Comparison of the reproducibility of quantitative cardiac left ventricular assessments in healthy volunteers using different MRI scanners: a multicenter simulation.

PURPOSE: To derive reproducibility assessments of ejection fraction (EF) and left ventricular mass (LVM) from short-axis cardiac MR images acquired at single and multiple time-points on different 1.5T scanner models.
MATERIALS AND METHODS: Images of 15 healthy volunteers were acquired twice using a Magnetom Avanto scanner (Siemens, Erlangen, Germany) and once using a Signa Excite scanner (General Electric, Milwaukee, WI, USA) over four months, and analyzed using ARGUS and MASS Analysis+ software, respectively. Two physicists independently segmented the myocardial borders in order to derive intra- and interobserver assessments of EF and LVM for single and multiple time-points on the same and different scanners.
RESULTS: For EF, the coefficient of repeatability (CoR) increased as different observers, multiple time-points, and different scanners were introduced. The CoR ranged from 2.8% (intraobserver measurements, single time-point, same scanner) to 10.0% (interobserver measurements, different time-points, different scanners). For LVM, intraobserver CoR parameters were consistently smaller than interobserver values. The CoR ranged from 7.8 g (intraobserver measurements, single time-point, same scanner) to 39.5 g (interobserver measurements, different time-points, different scanners).
CONCLUSION: Reproducible EF data can be obtained at single or multiple time-points using different scanners. However, LVM is notably susceptible to interobserver variation, and this should be carefully considered if similar evaluations are planned as part of multicenter or longitudinal investigations. (c) 2008 Wiley-Liss, Inc.