BIDS: How to label PURE bias-corrected T1w images?

Hi there,

We’re prepping a dataset for sharing, and our GE scanner gave us two sets of T1-weighted anatomical images for this scan: an uncorrected image, and one that’s gone through a “PURE” filter which does some form of bias correction.

My plan is to include both images in our shared dataset, and to indicate which is which with either the “acq-” k/v pair, or the “ce-” k/v pair. It’s one acquisition, so using “acq” seems weird. But there’s no contrast agent, so using “ce” seems weird.

Any takes on the best thing to do here?

Thanks!

Hi Nate,

I would say either of these options:

  1. Use the acq label. It’s not perfect as you note, but you can describe this designation in the BIDS readme file to clarify.

  2. Put the PURE T1 in derivatives. Since many T1 preprocessing workflows already do bias correction, you may not want a software such as fmriprep automatically taking your T1 and running it through the anatomical workflow, which would happen if the PURE T1 is in the BIDS root (though you can tell fmriprep to ignore the PURE T1 with a BIDS filter file).

My gut opinion is that option 1 would be better.

Best,
Steven

1 Like

Another suggestion would be:

  • use the rec-<label> label as the PURE method is part of the reconstruction pipeline, not the acquisition itself. The raw data is the same for the images reconstructed with or without PURE. It is a similar case with the PRESCAN NORMALIZE option with Siemens scanners.
    You may then need to use a bids filter option in your analyses to not process those two versions of your T1w images together:
    FAQ - Frequently Asked Questions — fmriprep version documentation
2 Likes

Thank you both! I think rec-<label> is the winner.