[Eeglablist] EMG artifact question

[Bachman, Peter]

Thank you very much, Scott.  That makes a lot of sense.

Peter


-----Original Message-----
From: eeglablist-bounces at sccn.ucsd.edu on behalf of Scott Makeig
Sent: Thu 7/30/2009 10:57 PM
To: Bachman, Peter
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] EMG artifact question
 
Peter - You have removed the major eyeblink artifact components (not shown)
-- Likely they contain no visible high-frequency broadband (EMG-like)
activity. You are left with the rest of the EEG projecting to the same
channels, including local scalp muscle signals, which are very likely
accounted for by a few other ICs. I see no evidence that the (probable)
muscle activity has increased -- though it has become more clearly visible
after the large eyeblink potentials have been removed.  Of course no spatial
filter, including those learned by ICA, is ever 'perfect,' but I see no
evidence here of the increase you are talking about.

Actually, however, the data that remain after removing one or more ICs may
well actually be larger (e.g., in variance) than the original data. This
happens when two source with opposite-phase waveforms (in some limited time
period) sum at a channel, nearly canceling each other out there. In this way
the summed variance of the individually back-projected ICs is (nearly
always, with probability approaching 1) larger than the variance of the sum
of their backprojections, the original data...

Scott Makeig

On Thu, Jul 30, 2009 at 9:35 AM, Bachman, Peter <bachman at psych.ucla.edu>wrote:

>  Hi everyone,
>
> In the course of using ICA in EEGLAB to remove artifacts from EEG data,
> I've noticed that what appears to be EMG-related contamination will
> sometimes not only persist in the dataset, but actually appear to become
> somewhat amplified and even distributed over channels in which it did not
> appear in the raw data.
>
> I can imagine why EMG activity may be less likely to collect on to a single
> component for easy removal - so why it would not be easily removed - but I'm
> not sure why it would appear to be more of a problem after the raw data is
> decomposed with ICA and then reconstructed back into EEG channel data (with
> amplitude scale and other parameters held constant).
>
> I've attached two .gif images with examples of this phenomenon.  In both
> instances, eye blinks are removed very nicely, but EMG activity seems worse
> after the cleaning.
>
> I'm wondering if anyone else has seen this before.  If so, do you
> understand why this might happen and how to avoid it?
>
> Thank you very much!
> Peter
>
> _______________________________________________
> 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
>



-- 
Scott Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0961, http://sccn.ucsd.edu/~scott

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20090731/a01c2ec2/attachment.html>