BIDS naming of T2w 2D TSE slab for hippocampal segmentation

Hi all!

We’re collecting high-resolution, anisotropic (0.4 x 0.4 x 2.0mm) 2D T2w TSE slabs, aligned orthogonal to the main hippocampal axis, to be used for hippocampal subfield segmentation and delineation of entorhinal cortex. What naming convention should one use for these in the BIDS structure?

My first thought was to use something like anat/sub-XX_acq-hipposlab_T2w.nii.gz, i.e., _T2w as modality label, however I’m afraid that they will then be picked up by fmriprep’s surface preprocessing routines.

To complicate things further, we also collect isotropic 3D T2w images (“T2-SPACE”) that ideally should be used by fmriprep (i.e., anat/sub-XX_acq-space_T2w.nii.gz).

Any suggestions are welcome!


Hi @markushs

You may consider anat/sub-XX_acq-highres_T2star.nii.gz (I considered T2star because it is a high-resolution T2 image, but T2w would work too. Adding hp to acq to become highreshp so you may see that it is specifically the high-resolution hippocampal scan)

Regarding the isotropic T2w image - is it simply a T2 image? Perhaps it would work to name it anat/sub-XX_T2w.nii.gz with the difference between these two names more apparent

Hi @franklin

Thanks for the input! Yes, using _T2star as modality for the HC slab should work (i.e., it won’t be picked up by fmriprep). However, this solution seems to somewhat defeat the purpose of the BIDS-naming, as the scans are not T2*-weighted.

I guess this is not really a BIDS-issue. As I read the specification there’s no problem in having several _T2w anatomicals associated with a participant, as long as they are named differently (e.g., using the _acq field).

Issues appear, however, when attempting to run these kinds of datasets through fmriprep, as fmriprep/smriprep apparently indiscriminately feeds any images marked as _T2w to the Freesurfer routines.

I see there’s an open issue over at smriprep touching upon this, As a workaround for now it looks like adding _run-01 to our isotropic T2w image (intended for freesurfer’s surface preprocessing) will ensure that only this one is picked up by fmriprep. I might add some further comments/suggestions over at github.

Again, thanks!


1 Like