面向BNCT硼剂量实时测量的新结构康普顿相机设计与图像快速重建研究

Real-time boron dose measurement is one of the challenges that restricts the development of boron neutron capture therapy (BNCT). Using the prompt γ rays from the neutron boron capture reaction to estimate the boron dose is one of the most promising techniques for online boron dose measurement. However, the implementation of this technology must seek for detection techniques with higher detection efficiency, higher spatial resolution, and faster reconstruction algorithms. On the basis of previous researches, this project proposes the idea of ​“single-stage Compton camera + artificial intelligence fast reconstruction algorithm”, for achieving real-time accurate measurement of 0.478 MeV prompt γ rays, which thereby can be used to obtain real-time boron dose distribution. To realize the specific purpose, the single-stage Compton camera based on drift strip CdZnTe detector will be used to detect the 0.478 MeV prompt gamma in the complex n/γ radiation field to improve the detection efficiency. The FPGA circuit will be used to obtain the three-dimensional position information to improve the spatial resolution. And optimized reconstruction algorithm based on artificial intelligence will be developed to achieve real-time rapid reconstruction of boron dose. Furthermore, the accuracy and robustness of the method will be evaluated in the real BNCT clinical environment. This project will help to further promote the development of real-time dose measurement technology for BNCT.

实时硼剂量测量是制约硼中子俘获治疗（BNCT）发展的难题之一，利用中子与B-10俘获产生的瞬发γ射线推算硼剂量是目前最有潜力实现在线硼剂量测量的技术手段之一。然而这一技术手段的实现必须寻求更高探测效率、更高空间分辨率的探测技术和更快速重建能力的数据分析技术。基于前期研究基础，本项目创新提出“单层康普顿相机+人工智能快速重建算法”的技术思路，以期实现0.478 MeV瞬发γ的实时精准测量，进而得到实时硼剂量分布。针对此目标，研究基于条形阵列CdZnTe的单层康普顿相机在n/γ辐射场探测0.478 MeV瞬发γ射线的结构性能优化以提高探测效率；结合FPGA电路获取γ射线三维位置信息提升单层康普顿相机空间分辨率；发展基于人工智能算法的优化重建技术实现硼剂量实时快速重建；结合真实束流环境评估在线硼剂量测量的稳定性。本项目研究将有助于进一步推动BNCT实时剂量精确测量技术的发展。

针对BNCT中硼剂量在线监测困难等问题，瞬发γ射线探测已成为最有潜力的在线监测方式。现有研究主要采用基于机械准直的PG-SPECT方法进行研究，但其受限于准直器使得探测效率较低。针对此问题，本项目提出利用单层康普顿相机开展BNCT中0.478 MeV瞬发γ射线的探测以实现BNCT治疗过程中硼剂量的实时监测。优化设计了面向BNCT应用的单层康普顿相机几何参数，分别探讨了探测器厚度、宽度以及放射源与探测器之间的距离对探测效率与成像质量的影响规律，实现了条形阵列CZT探测器结构性能优化；研究了探测器获取三维位置信息的信号读取及处理电路，实现单层康普顿相机模块的能谱探测及3D位置信息获取；发展了基于人工智能算法GAN网络的康普顿相机快速图像重建算法，实现了康普顿相机重建图像质量的快速提升，并研究了不同训练集对网络预测结果的影响；同时评估了单层康普顿相机探测模块对γ源重建结果的准确性，并结合南京中硼联康医疗科技有限公司所设计的加速器中子源评估实时硼剂量测量的稳定性。通过本研究工作，发展了基于CZT单层康普顿相机的PG-SPECT探测技术实现BNCT治疗过程中0.478 MeV瞬发γ射线的探测，实现高效率、高空间分辨的硼剂量实时测量，本项目研究将有助于进一步推动BNCT实时剂量精确测量技术的发展。