Study On The Techniques For3D Visualization Of Medical Images Based On CUDA

Medical image3D visualization is the process of constructing three-dimensional model of human physiological tissue and organ by the series of2D slice images scanned by the high technology imaging equipments X Radio、CT、MRI. It takes an important part in the field of biomedical, such as medical diagnosis、video teaching、surgical simulation. Currently, in order to achieve real-time3D rendering, the developing of most3D products relies on the expensive graphics display card or professional graphics workstation due to the great mount of calculation of3D reconstruction. The surface rendering technology based on fuzzy theory and CUDA parallel development platform is researched after deeply analyzing image segmentation and three-dimensional reconstruction associated with visualization technology. Details as follows:1The character of medical images is ambiguity and instability due to the noise caused by the random jitter of medical imaging equipment components, coupled with the complexity structure of human physiological tissues and organs. Although the standard fuzzy C-means clustering segmentation algorithm fully take into account the character of image, it is just good at to segment the image that is in normal distribution or ball distribution and its execution time is very long. In the paper, the fast fuzzy C-means clustering algorithm based on Gauss kernel function is achieved by combining Mercer theory and fast fuzzy C-means clustering algorithm. The algorithm effectively enhances the efficiency and accuracy of segmentation of image in any cluster distribution by fuzzy means clustering algorithm, highlighting the regions of interest of image, improves the visual effects of image.2Extract the boundary voxels by using parallelling computing before3D reconstruction using the paralleling Marching Cubes algorithm. After3D image reconstruction by parallelling Marching Cubes algorithm, two different patterns to store the equivalent points are researched. One is to store the equivalent point by CPU memory and the other is to store the equivalent point by CUDA buffer. According to the experiment result, the rate of image rendering is effectively enhanced in the way of using CUDA buffer to store equivalent points. Finally, the real-time3D rendering is achieved.3Three-dimension visualization system:A CUDA-based three-dimension visualization system is designed and finished, the system consists of image filtering、image segmentation、 image reconstruction and visualization modules. The mainly features of3D reconstruction module as following:illumination model and color settings, arbitrary isosurface extraction,3D section of tissues and organs、human-computer mouse interaction based on real-time.