Group-average response function in QSIRecon

[AmirHussein]

Hi,

I’m interested in using QSIrecon to generate a structural connectome from HCP-YA data. Specifically, I’d like to run the built-in “mrtrix_singleshell_ss3t_ACT-hsvs” pipeline, and have a few relevant questions in this regard:

1. Is it possible to customize the pipeline to limit the analysis to a specific shell (e.g., b = 3000) so that ss3t can run?
2. Given the importance of group-average FODs for later group analysis, can the pipeline be customized to include one additional step after dwi2response dhollander?
3. Should I run the hippocampal segmentation step in FreeSurfer before this pipeline, or will it do it automatically?

Thanks!

Best,
Amir

[Steven]

Hi @AmirHussein,

@AmirHussein

Is it possible to customize the pipeline to limit the analysis to a specific shell (e.g., b = 3000) so that ss3t can run?

Yes see example here (specifically the requested_shells parameter around line 30): qsirecon/qsirecon/data/pipelines/ss3t_fod_autotrack.yaml at main · PennLINC/qsirecon · GitHub

@AmirHussein

Given the importance of group-average FODs for later group analysis, can the pipeline be customized to include one additional step after dwi2response dhollander?

No, we don’t have any group average functionality at this time. Its on our long term roadmap to implement, but no idea when this feature will be added.

@AmirHussein

Should I run the hippocampal segmentation step in FreeSurfer before this pipeline, or will it do it automatically?

QSIPrep and recon do not run FreeSurfer or use hippocampus segmentations. It only uses the surface file for improved 5TT segmentation.

Best,

Steven

[AmirHussein]

Hi Steven,

Thank you for your reply, and thanks for the requested_shells snippet. Regarding group-average, is it theoretically possible to:

1. Run QSIRecon using a customized config file up to dhollander
2. Calculate group-average response functions using MRtrix functionalities outside QSIRecon
3. Run QSIRecon using a customized config file starting from SS3T-CSD, providing group-average files (or their values, if supported).

What do you think? Thanks in advance.

Best,
Amir

[Adam_Raikes]

@Steven and @AmirHussein. It is possible to precompute the response functions, average them for a group-averaged response, and feed them in. Example recon specs are here:

• Estimate
• Apply

The actual calls to QSIRecon should only differ in two ways:

1. The obvious is which recon spec you pass.
2. When you run apply, you also need –-recon-spec-aux-files /path/to/group_response_functions. As long as the naming for the response functions in that folder matches the response function names in your apply recon spec, it will ingress and use them.

Note that there is a bit of redundant processing that hasn’t been addressed yet. The estimate step will compute the FODs on the subject-specific RFs. These then get overwritten by the FODs from the group-averaged RFs.