Missing Tracts in pyAFQ Output

[emmarichard21]

Hello!

I recently ran AFQ using Docker (I’ve included the command used below), and I noticed that for 10 of my participants, there were very few tracts present in the whole-brain tractography output. I've included an example of someone's whole brain tractography file where there are missing tracts.

I checked the QSIPrep HTML reports for these participants, and everything appeared to be in good shape. Additionally, no errors were reported in the pyAFQ HTML pages. I’m not sure what might have gone wrong for these specific participants.

I was wondering if you might have any insight into what could cause this issue or suggestions for additional things I should check.

docker run -ti --rm
-v /data/BILCOR/BIDS/derivatives/qsiprep_dering/qsiprep/:/data
-v /data/BILCOR/BIDS/derivatives/pyafq_110425/:/output
-v /home/johnanderson/license.txt:/opt/freesurfer/license.txt
pennlinc/qsirecon:latest
/data /output participant
--fs-license-file /opt/freesurfer/license.txt
--recon-spec pyafq_tractometry

[arokem]

Yeah, that looks funky. A couple of things to look at:

• Did the whole brain tractography look OK for these subjects? If whole brain tractography looks unusual, there might be some things to investigate there, like whether there is something wrong with the b-vecs definitions, or something else unusual about the acquisition or processing up to that stage.
• Did registration of the subject's individual data to the MNI template fail for some reason? You can check that by looking at the file called something like "sub-01_ses-01_space-mni_b0ref.nii.gz" and comparing it to the MNI template, which you should be able to find in ".cache/templateflow/tpl-MNI152NLin2009cAsym/tpl-MNI152NLin2009cAsym_res-01_T2w.nii.gz"

[emmarichard21]

Thanks for the response!

Did the whole brain tractography look OK for these subjects?

• No it did not. The image in the initial message is an example of someone's whole brain tractography, where it did not run properly.

Did registration of the subject's individual data to the MNI template fail for some reason?

• I don't think so. I checked out the b0ref image and compared it to an MNI template and it was aligned good

[emmarichard21]

@arokem

For a participant, the raw whole-brain tractography (before AFQ cleaning) looks kinda reasonable, but after AFQ’s cleaning and bundle organization the tractogram becomes very sparse, and the resulting bundle segmentations contain very few streamlines. However, when I compare this subject to another participant, there are definitely fewer streamlines in the raw whole-brain tractography to begin with.

I’ve also rerun a newer version of QSIPrep and then rerun AFQ, but the outputs were still poor for this participant.

Any help on what I should try next would be appreciated! Thank you!

Image 1: example of raw whole-brain tractography (before AFQ cleaning) for a participant with poor tract bundles

Image 2: example of post AFQ cleaning whole-brain tractography for the same participant above

Image 3: example of raw whole-brain tractography (before AFQ cleaning) for a participant with good tract bundles

[36000]

Hello,
Thank you for the additional information. Is this the only subject with this problem? there are a few things to check. First, I would check the warped b0 in template space and see if it aligns with the MNI template. Then I would check the template warped into subject space and see if it aligns with the b0. QSIprep outputs both of these by default. I would also check that the brain mask and masked outputs produced by pyAFQ look roughly correct. You can also check the FA map and even threshold it by 0.2 to see that it is anisotropic in the white matter.

If everything still looks good, you can always increase the number of seeds! change n_seeds from 1 to 2 in the .yaml file. This should probably be the default anyways. You can even increase it to 3 if you want to wait a long time but generate a lot of streamlines, though at some point you get diminishing returns.

Let me know how it goes!