EEG data cleaning with few channels and without eog or ecg channels

Hi, I’m trying to implement artifact rejection in my EEG data, but I have certain limitations.

My data have a maximum of 8 channels, and I don’t have EOG or ECG channels to remove these artifacts. MNE has the option to artificially create these channels, but if you have a certain number of channels that I haven’t reached.

At the moment, my approach is to apply ICA and remove the component with the greatest variance. Additionally, in the case I’m working with ERPs, I remove epochs with a peak-to-peak distance greater than a certain limit.

Since I work with an average of 60 subjects, it’s not an option to implement manual artifact rejection.

Can you help me with alternatives to clean my data with MNE, please?

I’m thankful.

Hello @Jhoham ans welcome to the forum!

I would not recommend doing ICA with so few channels. You’re likely to remove large chunks of physiological activity.

@cbrnr may have some ideas on what would be more appropriate procedures here.

Best wishes,

1 Like

I agree with @richard that ICA requires more than eight channels (at least around 20, but ideally a lot more). Without dedicated EOG channels, you cannot perform regression either. IMO, there’s no way around looking at your data, even if you have many data sets. Blindly applying automatic artifact detection methods is rarely a good idea, but they can definitely supplement manual inspection. You could take a look at Autoreject, but there are many more alternatives available if this is not suitable for your use case.


This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.