基于CUDA架构与B样条的实时锥束CT重建算法  

Real-time cone beam CT reconstruction algorithm based on CUDA and B-splines

作　　者：秦安[1] 桂叶晨[1] 冯前进[1] 陈武凡[1] 

机构地区：[1]南方医科大学省医学图像处理重点实验室,广东广州510515

出　　处：《电路与系统学报》2010年第4期45-48,26,共5页Journal of Circuits and Systems

基　　金：国家重点基础研究发展规划973项目(2003CB716104);国家自然科学基金重点项目(30730036);广东产学研项目(cgzhzd0714)

摘　　要：随着X线探测板数据采集速度的快速发展,研究者开始利用C臂机采集投影数据并重建断层图像,用于手术导航或者放射治疗。但是普通PC的重建速度慢,很难匹配硬件数据采集速度,限制了其在实时临床环境中的应用。本文提出一种基于CUDA(Compute Unified Device Architecture)架构的改进FDK算法,利用GPU(Graphic Porcessing Unit)显卡的并行计算能力实现了实时CT重建,并通过B样条插值提高重建图像的质量,在实时临床环境中具有很好的应用价值。With the rapid speeding of X-ray detector plane acquisition,researchers begin to use the C-arms projections to reconstruct 3-D tomography images,which is useful in image guided surgery and radiotherapy.But the low reconstruction speed can hardly match the faster hardware acquisition.A CUDA based imporved FDK implementation is proposed in this paper.The real time CT reconstruction is achieved by taking use of the parallel computing power of commodity GPU graphic card.The quality of image is improved by using B-Splines for data interpolation.This method is valuable for real time clinical applications.

关 键 词：CUDA FDK算法 CT重建 GPU 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]