FBA analysis: reorienting + assigning tensor metrics fixels

Hi! I’m working through a multi-tissue FBA analysis. So far so good, but I’m unclear on how to handle tensor metrics (FA, RD, etc) I’m also interested in. Apologies in advance for the newbie questions.

My plan is detailed below, but I’m not sure if a) I should even be converting the tensor metrics to fixels (vs running mrclusterstats), and b) if I should, then do I have the correct steps below (e.g, not sure if I need to reorient and assign these fixels).

  1. Calculate tensors (dwi2tensor)
  2. Calculate tensor metrics (tensor2metric)
  3. Register metrics to population template:
mrtransform \
-reorient_fod no \
./tensor_metrics/${sub}_${ses}-${metric}.mif \
-warp ./template/FODs_in_template_space/${sub}_${ses}-to-template_warp.mif \
  1. Continue with conversion to fixels or just run mrclusterstats on the registered output from step 3?
voxel2fixel \
./tensor_metrics/${sub}_${ses}-${metric}_in_template_space.mif  \
./template/fixels/${sub}_${ses}-fixel_in_template_space \
./template/tensor_metrics_fixels/${metric}/${sub}_${ses}-${metric}-fixel \
  1. Reorient:
fixelreorient \
./template/tensor_metrics_fixels/${metric}/${sub}_${ses}-${metric}-fixel \
./template/FODs_in_template_space/${sub}_${ses}-to-template_warp.mif \
  1. Assign:
fixelcorrespondence \
./template/tensor_metrics_fixels/${sub}_${ses}-${metric}-reoriented/${sub}_${ses}-${metric}-fixel.mif \
./template/fixels/fixel_mask \
./template/tensor_metrics_fixels/${metric}-reoriented-assigned \
  1. Run fixelcfestats.


Hi ! Wondering if I should backtrack here to my primary question:

If i have tensor metrics in voxels (FA, RD, etc), is there any advantage to converting them to fixels (voxel2fixel) to run fixelcfestats over simply running voxel-wise stats with mrclusterstats? I’m unclear as to what I would gain by going to fixels if the same FA value is applied to all fixels in a voxel.

(I’ve already created a population template following the FBA pipeline so I can test the fixel-based metrics (FD, FDC, etc).


I found the paragraph in Raffelt et al 2015 that answers my own question :slight_smile:

We also note that while the proposed method was designed to investigate fixel-specific measures, the CFE method could also be used to investigate voxel-average quantities (e.g. myelin water fractions).This would be achieved by mapping the value at each voxel to all fixels within that voxel, then using CFE as described in this work. While this is not optimal since the quantitative measure is not fixel-specific in regions with crossing fibres, smoothing and enhancement would still be more tract-specific than if performed using traditional 3D smoothing and clustering, and the estimated p-value of fixels within the same voxel will differ based on the different connectivity-derived neighbourhoods

1 Like