Motivation: MRI quantifiability and repeatability is hindered by non-biological sources of variability (e.g. scanner hardware/software). Approaches to harmonising acquisition and data/features exist, but objective evaluation is lacking. Goal(s): To develop one of the most comprehensive brain multi-modal MRI harmonisation resources. Approach: We ran a travelling-heads study consisting of 20 participants, 8 3T scanners across major MRI vendors (GE/Siemens/Philips), with 5 MRI modalities (T1w/T2w/dMRI/fMRI/SWI). Nine participants had 5 additional within-scanner repeats. Results: We showcase how the resource can be used to map between-scanner, within-scanner, between-subject variability for hundreds of neuroimaging-derived features and to evaluate efficacy of explicit and implicit harmonisation approaches. Impact: Our openly-released multimodal brain MRI harmonisation resource considers a range of vendors and scanner generations (wide/narrow bore, low/high gradients, different coil numbers). It can therefore enable powerful studies of between-scanner effects and provide a testbed for evaluation of harmonisation approaches.
Warrington et al. (Tue,) studied this question.