# [Eeglablist] ICA and signal phase content

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Thu Sep 18 18:41:47 PDT 2014

Dear Iman and Georges,

> So, if some sources are removed because of artifact removing procedure
then the summation of the phase values of the  remainder sources may not be
matched to the initial un-cleaned EEG.

I agree with you Iman. Thank you.

> As a result the component rejection procedure  MAY change the phase
value/content at any given channel.

It's not MAY, but SHOULD, because can you think of a situation where you
keep the phase information identical before and after the IC rejection?

EEG.data, EEG.icaweight, EEG.icasphere, and EEG.icainv are all real.

> I would like to understand the mathematics and can only come up with this
explanation but maybe I see it wrongly.

Sorry I have no math skill to show why phases do not change, but here is my
best explanation. There are two cases phases remain unchanged after
component rejection.
1) rejected component is DC across time.
2) rejected component differs from the original signal only by a fixed
scale across time.
However such component cannot be independent of other components.

Makoto

On Thu, Sep 18, 2014 at 12:54 PM, Iman M.Rezazadeh <irezazadeh at ucdavis.edu>
wrote:

> Hi,
>
> The EEG reconstruction after removing bad components/sources MAY change
> the phase value of the signal at any electrode. Each source and its back
> projection into the channel space at any time point have individual phase
> value; however, the EEG phase value is a “summation” of all the phase
> values from each individual sources ( in other words, the EEG phase value
> at any given channel is the summation of phase values of each
> source/component reconstructed signal at that given channel). So, if some
> sources are removed because of artifact removing procedure then the
> summation of the phase values of the  remainder sources may not be matched
> to the initial un-cleaned EEG.
>
>
>
> Also, each source has a specific IC map ( spatial filter) so the effect of
> that source is distributed/propagated throughout all channels. The
> magnitude of the effect at any channel is proportional to the IC map value
> at that channel ( in EEGLAB: if the IC Map is dark red/blue, its effect is
> high; the greenish color means the effect is low) . So, by removing a
> source which may correspond to eye blink , the effect of that
> source/component  to other channels (other than ‘Fz’, for example)  will be
> also eliminated. As a result the component rejection procedure  MAY change
> the phase value/content at any given channel.
>
>
>
> Best
>
> Iman
>
> ============================================
>
> *Iman M.Rezazadeh, Ph.D. , M.Sc., B.Sc.*
>
> UCLA David Geffen School of Medicine
>
> Semel Institute for Neuroscience and Human Behavior
>
> 760 Westwood Plaza, Ste 47-448
>
> Los Angeles, CA  90095
>
>
>
>
>
>
>
>
>
>
>
>
> *From:* eeglablist-bounces at sccn.ucsd.edu [mailto:
> eeglablist-bounces at sccn.ucsd.edu] *On Behalf Of *otte georges
> *Sent:* Thursday, September 18, 2014 9:26 AM
> *To:* mmiyakoshi at ucsd.edu
> *Cc:* eeglablist at sccn.ucsd.edu
> *Subject:* Re: [Eeglablist] ICA and signal phase content
>
>
>
> Dear Makoto
>
>
>
> Thank You very much for the quick reply. I carefully checked the material
> You indicated in the link to the 2013 workshop (quite extensive but also
>  very interesting material) but could not find much specfic on the phase
> content (imaginary part) of the reconstructed EEG signals. In your mail You
> wrote to me  “of course it will change phase”. Is that “of course” because
> of a regression process in reconstructing the time points from the
> remaining components (after the artefact components have been removed) or
> is there another reason ? I would like to understand the mathematics and
> can only come up with this explanation but maybe I see it wrongly.
>
>
>
> I read the article of Roberto Montefusco who concluded ( on both reaml EEG
> and simulated signals) that the phase changes after reconstructing the
> signal s are non stationary in time , location and frequency. For instanced
> he found major changes (after removing blink components ) at Fz, around 6
> hz. But they were also present at other sensor locations and frequencies.
>
> Phase was determined by using the Hilbert transform.
>
>
>
> All help and explanations would be very much appreciated.
>
>
>
> -Sincerely
>
>
>
> Georges Otte
>
>
>
> *De :* Makoto Miyakoshi [mailto:mmiyakoshi at ucsd.edu <mmiyakoshi at ucsd.edu>]
>
> *Envoyé :* dinsdag 16 september 2014 4:09
> *À :* otte georges
> *Cc :* EEGLAB List
> *Objet :* Re: [Eeglablist] ICA and signal phase content
>
>
>
> Dear Georges,
>
>
>
> If you remove IC and reconstruct channel EEG by backprojecting the
> remaining ICs, of course it changes channel EEG phase!
>
>
>
> Here is our EEGLAB workshop 2013 materials. You can find several
> powerpoint/pdf files that explains how ICA works on data cleaning.
>
> http://sccn.ucsd.edu/wiki/EEGLAB_2013_UCSD
>
>
>
> Makoto
>
>
>
> On Sat, Sep 13, 2014 at 7:50 AM, otte georges <georges.otte at pandora.be>
> wrote:
>
> Dear ICA experts
>
>
>
> There is some discussion going on about the evidence that reconstructing
> an multichannel EEG after ICA decomposition and removing one or two
> artefact components could result in  scrambling the phase content (and
> coherence) of the reconstructed signals. Fi starting from a 19 ch EEG,
> removing two components with fi electrode pops and sweat artefacts ,
> reconstructing the 19 ch EEG from 17 components (taking out the artefact
> containing ones)  without those artefacts.
>
>
>
> Others say that this is not the case and that ICA respects the phase.
> However many authors (Zeman, Thatcher..) have published very convincing
> data to the contrary.
>
>
>
> Any suggestions how to evaluate this ? Is it known from the algorithm what
> the reconstruction  will do to the phase content (imaginary part of the
> signal) ?
>
>
>
> All help is very welcome
>
>
>
> Sincerely
>
>
>
> Georges
>
>
>
>
>
>
> _______________________________________________
> 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
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
>

--
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140918/1027ca4d/attachment.html>