Dynamic contrast-enhanced MRI in head-and-neck cancer: the impact of region of interest selection on the intra- and interpatient variability of pharmacokinetic parameters.

PURPOSE: Dynamic contrast-enhanced (DCE) MRI-extracted parameters measure tumor microvascular physiology and are usually calculated from an intratumor region of interest (ROI). Optimal ROI delineation is not established. The valid clinical use of DCE-MRI requires that the variation for any given parameter measured within a tumor be less than that observed between tumors in different patients. This work evaluates the impact of tumor ROI selection on the assessment of intra- and interpatient variability. METHOD AND MATERIALS: Head and neck cancer patients received initial targeted therapy (TT) treatment with erlotinib and/or bevacizumab, followed by radiotherapy and concurrent cisplatin with synchronous TT. DCE-MRI data from Baseline and the end of the TT regimen (Lead-In) were analyzed to generate the vascular transfer function (K(trans)), the extracellular volume fraction (v(e)), and the initial area under the concentration time curve (iAUC(1 min)). Four ROI sampling strategies were used: whole tumor or lymph node (Whole), the slice containing the most enhancing voxels (SliceMax), three slices centered in SliceMax (Partial), and the 5% most enhancing contiguous voxels within SliceMax (95Max). The average coefficient of variation (aCV) was calculated to establish intrapatient variability among ROI sets and interpatient variability for each ROI type. The average ratio between each intrapatient CV and the interpatient CV was calculated (aRCV). RESULTS: Baseline primary/nodes aRCVs for different ROIs not including 95Max were, for all three MR parameters, in the range of 0.14-0.24, with Lead-In values between 0.09 and 0.2, meaning a low intrapatient vs. interpatient variation. For 95Max, intrapatient CVs approximated interpatient CVs, meaning similar data dispersion and higher aRCVs (0.6-1.27 for baseline) and 0.54-0.95 for Lead-In. CONCLUSION: Distinction between different patient's primary tumors and/or nodes cannot be made using 95Max ROIs. The other three strategies are viable and equivalent for using DCE-MRI to measure head and neck cancer physiology.

Dynamic contrast enhanced-MRI in head and neck cancer patients: variability of the precontrast longitudinal relaxation time (T10).

PURPOSE: Calculation of the precontrast longitudinal relaxation times (T10) is an integral part of the Tofts-based pharmacokinetic (PK) analysis of dynamic contrast enhanced-magnetic resonance images. The purpose of this study was to investigate the interpatient and over time variability of T10 in head and neck primary tumors and involved nodes and to determine the median T10 for primary and nodes (T10(p,n)). The authors also looked at the implication of using voxel-based T10 values versus region of interest (ROI)-based T10 on the calculated values for vascular permeability (K(trans)) and extracellular volume fraction (v(e)). METHODS: Twenty head and neck cancer patients receiving concurrent chemoradiation and molecularly targeted agents on a prospective trial comprised the study population. Voxel-based T10's were generated using a gradient echo sequence on a 1.5 T MR scanner using the variable flip angle method with two flip angles [J. A. Brookes et al., "Measurement of spin-lattice relaxation times with FLASH for dynamic MRI of the breast," Br. J. Radiol. 69, 206-214 (1996)]. The voxel-based T10, K(trans), and v(e) were calculated using iCAD's (Nashua, NH) software. The mean T10's in muscle and fat ROIs were calculated (T10(m,f)). To assess reliability of ROI drawing, T10(p,n) values from ROIs delineated by 2 users (A and B) were calculated as the average of the T10's for 14 patients. For a subset of three patients, the T10 variability from baseline to end of treatment was also investigated. The K(trans) and v(e) from primary and node ROIs were calculated using voxel-based T10 values and T10(p,n) and differences reported. RESULTS: The calculated T10 values for fat and muscle are within the range of values reported in literature for 1.5 T, i.e., T10(m) = 0.958 s and T10(f) = 0.303 s. The average over 14 patients of the T10's based on drawings by users A and B were T10(pA) = 0.804 s, T10(nA) = 0.760 s, T10(pB) = 0.849 s, and T10(nB) = 0.810 s. The absolute percentage difference between K(trans) and v(e) calculated with voxel-based T10 versus T10(p,n) ranged from 6% to 81% and from 2% to 24%, respectively. CONCLUSIONS: There is a certain amount of variability in the median T10 values between patients, but the differences are not significant. There were also no statistically significant differences between the T10 values for primary and nodes at baseline and the subsequent time points (p = 0.94 Friedman test). Voxel-based T10 calculations are essential when quantitative Tofts-based PK analysis in heterogeneous tumors is needed. In the absence of T10 mapping capability, when a relative, qualitative analysis is deemed sufficient, a value of T10(p,n) = 0.800 s can be used as an estimate for T10 for both the primary tumor and the affected nodes in head and neck cancers at all the time points considered.