Figure 9.

Amount of physically implausible FA values (Verio and Skyra, MPIL sequence, n = 23 + 28) is most reduced after unringing (red) with the Kellner Method (paired Bayesian t-test: BF[unfiltered ~ Kellner Method] > 8 × 10⁹) but does not differ significantly between scanners neither before nor after differing preprocessing pipelines (Bayesian linear modelling, full/null model comparison: $BF\left[\frac{\left(voxels with FA > 1\right)\sim scanner \times preprocessing pipeline}{\left(voxels with FA > 1\right)\sim preprocessing pipeline}\right]$ = 0.05 ± 1.09%).