Add support for applying an EEG template montage

[hoechenberger]

For datasets where no digitizer information (i.e., measured sensor locations) are available, users may wish to use the template montage supplied by the manufacturer (or simply calculated according to the 10/20 etc. systems). This PR adds the eeg_template_montage configuration setting that allows to use MNE-Python's built-in templates.

---

IMHO, this again highlights an issue with the current BIDS specification: that it's not allowed to store montage / channel locations in the dataset if the locations are from a template and not actually measured. For the dataset I'm using for testing here, ds001810, I had to dig into *_eeg.json to find:

        "CapManufacturer": "Biosemi",
        "CapManufacturersModelName": "CAP 64 10/20",

and then I had to start searching for the template – fortunately, it's included in MNE-Python :) But this brings me to the next issue with not including the template / template locations in BIDS: if MNE hadn't included this particular template, I would have had to juggle yet another file that I have to store somewhere (since it's not included in the original dataset). This is just asking for problems… :)

cc @sappelhoff

[sappelhoff]

I would have had to juggle yet another file that I have to store somewhere

It'd have been a small PR to MNE-Python I guess 🤷 but I am not a dictator, you can always open a new issue or comment on the old ones, and convince other BIDS people 👍

edit: for example you could lobby that a new field is included in electrodes.tsv or coordsystem.json that unambiguously tells whether locations are template or measured (that was proposed multiple times by several people - including yourself, I think? I lose track 😬 )

[hoechenberger]

It'd have been a small PR to MNE-Python I guess 🤷

But then still one somehow has to correctly map from the BIDS data to the respective MNE template… which, I'm afraid, cannot be automated, right?

edit: for example you could lobby that a new field is included in electrodes.tsv or coordsystem.json that unambiguously tells whether locations are template or measured (that was proposed multiple times by several people - including yourself, I think? I lose track 😬 )

Yeah I don't remember my vote on this anymore either, but I do remember that we once touched upon this in a discussion, yes :)

[sappelhoff]

Let me prepare a starter kit for you 😆

• Encoding electrode positions in EDF, EDF+, BDF bids-standard/pyedf#7
• Discussion: BIDS-RAW specification --> Sharing of template data ... yes/no? bids-standard/bids-specification#318
• Template versus True digitization data mne-bids#264

these three issues and all crosslinks should cover 90% of what has been publicly discussed on this issue. It's a lot to read, but maybe worth it :)

[hoechenberger]

Oh my!! Thanks @sappelhoff

[hoechenberger]

But then still one somehow has to correctly map from the BIDS data to the respective MNE template… which, I'm afraid, cannot be automated, right?

I'm wondering if valid CapManufacturersModelName entries should be a part of the specification, and only a restricted list of values should be accepted. That way, one could be certain that two entries with the same name actually refer to the exact same cap model… And mapping from BIDS to MNE could safely be automated.

[hoechenberger]

@jasmainak Any more thoughts on this one? Otherwise it's good to merge from my end!

[jasmainak]

Looks good to me, merged. Thanks @hoechenberger !

[hoechenberger]

Thanks!