计算物理 ›› 2020, Vol. 37 ›› Issue (2): 153-162.DOI: 10.19596/j.cnki.1001-246x.8037

• • 上一篇    下一篇

康普顿相机图像重建中的快速滤波反投影算法

严小松, 杨建伦, 羊奕伟   

  1. 中国工程物理研究院核物理与化学研究所, 四川 绵阳 621999
  • 收稿日期:2019-01-07 修回日期:2019-02-01 出版日期:2020-03-25 发布日期:2020-03-25
  • 作者简介:严小松(1986-),男,助理研究员,博士,主要研究方向为探测器成像,E-mail:yanxs05@yeah.net
  • 基金资助:
    中国工程物理研究院院长基金(YZJJLX2016003)资助项目

A Fast Filtered-Back-Projection Algorithm for Image Reconstruction of Compton Camera

YAN Xiaosong, YANG Jianlun, YANG Yiwei   

  1. Institute of Nuclear Science and Chemistry, China Academy of Engineering Physics, Mianyang, Sichuan 621999, China
  • Received:2019-01-07 Revised:2019-02-01 Online:2020-03-25 Published:2020-03-25

摘要: 提出一个由CZT像素探测器与NaI闪烁体阵列组成的康普顿相机,计算探测器系统的几何参数、散射体探测器的能量分辨率及多普勒展宽等因素对相机空间分辨率的影响.相机的总体角分辨率为5.1°~5.6°,主要影响因素为探测器系统的几何参数.首先用解析方法研究图像重建中的滤波反投影算法,不考虑展宽影响时点源重建图像的角分辨率可达1.7°,考虑展宽影响时角分辨率可达4.2°;然后利用Geant4程序进行模拟,用实际应用方法对模拟数据进行重建,不考虑展宽影响时的角分辨率为1.7°,考虑展宽影响时角分辨率可达5.1°.采用滤波反投影算法进行图像重建时,在OpenCL框架下利用GPU实现并行加速,加速比为79倍.

关键词: 康普顿相机, 滤波反投影算法, Geant4模拟, OpenCL并行计算

Abstract: A Compton camera consisting of a CZT pixel detector and a NaI scintillator array was proposed. Factors that affect angular resolution of the camera, including geometry of detectors, energy resolution and Doppler effect of CZT detector, were analyzed. Overall angular resolution was between 5.1° and 5.6° and geometry of detectors was confirmed as main factor that limits angular resolution of the camera. A filtered-back projection algorithm was analytically deduced. The angular resolution was 1.7° as broadening was omitted, and 4.2° as broadening was considered. Geant4 simulations were performed and the data were reconstructed with a practical method of filtered-back projection algorithm. Angular resolution was 1.7° as broadening was omitted and 5.1° as practical broadening factors of Compton camera was considered. Parallel acceleration with GPU in the framework of OpenCL was applied in the reconstruction and the acceleration ratio was 79.

Key words: Compton camera, filter-back-projection algorithm, GEANT4 simulation, paralleled computing with OpenCL

中图分类号: