[Eeglablist] One or two dipoles

[Scott Makeig]

A reason for leaving the symmetry constraint on when fitting a two
equivalent dipole model is that fitting even two unconstrained dipoles has
local minima, so the model you get may depend on your starting point. This
is not a problem for the symmetry constrained case -- note that dipfit()
does not require directional or amplitude symmetry, just positional. We
know, however, that cortex is rarely perfectly symmetric (e.g.,
callosal-coupled left and right cortical areas are rarely in perfectly
symmetric locations). More research is needed to safely relax the exact
symmetry constraint.

Recall that ICA is decomposing the ongoing EEG, including any sensory ERPs
and etc. Thus, a dual-symmetric dipolar IC reflects some sort of distinct,
synchronous activity occurring in both left and right cortical areas.  This
could be alpha/mu bursts, symmetric sensory ERPs,  or other phenomena. For
example, in minimalistic visual paradigms in which the only stimuli are
small symbols displayed at fixation, a bilateral ERP in or near lateral
occipital cortex can be accounted for by a dual-symmetric IC.

Also, going a bit deeper:  A dual-dipole IC may actually be one basis of a
multi-IC dependent subspace.  The pairwise mutual information function
(pmi) in EEGLAB (in the gui only under the AMICA toolbox) can be used to
identify such subspaces in ICA decompositions.  Under many conditions, ICA
will separate them from the rest of the data and find a basis (ICs) whose
activities are 'most nearly' independent from one another -- without being
able to reduce dependence completely.  However, the dependent subspace, and
thus its basis ICs, will be near-independent from the other components.

>From this viewpoint, an IC that is (nearly) independent from any other IC
in the data can be defined as a 'one-dimensional dependent subspace';  then
we can say that ICA separates the data into a sum of dependent subspaces
(of from 1 to nchans dimensions).

Scott Makeig

On Mon, Feb 11, 2013 at 9:25 AM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>wrote:

> Dear Pal,
>
> I agree with you that (slight) asymmetrical dipole fitting is more
> realistic. It would be nice to have such 'almost symmetrical'
> constraint rather than that of complete symmetry.
>
> > you should take advantage of its superiority - the time
> > resolution. I think you should follow your asymmetrical findings.
>
> Just a note- EEGLAB primarily assumes that dipoles are fit to scalp
> topos obtained from ICA. These IC scalp topos are time-invariant.
>
> > I have been doing dipoling since the early 1990s (mainly in epilepsy).
> > Symmetrical dipoles are rare, but they excists.
>
> I respect your experience in the field and appreciate you share it
> with us. Thank you.
>
> Makoto
>
>
>
> 2013/2/10 Pål Gunnar Larsson <pall at ous-hf.no>:
> > Dear Makoto
> >
> >
> >
> >
> >
> > By default symmetry constraint is on (checked). You can uncheck it to see
> > what happens. The theoretical reason for this is that synchronization of
> > visual and motor cortices is mediated by corpus callosum and known to
> have
> > symmetrical sources (see fMRI studies...) Interestingly, the auditory
> cortex
> > seems to be somewhat different; ICA never gave me a paired auditory
> sources.
> >
> >
> >
> > The propagation time through CC is about 10-15ms. This is way outside the
> > time resolution in fMRI. However, fMRI has a superior space resolution,
> so
> > using EEG, you should take advantage of its superiority - the time
> > resolution. I think you should follow your asymmetrical findings.
> >
> >
> >
> > I have been doing dipoling since the early 1990s (mainly in epilepsy).
> > Symmetrical dipoles are rare, but they excists.
> >
> >
> >
> >
> >
> > Pål
> >
> >
> >
> > 2013/2/7 Pål Gunnar Larsson <pall at ous-hf.no>
> >
> > It is one major source here - left parietal. Then there is a weak right
> > parietal field. If it is a source of interest or just reflect
> uncorrelated
> > activity has to be decided from the signal and the situation.
> >
> >
> >
> > By the way, it seems like the two dipoles are placed symmetrically. Why?
> >
> >
> >
> >
> >
> > Pål
> >
> >
> >
> >
> >
> > Pål G. Larsson M.D., Dr. philos.
> >
> > Head of Clinical Neurophysiology
> > Department of Neurosurgery
> >
> > Division of Surgery and Clinical Neuroscience
> >
> >
> > Oslo University Hospital
> > Po.box 4950 Nydalen
> >
> > 0424 Oslo
> >
> > Norway
> >
> > Tel:  (+47) 23074407
> > Mobile: (+47) 93429791
> > E-mail: pal.gunnar.larsson at ous-hf.no
> >
> > non-sensitive
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > Dear Bernhardsson,
> >
> >
> >
> > Definitely two dipoles. It's not primarily because much smaller r.v., but
> > rather because your scalp topo shows bilateral foci.
> >
> >
> >
> > Makoto
> >
> > 2013/2/7 Bernhardsson Jens <Jens.Bernhardsson at miun.se>
> >
> > Hello,
> >
> >
> >
> >
> >
> > How would you dipole fit this component?
> >
> >
> >
> > I do not find it sufficient to fit this component with only one dipole.
> > However, the symmetrical dipole fitting is not as clear cut as it is with
> > the m-shaped bilateral occipital component. The corresponding right sided
> > topographical map (not attached) do not have this bilateral activation,
> and
> > fits fine with one dipole (rv 1.7%).
> >
> >
> >
> > So, one or two dipoles?
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > Best
> >
> > Jens
> >
> >
> >
> >
> > _______________________________________________
> > 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
> >
> >
> >
> >
> >
> > --
> > Makoto Miyakoshi
> > JSPS Postdoctral Fellow for Research Abroad
> > Swartz Center for Computational Neuroscience
> > Institute for Neural Computation, University of California San Diego
> >
> >
> >
> >
> >
> > --
> > Makoto Miyakoshi
> > JSPS Postdoctral Fellow for Research Abroad
> > Swartz Center for Computational Neuroscience
> > Institute for Neural Computation, University of California San Diego
>
>
>
> --
> Makoto Miyakoshi
> JSPS Postdoctral Fellow for Research Abroad
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
>
> _______________________________________________
> 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 92093-0559, http://sccn.ucsd.edu/~scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130224/3e6f6504/attachment.html>