[ENH] channels.tsv and electrodes.tsv should have a clear relationship

[neuromechanist]

Describe your problem in detail.

EEG-BIDS and iEEG-BIDS correctly differentiate between channels and electrodes. The list of channels goes into the channels.tsv and the list of electrodes goes into the electrodes.tsv.

While electrophysiological channels are created from the potential difference between two or more electrodes (that is electrophysiology channels are a function of the electrodes), the electrodes.tsv and channels.tsv seem somehow independent and do not have a mechanism to indicate the relation. Here is what the EEG-BIDS specifications text for this:

the entries in *_electrodes.tsv and *_channels.tsv do not have to match exactly, as for example in the case of recording a single EOG channel from a bipolar referencing scheme of two electrodes, or a data channel originating from an auxiliary, non-electrode device. That is, in most cases *_electrodes.tsv will have more entries than *_channels.tsv.

This lack of indicating relationship can create ambiguity. Overall there might be at least four scenarios that tools should check to determine how the channels and electrodes are related:

1. have the same order of channels and electrodes, respectively
2. have the same channels and electrodes, but out of order
3. there are more channels than electrodes, like this issue (BIDS electrodes.tsv import leads to incorrect chanlocs when channels.tsv and electrodes.tsv include different extra entries sccn/EEG-BIDS#236) with audio and pressure sensor channels (or in general, non-electrophysiological channels).
4. there are more electrodes than channels, for example, with some bipolar electrodes like EOG, creating fewer channels (the case that the specification was anticipating).

IMO, this ambiguity hinders the reuse and proper analysis of the datasets, and should be addressed at the specifications level, not the tools level.

Describe what you expected.

We have been discussing this issue at #1371 with @bids-standard/bep042, as EMG channels may have several electrodes associated with them (for example monopolar, bipolar, or double bipolar setups). The current suggestion (which is far from final) is to add a signal_electrode column to the channels.tsv so that the author can indicate how each channel is related to the elements of the electrodes.tsv, see examples of this implementation in bids-standard/bids-examples#480