Quantitative, Organ-Specific Interscanner and Intrascanner Variability for 3 T Whole-Body Magnetic Resonance Imaging in a Multicenter, Multivendor Study

Introduction Whole-body magnetic resonance (MR) imaging is increasingly implemented in population-based cohorts and clinical settings. However, to quantify the variability introduced by the different scanners is essential to make conclusions about clinical and biological data, and relevant for internal/external validity. Thus, we determined the interscanner and intrascanner variability of different 3 T MR scanners for whole-body imaging. Methods Thirty volunteers were enrolled to undergo multicentric, interscanner as well intrascanner imaging as part of the German National Cohort pilot studies. A comprehensive whole-body MR protocol was installed at 9 sites including 7 different MR scanner models by all 4 major vendors. A set of quantitative, organ-specific measures (n = 20; eg, volume of brain's gray/white matter, pulmonary trunk diameter, vertebral body height) were obtained in blinded fashion. Reproducibility was determined using mean weighted relative differences and intraclass correlation coefficients. Results All participants (44 ± 14 years, 50% female) successfully completed the imaging protocol except for two because of technical issues. Mean scan time was 2 hours and 32 minutes and differed significantly across scanners (range, 1 hour 59 minutes to 3 hours 12 minutes). A higher reproducibility of obtained measurements was observed for intrascanner than for interscanner comparisons (intraclass correlation coefficients, 0.80 ± 0.17 vs 0.60 ± 0.31, P = 0.005, respectively). In the interscanner comparison, mean relative difference ranged from 1.0% to 53.2%. Conversely, in the intrascanner comparison, mean relative difference ranged from 0.1% to 15.6%. There were no statistical differences for intrascanner and interscanner reproducibility between the different organ foci (all P ≥ 0.24). Conclusions While whole-body MR imaging-derived, organ-specific parameters are generally associated with good to excellent reproducibility, smaller differences are obtained when using identical MR scanner models by a single vendor.