PARALLEL IMAGE RECONSTRUCTION FOR 3-D POSITRON CT

发展三维正电子并行处理方案的研究 提出了计算机层析成像(P-CT)技术。计算机化断层扫描 成像技术在医学上扮演着重要的角色 20世纪70年代初的S，有望成为一种不可避免的诊断技术 在未来。然而，较长的图像重建时间一直是其中之一 与这项技术相关的主要缺点。计算需求 对于未来的3-D正电子CT系统来说，由于巨大的 投影数据量大，重建算法耗时长。它 人们都认识到传统顺序计算机不能提供 如此巨大的计算能力。因此，使用多个处理 加快重建势在必行。这样做的目的是 研究目的是开发高效的并行计算体系结构和 专用于3-D P-CT的算法，可以从 投影数据的速度足以满足实际应用的需要 成本。 CT图像重建由滤波和反投影两部分组成 解析算法，以及估计和修正，这基本上有 类似于反投影的计算结构，在迭代中 算法。这将是这个项目的第一步，以确定潜在的 算法的每个部分内部和不同部分之间的并行性。 然后，一个平衡良好的多处理器系统将被设计成最佳地 用高效的并行重构开发并行性 算法。在整个拟议的研究中，将非常强调 实现处理元素之间的数据流顺畅。

A study on developing parallel processing schemes for -3-D positron computerized tomography (P-CT) is proposed. Computerized tomographic imaging techniques have played an important role in medicine since the early 1970's and are expected to become an inevitable diagnostic technique in the future. However, the long image reconstruction time has been one of the major drawbacks associated with the technique. The computational demand for future 3-D positron CT systems would be formidable due to the enormous amount of projection data and time-consuming reconstruction algorithm. It is well recognized that conventional sequential computer cannot provide such an enormous computing power. Therefore, employing multiple processing elements is inevitable to speed up reconstruction. The goal of this research is to develop efficient parallel computing architecture and algorithm dedicated to 3-D P-CT, which can reconstruct images from projection data fast enough for practical applications at a reasonable cost. CT image reconstruction consists of filtering and backprojection in an analytic algorithm, and estimation and correction, which basically have the similar computational structure to that of backprojection, in an iterative algorithm. It will be the first step in this project to identify potential parallelism within each part and among different parts of the algorithm. Then, a well-balanced multiprocessor system is to be designed to optimally exploit the parallelism with the efficient parallel reconstruction algorithms. Throughout the proposed study, great emphasis will be laid on achieving smooth data flow among processing elements.