ENH: add interpolate_to method

[antoinecollas]

Reference issue (if any)

Closes #12486

What does this implement/fix?

Implements interpolate_to next to interpolate_bads to interpolate EEG data to a given montage

Additional information

Interpolating channels using this implementation has shown to be effective in

Mellot, A., Collas, A., Chevallier, S., Engemann, D. and Gramfort, A., 2024. Physics-informed and Unsupervised Riemannian Domain Adaptation for Machine Learning on Heterogeneous EEG Datasets. EUSIPCO 2024

[welcome]

Hello! 👋 Thanks for opening your first pull request here! ❤️ We will try to get back to you soon. 🚴

[drammock]

I see three options:

option 4: you add @antoinecollas's fork as a new remote, create a branch based off this antoinecollas/interpolate_to branch, push your changes to your own MNE-Python fork on GitHub, then open a PR from your fork to Antoine's fork. That method (often called "PR into your PR") doesn't require Antoine to grant you any special permissions, but it does require that they review & merge your "sub-PR" in order for your commits to show up in this PR.

$ # from within your MNE-Python folder:
$ git remote add antoinecollas git@github.com:antoinecollas/mne-python.git
$ git fetch antoinecollas
$ git checkout -b interpolate_to antoinecollas/interpolate_to
$ # make some changes, make some commits
$ git push -u origin interpolate_to

then on GitHub, open a PR with antoinecollas/mne-python as base repository, interpolate_to as base, contsili/mne-python as head repository, and interpolate_to as head. (see steps 4 & 5 here: https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/proposing-changes-to-your-work-with-pull-requests/creating-a-pull-request-from-a-fork)

[larsoner]

I think this is actually pretty close to being done, just a few little ideas for improvements and more complete testing. So maybe worth waiting a few days to get this in then continue work with extending it?

[antoinecollas]

Thanks for the review, @larsoner. I have taken it into account.

[larsoner]

Does the CI error make sense to you?

E   AssertionError: 2 nesting errors:
E   Line 997:  ("mne/channels/channels.py", "        from .._fiff.proj import _has_eeg_average_ref_proj", "hierarchy: must not nest _fiff"),
E   Line 1003: ("mne/channels/channels.py", "        from . import DigMontage", "non-explicit relative import"),

if not then we should maybe improve the error messages

[antoinecollas]

not super clear...
I changed the imports, is it better now?

[larsoner]

not super clear...
I changed the imports, is it better now?

Must have been clear enough, because you did fix it 😆 !

I'll do one last review and mark for merge-when-green assuming it looks okay 👍

[larsoner]

Just some tiny tweaks I'll commit, thanks in advance @antoinecollas !

[welcome]

🎉 Congrats on merging your first pull request! 🥳 Looking forward to seeing more from you in the future! 💪

[agramfort]

congrats @antoinecollas for nailing this one !

[mh1300]

Does the interpolate_to() method assume that the target montage has already been coregistered on the same headspace as the EEG data? If the method does not assume this, where is this coregistration performed inside of the method or is the coregistration not required? Finally, if the method assumes that the target montage has already been coregistered to the headspace, should that be added to the documentation?

Thanks a ton for the help!

[larsoner]

EEG montages in MNE-Python are always operated on in Neuromag head coordinates. If they don't start there but there is LPA/Nasion/RPA present, they should in theory be converted to that coord frame before anything happens like interpolation (this happens internally in set_montage for example). In theory the interpolate_to should take the montage, transform it to head coords if it's not already there, then do the interpolation. We should probably make sure that this is the case, though!