Interpolate or remove bad channels before EEG source reconstruction

Discussions
1

Dear BST Masters,

I have an EEG dataset with some bad channels (on average, 4-5 out of 64 per subject), already preprocessed in EEGLAB with the bad channels interpolated and then average-referenced. I imported it into BST for source reconstruction.

My key question: Is it correct to do source reconstruction on such data or should I mark the bad channels as bad in BST before doing it (as I guess interpolation at the scalp level may add noise in the source reconstruction)?

In the latter case, as average-reference was computed including the interpolated channels, should I import the data before average-reference and then average-reference in BST after marking the bad channels (though I understand that this last step is irrelevant for source reconstruction)?

Thanks!

Best,

Marco

2

Dear all,

While waiting for a response, I reasoned that the most probable correct answer is: Yes, I should remove the bad channels instead of interpolating them before doing source reconstruction.

Now my question is: how should I do that? Should I

  1. mark the bad channels on each single recording, including resting state, and then do source reconstruction?
  2. Or should I remove the channels from the electrode channel file?

In case 1), if I already did source reconstruction with all the channels, is it sufficient to mark the channels as bad in the recordings to automatically have the corrected source reconstruction or should I run the head model and source reconstruction again?

Thanks,

Best,

Marco

3

That's right, the interpolated channels will not provide additional information to the source estimation.

You would need to run the source estimation again, the head model is still valid.

@tmedani, you may want to take a look to this post

1 Like
4

Thank you for your reply Raymundo, this is clear now.

Best,

Marco