<html>
<head>
</head>
<body bgcolor="#FFFFFF" text="#000000">
Dear Baris/ list members,<br>
Is it actually correct to run an ICA more than once on the same
dataset? So I could run an ICA, remove all the artifacts I can find,
then run another ICA on the same dataset which presumably would find
artifactual components that weren't found by the first one? Wouldn't
that be a very promising approach to thoroughly clean the data?<br>
(sorry for the hijack..)<br>
Cheers, Ulrich<br>
<br>
<br>
On 02.02.2012 06:14, Baris Demiral wrote:
<blockquote
cite="mid:CAPMEqvamBej1Sq4GCzCoDVGk6JJ08gv6Wn3-XLy=XL+gVz1tog@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html;
charset=ISO-8859-9">
<meta name="Generator" content="MS Exchange Server version
08.03.0213.000">
<title>Re: [Eeglablist] ICA question</title>
<!-- Converted from text/plain format -->
<p><font size="2">I think your suggestion that concatenating
everybody and at least find<br>
the weak artifactual IC activity of the clean (least affected)<br>
subjects, and taking this IC out may lead to, again, some
spurious<br>
effect, such that the weights you find will not exactly be the
IC<br>
weights related to the artifacts of the clean subject.<br>
<br>
Your hypothesis is a bit hard to follow. You assume that there
are<br>
artifacts which are not captured by ICA for those clean
subjects. If<br>
you are so sure of that do the following: After the first ICA
run,<br>
take the observed artifactual ICs. Then run ICA again. And if
there<br>
are no more artifacts, then you are wrong.<br>
<br>
In parallel, find the dipoles of the ICs with dipfit plugin.
If a<br>
component is really an artifactual component, the dipole will
be near<br>
the scull or out of the brain (use BESA 4-layer, with no<br>
co-registration, will give you good estimations).<br>
<br>
If you run ICA separately for the two groups twice (rejecting<br>
artifacts after the first one), you will end up with same
number of<br>
ICs for the two groups.<br>
<br>
Baris<br>
<br>
On Wed, Feb 1, 2012 at 8:44 AM, Enrico Schulz
<a class="moz-txt-link-rfc2396E" href="mailto:enrico.schulz@gmail.com"><enrico.schulz@gmail.com></a> wrote:<br>
> Dear EEGlab list,<br>
><br>
> I have a problem with the ICA-based artefact reduction
that is actually not<br>
> just restricted to the EEGlab software.<br>
><br>
> I'm struggling with a lot of high frequency- artefacts at
frontal and<br>
> inferior electrodes around the head exhibiting a much
higher amplitude than<br>
> the cortical gamma band activity I'm interested in.
Although it is possible<br>
> to remove the strongest artefacts, some muscle activity
could not be removed<br>
> in my data sets because some of the artefacts do not give
rise to a separate<br>
> component.<br>
><br>
> In my naive view, in addition to the fact that there are
still artefacts in<br>
> the data set, this could lead to a bias for some
subjects. In theory, if a<br>
> strong artefact gives rise to an independent component
and can, hence, be<br>
> removed, the amount of artefacts in that data set is now
lower than in a<br>
> different data set, where that artefact is not strong
enough for a distinct<br>
> component.<br>
><br>
> The problem is even more complicated if an experimental
group (e.g. pain<br>
> patients) has stronger muscle artefacts than a healthy
control group.<br>
><br>
> Sorry for the long introduction, but my actual question
is, whether it is<br>
> possible to concatenate all single subject files and
doing the ICA for that<br>
> big file.<br>
> I'm aware that this approach has other disadvantages,
e.g. it requires a<br>
> similar topography for each artefact across all subjects
and a fast<br>
> machine.<br>
><br>
> Any help/opinion is highly appreciated!<br>
><br>
> Best regards,<br>
> Enrico<br>
><br>
><br>
><br>
> _______________________________________________<br>
> Eeglablist page: <a moz-do-not-send="true"
href="http://sccn.ucsd.edu/eeglab/eeglabmail.html">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
> To unsubscribe, send an empty email to
<a class="moz-txt-link-abbreviated" href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
> For digest mode, send an email with the subject "set
digest mime" to<br>
> <a class="moz-txt-link-abbreviated" href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br>
<br>
<br>
<br>
--<br>
Ş. Barış Demiral, PhD.<br>
Department of Psychiatry<br>
Washington University<br>
School of Medicine<br>
660 S. Euclid Avenue<br>
Box 8134<br>
Saint Louis, MO 63110<br>
Phone: +1 (314) 7477 1603<br>
<br>
_______________________________________________<br>
Eeglablist page: <a moz-do-not-send="true"
href="http://sccn.ucsd.edu/eeglab/eeglabmail.html">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
To unsubscribe, send an empty email to
<a class="moz-txt-link-abbreviated" href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
For digest mode, send an email with the subject "set digest
mime" to <a class="moz-txt-link-abbreviated" href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br>
</font>
</p>
</blockquote>
<br>
</body>
</html>