Research On Accelerating Three Dimensional Medical Image Reconstruction

Visualization of medical data is the most triumphant area for ViSC (Visualization in scientific computing).It has been an important method of assistant therapy in these years , and definitely it will be applied to many aspects of medicine. Volume Visualization is an important branch of ViSC. Its processing procedures include volume data acquisition,model construction , data mapping and rendering. Volume data is mainly acquired by medical instrument, such as CT,MRI,MRA,US,PET,SPECT. The object of Volume Visualization is to create accurate image of human body.This paper will study 3D Volume Reconstruction algorithm,discuss Surface Rendering and Volume Rendering algorithm, analysis the classic algorithms, and advance a new algorithm which is based on raycasting .The algorithm will be optimized aiming at specific hardware.This paper presents an accelerated ray-casting algorithm using cache efficiently. The algorithm is developed by C++ programming language and the performance critical section is implemented by assembly language. According to the specialty of the CPU cache, the algorithm interleave the memory address of the volume data to enhance cache hit rate by using the hardware prefetch.Furthermore, this paper uses several optimizing methods, such as empty space skipping and early ray termination, and writes the critical code with MMX instructions. At the same time, this paper uses post-classification and tri-linear interpolation to ensure the quality of the final image. The result shows that this algorithm inherits the advantage of the traditional ray-casting algorithm and produces high quality images while it achieves higher speed.The last chapter discusses 3D Volume Reconstruction algorithm based on programmable graphics process unit(GPU). This chapter summarizes the development of up-to-date GPU and expounds how the development of GPU impacts the 3D Volume Reconstruction algorithm.