[ENH] Add API for bids pipeline annotating

[adam2392]

PR Description

Addresses the API needed for the annotation GUI to interface with BIDS files of #12

Merge checklist

Maintainer, please confirm the following before merging:

• All comments resolved
• This is not your own PR
• All CIs are happy
• PR title starts with [MRG]
• whats_new.rst is updated
• PR description includes phrase "closes <#issue-number>"

[codecov]

Codecov Report

Merging #60 (a2276de) into main (cec4921) will decrease coverage by 1.58%.
The diff coverage is 83.05%.

❗ Current head a2276de differs from pull request most recent head 531b9a9. Consider uploading reports for the commit 531b9a9 to get more accurate results

@@            Coverage Diff             @@
##             main      #60      +/-   ##
==========================================
- Coverage   97.80%   96.22%   -1.59%     
==========================================
  Files          12       15       +3     
  Lines         501      556      +55     
==========================================
+ Hits          490      535      +45     
- Misses         11       21      +10     

Impacted Files	Coverage Δ
mne_icalabel/label_components.py	95.83% <80.00%> (-4.17%)	⬇️
mne_icalabel/annotation/bids.py	82.00% <82.00%> (ø)
mne_icalabel/annotation/__init__.py	100.00% <100.00%> (ø)
mne_icalabel/config.py	100.00% <100.00%> (ø)
mne_icalabel/iclabel/config.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cec4921...531b9a9. Read the comment docs.

[mscheltienne]

So simple, I love it :) The API from mne-bids looks super easy to re-use!

[mscheltienne]

Does this allow for a component to be labeled with different labels by different authors?

[adam2392]

Unfortunately no...

So I would say perhaps the tsv file stores the "final annotation" and optionally a score produced if it was an automated method

Perhaps what is better is a JSON, where there is a dict of lists with keys:

• component
• author
• label
• date
• (optional but idk how to get it) expertise

And then someone can leverage the JSON to formulate the "labels" training/testing dataset.

How does this sound?

[mscheltienne]

Yes, I think that would be better 👍

[adam2392]

@sappelhoff and @hoechenberger I know BIDS isn't set yet, but wdyt about this for ICA derivatives to handle the problem of annotating labels manually? Anything you think we might be missing?

If this works well, perhaps it'll be a good explicit use case for BIDS-EEG-Derivatives to see and use for their specification for ICA.

[sappelhoff]

I think as long as you mark this as "experimental / development" (with regards to BIDS) it can spark further discussion in bids-standard/bep021#5 and https://bids.neuroimaging.io/bep021 👍