Fmriprep fsLR output space den-91k

Hello - we used fmriprep 20.1.1 with the --output-spaces fsLR, among other spaces. We also used the --cifti-output flag but that might not be necessary since (I think) fmriprep automatically adds this if you request fsLR space.

We have space-fsLR_den-91k outputs and I’m wondering why 91k? Isn’t it convention to have the XXk be the mesh resolution, similar to how TemplateFlow organizes the different resolutions. If that’s the case, shouldn’t the outputs be den-32k? I used wb_command to look at the file dimensions and the data are definitely on the 32k mesh.

I’m wondering if this is a bug or if this was intended? Yeah there are 91282 rows in the cifti but not all of them are on the mesh. If this is the intended behavior, then how would it change if I specified a fsLR:den-32k output space? Shouldn’t the output file names be in the format that I request? It seems confusing to me to have a den-91k since there is no template with this density in TemplateFlow.

fMRIPrep doesn’t produce standalone fsLR surfaces, and instead creates the full CIFTI-2 file in fsLR space. We use den-91k to denote that the output is in the 91282 Greyordinates space - since this file includes surface vertices (left/right cortex) as well as subcortical volumetric voxels.

If you wish to extract just the surface information, you can use the Connectome Workbench. I think the command to do it would be:

wb_command -cifti-separate <cifti-file> COLUMN -metric CORTEX_LEFT <output-gifti0> -metric CORTEX_RIGHT <output-gifti1>