Study Of 3D Image Reconstruction Algorithms Based On Graphic Processing Unit For Hardwave Acceleration

Rapid advance in imaging technology makes digital analysis technology of medical image make new progress. Relying on the radiologists’ subjective experience makes it difficult to extract information. Reconstructing a three-dimension image from a series of consecutive two-dimension image provides doctors more clear and vivid images of organs or tissues. Then the accuracy of medical diagnosis and treatment plan will be improved.Visualization can be divided into surface rendering and volume rendering. This study was focused on improving medical image reconstruction in real-time and interaction. For surface rendering, the following four aspects of work were done. Firstly, based on the understanding of Marching Cubes algorithm, optimization algorithm which is the combination of the region growing and standard MC algorithm was done. Secondly, to assess image quality and rendering time, three-dimension reconstruction was implemented in VTK visualization platform. The data flow of GPU and MC algorithm were combined. Hardware acceleration was completed by Shader Model 4(SM4) programmable rendering pipeline. Finally, the developed algorithm was integrated into Voreen algorithm development platform. For volume rendering, the following three aspects of work were done. First, three-dimension reconstruction was implemented in VTK visualization platform. Next, hardware acceleration was completed by programming interfaces provided by VTK itself. Finally,multiple organizations was reconstructed by adjusting the color scheme and the transfer function.Experiments show that the optimization algorithms and hardware acceleration can both shorten the time used for reconstruction. According to the real-time, hardware acceleration is better than optimization algorithms. Integration with medical visualization platform can be an effective method to improve interactive manipulation capability of reconstruction.