Optimal fmls_peak_value for FBA

Hi all,

I know that there is no ground truth for fmls_peak_value in fod2fixel, however, I’d be glad to know your opinion on which value is optimal for my data.

The data I’m working on has b=0,700,1000,2000 in 64 directions. I did dwi2response dhollander and ss3t_csd_beta1 on b=0,2000.

Here are different fmls_peak_value index.mif files overlaid on the wmfod_template.mif:

for 0.10 (default):

for 0.08:

for 0.06 (suggested by MRtrix’s FBA guideline):

The lower the value, the more “bleeding” is observed into the non-white matter tissue (such as basal ganglia or maybe CSF). The higher it is, you will lose fixel data in cross-fiber regions (even with fmls_peak_value = 0.1).

My intuition was to go with the middle value, 0.08, however, I don’t know if that “bleeding” by choosing the suggested 0.06 value may be problematic in the next steps.


p.s. another option is to go with a “multi-verse” kind of analysis, and reporting all cut-off values, which computationaly may not be really what I’m looking for.