[Eeglablist] ICA preprocessing and bipolar montage configuration

[Scott Makeig]

Jason -

Thanks - exact, as usual for you ... I am looking to try it - then put it
into EEGLAB.

Scott

On Wed, Jun 1, 2022 at 1:13 PM Jason Palmer <japalmer29 at gmail.com> wrote:

> ICA will try to produce independent sources (the s below) regardless of
> how they are linearly mixed.
>
>
>
> A bipolar channel matrix looks like:
>
>
>
> B = [ 1 -1 0 0 0 0 0 …. ;
>
>          0 1 -1 0 0 0 0 …. ;
>
>          0 0  1 -1 0 0 0 … ;
>
>          ...
>
>          ];
>
>
>
> Or similar, depending on how you define the bipolar channels.
>
>
>
> And the raw data x, after bipolar channel extraction is B*x = B*A*s, where
> A is mixing matrix topo maps, and s is the source vector. If you run ICA on
> B*x, then you get a decomposition with a mixing matrix M. This is not
> necessarily interpretable with topoplot. To get the actual topo maps, you
> need to invert the bipolar matrix:
>
>
>
>                B*A = M    =>    A = pinv(B)*M
>
>
>
> where again, B is the custom bipolar channel difference matrix, and M is
> the EEG.icawinv after running ICA on B*x (where x is EEG.data). The columns
> of this A matrix should be as usual. Using the bipolar differences might
> remove more far field noise and components vs usual raw channels.
>
>
>
> Jason
>
>
>
> *From:* Scott Makeig [mailto:smakeig at gmail.com]
> *Sent:* Wednesday, June 1, 2022 10:01 AM
> *To:* Jason Palmer <japalmer29 at gmail.com>
> *Cc:* Velu Prabhakar Kumaravel <velu.kumaravel at unitn.it>; EEGLAB List <
> eeglablist at sccn.ucsd.edu>
> *Subject:* Re: [Eeglablist] ICA preprocessing and bipolar montage
> configuration
>
>
>
> Jason -  But the mapping problem remains, no? Will ICA make the IC maps
> (including somehow the bipolar channel) smooth even when the data contains
> a bipolar channel?  What about this argument:  Making such a dataset
> zero-mean does not necessarily mean that the difference between the common
> reference and each electrode in the bipolar channel is 0.  If the sources
> were [equally everywhere], then this might be the case, but in actual fact,
> is it?
>
>
>
> Scott
>
>
>
> On Wed, Jun 1, 2022 at 12:07 PM Jason Palmer <japalmer29 at gmail.com> wrote:
>
> A bipolar montage is basically just a linear transformation, and assuming
> you use at most nbchan number of bipolar channels, it is invertible.
>
>         x = A*s
>         B*x = B*A*s = M*s
>         A = inv(B) * M
>
> Where B is the bipolar montage with 1 -1 in the columns corresponding to
> the
> bipolar difference for each row, and M is the EEG.icawinv after running ica
> on the bipolar transformed data.
>
> Jason
>
> -----Original Message-----
> From: eeglablist [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of
> Scott Makeig
> Sent: Wednesday, June 1, 2022 8:52 AM
> To: Velu Prabhakar Kumaravel <velu.kumaravel at unitn.it>
> Cc: EEGLAB List <eeglablist at sccn.ucsd.edu>
> Subject: Re: [Eeglablist] ICA preprocessing and bipolar montage
> configuration
>
> Velu -
>
> ICA is a linear decomposition, so should be able to decompose bipolar and
> common-reference channels in the same session together. However, the
> bipolar
> channels will be 'floating' with respect to the common-reference channels.
> This might interfere with the decomposition (I have no practical experience
> here) - but even if not it will mean that the ICA scalp maps should not be
> plotted to include the bipolar channels. Here I imagine you might try to
> find fixed offsets representing a 'standing difference'
> between each bipolar channel and the common reference channel that would
> produce max smooth  IC maps -- but are these differences reliably
> stationary? I'd be interested to see a result of attempting this ...
>
> Scott Makeig
>
> On Wed, Jun 1, 2022 at 11:42 AM Velu Prabhakar Kumaravel <
> velu.kumaravel at unitn.it> wrote:
>
> > Dear EEGLABers,
> >
> > Does anyone know the effects of ICA preprocessing on EEG acquired
> > using bipolar configuration?
> > I tried on a few datasets and it looks like the decomposition is not
> > effective. Classifying using ICLabel results in more number of "Other"
> > category.
> >
> > Could someone provide insights on this?
> >
> > Best regards,
> >
> > Velu Prabhakar Kumaravel, Ph.D. Student Center for Mind/Brain
> > Sciences, University of Trento, Italy
> > _______________________________________________
> > 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
> _______________________________________________
> 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
>


-- 
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