Prism-PET: A TOF-DOI-Compton PET detector technology for total-body PET imaging

摘要 几乎所有PET扫描仪中相对较小的轴向视场（FOV）严重限制了其灵敏度，并导致图像信噪比（SNR）、采集时间和输送剂量之间的重大权衡。解决方案是使用大轴向FOV进行全身成像以改善几何覆盖，与现有的商业全身PET扫描仪相比，其具有将灵敏度提高约40倍的潜力。世界上第一台全身PET/CT扫描仪，称为EXPLORER，具有194 cm轴向FOV，最近已经完成。然而，与具有200 ps定时分辨率的最先进的全身PET扫描仪（Siemens Biograph Vision）相比，它具有430 ps的较差的飞行时间（TOF）分辨率。此外，在该宽接受角内检测到的倾斜入射伽马光子的轴向检测器穿透将引入显著的相互作用深度（DOI）视差误差，导致空间分辨率降低。与Biograph Vision PET扫描仪相比，EXPLORER扫描仪的这两个主要缺陷抵消了其有效的灵敏度增益，特别是在成像单个器官（如人脑）时。对于这项研究，我们建议使用单端读出和机器学习后处理引擎构建具有组合的最高空间分辨率（2 mm）、最高TOF分辨率（<120 ps和潜在100 ps）、最佳DOI定位（1.85 mm）和最高晶间康普顿散射恢复能力（88%）的具有成本效益的块检测器。我们的新型探测器模块称为Prism-PET，通过非常接近地模仿双端读出的行为，使用直角棱镜-反射镜光导的分段阵列来增强和控制光共享，从而实现这些功能。在此基础上，本文完成了以下工作：（1）利用GATE Monte Carlo模拟建立TOF-DOI PET扫描仪的模型，并利用其列表模式数据进行TOF-DOI图像重建;（2）构建Prism-PET探测器模块并表征符合探测器块的DOI校正的空间分辨率、能量分辨率，和时间分辨率，并使用机器学习来恢复晶间康普顿散射事件，以在不牺牲灵敏度的情况下达到2-mm的固有空间分辨率;（3）构建单环原型Prism-PET扫描仪;（4）进行归一化扫描，使用实验符合事件优化我们的图像重建，并使用热点体模验证最佳成像性能。我们提出的TOF-DOI-Compton探测器和系统技术最大限度地提高了全身PET扫描仪在整个FOV中的几何覆盖率。

Abstract The relatively small axial field-of-view (FOV) in almost all PET scanners has severely limited their sensitivity and resulted in significant tradeoffs between image signal-to-noise ratio (SNR), acquisition time, and delivered dose. The solution is to improve geometric coverage using large axial FOV for total-body imaging which has the potential to improve sensitivity by about 40 times compared to existing commercial whole-body PET scanners. The construction of the world's first total-body PET/CT scan-ner, called EXPLORER, with 194-cm axial FOV has recently been completed. However, it has a poor time-of-flight (TOF) resolution of 430 ps, compared to the state-of-the-art whole-body PET scanner with 200 ps timing resolution (Siemens Biograph Vision). In addition, axial detector penetration of obliquely incident gamma photons detected within this wide acceptance angle will introduce significant depth-of-interaction (DOI) parallax error, leading to degraded spatial resolution. These two major defi-ciencies of the EXPLORER scanner offset its effective sensitivity gain, specially for imaging single organs such as the human brain when compared to the Biograph Vision PET scanner. For this research, we propose to build cost-effective block detectors with combined highest spatial resolution (2 mm), highest TOF resolution (<120 ps and potentially 100 ps), best DOI localization (1.85 mm), and highest inter- crystal Compton scatter recovery capabilities (88%) using single-ended readout and machine learning post-processing engines. Our novel detector module, called Prism-PET, enables these capabilities by mimicking very closely the behavior of dual-ended readout with enhanced and controlled light sharing using a segmented array of right angle prism-mirror light guides. Given our promising preliminary ex-perimental results, we aim to accomplish the following tasks: (1) model a prototype TOF-DOI PET scanner using GATE Monte Carlo simulations and use its list-mode data to develop our TOF-DOI image recontruction; (2) build Prism-PET detector modules and characterize coincidence detector blocks for DOI-corrected spatial resolution, energy resolution, and timing resolution and use machine learning to recover inter-crystal Compton scattered events to reach the 2-mm intrinsic spatial resolution without sacrificing sensitivity; (3) Build a one-ring prototype Prism-PET scanner; (4) perform normalization scan, optimize our image reconstruction using experimental coincidence events and validate the optimal imaging performance using Hot Spot phantoms. Our proposed TOF-DOI-Compton detector and system technologies maximize the sensitivity benefits of improved geometric coverage in total-body PET scanners across the entire FOV.