RFC: Discrete aparc/aseg segmentations - naming issues

[oesteban]

Right now, we generate files with names like GLS005_ses-2_task-graphlocalizer_space-MNI152NLin2009cAsym_desc-aparcaseg_dseg.nii.gz. However, the contents of the segmentation are the same regardless of the task, because this is just the aseg/aparc resampled into standard space with the resolution of the BOLD, which is generally homogeneous.

Should we leverage res- here and minimize the number of different aparc/aseg conversions in standard spaces? (if all tasks have exactly the same shape & zooms, then generate just one conversion without res entity).

[effigies]

Do they always have the same shape and zooms?

[oesteban]

(if all tasks have exactly the same shape & zooms, then generate just one conversion without res entity).

As I mentioned (see above) - no, they don't have to be matched. But it is the most common case.

[ariekahn]

Not only would this reduce duplication, but it would simplify figuring out which volumes are the same down the road.

For myself, I've definitely compared the aparc/aseg volumes manually to verify whether an ROI extraction, for instance, needs to use the respective segmentation for each run, or if I can just load the aparc for the first run and use that for the rest of the BOLD data.

If this is indeed always consistent for a combination of space and res, it might reduce some potential confusion.