[Eeglablist] FW: Beyond good and evil of ICA

Robert Thatcher rwthatcher2 at yahoo.com
Tue Aug 1 14:42:13 PDT 2017


Makoto,   The math that you show is not in dispute - it is simple linear algebra for separating two mixed signals.  The problem is that it does not address invalid assumptions about ICA decomposition and sources.
1- ICA invalid assumption #1 - EEG sources are independent.
2- ICA invalid assumption #2 - The Central Limit Theorem regarding Gaussian distributions is invalid.
3- ICA invalid assumption #3 - The temporal complexity of any signal mixture is greater than that of its simplest constituent source signal.

Those principles contribute to the basic establishment of ICA.If the signals we happen to extract from a set of mixtures are independent likesources signals, or have non-Gaussian histograms like source signals, or havelow complexity like source signals, then they must be source signals.

 

Now here is the problem in order for the ICA to identifyanything: THE SOURCES HAVE TO BE INDEPENDENT. Moreover when the people who createICA talk about sources they do not seem to address the inversesolution (von Helmholtz, Maxwell, Poisson, etc)  and that is a mistake in our opinion. Let's begin, with the inversesolution and find brain areas that can be:

 

1. Non-independent and what is more correlated.

 

2. EEG sources are highly temporal correlated because information is flowing from one area to another and this means temporal correlations. 

 

3. The number of independent components ICA are correlated with the number of sensors.

 

So the problem is not a mathematical problem itself, your math is correct but it does not solve the problem. The main problem isthat ICA cannot find sources that are correlated and for that reason it puts them under the same component and when one removes that component then one removes all of that information. Of course by removing all the information you alsowill remove the phase information that goes with it and that is the phasedistortion that every one can demonstrate. This problem is worse because if we take into accountthat we can't have more IC's components than sensors when you have 19 channelsthen the ICA method is forced to put things together and there is no way around this problem because one needs to explain the data, and by doing that if you remove a single componentyou will remove a lot of information and that is what you may be missing because you are only focusing on the math and forgot 2 things:

1. The ICA assumptions 
2. The reality of the physiology and physics of the signals that one is looking at.
Kind regards,
Robert

On Tuesday, August 1, 2017, 5:19:10 PM EDT, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:

Dear Robert,
I paste a link below to show mathematical process of how phase CHANGES after rejecting a component obtained by a linear method.You are calling the difference between 18 Bitcoins and 19 Bitcoins 'distortion'. It's a due change.https://sccn.ucsd.edu/wiki/ How_phase_is_calculated_in_ linear_decomposition
If your understanding is different, please show it in math.
Makoto
On Thu, Jul 27, 2017 at 12:30 PM, Robert Thatcher <rwthatcher2 at yahoo.com> wrote:

Dear Makoto,
You stated: "So far I know Montefusco-Siegmunt et al. (2013) is the only paper that makes this invalid claim. If you know other papers, please let me know."

The phase distortion by ICA reconstruction was only discovered in 2014 so there are not a lot of publications on this topic.  However, you are the author of one publication yourself on this Eeglablist.
For example: “If you remove IC andreconstruct channel EEG by back projecting the remaining ICs, of course itchanges channel EEG phase!” (Makoto Miyakoshi, Eeglablist ICA and signal phasecontent, Sept. 16, 2014) 
The proof of the distortion was discovered and validated by comparing the phase differences of the original EEG to the ICA reconstruction time series thereby invalidating the cross-spectrum which is essential for network analyses and also inverse source solutions.  The proof is by observation and mathematics for example by yourself and the following other Eeglablist publishers:

“The EEGreconstruction after removing bad components/sources MAY change the phase valueof the signal at any electrode.” (M. Rezazadeh Eeglablist ICA and signal phasecontent, Sept. 18, 2014).

 

“Thereconstructed data after removing spurious ICA components differs from theoriginal time series, and because of that there are phase differences.” (ArnaudDelorme, EeglablistICA misinformation, June 10, 2017).

  “I first noticed the problem with phase distortion more than adecade ago” (Robert Lawson, Eeglablist ICA misinformation, June 14, 2017).

“Ithink Bob is right that the relative phase will be changed by deleting 1 or 2artifact components.” (Ramesh Srinivasan, Eeglablist ICA misinformation, June 14, 2017).

“Wefound phase distortions in the 8-10 Hz alfa band (greatest near the source ofartefact) but also on more remote electrodes such as occipital and also inartefact free strokes of EEG.” (Georges Otte, Eeglablist ICA misinformation, June 15,2017).

Additional proof is by direct comparisons like Arno did showing about 98% of the phase differences are statistically significantly altered at P < 0.0001. Here is a url to some of the statistics and tutorial demonstrations that allow one to verify for themselves:http://www. appliedneuroscience.com/ Tutorial_Adulteration_Phase_ Relations_when_using_ICA.pdf

Myself and colleaques will be publishing more statistical comparisons and also show how ICA reconstruction distorts other network measures such as the Phase Slope Index and phase shift and phase lock duration and phase-amplitude coupling and cross-frequency coupling, etc.
Other publications are:  Bridwell et al (2016) Spatiospectral Decomposition ofMulti-subject EEG: Evaluating Blind Source Separation Algorithms on Real andRealistic Simulated Data. Brain Topogr DOI 10.1007/s10548-016-0479-1 Feb2016 - see page 13 where they state: “The current group spatiospectral BSS approachdiscards phase information …” (Pg 13).
R.W. Thatcher (2012) "Handbook of QEEG and EEG Biofeedback" , Anipublishing Co., St. Petersburg, Fl
Otte, G. "ICA Reconstruction"  Presented at the ANT workshop, Beaune, France
I hope that this is helpful.
Best wishes,
Robert
On Thursday, July 27, 2017, 1:58:05 AM EDT, otte georges <georges.otte at pandora.be> wrote:


Dear Bob

  

I reposted thismessage below  to the EEGLablist and asked Makoko what caused his opinion switch since 2014 ….

  

Maybe another mail that will get “lost”….   ?  We’ll wait and see….

  

Sincerely

  

  

Georges

  

  

  

From: otte georges [mailto:georges.otte at pandora. be] 
Sent: Thursday, July 27, 2017 7:55 AM
To: 'mmiyakoshi at ucsd.edu' <mmiyakoshi at ucsd.edu>
Subject: RE: Beyond good and evil of ICA

  

Dear Makoto

  

Below is the mail I have send to the EEGLab list and that could maybe also be relevant as a reply to Mr. Andrew Smart. 

  

I can imagine that managing a busy list in extra time is quite a hard task so therefore I can understand that messages get lost in transit. No offense taken.

  

PS the bitcoin image (a string of chars) is mine. It reflects to the fact that if one has 19 strings or components and omits 4  or 5 the reconstructed ones will not be accepted as true bitcoins. In case I am wrong I will send you my sincere apologies and some char strings (just joking)

  

Sincerely

  

PS: (no joke) in a mail You send me in 2014 when we had this discussion again you did state that ICA reconstruction does indeed change phase relations between channels. What causes Your switch of opinion ?

  

Sincerely

  

Georges

  

Mail of july

  

There is a major evolution in modern neuropsychiatry that aims at linking clinical symptoms to brain network dysfunctions. While this approach was successful in grounding neurological symptoms to structural pathologic alternations in brain networks, in psychiatry the main momentum was not structural but functional network dysfunctions. While fMRI was the pioneer, the much better time resolution of MEG and EEG made them the preferred tools. Their output ( time series) is but a means to further construct a functional image of the networks involved where phase dynamics teach us the directionality of the information flow in the network nodes and allows us by comparison with a database of normal values what functional abnormalities can be detected. For me phase integrity in the data is thus very important to be able to construct valuable graph theory models of those networks be them dysfunctional or compensatory. Much work has been devoted on this topic since many decades by DrThatcher but also by many other authors such as Vinod Menon ( Stanford) linking psychiatric symptoms to specific network dysfunctions. For us, clinicians this introduces a new approach to neuroscientific psychiatry that links psychiatry back to it's neurobiological roots and can hopefully one day send the DSM categorization to the museum of the history of psychiatry.

As phase is IMHO a most important parameter in order to establish the network internode information flow, it should not come as a surprise to hear that some find phase unimportant as contaminated by continuous artefact or hear about ICA’ s a signal reconstruction method that presents the danger of changing the phase dynamics in the original time series especially in low channel (19ch) recordings with perhaps more prominent effect due to overcompleteness. 

If in a 19 ch. EEG a clinician rejects (nulls out the rows of the mixing matrix ) ICAas components for blinks EMG, pletysmo and ecg ( 4 ) and then does a "reconstruction"  ( creating 19 channels  out of 15 ??) what we then get might look nice but is IMHO  not a valid base for a graph theoretical model of the underlying brain network.

I think this is the reason this discussion is important and certainly not a trivial pro or contra ICA pugilism. 

  

Sincerely

  

Georges

  

  

From: Makoto Miyakoshi [mailto:mmiyakoshi at ucsd.edu] 
Sent: Wednesday, July 26, 2017 11:40 PM
To: Robert Thatcher <rwthatcher2 at yahoo.com>
Cc: EEGLAB List <eeglablist at sccn.ucsd.edu>; Georges Otte <georges.otte at telenet.be>
Subject: Re: Beyond good and evil of ICA

  

Dear Robert,

  

I want to know all publications that makes a clear claim that 'ICA distorts phase'. I will include all of them for our clarification paper. So far I know Montefusco-Siegmunt et al. (2013) is the only paper that makes this invalid claim. If you know other papers, please let me know.

  

Again, you are calling the difference between 18 Bitcoins and 19 Bitcoins 'distortion'. It's a due change. See the pages below.

https://sccn.ucsd.edu/wiki/ How_phase_is_calculated_in_ linear_decomposition

https://sccn.ucsd.edu/wiki/ ICA_phase_distortion

  

Georges, Ramon told me that all the posts were published on the list. If otherwise, please let us know. Sorry for the trouble.

  

Makoto

  

On Wed, Jul 26, 2017 at 12:06 PM, Robert Thatcher <rwthatcher2 at yahoo.com> wrote:


Dear Makoto,

   I think your criticisms are important and note that there are traveling waves in the EEG and also there is nonlinearity in the form of wave dispersion as noted by Nunez, 1981 and demonstrated in the paper that can be downloaded at this url (see Table IV):

  

http://www. appliedneuroscience.com/TWO- COMPARTMENTAL_MODEL_EEG_ COHERENCE.pdf

  

It seems that your 3rd criticism does not recognize that ICA reconstruction of a new time series violates the "Reciprocity" theorem of Helmoltz and the "Lead Field" necessary for a valid inverse solution.

  

You mentioned a recent criticism on ICA that you stated is "technically invalid".   I doubt that you are referring to the criticism about ICA reconstruction adulterating phase differences between EEG channels?   The issue of ICA reconstruction and phase alteration is a settled issue based on math (not the separation of mixtures of phase or frequencies but rather the cross-spectrum at the same frequency at different locations) as well as multiple empirical demonstrations and tutorial demonstrations that anyone can verify for themselves.  Also, I am copying from the Eelablist statements by yourself and five others agreeing that ICA reconstruction alters phase differences.  

“If you remove IC and reconstruct channel EEG by back projecting the remaining ICs, of course it changes channel EEG phase!” (Makoto Miyakoshi, Eeglablist ICA and signal phase content, Sept. 16, 2014) 

 

“The EEG reconstruction after removing bad components/sources MAY change the phase value of the signal at any electrode.” (M. Rezazadeh Eeglablist ICA and signal phase content, Sept. 18, 2014).

  

“The reconstructed data after removing spurious ICA components differs from the original time series, and because of that there are phase differences.” (Arnaud Delorme, Eeglablist ICA misinformation, June 10, 2017).

 

  “I first noticed the problem with phase distortion more than a decade ago” (Robert Lawson, Eeglablist ICA misinformation, June 14, 2017).

 

“I think Bob is right that the relative phase will be changed by deleting 1 or 2 artifact components.” (Ramesh Srinivasan, Eeglablist ICA misinformation, June 14, 2017).

  

“We found phase distortions in the 8-10 Hz alfa band (greatest near the source of artefact) but also on more remote electrodes such as occipital and also in artefact free strokes of EEG.” (Georges Otte, Eeglablist ICA misinformation, June 15, 2017).

Best regards,

  

Robert

  

On Wednesday, July 26, 2017, 2:01:59 PM EDT, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:

  

  

Dear List,

  

Recently there was a criticism against ICA on the list. Unfortunately it is technically invalid so I remained unsatisfied. Let me share real problems of the ICA model (Onton and Makeig 2006) to re-do it. This is a continued discussion from the one titled 'How phase is calculated in linear decomposition' and now this is my turn to criticize ICA!

  

As far as I know, there are three known limitations in ICA model.
   
   - Spatial stationarity. I have seen a nice traveling waves in ECoG grid data during Joaquin Repela's presentation at SCCN. This clearly violates the assumptions of spatial stationarity in ICA.
   - Temporal stationarity. Shawn Hsu at SCCN presented time-series data of ICA model likelihood during drowsy driving task. Also, Jason Palmer's AMICA also demonstrated temporal changes in model likelihood. So one model per data does not fit the truth (unless the task has a strong control over a subject's cognitive and behavioral states).
   - Dipolar source model. Although most of ICA results are fit with dipole models, it seems ICA also returns (probably) non-point sources. When one fits a dipole model to such a non-point source, the location tend to end up with physiologically invalid depth (this is the most annoying thing about ICA today)

I'd like to hear detailed criticism about these points. Note I saw these critical counterevidence in SCCN; we are not a boring ICA cult who have blind belief in it.

  

Nonetheless, ICA model has a critical merit. I named it Independence-Dipolarity Identity (I-D Identity, or IDId). I-D Identity means that when ICA solves temporally problem, it also solves spatial problem at the same time without using ANY spatial constraint. Dipolarity can be thought of, in short, biophysical origin-ness. Hence I believe that this is evidence that ICA hits some physiological truth of EEG generation.

  

There could be multiple criticisms against the limitations of ICA model, but at the same time any criticism, at least so far, was NOT strong enough to deny I-D Identity of the ICA model. After all, because of this I-D Identity, I still advocate ICA (but similar dipolarity can be achieved by using very different approach, such as SOBI... so independence is not the only requirement to reach the biophysical validity. It's still a mystery to me.)

  

All models are wrong, but some are useful... but I want to go beyond this statement to reach the ground truth of EEG!

  

-- 

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

______________________________ _________________
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170801/16704519/attachment.html>


More information about the eeglablist mailing list