Fixel2voxel dimension mismatch

Hi all,

I’m new to MRtrix, so I hope my issue can be addressed easily (due to my misunderstanding). I ran into the following issue:

I have MRtrix FOD data, as generated by qsiprep. I segmented the FOD to fixels with:

$ fod2fixel sub-003_ses-mri01_acq-diffusion1p6mm16b0120b1250120b2500mb3x2_space-ACPC_model-msmtcsd_param-fod_label-WM_dwimap.mif.gz fixels -nii -maxnum 2
fod2fixel: \[100%\] uncompressing image “sub-003_ses-mri01_acq-diffusion1p6mm16b0120b1250120b2500mb3x2_space-ACPC_model-msmtcsd_param-fod_label-WM_dwimap.mif.gz”
fod2fixel: \[100%\] segmenting FODs

I then converted the output to (sparse) volumetric nifti format with (the NIFTI error is just FYI, all outcomes are the same if I use mif format):

$ fixel2voxel fixels/directions.nii none directions_2.nii
fixel2voxel: \[ERROR\] image “fixels/directions.nii” is not in NIFTI-1.1 format (sizeof_hdr != 348)
fixel2voxel: \[100%\] determining largest fixel count
fixel2voxel: \[100%\] converting sparse fixel data to scalar image

If I look at the dimensionality of the fixel direction-image I get the expected dim2 = 3

$ fslinfo fixels/directions.nii
data_type	FLOAT32
dim1		472584
dim2		3
dim3		1
dim4		1

However, for the same data converted to volume, I get an unexpected dim4 = 2 (I expected it to be 6, i.e. 3 coordinate-values for fixel 1 and another 3 for fixel 2). How should I interpret this?

mentat001/dwi\_$ fslinfo directions_2.nii
data_type	FLOAT32
dim1		101
dim2		123
dim3		107
dim4		2

(If I leave out -maxnum 2 in fod2fixel then I get dim4 = 11, which is just as weird)