Clinical Evaluation of GPU-Based Cone Beam Computed Tomography
Clinical Evaluation of GPU-Based Cone Beam Computed Tomography
复制标题
基于 GPU 的锥形束计算机断层扫描的临床评估
DOI：
--
复制
发表时间：
2008
期刊：
影响因子：
--
通讯作者：
S. Schafer
中科院分区：
文献类型：
--
作者：
P. Noël;A. Walczak;K. Hoffmann;Jinhui Xu;Jason J. Corso;S. Schafer
锥形束计算机断层扫描（CBCT）在临床领域的应用越来越广泛，因为它能够在干预过程中提供三维信息，诊断质量高（亚毫米分辨率），扫描时间短（60秒）。在许多情况下，CBCT扫描时间短，随之而来的是耗时的三维重建。CBCT数据的标准重建算法是过滤后的反向投影，对于大小为256的体积，在标准系统上需要长达25分钟的时间。图形处理单元（gpu）领域的最新发展使得以低成本获得高性能计算解决方案成为可能，从而允许在许多科学问题的应用中使用。我们使用NVIDIA （NVIDIA Cor., Santa Clara, California）提供的计算统一设备架构（CUDA）实现了CBCT数据的三维重建算法，该算法在NVIDIA GeForce 8800GT上执行。我们的实现将重建时间从几分钟，甚至几小时缩短到几秒钟，同时还使临床医生能够以更高的分辨率查看3d体积数据。我们在10个临床数据集和一个虚拟数据集上评估了我们的实现，以观察基于CPU和GPU的重建之间可能发生的差异。通过使用我们的方法，256的计算时间从CPU上的25分钟减少到GPU上的4.8秒。512的重构时间为11.3秒，1024的重构时间为61.4秒。
The use of cone beam computed tomography (CBCT) is growing in the clinical arena due to its ability to provide 3-D information during interventions, its high diagnostic quality (sub-millimeter resolution), and its short scanning times (60 seconds). In many situations, the short scanning time of CBCT is followed by a time consuming 3-D reconstruction. The standard reconstruction algorithm for CBCT data is the filtered backprojection, which for a volume of size 256 takes up to 25 minutes on a standard system. Recent developments in the area of Graphic Processing Units (GPUs) make it possible to have access to high performance computing solutions at a low cost, allowing for use in applications to many scientific problems. We have implemented an algorithm for 3-D reconstruction of CBCT data using the Compute Unified Device Architecture (CUDA) provided by NVIDIA (NVIDIA Cor., Santa Clara, California),which was executed on a NVIDIA GeForce 8800GT. Our implementation results in improved reconstruction times from on the order of minutes, and perhaps hours, to a matter of seconds, while also giving the clinician the ability to view 3-D volumetric data at higher resolutions. We evaluated our implementation on ten clinical data sets and one phantom data set to observe differences that can occur between CPU and GPU based reconstructions. By using our approach, the computation time for 256 is reduced from 25 minutes on the CPU to 4.8 seconds on the GPU. The GPU reconstruction time for 512 is 11.3 seconds, and 1024 is 61.4 seconds.