[Eeglablist] Drift(?) IC in Biosemi EEG systems?

[Baris Demiral]

Hi Jan,

How many electrodes are you using in your BP system? I don't think the
look of the IC you showed us would be due to the BioSemi recording (I
have used biosemi for 3 years), but maybe due to the number of
electrodes and correspondingly the number of ICs found to operate. I
suspect if you had 64 electrodes, ICA might have shown you 2 or 3
different ICs instead of this weird IC (maybe the combination of
non-relevant ICs plus most representative IC).

Recently I was using ICA on a 30-channel neuroscan recording, and I
noticed that the way the electrodes are configured on the cap might be
an important issue for the quality of the ICA, particularly on a
low-density electrode system. For instance, if the 20th and 30th
percentile electrodes from the nasion are missing in a 30-electrode
cap (using a 10-20), eye blink or eye movement ICs may not be
perfectly detected, and ICA based artifact corrections might mass up
your data. Maybe due to the non-homogenous distribution of the
electrodes on the frontal sites.

Well, if any one has any experience in those lines, please, let me
know. I am wondering two issues: 1) what is the effect of the number
of electrodes on the quality of ICA? 2) what is the effect of the
electrode montage/orientation on the quality of the ICA (electrode
non-homogeneity)?

Thanks,
Baris


On Wed, Nov 30, 2011 at 2:10 PM, Jan R. Wessel <Jan.Wessel at nf.mpg.de> wrote:
> Hey fellow EEGLAB'ers,
>
> After exclusively using BrainProducts EEG Hardware for the last couple
> of years, I have recently started looking at a colleagues dataset that
> was acquired on a BioSemi System.
> He keeps getting components like the following one (see download link),
> that have very "non-dipolar" topographic maps, but explain a pretty good
> part of the data:
> http://img15.imageshack.us/img15/2390/component11.png
> Can anybody tell me what process / artifact (or what have you) these
> components represent? It keeps occurring pretty independently of how I
> preprocess his data.
>
> Looking at the component ERPimage (non-baseline corrected epoched data),
> I suspected it might be a drift component, since I think BioSemi doesn't
> use DC cutoff during recording.
>
> However, two things seem to countermand this interpretation: a) HP
> filters as drastic as 2Hz did not prevent these components from being
> present in the ICA model (the screenshot is from 2Hz HP-filtered
> (two-way FIR) data), and b) there does seem to be some event-related
> activity in that component as well (there's a visual event at 1.5 secs
> in the dataset underlying this IC).
>
> Now the latter point could in principle be due to an under-fitting ICA
> model, since the data dimensionality was very low to begin with (only 34
> electrodes were recorded), and overall data quality w.r.t.
> spurious/non-stereotyped artifacts was sub-standard (even though I've
> been pretty conservative on the pre-ICA data rejection). But that still
> leaves me curious as to what this component actually is.
>
> I'm pretty sure that this might be a pretty standard thing that has to
> do with the BioSemi that I am missing here, but I could be mistaken. In
> an case, I would appreciate any help in this matter, because I'm just
> curious w.r.t. what I am looking at.
>
> Thanks very much!!
> Jan
>
> --
> Dr. Jan R. Wessel
> Department of Psychology
> University of California, San Diego
> &
> Cognitive Neurology Research Group
> Max Planck Institute for Neurological Research, Cologne
> http://www.nf.mpg.de/cv/jan-wessel.html
>
>
> _______________________________________________
> 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



-- 
Ş. Barış Demiral, PhD.
Department of Psychiatry
Washington University
School of Medicine
660 S. Euclid Avenue
Box 8134
Saint Louis, MO 63110
Phone: +1 (314) 7477 1603