2026年“核理安邦”联合博士生学术论坛

小鼠体内¹⁷⁷Lu-FAP临床前吸收剂量评估：基于SPECT/CT图像的GATE蒙特卡洛全粒子输运与MIRD近似模型物理差异

Not scheduled

1h

地址：清华大学校内

海报展示 海报展示 海报展示

Speaker

Mr 庞 瑞磊 (中国药科大学)

摘要

摘要

背景：成纤维细胞激活蛋白（FAP）是放射性药物治疗（RPT）的重要靶点，在多种实体瘤中高表达。随着¹⁷⁷Lu标记FAP药物的快速发展，准确评估临床前模型中的吸收剂量对疗效、安全及临床转化至关重要。目前，基于MIRD的器官水平剂量估算与基于蒙特卡罗（MC）的体素水平剂量计算在物理模型上存在本质差异，其一致性尚需系统比较。本研究旨在比较GATE蒙特卡罗模拟与MIRD方法在荷瘤小鼠中的剂量计算结果，重点探讨MC在粒子输运、次级电子沉积、组织非均匀性及边界效应等物理过程中的优势。

方法：基于U87荷瘤BALB/c裸鼠模型，尾静脉注射三种177Lu-FAP药物。于注射后进行小动物SPECT/CT显像，手动勾画主要器官VOI，获取各时间点器官放射性活度。采用OLINDA/EXM、Odam及PKAD算法计算TIAC并归一化，结合S值获得MIRD体系器官吸收剂量。同时，将SPECT/CT图像导入3D Slicer进行格式转换，利用Python进行空间坐标变换，作为GATE模拟输入。基于GATE v9.0（Geant4）进行体素级MC模拟，以CT图像（0.25 mm体素）构建体素化体模，SPECT图像定义体素化源。物理过程包含光电效应、康普顿散射、轫致辐射，无能量削减及方差减少。放射源为177Lu，DoseActor输出能量沉积与剂量分布。模拟时间为实际采集的1/10–1/100，统计不确定性<5%，每只小鼠耗时3.5小时。基于VOI计算各器官剂量率，积分拟合得总吸收剂量，并分析能量剖面及剂量等高线。

结果：三种放射性药物在肿瘤中均显示出较高的放射性摄取与滞留。在器官吸收剂量方面，基于MIRD方法（OLINDA/EXM、Odam、PKAD）计算得到的肾脏和肿瘤吸收剂量分别为19.78—45.94Gy/GBq和670.27—744.43 Gy/GBq，不同计算工具之间结果较为接近/存在一定差异。GATE蒙特卡罗模拟获得的体素级剂量分布显示，肿瘤内剂量呈现明显异质性，平均吸收剂量为713.51—840.10Gy/GBq，与MIRD方法相比差异为12%—25%。肾脏剂量在体素水平同样表现出局部热点，最高与平均剂量比值可达2.5。

结论：MIRD方法提供器官平均水平剂量，适用于快速评估，但无法反映内部剂量异质性。GATE-MC通过精确模拟粒子输运、次级电子沉积及边界效应，揭示肿瘤及肾脏内部的剂量异质性，更接近真实物理分布。两者物理本质互补，联合使用可为RPT药物临床前剂量学表征及临床转化提供更全面的物理依据。

Abstract

Abstract
Background: Fibroblast activation protein (FAP) is an important target for radionuclide radiotherapies (RPT) and is highly expressed in various solid tumors. With the rapid development of ¹⁷⁷Lu-labeled FAP agents, accurate assessment of absorbed dose in preclinical models is essential for efficacy, safety, and clinical translation. Currently, MIRD-based organ-level dosimetry and Monte Carlo (MC)-based voxel-level dose calculations have fundamental differences in physical models, and their consistency requires systematic comparison. This study aims to compare GATE Monte Carlo simulations with MIRD-based methods for dose calculation in tumor-bearing mice, with an emphasis on the advantages of MC in physical processes such as particle transport, secondary electron deposition, tissue heterogeneity, and boundary effects.

Methods: Based on a U87 tumor-bearing BALB/c nude mouse model, three ¹⁷⁷Lu-FAP agents were injected via the tail vein. Small-animal SPECT/CT imaging was performed at multiple time points post-injection, and VOIs of major organs were manually delineated to obtain organ radioactivity at each time point. OLINDA/EXM, Odam, and the PKAD algorithm were used to calculate TIACs followed by normalization, and organ absorbed doses were obtained based on the MIRD framework combined with S-values. Meanwhile, SPECT/CT images were imported into 3D Slicer for format conversion, and spatial coordinate transformation was performed using Python to generate input for GATE simulations. Voxel-level MC simulations were performed using GATE v9.0 (Geant4). CT images (0.25 mm voxel size) were used to construct a voxelized phantom, and SPECT images were used to define a voxelized source. The physical processes included photoelectric effect, Compton scattering, and bremsstrahlung, without energy cuts or variance reduction. The radioactive sources were ¹⁷⁷Lu. DoseActor was used to output energy deposition and dose distributions. The simulation time was set to 1/10–1/100 of the actual acquisition time, with statistical uncertainty <5%. Each mouse simulation took 3.5 hours on a computing cluster. Based on the delineated VOIs, dose rates of each organ were calculated, and total absorbed doses were obtained by integral fitting. Energy profiles and dose contour lines were also analyzed.

Results: All three radiopharmaceuticals showed high radioactive uptake and retention in tumors. In terms of organ absorbed doses, the kidney and tumor absorbed doses calculated by the MIRD-based methods (OLINDA/EXM, Odam, PKAD) were 19.78–45.94 Gy/GBq and 670.27–744.43 Gy/GBq, respectively. The results from different calculation tools were relatively close / showed some differences. Voxel-level dose distributions obtained from GATE Monte Carlo simulations revealed significant intratumoral dose heterogeneity, with mean absorbed doses ranging from 713.51 to 840.10 Gy/GBq, showing a difference of 12%–25% compared to the MIRD-based methods. At the voxel level, the kidneys also exhibited local hotspots, with a maximum-to-mean dose ratio reaching 2.5.

Conclusion: The MIRD-based methods provide organ-level mean doses suitable for rapid assessment but fail to reflect internal dose heterogeneity. GATE-MC, by accurately simulating particle transport, secondary electron deposition, and boundary effects, reveals dose heterogeneity within tumors and kidneys, which is closer to the true physical distribution. The two approaches are physically complementary, and their combined use can provide a more comprehensive physical basis for preclinical dosimetry characterization and clinical translation of RPT agents.