[Eeglablist] ICA identification of sub-gaussian components

Thomas Witzel witzel at MIT.EDU
Thu Jul 28 13:18:00 PDT 2005


I have been using runica to identify artifact components from MEG data. This
works fairly well, however I'd be very interested in extracting the 60 Hz
component. I used runica with 'extended',-1, but this resulted only in slower
convergence, but no 60 Hz component was extracted. The 60 remains in
essentially all components the same way. Is there anything I can do to tackle
this problem ?

I'm doing the ICA separately on the magnetometer and gradiometer data and I
perform a PCA in my own software first. I tried using 20 and 40 PCA components
for the ICA, both with no success. Lowpass on the data was 200 Hz.

Thank you very much,
Thomas Witzel

More information about the eeglablist mailing list