[Eeglablist] Decay in frequency contents when ICA is applied

Mikołaj Magnuski imponderabilion at gmail.com
Wed May 21 16:26:33 PDT 2014


Dear Haj,

I am not an expert on ICA but I don't think averaging across components is
a fair comparison to mean over electrodes. First of all the components
differ in the amount of variance they explain in the signal - so not taking
this into account will result in treating very weak components (that may
represent rather 'noise' than signal, variance unexplained by other,
physiologically realistic components) with the same weight as 'good'
components explaining a lot of the data. Also keep in mind that components
that explain a lot of the eeg signal may be far less numerous than the
'weak' components (for example you may have 6 components [out of 64] that
explain most of the alpha activity in the eeg).
Then, some of the 'strong' components do not represent brain activity at
all - like blinks, heart or muscle activity. Some of alpha activity may
leak into these components but in ideal situation such components should
represent only the artifactual signal.
Lastly - the magnitude of component signal is not as informative as
electrode signal. This is because the magnitude of IC signal is distributed
between the component signal and component weights - you can search this
list for more information on this, this topic was discussed here lately.
21 maj 2014 19:02 "Haj Morez" <hajmorez at gmail.com> napisał(a):

> I am using AMICA ICA for decomposing EEG signals and study behavior of the
> EEG signals before and after applying ICA.
> At the moment I am trying  to compare spectral content of raw EEG signals,
> with spectral content of the ICA decomposition.
> I am using the "spectopo()" function from EEGLAB, and calculate the peak
> amplitude at 10 Hz averaged over all channels. Then trying to compare it
> with peak amplitude at 10 Hz averaged over all components (ICs).
> I noticed, there is a dramatic decay in the spectral content of the signal
> when ICA is applied. The mean peak amplitude of the ICs is dramatically
> lower than channels! (there was no IC rejection).
> The 10 Hz peak amplitude averaged over channels is ranging between 20-50
> dB, however it ranges between -20 to -5 dB when it is calculated over ICs.
> I really appreciate it if you could help me to explain the reason why it
> happens.
>
> P.S. I had same results when used EX-infomax ICA.
>
> Kindest regards,
> Haj
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140522/9528b35c/attachment-0001.html>


More information about the eeglablist mailing list