The repeatability and inter-scanner reproducibility of fitted parameters from mono-exponential and non-mono-exponential (stretched exponential, kurtosis and bi-exponential) models of diffusion-weighted MRI signal attenuation were assessed in healthy volunteers and patients with advanced ovarian cancer imaged using MRI scanners from three manufacturers.
Repeatability of ADC estimates, evaluated in abdominal organs in healthy volunteers, was good on all three scanners. Estimates of DDC and α from the stretched exponential model, Dk from the kurtosis model and D from the bi-exponential model showed comparable repeatability to ADC on all three scanners; the standard deviation of differences in k was comparable across three scanners. Repeatability of f, D* and fD* was poor on all three scanners.
ADC estimates showed no significant differences between the three scanners in data from patients or healthy volunteers. Significant differences were observed between scanners in α, k, D and f in data from healthy volunteers. In lesions, there was a significant difference in k between scanners. Differences between parameters estimated from different scanners should be considered in multi-centre studies.