Limit Epochs rejection to subset of channels / exclude channels from rejection

I’m working with EEG recordings that include data from electrodes placed very close to the eyes, capturing lots of EOG activity. I use these channels to identify and auto-clean EOG artifacts using ICA. I have ICA operate on Epochs.

Because I want to avoid passing any Epochs with spuriously large signal amplitudes to ICA, I set a reject parameter for my epochs. This successfully removes a bunch of epochs where the signal indeed jumped way out of bounds for an electrode or two.

However, the EOG amplitudes I get in Fp1 and Fp2 are so big that in order to not remove a large number of epochs based on the EOG signal alone, I’d have to set the rejection threshold so high that I’d then be unable to reject epochs where some of the other electrodes went rogue. So this is a problem for me: either I set the threshold so high that it more or less becomes useless; or I lower it and sacrifice a lot of data due to the EOG signal.

What I’d like to do is specify “drop epochs as bad if amplitude exceeds X, unless this occurs in Fp1 and Fp2”.

Wanted to discuss with you

• if you think this would be good to have (I very much do!)
• and in a second step, what the API could look like

Thanks!

cc @cbrnr @agramfort

[hoechenberger]

Ok, everybody cool with going with

reject={"eeg": 100e-6, "Fp1": 2e-5, "Fp2": 2e-5}

then…?

[drammock]

thinking about your particular use case @hoechenberger, would it work to mark the frontal channels as “bad”, run the epoching, then set them back to “good” before running ICA?