摘要
介入诊疗的工作人员通常暴露在X射线辐射场中,而且操作时间长,因此,研究介入诊疗的旁散射分布和强度计算对其射线屏蔽和工作人员主动防护有重要意义。然而旁散射并非X射线射野的原发射线,模拟效率较低,所以提高模拟效率对旁散射相关研究具有重要作用。该文从抽样原理层面提出一种基于任务驱动的旁散射GPU智能模拟技术。首先,推导出旁散射光子路径输运方程,应用Metropolis-Hasting(M-H)抽样方法进行光子路径的自动重要性抽样,从而实现光子散射阶数和类型可控的任务驱动型GPU智能模拟方法。在模拟实验中,相同时间内,该方法的信噪比(SNR)相对于常规方法提升了4.5 dB,同时均方根误差相比下降了25.32。在机器数据实验中,该方法均方根误差比常规方法平均降低了5.20,旁散射模拟的时间从250 min缩短到5 min。该方法完成了基于任务驱动的光子输运抽样,可快速准确实现DSA旁散射的计算机GPU模拟计算。
The staff of interventional diagnosis and treatment are usually exposed to the X-ray radiation field,and the operation time is long,so the calculation of its side scattering distribution and intensity is of great significance to the radiation shielding of interventional diagnosis and treatment and the staff’s active protection.However,the side scattering is not the primary ray of the X-ray field,and the simulation efficiency is low,so it is very important to improve the simulation efficiency in the research of side scattering.Based on the sampling principle,this paper proposes a task driven GPU intelligent simulation technology of side scattering.Firstly,the transport equation of side scattering photon path is derived,and the Metropolis Hasting(M-H)sampling method is used to sample the photon path automatically via importance sampling,so as to realize the task driven GPU intelligent simulation method with controllable order and type of photon scattering.In the simulation experiment,the signal to noise ratio(SNR)of this method is improved by 4.5 dB compared with the conventional method in the same time,and the root mean square error is reduced by 25.32.In the machine data experiment,the root mean square error of the proposed method is 5.20 lower than that of the conventional method,and the simulation time of side scattering is shortened from 250 min to 5 min.This method completes the task driven photon transport sampling,and can realize the computer GPU simulation of DSA side scattering quickly and accurately.
作者
龙超
徐美奕
熊刚强
LONG Chao;XU Meiyi;XIONG Gangqiang(School of Biomedical Engineering,Guangdong Medical University,Zhanjiang 524023,China;Asset and Laboratory Management Department,Guangdong Medical University,Dongguan 523808,China;School of Biomedical Engineering,Guangdong Medical University,Dongguan 523808,China)
出处
《实验技术与管理》
CAS
北大核心
2022年第7期110-114,共5页
Experimental Technology and Management
基金
国家自然科学基金面上项目(61170320)
湛江市科技攻关计划项目(2020B01193)。
关键词
DSA
介入诊疗
旁散射
智能模拟
任务驱动
GPU
DSA
interventional diagnosis and treatment
side scattering
intelligent simulation
task driven
GPU