光学学报, 2015, 35 (11): 1117003, 网络出版: 2015-10-15
显微CT血管系统三维结构的骨架细化算法并行化设计实现 下载: 587次
Parallelization of 3D Thinning Algorithm for Extracting Skeleton of Micro-CT Vasculature
摘要
提取骨架是计算机断层扫描(CT)三维(3D)血管图像定量分析中的关键步骤,通常耗费数小时,直接制约了图像分析的定量研究。分析串行骨架细化算法各步骤中包含的可并行化操作,对其进行并行化设计,提出的算法通过OpenMP多线程技术实现,并采用不同大小的三维CT血管图像进行分析和测试。根据测试结果,改进后的算法获取到的骨架准确可靠,对于1.95 GB大小的三维血管图像,使用16个线程进行并行运算时,可将运算时间由176 min缩短到13 min,时间消耗上降低了一个数量级。因此,提出的方法可实现大型血管骨架的准确、高效提取,解决了大型三维图像分析问题中运算效率低这一瓶颈问题。
Abstract
The extraction of skeleton is often a crucial step in quantitative assessment of three-dimensional (3D) vasculature computed tomography (CT) image. The process always consumes several hours to get the skeleton, which confines the efficiency of quantitative analysis. By means of OpenMP, a parallel designing method based on sequential thinning is proposed to improve the computational time of the skeletonization. The implemented method is utilized for different sizes of real 3D vascular CT images in order to evaluate its performance and efficiency. The testing results show that the proposed method, which is implemented in 16-thread parallel, does not only extract precise skeleton, but also conspicuously reduces the processing time to an acceptable scale. The corresponding computational time is reduced from 176 min to 13 min. Therefore, the time efficiency of quantitative assessment is no longer an obstruction for the analysis of the large scale 3D vasculature CT images.
谭海, 王大东, 薛艳玲, 王玉丹, 杨一鸣, 肖体乔. 显微CT血管系统三维结构的骨架细化算法并行化设计实现[J]. 光学学报, 2015, 35(11): 1117003. Tan Hai, Wang Dadong, Xue Yanling, Wang Yudan, Yang Yiming, Xiao Tiqiao. Parallelization of 3D Thinning Algorithm for Extracting Skeleton of Micro-CT Vasculature[J]. Acta Optica Sinica, 2015, 35(11): 1117003.