select which T1 to use when multiple _T1w.nii.gz are present

[bpinsard]

For instance if you acquire in the same session a brain T1w which outputs ..._T1w.nii.gz and ..._rec-norm_T1w.nii.gz (scanner normalization) and also a cervical spinal cord T1w named ....loc-spine_T1w.nii.gz, fmriprep seems to try making an average of all these.
Is there an option to select from a BIDS anat directory, the specific T1w to select for anatomical processing using a glob or else?
Thanks!

[oesteban]

Yes, this is indeed a problematic issue. I'm transferring this issue over to poldracklab/smriprep though, because the anatomical processing has been outsourced.

[bpinsard]

Ok thanks!
I guess it would be useful to have options --anat_bids_filter and --func_bids_filter to pass to pybids in fmriprep and smriprep, and maybe deprecate the -T option.

[oesteban]

Related - #31

[patrick-g-h]

Also interested in this issue! i.e. where there are multiple versions of the same T1w scan due to different gradient correction methods applied.

Happy to help as well. In our case, the file names look like
sub-xx_acq-yy_rec-DIS3D_run-01_T1w
sub-xx_acq-yy_rec-DIS2D_run-01_T1w
so maybe just parsing for the tag and adding a command line argument --t1w_select to specify?

[utooley]

I've been watching the issues/PRs about this since asking about it on Neurostars a long time ago, and was currently going to stick with simply adding the anatomicals that I don't want to use to the .bidsignore (as opposed to removing them entirely), in line with the suggestion by @emdupre on the fmriprep PR. I think this will work generally (but haven't tested it yet), but if this feature is close to being implemented in a new release, I might wait on the trouble of doing that... Any updates?

[effigies]

Adding things to .bidsignore only affects the validator, not pybids or fMRIPrep.

I think the selection PRs are nearly there... I can try to have a look this week.

[emdupre]

Adding things to .bidsignore only affects the validator, not pybids or fMRIPrep.

Sorry @effigies: Is that an expected situation ? I didn't think so based on discussion in bids-standard/bids-specification#131 and bids-standard/pybids#378. Or is this something particular to *MRIPrep ?

[effigies]

It's what I expect... I can read through those threads to see if there was an alternative consensus, though. Not likely to be tonight.

[effigies]

There doesn't seem to be a clear consensus that came out, but I can't pin down a population vector of any magnitude.

I think one problem we'll face in fMRIPrep is that "ignore" is a blunt instrument and it's encoded in the dataset, not in the execution inputs. I may have two perfectly BIDS-valid T1w images but only want to pass one to fMRIPrep. .bidsignore feels like the wrong way to handle that. Second, ignore in the BIDS-validator context means "do not error on", but in pybids it sounds like we want it to mean "do not index". BUT, we use BIDSValidator errors to mean "do not index", so suppressing errors might intuitively mean "index in spite of errors".

The conceptual ambiguity is probably part of why we haven't moved toward making .bidsignore official for anything but bids-validator. And I think this input form is much more clear. I will see if we can push on this and aim for inclusion in fMRIPrep 20.1.