Volume Rendering Algorithm And Simulation Method Of Projection Data For Tomosynthesis

With the extensive application of medical imaging technology such as CT, MRI and US, a large number of medical image data have been obtained. Clinical diagnosis based on 2D image of focus requires doctors to have rich clinical experience.3D visualization technology products 3D images based on 2D slices of object, provides richer and more accurate information for doctors and has practical application value.Common methods of visualization are surface rendering and volume rendering. The former can show surface of interested tissue but not interior structure and is quick in rendering. The later can show the surface and interior of interested tissue with high image quality and is slow in rendering. As expensive computation, volume rendering based on CPU is not real-time rendering, which exercise restraint in its application. With the rapid development of GPU in recent years, volume rendering based on GPU becomes a hotspot. This paper deals with Raycasting and 3D texture mapping and realizes Raycasting and 3D texture mapping based on GPU which can render small and medium-scale data sets in real time on consumer PC. In addition, the paper has implemented Raycasting on GPU using CUDA that attains high-quality image.Bounding box is used to calculate parameters of the ray that casts the data volume in volume rendering, tight-fitting bounding box can accelerate the rendering process by leapping lots of empty voxels. In this paper a tight-fitting bounding box based on back-step is proposed, which can search the boundary of tight-fitting bounding box quickly. This method needs no volume data preprocessing and additional memory, which has no loss in image quality. Under CUDA the paper has implemented the acceleration of Raycasting with tight-fitting bounding box based on back-step and achieved good performance.Mapping scalar value into a color specification and an opacity value, transfer function could highlight specific structure. A transfer function based on different regions of image plane is proposed in this paper. We set different transfer function to different regions of image plane so that we can show different tissue in single-pass rendering, which is convenient for contrasted observation. But it is not enough to show all the information of the data set only by the transfer function based on different regions of image plane. This paper has implemented transfer function based on different regions of object space, which highlights the interested region in object space. Rotating and zooming the interest object, we can observe the interested region with any size from arbitrary angle. In addition, a fast volume clipping is proposed in this paper, which allows the users to cut away selected parts of volume based on the position of voxels in the data set so that we can observe the occlusion.Tomosynthesis can reconstruct any sectional image of object using limited-angle projection. Projection simulation is an important step for algorithm research. Taking account of geometry comparability between Raycasting and Tomosynthesis, a acceleration technology of volume rendering has been applied to Tomosyhthesis based on GPU acceleration. Comparing to CPU simulation method, GPU simulation method is faster in calculation and higher in accuracy.