Improving EDF reading in line with specification

Describe the new feature or enhancement

In https://github.com/mne-tools/mne-python/pull/9694#issuecomment-908135503, I realized that the EDF specification can “encode” the channel type. One can have a label string in EDF be:

EEG Fz, implying that it is the Fz channel of eeg type. However, it’s “optional”.

For exporting EDF data, I would like to augment the export to encode the channel type, such that if you have channels:

EOG1, ECG1, A1, A2, A3, Fz, Cz, then they would be written to EDF like:

EOG EOG1
ECG ECG1
sEEG A1
sEEG A2
sEEG A3
EEG Fz
EEG Cz

Describe your proposed implementation

First, I would modify these lines: https://github.com/mne-tools/mne-python/blob/22cd7747529db7895345917c4d3884a482ed8ab3/mne/io/edf/edf.py#L387-L416

to allow for backwards-compatible behavior that also then sees if there is the “channel type” encoded in the label string.

Additional comments

This addition would not “hurt” our codebase, since it really just adds backwards-compatible behavior to allow for additional parsing of the channel type upon reading the EDF file. In addition, it would add more robust exporting. Overall, it is an improvement in our roundtrip of compliant EDF files.

That way I can modify the string of PRs:

• https://github.com/mne-tools/mne-python/pull/9694
• #9643

See:

• https://edfplus.info/specs/edftexts.html#signals for description of string channel types in EDF
• https://edfplus.info/specs/edftexts.html#label for the specification of the channel label in EDF

[adam2392]

So if you have the following channels:

{
   A1 = 'eeg',
   A2 = 'eeg',
   ECG1 = 'ecg',
   EOG1 = 'eog',
   LMT1 = 'seeg'
}

Then a parameter would either let their labels be written as the following in the EDF file?

A1, A2, ECG1, EOG1, LMT1

or

EEG A1, EEG A2, ECG ECG1, EOG EOG1, SEEG LMT1

Is there any reason to want the labels be the first way, rather than the second? Mistakes by user?

[larsoner]

Indeed I would make a list of all the channel names that had a type that we didn’t know and say we mapped them to EEG at the end, something like:

bad_map = dict()
for ci in range(channels):
    if this_ch_type not in ch_type_dict:
        bad_map[ch_name] = this_ch_type
        this_ch_type = 'eeg'
if len(bad_map):
    bad_map = '\n'.join(f'{ch_name}: {ch_type}' for ch_name, ch_type in bad_map.items()'
    warn(f'Found the following unknown channel type mapping(s), setting the channel type to EEG:\n{bad_map})