Easiest way to combine channels while retaining other existing channels?

đź’¬ Support & Discussions
,
1
  • MNE-Python version: 0.22

Hello, I’m looking for a way to easily

  • combine two channels into a new, virtual channel,
  • add the virtual channel to the existing data, and
  • drop the two combined channels from the data.

Concrete application: I have EEG data with two HEOG and two VEOG electrodes. I’d like to combine the two electrodes into a bipolar channel (i.e., calculate the differences) for HEOG and VEOG, respectively, for use in further processing.

Is there any convenient way to do this?

I do know about mne.channels.combine_channels(), but it will only return the combined channels and none of the other existing channels. I could take the combined channels and add them back to my original data using Raw.add_channels(), and eventually drop the source channels via Raw.drop_channels(). But this feels very cumbersome and hacky. :exploding_head:

Does anybody have a better idea how to achieve the same result more easily?

Thanks,

–Richard

2

I don’t know a better way, but it’s the first time I’ve heard of mne.channels.combine_channels. How would you use it to compute a bipolar channel? Like this (assuming that HEOG and VEOG have indices 64 and 65, respectively)?

mne.channels.combine_channels(raw,
                              {"eog": [64, 65]},
                              method=lambda x: np.diff(x, 0))
3

Yes that’s how I thought I could do it…

Note, however, that I have two channels for each EOG, i.e. I’d like to have VEOG = VEOG1 – VEOG2, and the equivalent for HEOG. But yes you got the principle right.

… But I can tell you’re not really convinced either! :slight_smile:

4

I think what I want is what set_bipolar_reference() does, just for VEOG and HEOG.

5 
import mne
path = mne.datasets.sample.data_path()

raw = mne.io.read_raw_fif(f'{path}/MEG/sample/sample_audvis_raw.fif', preload=True)

# Copy the EOG channel for demonstration purposes
single_eog_channel = raw.copy().pick_channels(['EOG 061'])
single_eog_channel.rename_channels({'EOG 061': 'EOG 062'})
raw.add_channels([single_eog_channel])

# Create a bipolar EOG channel.
# The resulting signal should be EOG 061 - EOG 062, so all zeros, because in
# this example we just copied the channel.
raw = mne.set_bipolar_reference(raw, 'EOG 061', 'EOG 062', ch_name='HEOG')

# There should now be an all zeros HEOG channel
raw.plot()
1 Like
6

By default, the original EOG 061 and EOG 062 channels are dropped.

7

Hello @wmvanvliet, thank you for your example!

But wouldn’t this reference all my signals to the new, virtual channel? (I haven’t tried it yet, will do later)

8

No. Unlike functions such as set_eeg_reference, set_bipolar_reference just takes the channels you give it, subtracts them from each-other and writes the result to a new channel. Then (optionally) removes the original channels. It was designed with exactly your use case in mind.

1 Like
9

Interesting! Then I wonder if the name isn’t a bit misleading? Because … you know, set_*_reference(), I expect it to set a reference :slight_smile: I will try it out, thank you so much!!

10

Reading the doctoring again, yes, it all makes sense the way you explained it!!

Re-reference selected channels using a bipolar referencing scheme.

A bipolar reference takes the difference between two channels (the anode
minus the cathode) and adds it as a new virtual channel. The original
channels will be dropped.

I think the function name is kind of ill-chosen, though… Anyway, I’ll give it a shot and report back! Thank you so much for your suggestion, @wmvanvliet

12

Thanks @wmvanvliet, it’s doing what I want! Awesome!

1 Like