Following on the suggestion by @Lestropie (this commit):
RS: Currently missing from here is the mtnormalise command. I do not know the extent to
which the UKB pre-processing deals with the B1 bias field, as many models perform all
computations relative to the b=0 image intensity, and therefore a multiplicative bias
field is irrelevant; but for AFD it is highly critical. I would suggest just using
the mtnormalise command as-is here. It is however worth checking the quality of the
brain masks in some exemplar subjects: inclusion of non-brain voxels in the mask here
can be highly detrimental:
https://github.com/MRtrix3/mrtrix3/issues/1973
So unlike the mask used for CSD above, where a generous mask is preferable, here a more
conservative mask is preferable; indeed it may be preferable to erode the mask slightly
for mtnormalise.
Following on the suggestion by @Lestropie (this commit):
RS: Currently missing from here is the mtnormalise command. I do not know the extent to which the UKB pre-processing deals with the B1 bias field, as many models perform all computations relative to the b=0 image intensity, and therefore a multiplicative bias field is irrelevant; but for AFD it is highly critical. I would suggest just using the mtnormalise command as-is here. It is however worth checking the quality of the brain masks in some exemplar subjects: inclusion of non-brain voxels in the mask here can be highly detrimental: https://github.com/MRtrix3/mrtrix3/issues/1973 So unlike the mask used for CSD above, where a generous mask is preferable, here a more conservative mask is preferable; indeed it may be preferable to erode the mask slightly for mtnormalise.