Abstract #1638

Gandy S, Waugh S, Nicholas R, Rylott T, Milne W, Houston J, Simpson H

In this study multicentre reproducibility of left ventricular cardiac MRI parameters was investigated. Volunteers were imaged twice on a Siemens scanner and once on a GE scanner over two months, and short-axis myocardial borders were segmented by two observers. Scan-to-scan parameter variations (e.g. ejection fraction, left ventricular mass) were similar whether derived from the same or different scanners. However, scan-to-scan variations were larger than single timepoint test-retest variations, and interobserver variability was larger than intraobserver variability. It is possible to simulate longitudinal multicentre conditions and obtain reproducible left ventricular measurements, but single or multiple segmenter use should be carefully considered.