TractSeg Producing Empty Tractograms from QSIPrep Outputs

emma_richard · November 19, 2025, 5:05pm

Summary of what happened:

I’m trying to use TractSeg (GitHub - MIC-DKFZ/TractSeg: Automatic White Matter Bundle Segmentation) to segment white matter tracts from diffusion data that was preprocessed with QSIprep.

Initially, I wasn’t sure whether I needed to convert my data to MNI space, since the TractSeg documentation states:

“For best results the input image must have the same orientation as the HCP data (MNI space)… If the image orientation and the gradient orientation of your data is the same as in examples/Diffusion.nii.gz you are fine.”

My diffusion image has the same orientation as the example file, so I assumed a transformation to MNI space wasn’t necessary. However, I tested both approaches:

Running TractSeg with the MNI transformations
Running it without any MNI transformation

In both cases, my commands ran without errors, but when generating bundles with --nr_fibers, TractSeg reported that the tract masks were empty, and thus produced empty tractograms, even though the masks are visible when I open them in FSLeyes (see image).

I’m wondering if this issue is related to the space of the QSIprep outputs, or possibly to my bval/bvec transformations not being applied correctly. I’m not fully sure where the pipeline is breaking.

Any advice or suggestions would be greatly appreciated!

Command used (and if a helper script was used, a link to the helper script or the command generated):

#Transformation to MNI space for TractSeg

calc_FA -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.nii.gz -o .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/FA.nii.gz --bvals .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.bvals --bvecs .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.bvecs \
--brain_mask .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/nodif_brain_mask.nii.gz


flirt -ref .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/MNI_FA_template.nii.gz -in .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/FA.nii.gz \
-out .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/FA_MNI.nii.gz -omat FA_2_MNI.mat -dof 6 -cost mutualinfo -searchcost mutualinfo


flirt \
  -ref .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/MNI_FA_template.nii.gz \
  -in .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.nii.gz \
  -out .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/Diffusion_MNI.nii.gz \
  -applyxfm -init FA_2_MNI.mat \
  -dof 6

cp \
  .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.bvals \
  .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/Diffusion_MNI.bvals

rotate_bvecs \
  -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/sub-02_acq-b0b1000b2000_dir-AP_space-T1w_desc-preproc_dwi.bvecs \
  -t FA_2_MNI.mat \
  -o ...BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output/Diffusion_MNI.bvecs

# Commands to run TractSeg
TractSeg -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/Diffusion_MNI.nii.gz -o .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/ --raw_diffusion_input --output_type tract_segmentation

TractSeg -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/peaks.nii.gz -o .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/ --output_type endings_segmentation


TractSeg -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/peaks.nii.gz -o .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/ --output_type TOM


Tracking -i .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/peaks.nii.gz -o .../BIDS/derivatives/qsiprep_dering/qsiprep/sub-02/dwi/tractseg_output_MNI/tractseg_output/tractseg_output/ --nr_fibers 5000

Environment (Docker, Singularity / Apptainer, custom installation):

Docker from GitHub:

 sudo docker run -v /absolute/path/to/my/data/directory:/data \
-t wasserth/tractseg_container:master TractSeg -i /data/my_diffusion_file.nii.gz -o /data --raw_diffusion_input

Screenshots / relevant information:

Steven · November 19, 2025, 11:18pm

Hi @emma_richard,

What if you make your own FODs from QSIPrep outputs (using the .b MRTrix-style gradient file) and pass that into TractSeg? The bval/bvec files from QSIPrep are in FSL format (image space), which differ from MRTrix (world space). You can also use QSIRecon to get FODs (and do all of tractography for that matter, if you aren’t tied to TractSeg).

Best,
Steven