Ringing in FA images after TBSS registration

Software Support
1

Summary of what happened:

Hello everyone!
I am trying to analyse DTI data using FSL. I preprocessed the data using QSIPrep and output looks good. However, when running TBSS I see some ringing appear in the all_FA.nii along the corpus callosum for most of the subjects. My guess is that it is a registration problem, but I don’t see how to fix it. Would anyone have any insights?
Thank you in advance!

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

tbss_1_preproc
tbss_2_reg -T
tbss_3_postreg -S

Version:

FSL 6.0.4

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

Data formatted according to a validatable standard? Please provide the output of the validator:

BIDS

Relevant log outputs (up to 20 lines):

PASTE LOG OUTPUT HERE

Screenshots / relevant information:

ringing
ringing925×746 18 KB

2

Hi @jruizfdez,

Not a direct answer, but if you just need metrics in a common space, I might recommend using QSIRecon for that. Ringing could be due to the interpolation method. Sometimes using something simple like nearest neighbor can mitigate it.

See an example recon spec with spatial normalization here, which you can tailor for your needs (in particular thetemplate_map node): qsirecon/qsirecon/data/pipelines/abcd_recon.yaml at main · PennLINC/qsirecon · GitHub

Best,
Steven

3

Hi @Steven, thank you for your fast response! I have looked into it and tried to run it but it crashes eventually - my data has a single-shell acquisition scheme. Would you recommend a specific pipeline suited for that? For now, I am interested in FA, MD values.
Thank you!
Best,
Julia

4

Hi @jruizfdez,

Yes, you cannot use the ABCD recon spec on singleshell acquisitions. But you can use the metrics from the DSI Studio GQI modeling (which includes tensor metrics).

Yes, we have several pipelines here qsirecon/qsirecon/data/pipelines at main · PennLINC/qsirecon · GitHub. In particular, you would be interested in qsirecon/qsirecon/data/pipelines/ss3t_fod_autotrack.yaml at main · PennLINC/qsirecon · GitHub, but editing it to uncomment the last block to enable MNI mapping:

# Uncomment to map scalar maps to template space
# -   action: template_map
#     input: qsirecon
#     name: template_map
#     parameters:
#         interpolation: NearestNeighbor
#     scalars_from:
#     - gqi_scalars
#     software: qsirecon

This pipeline runs GQI modeling (which includes tensor metrics), runs tractography on SS3T FODs, and maps all the microstructural metrics to bundles and to MNI (when you uncomment the last block).

Best,
Steven

5

Hi @Steven,
Thank you for your help! That pipeline did run successfully.
Best,
Julia

6

Hi @Steven,

I do have a follow-up question. To do (wholebrain) voxel wise analyses, would the following pipeline be sound?

fslmerge -t all_FA.nii.gz sub-*_space-MNI152NLin2009cAsym_model-tensor_param-fa_dwimap.nii.gz

fslmaths all_FA -Tmean mean_FA

fslmaths mean_FA -thr 0.2 -bin mean_FA_mask     

randomise -i all_FA.nii.gz -o FA_stats -d design.mat -t design.con -m mean_FA_mask.nii.gz -n 5000 -T

Thank you for your help!
Best,
Julia

7

Hi @jruizfdez,

You could do this, or you could use the white matter segmentatino from MNI space as your mask as opposed to getting it from the FA. Also, I do not know what is in your design matrix / contrast file. It could be good to include a quality matrix from QSIPrep (e.g., raw_neighbor_corr or t1post_dwi_contrast).

Best,
Steven