Department of Oncologic Imaging, National Cancer Centre Singapore, Singapore.
Radiology. 2013 Apr;267(1):145-54. doi: 10.1148/radiol.12120186. Epub 2013 Jan 7.
To compare four different tracer kinetic models for the analysis of dynamic contrast material-enhanced computed tomographic (CT) data with respect to the prediction of 5-year overall survival in primary colorectal cancer.
This study was approved by the ethical review board. Archival dynamic contrast-enhanced CT data from 46 patients with colorectal cancer, obtained as part of a research study, were analyzed retrospectively by using the distributed parameter, conventional compartmental, adiabatic tissue homogeneity, and generalized kinetic models. Blood flow, blood volume, mean transit time (MTT), permeability-surface area product, extraction fraction, extravascular extracellular volume (v(e)), and volume transfer constant (K(trans)) were compared by using the Friedman test, with statistical significance at 5%. Following receiver operating characteristic analysis, parameters of the different kinetic models and tumor stage were compared with respect to overall survival discrimination, with use of Kaplan Meier analysis and a univariate Cox proportional hazard model, with additional cross-validation and permutation testing.
Blood flow was lower with the distributed parameter model than with the conventional compartmental and adiabatic tissue homogeneity models (P < .0001), and blood flow values determined with the conventional compartmental and adiabatic tissue homogeneity models were similar. Conversely, MTT was longer with the distributed parameter model than with the conventional compartmental and adiabatic tissue homogeneity models (P < .0001). Blood volume, permeability-surface area product, and v(e) were higher with the conventional compartmental model than with the adiabatic tissue homogeneity, distributed parameter, or generalized kinetic models (P < .0001). The extraction fraction was higher with the distributed parameter model than with the adiabatic tissue homogeneity model. With respect to 5-year overall survival, only the distributed parameter model-derived v(e) was predictive of 5-year overall survival with a threshold value of 15.48 mL/100 mL after cross-validation and permutation testing.
Parameter values differ significantly between models. Of the models investigated, the distributed parameter model was the best predictor of 5-year overall survival.
http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.12120186/-/DC1.
比较分析动态对比增强计算机断层扫描（CT）数据分析的 4 种不同示踪剂动力学模型，以预测原发性结直肠癌患者的 5 年总生存率。
本研究经伦理审查委员会批准。回顾性分析 46 例结直肠癌患者的动态对比增强 CT 数据，这些数据是作为一项研究的一部分获得的，使用分布参数、常规室分模型、绝热组织均一性模型和广义动力学模型进行分析。采用 Friedman 检验比较血流、血容量、平均通过时间（MTT）、渗透率-表面积乘积、提取分数、细胞外间隙体积（v(e)）和容积转移常数（K(trans)），P 值<0.05 为差异有统计学意义。采用受试者工作特征曲线分析，比较不同动力学模型的参数和肿瘤分期对总生存的区分能力，采用 Kaplan-Meier 分析和单变量 Cox 比例风险模型，同时进行额外的交叉验证和置换检验。
与常规室分模型和绝热组织均一性模型相比，分布参数模型的血流较低（P<0.0001），而常规室分模型和绝热组织均一性模型的血流值相似。相反，分布参数模型的 MTT 比常规室分模型和绝热组织均一性模型长（P<0.0001）。与绝热组织均一性模型、分布参数模型和广义动力学模型相比，常规室分模型的血容量、渗透率-表面积乘积和 v(e)较高（P<0.0001）。分布参数模型的提取分数高于绝热组织均一性模型。在 5 年总生存方面，只有经过交叉验证和置换检验后，分布参数模型得出的 v(e)阈值为 15.48 mL/100 mL 时，才具有预测 5 年总生存的能力。
各模型之间的参数值差异显著。在所研究的模型中，分布参数模型是预测 5 年总生存的最佳模型。
http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.12120186/-/DC1.
