Misunderstanding in iEEG electrodes.tsv description of "name" column?

[sappelhoff]

For the name column in an electrodes.tsv file, the spec says:

The label must correspond to _electrodes.tsv name and all ieeg type channels are
required to have a position.

This doesn't make sense technically. An electrode is an electrical contact, a channel is a potential measurement between two electrodes. When applying an average reference, or a dedicated (common to all) reference electrode, it is kind of sensible to name a channel like an electrode, and this is arguably the most common practice, and where a lot of confusion between electrode vs channel comes from.

The concept breaks apart for bipolar recordings: Imaging an EOG recording with an electrode named VEOG+ above the left eye, and its counterpart named VEOG- below the left eye. Both these electrodes together form the channel named VEOG ... however, in electrodes.tsv, this bipolar recording would need two rows, not a single one. Then again in channels.tsv, there would only be VEOG, a single row.

Originally posted by @sappelhoff in #827 (comment)

--> We don't have such a description in MEG or EEG. Is this a small bug that needs to be corrected @dorahermes? Or am I not seeing the full picture here?

I would cut out the requirement "The label must correspond to _electrodes.tsv name and all ieeg type channels are
required to have a position.", because I think it doesn't make sense in all situations.

[sappelhoff]

Tagging some more people, because I think the sooner we clear this up, the better.

@robertoostenveld @rob-luke @dorahermes @hoechenberger @adam2392

[robertoostenveld]

I agree with your description. Electrodes in/on the tissue relate to, but are not the same as channels in the amplifier and ADC chain. Also, there can be electrodes for which the position is not specified (e.g. the shaft of a multicontact LFP probe in animals) because it is extended, and/or because the position is not precise (e.g. large stickers on the left and right of the eyes for EOG). Vice versa, there can also be electrodes of which the position is specified but for which there is no channel, such as REF and GND.

The specification for name in _channels.tsv now states

Label of the channel. The label must correspond to _electrodes.tsv name and all ieeg type channels are required to have a position. The reference channel name MAY be provided in the reference column.

I suggest to change it to

Label of the channel. When a corresponding electrode is specified in _electrodes.tsv, the name that electrode can be specified here and the reference electrode name MAY be provided in the reference column.

Here I also dropped the mentioning of the position, as that is something for the electrodes file, not the channels file.

[robertoostenveld]

I think there are two aspects to consider in the current text. Let me highlight them in bold:

Label of the channel. The label must correspond to _electrodes.tsv name and all ieeg type channels are required to have a position. The reference channel name MAY be provided in the reference column.

The second one is easy: channels don't have a position (but electrodes have).

The first one probably was initially inspired to have an unambiguous link between channels and electrodes. Having two columns, each with one electrode label in it, allows mapping the channel to the electrodes. If a channel were named "ch1", then you would not know to which electrodes (the one of interest and the reference) it pertains. If a channel were named "Cz-M1" (using EEG here, as more people are probably familiar with the electrode naming), then a single name disambiguates the link between channel and electrodes. If name=Cz and referenmce=M1, then it is also unambiguous.

[dorahermes]

I agree that the 'channels are required to have a position' is not correct. The first version was indeed inspired to have an unambiguous link between electrodes and channels, to ensure good practices, but this already creates an issue if you have a trigger channel.