基于GPU加速的医学图像配准技术进展Developmentofmedical.PDFVIP

JournalofComputerApplications ISSN1001-9081 2015- 【-10 ,2015, ( ): - CODENJYIIDU 计算机应用 【【 【 【 :1001-9081(2015) -0000-00 doi:10.11772/j.issn.1001-9081.2015. .0000 文章编号 【 【 基于GPU加速的医学图像配准技术进展 查珊珊，王远军 ，聂生* (上海理工大学 医学影像工程研究所，上海 200093) (*通信作者电子邮箱 yjusst@126.com) 摘 要 针对目前医学图像配准技术无法满足临床实时性需求问题，对基于: GPU加速的医学图像配准技术进行综述探讨。 首先对GPU通用计算进行概述，再以医学图像配准基本框架为主线，对近年来基于GPU加速的医学图像配准技术在国内外 发展现状进行深入研究，并针对PET和CT数据的非线性配准问题，分别基于CPU和GPU平台进行配准实验，通过实验结 果的对比，体现GPU加速配准技术的优越性。基于GPU加速的FFD和NMI结合的非线性配准方法配准后互信息值略低于 CPU平台的配准结果，但其配准速度是CPU平台的12倍。基于GPU加速的配准算法在保持配准精度的基础上，配准速度都 得到了很大的提升。 关键词: 医学成像；图像配准；图形处理器；加速；空间变换 中图分类号: TP391.41 文献标志码:A Developmentofmedicalimageregistrationtechnologyusing GPU ZHAShanshan,WANGYuanjun ,NieShengdong* (InstituteofMedicalImagingEngineering,UniversityofShanghaiforScienceandTechnology,Shanghai200093,China) Abstract: Thecurrentmedicalimageregistrationtechnologycouldn’tmeetthereal-timerequirements forclinicaldiagnosisand treatment. GPU accelerated medical image registration technology is reviewed and discussed for this problem in this paper. Summarized GPU general purpose computation, studied current technology of medical image registration which based on GPU accelerationwiththeessentialframeworkofmedicalimageregistrationasmainline,andimplementedPETandCTimageregistration experiment respectively on CPU and GPU computing platforms. The NMI value of GPU accelerated medical image registration is slightly smaller than that of CPUmethod, but the registration efficiency increases 12times than CPUmeth