Apparent diffusion coefficient measurements in diffusion-weighted magnetic resonance imaging of the anterior mediastinum: inter-observer reproducibility of five different methods of region-of-interest positioning.

Abstract

To investigate inter-reader reproducibility of five different region-of-interest (ROI) protocols for apparent diffusion coefficient (ADC) measurements in the anterior mediastinum. In eighty-one subjects, on ADC mapping, two readers measured the ADC using five methods of ROI positioning that encompassed the entire tissue (whole tissue volume [WTV], three slices observer-defined [TSOD], single-slice [SS]) or the more restricted areas (one small round ROI [OSR]), multiple small round ROI [MSR]). Inter-observer variability was assessed with interclass correlation coefficient (ICC), coefficient of variation (CoV), and Bland-Altman analysis. Nonparametric tests were performed to compare the ADC between ROI methods. The measurement time was recorded and compared between ROI methods. All methods showed excellent inter-reader agreement with best and worst reproducibility in WTV and OSR, respectively (ICC, 0.937/0.874; CoV, 7.3 %/16.8 %; limits of agreement, ±0.44/±0.77 × 10^-3 mm²/s). ADC values of OSR and MSR were significantly lower compared to the other methods in both readers (p < 0.001). The SS and OSR methods required less measurement time (14 ± 2 s) compared to the others (p < 0.0001), while the WTV method required the longest measurement time (90 ± 56 and 77 ± 49 s for each reader) (p < 0.0001). All methods demonstrate excellent inter-observer reproducibility with the best agreement in WTV, although it requires the longest measurement time. • All ROI protocols show excellent inter-observer reproducibility. • WTV measurements provide the most reproducible ADC values. • ROI size and positioning influence ADC measurements in the anterior mediastinum. • ADC values of OSR and MSR are significantly lower than other methods. • OSR and WTV methods require the shortest and longest measurement time, respectively.