GPU-based Medical Images Three-dimensional Reconstruction Algorithms And Applications

Medica l ima ges three-dimensiona l reconstruction whose essence is the volume datavisua lization could figure out three-dimensiona l (3D) anatomy structures of the huma nbodies from a series of two-dimensiona l digit ima ges and pla ys an important part inmed ical clinica l applica tions. The conventiona l central processing unit (CPU) is verydifficult to meet the dema nds of med ical ima ges visua lization for real-time, so graphicsprocessing unit (GPU) technology was employed to study the algorithms of volumevisua lization and their applica tions for med ical ima ges 3D reconstruction in this paper.Marching tetrahedron is a voxel-based surface rendering algorithm. The basic principleof marching tetrahedron algorithm was introduced first, next the problems such as divid ingthe cube unit, configuring out the tetrahedron unit, and calculating the intersection pointsand norma l vectors for every triangle generated were discussed, then the algorithm wascarried out on GPU, and the details of processing tetrahedron unit and generating triangleson geometry shader were expla ined.Ray casting is an ima ge-order-based direct volume rendering algorithm. The basicprinciple of the ray casting algorithm was introduced first, next the problems such as colormapping, color blend ing, norma l calculation, the effect of projection mapping and the wayof sampling were discussed, then the algorithm was carried out on GPU, and the details ofray casting and color blend ing on fragment shader were expla ined.In order to develop a med ical ima ges 3D reconstruction applica tion, the implementationof DICOM (Digital Ima ging and Communica tions in Medicine) file interface, the dataflow-based applica tion framework, the way of mouse control and the design of userdefinedcolor look-up table were introduce d in the paper.The applica tion using GPU-based visua lization algorithms can be used in the commonpersona l computers with sufficient resolutions and interactive frame ima ges, which couldmeet the dema nds of rapid med ical ima ges 3D reconstruction for clinica l applica tions.