reject epochs igoring EOG

šŸ’¬ Support & Discussions
,
1

Dear MNE community:

i am new in the field and i just started to learn how to analyze ERP data. I have a question about epoch rejection:

I did ICA on the continuous data and it corrected eye blinks successfully. And i did epoching and baseline correction. I used autoreject package to clean my epochs. It rejected 36.2% of my critical data, but almost all of the rejected epochs were rejected based on EOG/T. Actually other channels were pretty good. Considering my regions of interest is mainly on central and posterior regions. I feel it is wasteful to reject these many epochs. Is there anyway to ignore the EOG here to save these epochs? Or should i mark them as bad channels, but i feel weird because they are not ā€˜badā€™ channelsā€¦ If I marked EOG as bad channels, would it influence my following analysis, like averaging?

My systems are:

  • MNE version: 1.0.3
  • operating system: Windows 10

My lines of code:

tmin = tmin_epoch # start point of epoch
tmax = tmax_epoch # end point of epoch
baseline = (tmin, tmin+0.1)

target_events = mne.pick_events(events, include=[12,22,32])
event_id = {
ā€˜related/targetā€™: 12,
ā€˜unrelated/targetā€™: 22,
ā€˜nonword/targetā€™: 32}

epochs = mne.Epochs(raw_ica,
events=target_events,
event_id=event_id,
tmin=tmin,
tmax=tmax,
baseline=baseline,
preload=True)

Then i did autoreject, here is the screenshot of results:

% of rejection
% of rejection678Ɨ696 26.3 KB

I also tried to set reject_criteria values, the results were similar to the above.

Could you please give me some suggestions how to save more epochs?

Thank you very much!!

Yunr

2

You should set the channel type of the EOG channel to eog

It appears right now itā€™s treated like an ordinary EEG channel?

3

Hiii. I tried exclude EOGs and created a new epoch group:

epochs_noEyeCh = epochs.copy().pick_types(meg=False,
eeg=True,
eog=True,
emg=True,
stim=True,
exclude=[ā€˜EOG/Lā€™,ā€˜EOG/Rā€™,ā€˜EOG/Tā€™,ā€˜EOG/Bā€™])

Is this OK, or it may influence the following analysis?

4

Thank you very much for your reply!

I did this at the beginning. This screenshot shows the info of data on which i did epoching.
image

5

So it looks like your EOG channels do have the correct type. Can you share the code you used to do the rejection? There is probably a ch_type or picks argument that you can set to 'eeg'. For instance, if you used autoreject.get_rejection_threshold, a threshold for the EOG channel will be computed by default. If you want to reject only based on EEG channels, you have to provide the argument ch_type="eeg".

6

Also, yes you could just drop the EOG channels for the rest of the analysis. Taking your code:

epochs_noEyeCh = epochs.copy().pick_types(
    meg=False,
    eeg=True,
    eog=True,
    emg=True,
    stim=True,
    exclude=[ā€˜EOG/Lā€™,ā€˜EOG/Rā€™,ā€˜EOG/Tā€™,ā€˜EOG/Bā€™]
)

That would work, but there is simpler. If you set eog to False, you are done!

.pick_types(eeg=True, eog=False, emg=True, stim=True)

Or even: .pick_types(eeg=True, emg=True, stim=True) as all arguments of pick_types are set to False by default.

7

Yes! I tried it, and it saved many epochs!
Thank you very much!!! :heart: