[Eeglablist] [Fwd: Eye Movement Artifacts and Filtering]
Timothy O'Keefe [BROWN]
timothy_okeefe at brown.edu
Thu May 15 05:56:16 PDT 2008
EEGLAB Users,
I might try to "inject" my opinion here, for whatever it's worth.
Among other endeavors involving ICA, I've had significant dealings with
ocular artifact "correction" using EEGLAB/ICA. Now, while I'm not an
"expert" by any means, I would like to provide some food for thought.
What it will feed has yet to be determined :)
So far as introducing bipolar electrodes - or anything for that matter -
into the input matrix, I was always under the assumption that it should
not matter as ICA should not determine any relationships which include
these nonsense(?) data.
As for "cleaning" EEG data. This is one problem that I've found
particularly frustrating. My inner demon has me questioning *exactly*
which components EOG - or the influences of EOG - manifest?
How does one test this with a measurable degree of certainty? Better
yet, how does one test this without making assumptions? How good of an
indicator of "cleanliness" is a "flat" EOG ERP anyways? Pretty good?
Good enough? Is it enough to bet your life on (where some of us might
be!!)?
Best,
Tim
Scott Makeig wrote:
> ICA can decompose common reference and bipolar channel pairs
> simultaneously, since the sources sum linearly at all these channels
> proportional to the difference in source projection strengths to the
> 'active' and 'reference' channels. However, plotting the scalp maps of
> ICA components including both types of channels is not apt to be wholly
> accurate, since the baselines for the two channel types are
> different.However, iIt should be possible to reconstruct the full
> component scalp maps (at least for 'dipolar' components) using a forward
> (BEM) head model... In the absence of such a procedure, one can always
> remove the bipolar channels *after* running ICA decomposition, using the
> EEGLAB 'remove channels' function. This will allow plotting the
> component maps based on the (common reference) montage.
>
> Scott Makeig
>
> On Sun, Apr 20, 2008 at 3:45 PM, Nicolas Robitaille
> <nicolas.robitaille at umontreal.ca
> <mailto:nicolas.robitaille at umontreal.ca>> wrote:
>
> Dear Arno,
>
> Just a related idea:
>
> Would it be a better practice to simply include the EOG channels as
> if they were just others EEG electrodes (if they were all recorded
> with say a left-earlobe reference, as the others EEG channels, or
> reference-free using a Biosemi system) instead of bipolar channels,
> when possible? In my mind, this would help ICA to isolate these
> signals (which is not always trivial for horizontal eye movements,
> which are smaller and can be less frequent than eye blinks),
> according that at least some channels would have clear eye-related
> signals. This would also avoid the difference in referencing methods
> across channels you mentionned.
>
> Nicolas
>
> > Date: Sun, 20 Apr 2008 23:13:52 +0200
> > From: arno at ucsd.edu <mailto:arno at ucsd.edu>
> > To: modestino at ccs.fau.edu <mailto:modestino at ccs.fau.edu>
> > CC: eeglablist at sccn.ucsd.edu <mailto:eeglablist at sccn.ucsd.edu>
> > Subject: Re: [Eeglablist] [Fwd: Eye Movement Artifacts and Filtering]
>
> >
> > Dear Justin,
> >
> > > I am planning to use ICA in EEGLAB to remove eye movement
> artifacts. I
> > > know that ICA does not use EOG channels locate the component
> associated
> > > with eye movements. However, my advisor will want me to show
> ERP averages
> > > of the EOG channels as evidence that the eye movements have
> been removed
> > > using ICA. So, I will be including the EOG channels (bipolar, but
> > > represented as individual difference waves as one channel, for
> HEO and
> > > VEO).
> > >
> >
> > It is OK to include bipolar channels for that purpose. ICA should be
> > able to detect the difference in reference (although you might
> have to
> > search for the component that accounts for that). Then you may
> compute
> > ERP for your EOG channels. A better idea though would be to do
> that for
> > frontal channels such as FPz.
> >
> > > Finally, I need to run an acausal FIR high pass filter at 1 Hz.
> [which
> > > will not change the temporal components of the EEG] on the data
> before
> > > running ICA, as there is an extreme trend in the data that made
> automated
> > > and even manual eye movement removal inadequate using the raw
> data outside
> > > of EEGLAB. Also, the trend needs to be removed anyway. However,
> removing
> > > the linear trend will not adequately alter the tend in the
> continuous data
> > > as the FIR filter will. So, I wish to use such a filter. [I may
> remove
> > > the linear trend once the data is epoched per condition, but
> not on the
> > > continuous data.] What sort of filters does EEGLAB have built
> in? Is this
> > > filter already available?
> > >
> >
> > Yes, there are 2 types of filters available in EEGLAB by default. I
> > would advise to use the non-linear IIR filter which has better
> design
> > than the linear FIR filter (and does not introduce phase distortion
> > because it is applied in both directions). There is also an
> additional
> > plugin that allows more advanced filtering.
> >
> >
> http://www.uni-leipzig.de/~biocog/content/widmann/eeglab-plugins/#firfilt
> <http://www.uni-leipzig.de/%7Ebiocog/content/widmann/eeglab-plugins/#firfilt>
> >
> > Sorry for the late answer. Answer do come in time if you are patient.
> >
> > Best regards,
> >
> > Arno
> >
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> <mailto: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
> <mailto:eeglablist-request at sccn.ucsd.edu>
>
>
> ------------------------------------------------------------------------
> Branchez-vous aujourd'hui. Jusqu'au 12 mai, lorsque vous vous
> connectez à Windows Live Messenger, vous courez la chance de gagner
> 1000$ à chaque jour. Plus d'info sur connectezetgagnez.ca
> <http://g.msn.ca/ca55/218>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> <mailto: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
> <mailto: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
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> 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
--
--------------------------
Timothy M. O'Keefe
Brown University
Department of Neuroscience
185 Meeting Street
Providence, RI 02912
Voice: +1.401.863.1258
--------------------------
More information about the eeglablist
mailing list