Dynamic contrast-enhanced magnetic resonance imaging measurements in renal cell carcinoma: effect of region of interest size and positioning on interobserver and intraobserver variability.

Abstract

The purpose of this study was to assess the influence of region of interest (ROI) size and positioning on perfusion and permeability parameters as well as on interobserver and intraobserver variability of dynamic contrast-enhanced (DCE-MRI) of primary renal cell carcinoma (RCC) and metastases. Thirty-nine DCE-MRI examinations of 34 patients with primary RCC and 20 examinations of 9 patients with RCC metastases obtained at 1.5 T were evaluated. Pretreatment and posttreatment analysis with antiangiogenic therapy was performed in 4 patients with primary RCCs and 5 patients with metastases. The ROIs of the whole tumor (wROI), the circular edge (cROI), a user-defined arbitrary small region (sROI), and a semiautomated segmented ROI were independently defined by 2 readers on 1 slice on arterial phase DCE-MRI images or on parametric plasma-flow maps. Analysis with a 2-compartment exchange model provided 4 parameters: plasma flow (FP), plasma volume (vp), permeability-surface product (PS), and extravascular-extracellular volume (ve). Interobserver and intraobserver parameter correlations were calculated using the intraclass correlation coefficient, and within-subject variability were considered on the basis of the coefficient of variation. Differences in measurement values of variable ROI size were assessed with paired t test. Mean values of FP and vp with sROIs were significantly higher than those with wROI, cROI, and semiautomated segmented ROI placement in tumor or metastases. Values of ve showed no significant difference between ROI sizes. The highest interobserver and intraobserver correlation with 0.99/0.98 for metastases and 0.97/0.98 for primary RCCs, respectively, was observed for all parameters when defining wROIs on dynamic images. Perfusion parameters of wROI measurements for FP (dynamic, 0.97; parametric maps, 0.96) and vp (0.95/0.89) showed higher interobserver correlation than did permeability parameters ve (0.64/0.6) and PS (0.79/0.5) in primary RCCs. The wROIs showed also the lowest within-subject coefficients of variation for perfusion parameters FP and vp compared with cROI and sROIs in primary RCCs and metastases. The ROI size and positioning do substantially influence quantitative perfusion and permeability parameters in DCE-MRI. The best interobserver and intraobserver correlation can be obtained when defining a whole-tumor ROI. The perfusion parameters are the most reliable, whereas the permeability parameters are more susceptible to interobserver variability. No significant differences between placing ROIs on morphological or parametric images were observed.