基于蒙特卡洛的医学图像重建体积计算算法GPU加速研究  

GPU acceleration for Monte Carlo algorithm-based volume calculation of medical image reconstruction model

作　　者：何冬林 勾成俊[1] 文玉梅 陈昭 雷琴[1] 杨鹏[1] 吴章文[1] HE Donglin, GOU Chengjun, WEN Yumei, CHEN Zhao, LEI Qin, YANG Peng, WU Zhangwen(Key Laboratory for Radiation Physics and Technology of Ministry of Education, Institute of Nuclear Science and Technology, Sichuan University, Chengdu 610064, Chin)

机构地区：[1]四川大学原子核科学技术研究所/辐射物理及技术教育部重点实验室,四川成都610064

出　　处：《中国医学物理学杂志》2018年第3期260-264,共5页Chinese Journal of Medical Physics

基　　金：国家重点研发计划(2016YFC0105103)

摘　　要：目的:确定基于蒙特卡洛算法计算体积的精度及提高其速度,为医学图像的应用领域快速提供可靠的数据。方法:在Fonics Plan计划系统平台上,实现基于蒙特卡洛方法的体积计算,并使用C++AMP对算法做GPU并行加速,然后对体积计算结果在精度和速度上进行比较分析。结果:与像素累加法、体元累加法相比,蒙特卡洛算法的准确性最高但其算法用时也最长。通过充分利用计算机的显卡计算性能,可将计算速度平均提高50倍。结论:经GPU加速后的蒙特卡洛算法在计算体积的速度和精度两方面都能满足临床要求,在医学图像处理及临床诊疗具有较高的应用价值。Objective To ascertain the accuracy and improve the speed of Monte Carlo （MC） algorithm for volume calculation and provide reliable data for the clinical application of MC algorithm with GPU acceleration. Methods MC algorithm-based volume calculation was implemented on FonicsPlan planning system platform, meanwhile, C＋＋ accelerated massive parallelism was applied to MC algorithm for parallel GPU acceleration. The accuracy and the speed of algorithms for volume calculation were compared and analyzed. Results Compared with pixel accumulation method and voxel accumulation method, MC algorithm has the highest accuracy and the longest elapsed time. The speed of MC was substantially increased to an average of 50 times by making full use of GPU performance. Conclusion The MC algorithm with GPU acceleration for volume calculation has a highly application value on medical image process and the clinical diagnosis and treatment for it achieves a higher accuracy and reduces elapsed time, which meets the clinical requirements.

关 键 词：蒙特卡洛算法 体积测量 C++AMP GPU加速 

分 类 号：R811[医药卫生—放射医学]