Is the NU-regularization in NU-refine carried out at every voxel in the box (with a sliding window controlled by the adaptive window factor)? Or just at voxels within a mask? And if the latter, which mask is used -
The reason I ask is that in high resolution NU-refine jobs, we see a certain (relatively minor!) degree of “ringing” artifacts around strong protein features, even when dynamic masking is disabled.
Consequently I’m wondering if this is due to some other mask being used for NU regularization, or possibly due to the order of the Butterworth filter used for smoothing (I don’t think it’s this as reducing the order doesn’t seem to make a big difference)…?