FD an NODDI icvf discrepancies


I have run an FBA analysis using SS3T with b-values 0 and 2500. The other aspects of analysis I followed exactly like in the multi-tissue CSD tutorial. As I have healthy people and multiple sclerosis patients as groups (n=20) I have chosen 10 MS patients with the least lesion volume and 10 HCs for the template. I don’t have any prior experience but it seems that the analysis went fine and now I have the results where in FD for example (HC-SM contrast) there seems to be no significant fwe corrected fixels (uncorrected p-vals show widespread differences). I do realize that this is probably the case and I have to live with it. However, I have also compared icvf NODDI parameter within the normal-appearing white matter in the same cohort and I do see a change with decent effect size. So, my questions are:
From you experience, how does FD relate to NODDI icvf?
Is it sensible to try and transform normal appearing tissue masks to fod_template to create a NAWM group mask and run the fixelstats using this mask? (and expect changes)

Thank you and all the best,

1 Like

Okay, so after literature review, I see that for this type of comparison (voxel measure to fixel measure) people usually use fixel2voxel for conversion. I would like to do that as well (convert FD, FC and FDC to scalar maps). However, I can’t find a definitive answer on how to do that properly. In your 2018 article Fibre-specific white matter changes in multiple sclerosis patients with optic neuritis - ScienceDirect you write “These comparisons required us to convert fibre density from a fixel-wise measure to a simplified voxel- wise measure, and thus reduced the amount of potential information encoded in the fixel-based analysis. The angular integral of the fibre orientation distribution in each voxel was considered as the total fibre density.”. Unfortunately I am to much of a dummy to understand that in terms of mrtrix command. The fixel2voxel documentation states an example where scalar map is made from FC by -mean and -weighted option and I am thinking about using this. In other articles and 1 thesis that I found there was no exact information on how fixel2voxel conversion was done. I would also like to ask if this should be done on tractography smoothed data or not?