Inter- and intra-rater reproducibility of quantitative dynamic contrast enhanced MRI using TWIST perfusion data in a uterine fibroid model.

PURPOSE To determine the reproducibility of TWIST-derived (Time-Resolved Angiography with Interleaved Stochastic Trajectories) quantitative dynamic contrast enhanced (DCE) MRI in a uterine fibroid model. MATERIALS AND METHODS The institutional review board approved this retrospective study. Dynamic contrast-enhanced TWIST datasets from 15 randomly selected 1.5 Tesla pelvic MR studies were postprocessed. Five readers recorded kinetic parameters (K(trans) [volume transfer constant], ve [extracellular extravascular space volume], kep [flux rate constant], iAUC [initial area under the gadolinium-time curve]) of the largest uterine fibroid using three region-of-interest (ROI) selection methods. Measurements were randomized and repeated three times, and measures of reproducibility were calculated. RESULTS The intra-rater coefficients of variation (CVs, brackets indicate 95% confidence intervals) varied from 4.6% to 7.6% (K(trans) 7.6% [6.1%, 9.1%], kep 7.2% [5.9%, 8.5%], ve 4.6% [3.8%, 5.4%], and iAUC 7.2% [6.1%, 8.3%]). ve was the most reproducible (P < 0.05). Inter-rater reproducibility was significantly (P < 0.05) greater for the large ROI method (range of intraclass correlation coefficients [ICCs] = 0.80-0.98 versus 0.48-0.63 [user-defined ROI] versus 0.41-0.69 [targeted ROI]). The uterine fibroid accounted for the greatest fraction of variance for the large ROI method (range across kinetic parameters: 83-98% versus 56-69% [user-defined ROI] versus 47-74% [targeted ROI]). The reader accounted for the greatest fraction of variance for the user-defined ROI method (0.4-14.1% versus 0.1-3.0% [large ROI] versus <0.1-1.5% [targeted ROI]). CONCLUSION Changes in TWIST-derived DCE-MRI kinetic parameters of up to 9-15% may be attributable to measurement error. Large DCE-MRI regions of interest are the most reproducible across multiple readers.