[Eeglablist] How to deal with interpolated channels when doing ICA artifact rejection

[Maitane Barrenetxea Carrasco]

Hi everyone,

When trying to reject the artifacts of a dataset, I am concerned about how
to deal with the previously interpolated channels.

In my preprocessing pipeline, I bandpass-filter the signal, then I reject
bad channels (if any) and I re-reference the signal to average. Finally, I
perform ICA on the non-interpolated channels (e.g. if 2 channels have been
previously rejected and interpolated, then the remaining 126 channels are
used for ICA decomposition) and reject the artifactual ICs.

It is unclear to me what to do afterwards, as the previously interpolated
channels are still contaminated with artifacts that were removed via IC
selection. One possible solution to this problem would be to perform a
second interpolation on these channels. Would this be an acceptable
solution?

I have also considered using PCA for dimensionality reduction, although it
is not recommended to use it prior to ICA (
https://sccn.ucsd.edu/pipermail/eeglablist/2010/003339.html)

Thank you very much in advance,

Maitane