Multi-center reproducibility of structural, diffusion tensor, and resting state functional magnetic resonance imaging measures

S. Deprez, Michiel B. de Ruiter, S. Bogaert, R. Peeters, J. Belderbos, D. De Ruysscher, S. Schagen, S. Sunaert, P. Pullens, E. Achten

The aim of this study is to assess multi-center reproducibility and longitudinal consistency of MRI imaging measurements, as part of a phase III longitudinal multi-center study comparing the neurotoxic effect following prophylactic cranial irradiation with hippocampal avoidance (HA-PCI), in comparison with conventional PCI in patients with small-cell lung cancer. Harmonized MRI acquisition protocols from six participating sites and two different vendors were compared using both physical and human phantoms. We assessed variability across sites and time points by evaluating various phantoms and data including hippocampal volume, diffusion metrics, and resting-state fMRI, from two healthy volunteers. We report average coefficients of variation (CV) below 5% for intrascanner, intravendor, and intervendor reproducibility for both structural and diffusion imaging metrics, except for diffusion metrics obtained from tractography with average CVs ranging up to 7.8%. Additionally, resting-state fMRI showed stable temporal SNR and reliable generation of subjects DMN across vendors and time points. These findings indicate that the presented multi-site MRI acquisition protocol can be used in a longitudinal study design and that pooling of the acquired data as part of the phase III longitudinal HA-PCI project is possible with careful monitoring of the results of the half-yearly QA assessment to follow-up on potential scanner-related longitudinal changes in image quality.