Department of Nuclear Medicine & PET Center, Aarhus University Hospital, 8000 Aarhus, Denmark. Author to whom any correspondence should be addressed.
Phys Med Biol. 2017 Dec 29;63(1):015032. doi: 10.1088/1361-6560/aa9475.
The standard compartment model (CM) is widely used to analyse dynamic PET data. The CM is fitted to time-activity curves to estimate rate constants that describe the transport of a tracer between well-mixed compartments. The aim of this study was to develop and validate a more realistic microvascular compartment model (MCM) that includes capillary tracer concentration gradients, backflux from cells into the perfused capillaries and multiple re-uptakes during the passage through a capillary. The MCM incorporates only parameters with clear physiological meaning, it is easy to implement, and it does not require numerical solution. We compared the MCM and CM for the analysis of 3 min dynamic PET data of pig livers (N = 5) following injection of C-methylglucose. During PET scans, the tracer concentrations in blood were measured in the abdominal aorta, portal vein and liver vein by manual sampling. We found that the MCM outperformed the CM and that dynamic PET data include information which cannot be extracted using standard CM. The MCM fitted dynamic PET data better than the CM (Akaike values were 46 ± 4 for best MCM fits, and 82 ± 8 for best CM fits; mean ± standard deviation) and extracted physiologically reasonable parameter estimates such as blood perfusion that were in agreement with independent measurements. The difference between model-independent perfusion estimates and the best MCM perfusion estimates was -0.01 ± 0.05 ml/ml/min, whereas the difference was 0.30 ± 0.13 ml/ml/min using the CM. In addition, the MCM predicted the time course of concentrations in the liver vein, a prediction fundamentally unobtainable using the CM as it does not return tracer backflux from cells to capillary blood. The results demonstrate the benefit of using models that include more physiology and that models including concentration gradients should be preferred when analysing the blood-cell exchange of any tracer in any capillary bed.
标准隔室模型（CM）广泛用于分析动态 PET 数据。CM 拟合时间-活性曲线以估计描述示踪剂在混合良好的隔室之间转运的速率常数。本研究旨在开发和验证一种更现实的微血管隔室模型（MCM），该模型包括毛细血管示踪剂浓度梯度、细胞内返回到灌注毛细血管的返流量以及通过毛细血管时的多次再摄取。MCM 仅包含具有明确生理意义的参数，易于实现，并且不需要数值解。我们比较了 MCM 和 CM 对注射 C-甲基葡萄糖后猪肝脏 3 分钟动态 PET 数据的分析（N=5）。在 PET 扫描期间，通过手动取样测量腹主动脉、门静脉和肝静脉中的血液示踪剂浓度。我们发现 MCM 优于 CM，并且动态 PET 数据包含使用标准 CM 无法提取的信息。MCM 比 CM 更好地拟合动态 PET 数据（最佳 MCM 拟合的 Akaike 值为 46±4，最佳 CM 拟合的 Akaike 值为 82±8；平均值±标准偏差），并提取了生理上合理的参数估计，如与独立测量一致的血液灌注。模型独立灌注估计值与最佳 MCM 灌注估计值之间的差异为-0.01±0.05 ml/ml/min，而使用 CM 的差异为 0.30±0.13 ml/ml/min。此外，MCM 预测了肝静脉中浓度的时间过程，这是使用 CM 无法获得的基本预测，因为它不会将细胞内返流量从细胞返回到毛细血管血液中。结果表明，使用包含更多生理学的模型的好处，并且在分析任何毛细血管床中任何示踪剂的血液-细胞交换时，应优先使用包含浓度梯度的模型。
