Optimal approach for intensity normalization in multi-shell data for quantitative analysis



Dear experts,

while reading

I was wondering how is current recommended approach for intensity normalization of multi-shell data for group-wise quantitative analysis (FBA).
As I understand, there are currently three commands for intensity normalization:

  1. dwibiascorrect - bias correction using ANTS or FSL (i.e. removing effect of spatially-varying intensity)
    According to documentation, ANTS give better performance than FSL, although some issues have been raised with ANTS which are probably solved now.
  2. dwiintensitynorm - global intensity normalization across subjects, the prerequisite is dwibiascorrect, requiring time consuming registration to common template
  3. mtnormalise - joint bias field and intensity normalization, avoiding registration to common template, can be performed independently on each subject. Does it, according to your experiences, effectively solve bias field correction and between-subject within group intensity normalization?

To be specific if I have multi-shell data (b=0, b=1000 64 directions, b=2500 64 directions), wanting to do FBA, time consuming registration to common template is not a problem, which way would be currently recommended?
a) dwibiascorrect (ANTS, with updated parameters) + dwiintensitynorm
b) dwibiascorrect (ANTS, with updated parameters) + mtnormalise
c) mtnormalise only
d) mtnormalise + dwiintensitynorm
e) dwibiascorrect (ANTS, with updated parameters) + mtnormalise + dwiintensitynorm (overkill?)

EDIT: never mind, I will stick with official FBA tutorial, using dwibiascorrect + mtnormalise, supposing dwiintensitynorm is not needed in this case.