[ENH] Specify how to share cross-talk and fine-calibration for Neuromag/Elekta/MEGIN data

[sappelhoff]

This is a proposal on how cross-talk and fine-calibration files can be shared using BIDS.

Both files are necessary to make Neuromag/Elekta/MEGIN data usable, more information can be found in these threads:

1. MEG fine-calibration and crosstalk files mne-tools/mne-bids#495 ... including a comprehensive discussion about implementation options
2. Non BIDS files in MEG example ds000248 bids-examples#56 ... basically pointing out the issue many months ago

The proposed solution would require validator adjustments:

1. .dat becomes a valid extension for MEG data ... but ONLY if acq-crosstalk is part of the file name
2. not sure if we need any adjustments for *_acq-calibration_meg.fif

Disclaimer: I never worked with Neuromag/Elekta/MEGIN data

Attention needed:

1. is what I am writing in the text correct? I basically took it from comments by @robertoostenveld @jasmainak and @hoechenberger
2. is it fine to simply rename the dat and fif files without breaking any internal structures?
3. do you agree that this is a reasonable way to enable sharing of these apparently crucial files?

@bids-standard/raw-electrophys-meg @agramfort @larsoner @guiomar

[jasmainak]

I'm wondering if ".dat" file should be a special case for fif files or whether fif files should also be stored in "father level" directories like KIT or 4D. That way you can add any other files that go with the recording. I realize this would break backwards compatibility though.

[hoechenberger]

Thanks for writing up this proposal, @sappelhoff!

1. .dat becomes a valid extension for MEG data ... but ONLY if acq-crosstalk is part of the file name

I'd like to point out that the fine-calibration file is really just an ASCII text file with columns. So if one has to rename the file anyway to adhere to the proposed BIDS naming scheme, one might as well change the extension to something that's already considered valid? The file I have here consists of multiple columns (without a column header); the first column is separated from the second via a single tab character; and all other columns are separated using a single space. So unfortunately this doesn't qualify as a TSV :(

not sure if we need any adjustments for *_acq-finecalibration_meg.fif

That's quite a long name, maybe *_acq-finecal_meg.fif would suffice?

1. is what I am writing in the text correct? I basically took it from comments by @robertoostenveld @jasmainak and @hoechenberger
2. is it fine to simply rename the dat and fif files without breaking any internal structures?

Yes, there are no cross-references embedded in those files. For using them, one points the processing software to the exact location of either file, and that's sufficient.

3. do you agree that this is a reasonable way to enable sharing of these apparently crucial files?

Yes I believe this should be sufficient.

Since the files won't change within sessions (i.e., they'll be the same for all tasks & runs), the full names could be:

• sub-<label>[_ses-<label>]_acq-crosstalk_meg.dat
• sub-<label>[_ses-<label>]_acq-finecalibration_meg.fif

[robertoostenveld]

the suggestec changes look good to me

[sappelhoff]

@hoechenberger just a note that we typically wait for 5 days until merging a PR so to give everybody a chance to review.

See also decision-making

Just as a friendly note for the future. :)

If @jasmainak or @guiomar disagree with the changes we'll have to make adjustments post-merge.

[hoechenberger]

Oh. Whoops. I'm sorry, I just waited for the Merge button to turn green :<