Design and evaluation of two multi-pinhole collimators for brain SPECT
Design and evaluation of two multi-pinhole collimators for brain SPECT
复制标题
DOI：
10.1007/s12149-017-1195-y
复制
发表时间：
2017-10-01
影响因子：
2.6
通讯作者：
Mok, Greta S. P.
中科院分区：
文献类型：
--
作者：
Chen, Ling;Tsui, Benjamin M. W.;Mok, Greta S. P.
客观 SPECT 是诊断或分期脑部疾病（例如阿尔茨海默病 (AD) 和帕金森病 (PD)）的强大工具，但因其分辨率和灵敏度较低而受到限制。同时，与用于小视场 (FOV) 成像的传统平行孔准直器相比，针孔 SPECT 提供了卓越的分辨率和检测效率权衡，适合脑成像的情况。在本研究中，我们建议开发和评估两种多针孔（MPH）准直器设计，以改善脑血流和纹状体的成像。方法我们将目标分辨率分别设置为12和8毫米，视场为200毫米，足以覆盖整个大脑。系统优化的约束包括最大和最小探测器到视场中心 (CFOV) 的距离分别为 344 毫米和 294 毫米，以及 135 毫米的最小旋转半径 (ROR) 以适应患者的肩部。根据目标视场、分辨率和约束条件，我们确定了使系统灵敏度最大化的针孔数、ROR、焦距、孔径接受角和孔径直径。然后，我们使用具有 Tc-99m-HMPAO/Tc-99m-TRODAT-1 分布的数字 NCAT 大脑模型的分析模拟来评估所提出的 MPH 和标准低能量高分辨率 (LEHR) 准直器的成像性能；热棒模型的蒙特卡罗模拟；以及使用 GATE v6.1 的 Defrise 模型。生成超过 360 度的投影，并使用 3D MPH/LEHR OS-EM 方法进行重建，最多可进行 720 次更新。分别计算从 Tc-99m-HMPAO 和 Tc-99m-TRODAT-1 模拟的重建图像中提取的大脑和纹状体区域的归一化均方误差 (NMSE)，并在选定的均匀 3D 区域上评估基于 20 个噪声实现的平均归一化标准偏差 (NSD) 作为噪声指数。对蒙特卡罗模拟的结果进行视觉评估和图像轮廓。结果 MPH准直器的优化设计参数分别为9个针孔，针孔直径分别为4.7和2.8 mm、73度接受角、127 mm焦距、167 mm ROR，分别用于12 mm和8 mm目标分辨率。根据优化结果，与 LEHR 相比，所提出的准直器的检测效率分别提高了 270% 和 40%。蒙特卡罗模拟表明，目标分辨率为 12 毫米和 8 毫米的 MPH 准直器可以区分 7.9 毫米和 6.4 毫米棒。 Defrise 体模的 8 个 12 毫米厚的圆盘也可以在轴向平面上清晰地解析，如 MPH 准直器生成的图像轮廓所证明的那样。 结论 与传统 LEHR 相比，两种准直器设计提供了卓越的图像质量，并显示出改善当前基于传统 SPECT 扫描仪的大脑 SPECT 成像的潜力。
Objective SPECT is a powerful tool for diagnosing or staging brain diseases such as Alzheimer's disease (AD) and Parkinson's disease (PD) but is limited by its inferior resolution and sensitivity. At the same time, pinhole SPECT provides superior resolution and detection efficiency trade-off as compared to the conventional parallel-hole collimator for imaging small field-of-view (FOV), which fits for the case of brain imaging. In this study, we propose to develop and evaluate two multi-pinhole (MPH) collimator designs to improve the imaging of cerebral blood flow and striatum.Methods We set the target resolutions to be 12 and 8 mm, respectively, and the FOV at 200 mm which is large enough to cover the whole brain. The constraints for system optimization include maximum and minimum detector-to-center-of-FOV (CFOV) distances of 344 and 294 mm, respectively, and minimal radius-of-rotation (ROR) of 135 mm to accommodate patients' shoulder. According to the targeted FOV, resolutions, and constraints, we determined the pinhole number, ROR, focal length, aperture acceptance angle, and aperture diameter which maximized the system sensitivity. We then assessed the imaging performance of the proposed MPH and standard low-energy high-resolution (LEHR) collimators using analytical simulations of a digital NCAT brain phantom with Tc-99m-HMPAO/Tc-99m-TRODAT-1 distributions; Monte Carlo simulations of a hot-rod phantom; and a Defrise phantom using GATE v6.1. Projections were generated over 360 degrees and reconstructed using the 3D MPH/LEHR OS-EM methods with up to 720 updates. The normalized mean square error (NMSE) was calculated over the cerebral and striatal regions extracted from the reconstructed images for Tc-99m-HMPAO and Tc-99m-TRODAT-1 simulations, respectively, and average normalized standard deviation (NSD) based on 20 noise realizations was assessed on selected uniform 3D regions as the noise index. Visual assessment and image profiles were applied to the results of Monte Carlo simulations.Results The optimized design parameters of the MPH collimators were 9 pinholes with 4.7 and 2.8 mm pinhole diameter, 73 degrees acceptance angle, 127 mm focal length, 167 mm ROR for 12 mm and 8 mm target resolution, respectively. According to the optimization results, the detection efficiencies of the proposed collimators were 270 and 40% more as compared to LEHR. The Monte Carlo simulations showed that 7.9 and 6.4 mm rods can be discriminated for the MPH collimators with target resolutions of 12 and 8 mm, respectively. The eight 12 mm-thick discs of the Defrise phantom can also be resolved clearly in the axial plane as demonstrated by the image profiles generated with the MPH collimators.Conclusion The two collimator designs provide superior image quality as compared to the conventional LEHR, and shows potential to improve current brain SPECT imaging based on a conventional SPECT scanner.