfaces and d surfaces.UTriangular surface is the traditional method to reconstruct surface from parallel ?~ slices of contours. The hot spot in recent years in this area is the methods based onskeletons or medical axis. The method based on implicit surfaces produces smooth surface (with minimal energy), thus becomes another hot research topic. There are several presentations appeared in S1GGRAPH in recent years. Many researchers are working on practical algorithms to processing large-scale datasets [CarrOll.Chapter 2 explores the techniques to extract contours from medical images. A method based on differential operators is presented for treating sophisticated contours. The algorithm of extracting directional contours from segmented images is also discussed. Semi-automatic scheme is proposed. When automatic contouring fails to get desired result, user has to modify it.in Chapter 3, we study the skeleton-based method to reconstruct surfaces from contours. Skeletons reflect the main features of the original object, whereas contain far less data points than contours. When treating complex contours, traditionaliiialgorithms may produces surface with antifacts or singular appearance. The matched pairs of characteristic skeleton points are used to guide the process of surface tiling in the new method. The information of skeleton is also used to deal with branching problem. Experiments show that the proposed algorithm can retrieve cases with complex contours.The technique of surface reconstruction using distance field is explored in Chapter 4. These kinds of 搒tronger?algorithms do not need to retrieve correspondence and branching problems. The contour-spreading algorithm is proposed to calculate distance field quickly within the complexity 0(n). The intermediate surface is generated by linear approximation, while in conventional methods, it is tiled between a designated layer of contours and the original lower slice or upper slice of contours. Thus this approach is dominant over conventional ones.In Chapter 5, accelerating techniques for the extraction of iso-surface is discussed. A modified MC algorithm is presented with cache mechanism, which can greatly reduce the redundant calculation and visit on grid points. Experimental results show that it runs with great efficiency.Th& technique of RBF-based implicit interpolation is applied to the surface recnnstruction from medical imaging in Chapter 6. This method minimize the energy of reconstructed surface, hence the resultant surface is smooth. RBF-based interpolation requires expensive computation, so we propose local RBF to reduce the scale and complexity of the problem. A fast algorithm for polygonizing reconstructed surface is also proposed.Finally, we conclude in Chapter 7.