<html><head></head><body><div style="font-family:Helvetica Neue, Helvetica, Arial, sans-serif;font-size:small;"><div><div>Ramesh,</div></div><div> This list group seems unique in believing that all EEG is so full of artifact that it justifies using methods like ICA reconstruction that create more artifact because whats wrong with adding more artifact to something that is already essentially 100% artifact? I have not come across this belief system in my career until engaging with the EEGlab list. Obviously, I do not agree with this opinion. In the short term it is ok to have such a belief because science relies upon hypothesis testing and cross-validation and verification and independent testing and therefore this belief system will not have a long-term impact on the science of EEG and hopefully will not waste tax payer money or set back the young minds of students trying to learn about EEG. </div><div><br></div><div>I encourage members of this forum to visit the National Library of Medicine database (Pubmed) and ask questions and challenge the belief that EEG is "chaos" or "random" or "full of artifact". Here is a url to the Pubmed website:</div><div><a href="https://www.ncbi.nlm.nih.gov/pubmed/" rel="nofollow" target="_blank">https://www.ncbi.nlm.nih.gov/pubmed/</a><br></div><div><br></div><div>For those interested, here is a url to a paper that I wrote titled: VALIDITY AND RELIABILITY OF
QUANTITATIVE
ELECTROENCEPHALOGRAPHY (qEEG) </div><div><br></div><div><br></div><div><a href="http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf" rel="nofollow" target="_blank">http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf</a><br></div><div><br></div><div><br></div><div>EEGlabs is a fantastic resource with very important contributions and I appreciate all of the dialog that we have had! I agree with Ramesh and Gedeon's suggestions to use this discussion to transform to positive outcomes and seek the best uses of QEEG and also ICA. If we can change the subject from ICA reconstruction to allegedly remove artifact to ICA feature detection (try to avoid the reconstruction) then great progress can be made. ICA feature detection (not reconstruction) helps identify "seeds" that can isolate critical parts of a dynamical system. Independent components help identify the core parts of a specialized hub in a network. Once the critical features are idenified then one can improve the signal-to-noise ratio and try to spatially localize the local features and then study their functional and effective connectivity between the local set of feature detectors to distant hubs. </div><div><br></div><div>I am copying a url to an exemplar study that ICA (without reconstruction) has promise: "The Code for Facial Identity in the Primate Brain" where: "Facial identity is encoded via a remarkably simple neural code that relies on the ability of neurons to distinguish facial features along specific axes in face
space, disavowing the long-standing
assumption that single face cells encode
individual faces. "</div><div><br></div><div>http://www.appliedneuroscience.com/Face recognition by 206 cells each representing a multi-dimensionl factors (Vector)-Le Chang.pdf<br></div><div><br></div><div>Best wishes,</div><div><br></div><div>Robert</div><div><br></div><div><br></div><div id="yahoo_quoted_8463533042" class="yahoo_quoted"><div style="font-family:'Helvetica Neue', Helvetica, Arial, sans-serif;font-size:13px;color:#26282a;"><div>On Saturday, June 24, 2017, 3:32:19 PM EDT, Robert Thatcher <rwthatcher2@yahoo.com> wrote:</div><div><br></div><div><br></div><div><div id="yiv3996281630"><html><head></head><div><div style="font-family:Helvetica Neue, Helvetica, Arial, sans-serif;font-size:small;"><div style="font-family:Helvetica Neue, Helvetica, Arial, sans-serif;font-size:small;"><div><div>Ramesh,</div></div><div> I have personnally examined over 20,000 EEGs and I do not agree that all EEG recordings and all digital time points include artifacts such as eye movement or EMG or heart beats or environmental noise, etc. I agree that measures of robustness are important and that is one of the reasons the standards for clinical applications of EEG and admission of the EEG in court cases are test re-test reliability > 0.90. Most of the time the test re-test reliability is > 0.98 (test re-test defined by using one a sample of EEG in a record to predict a different sample of EEG) later in the recording session. Similar test re-test values are obtained when repeating power spectra. If you use the search term "EEG and test retest reliability" in a search of the National Library of Medicine (Pubmed) then you will find 3,028 citations. A quick read of the abstracts commonly show test retest values > 0.90. If EEG is always full of artifact then such high reliability and robustness would not be published and low statistical effect sizes would be common which they are not and EEG would not be admitted in court and certainly there would not be over 100,000 peer reviewed QEEG studies. Also, I did not see any ICA reconstruction studies in this search, instead visual deletion of eye movement and other artifact and some used artifact detection routines are used to delete artifact without any type of reconstruction of an entire record and thereby altrering the original EEG recording. This is clear evidence of Robustness of the EEG where robustness is "the ability to withstand or overcome adverse conditions or rigorous testing". </div><div><br clear="none"></div><div>Robert</div><div><br clear="none"></div><div>I am copying one of the test re-test reliability studies that is commonly presented in court cases.</div><div>
<p class="yiv3996281630ydp979523eeMsoNormal"><b><span style="font-size:14.0pt;">Test-retest reliability in
EEG frequency analysis</span></b>.</p>
<p class="yiv3996281630ydp979523eeMsoNormal">Electroencephalography and Clin Neurophysiol 1991
Nov;79(5):382-92</p>
<p class="yiv3996281630ydp979523eeMsoNormal"> Salinsky MC, Oken BS, Morehead L</p>
<p class="yiv3996281630ydp979523eeMsoNormal">This study was performed to gain a better understanding of
EEG frequency analysis test-retest reliability in normal healthy adults, and to
evaluate factors which could influence the measured inter-record differences.
Nineteen subjects underwent serial EEG recordings at 5 min and 12-16 week
intervals. Records were visually edited using a standardized protocol, and FFT
frequency analysis performed on segments of 60, 40, or 20 sec total length.
Correlation coefficients for broad band features averaged 0.92 over the 5 min retest
interval and 0.84 over the 12-16 week interval. There was essentially no
difference between correlation coefficients of absolute and relative power features.
Coefficients based on 60 sec records were marginally higher than those of 40 or
20 sec records. On the other hand, test-retest percent differences were typically lower
for relative as opposed to absolute power features, and 60 sec records showed
consistently lower percent differences than did 40 or particularly 20 sec records. Peak
alpha frequency and mean frequency were the most stable EEG features at either
interval. </p>
</div><div><br clear="none"></div><div><br clear="none"></div></div><div class="yiv3996281630yqt2670707173" id="yiv3996281630yqt74290"><div class="yiv3996281630yahoo_quoted" id="yiv3996281630yahoo_quoted_8435431032"><div style="font-family:'Helvetica Neue', Helvetica, Arial, sans-serif;font-size:13px;color:#26282a;"><div>On Saturday, June 24, 2017, 2:58:02 PM EDT, Ramesh Srinivasan <r.srinivasan@uci.edu> wrote:</div><div><br clear="none"></div><div><br clear="none"></div><div><div id="yiv3996281630"></div></div></div></div></div></div></div><html><head></head><div class="yiv3996281630yqt2670707173" id="yiv3996281630yqt47747"><div><div>
<p>Hi Bob - <br clear="none">
</p>
<span style="font-size:12.0pt;line-height:107%;">It is my opinion that all EEG recordings
are a mixture of artifact and brain activity. The observation of
consistent and reliable EEG results in many experiments reflects
the robustness of EEG phenomena, in spite of the presence of
artifacts, which vary from laborastory to laboratory and person to
person. <br clear="none">
<br clear="none">
I think your are very mistaken to think the objective is to
measure something relative phase very precisely when in fact the
objective is to measure something robust in the face of a noisy
recording situation. <br clear="none">
<br clear="none">
I'd recommend this discussion, if it continues, focus on
robustness. In my own work, I like to see that the effect of
interest is present with and without ICA clean-up. <br clear="none">
<br clear="none">
ramesh <br clear="none">
</span>
<blockquote type="cite">
<div style="font-family:Helvetica Neue, Helvetica, Arial, sans-serif;font-size:small;">
<div>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12.0pt;line-height:107%;">There is another issue that arose
that needs to be
addressed, i.e., the belief by some that the original EEG
recording is not
valid to begin with and therefore this justifies
distorting the phase
differences. </span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">You stated: “Your
phrase "distortion of the original time series" does make
sense<span class="yiv3996281630ydp71640538apple-converted-space"> </span></span>only
if you believe that the
original time series represents somehow a<span class="yiv3996281630ydp71640538apple-converted-space"> </span>gold
standard, something special. In
my view this is not<span class="yiv3996281630ydp71640538apple-converted-space"> </span>justified,
for instance because it disregards the fact that recording<span class="yiv3996281630ydp71640538apple-converted-space"> </span>settings
will influence how the
data will be recorded, that is, they<span class="yiv3996281630ydp71640538apple-converted-space"> </span>determine
phase and amplitude!”</p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">This is simply not
true
and the reason we know it is not true is that we use
microvolt calibration sine
waves at different frequencies and different amplitudes
and different phase
differences injected into the inputs of over 40 different
amplifiers. 20
microvolts in gives 20 microvolts out and a 30 degree
phase shift in gives 30
degree phase shift out, etc. That is the
amplifiers that we test as part of the FDA quality control
process are
excellent amplifiers that faithfully map one to one the
input to the
output. Also, the vast majority of EEG
amplifiers cannot be sold unless they past rigorous FDA
quality control tests,
for example, Cadwell, Nihon Khoden, Biosemi, Neuroscan,
Nexus, Deymed, ANT,
Nicolet. None of these amplifiers
distort phase differences or amplitudes.
If they did then FDA would fine the manufacturers.</span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);"> </span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">You stated: “The
second
flaw in your reasoning is that you believe there are<span class="yiv3996281630ydp71640538apple-converted-space"> </span></span>"artifact-free"
intervals.<span class="yiv3996281630ydp71640538apple-converted-space">”</span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span class="yiv3996281630ydp71640538apple-converted-space"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);"> </span></span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span class="yiv3996281630ydp71640538apple-converted-space"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">The
standard definition of artifact free are electrical
potentials recorded from the scalp that were generated
by the brain and only by
the brain. There have been millions of
EEG recordings over the last 50 years and it is a false
belief that 100% of EEG
recordings are artifact and that the over 100,000 peer
reviewed studies cited
in the National Library of Medicine only involve
artifact.</span></span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span class="yiv3996281630ydp71640538apple-converted-space"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">The
skull does not generate electrical potentials and there
is attenuation by the low conductivity but there is no
fundamental change in
the phase differences that are produced by synchronous
synaptic potentials and
various differences in white matter conduction
velocities. As the project manager at NIH for the
development
of the first 128 channel system there were occasions
that we recorded EEG
before and after a craniotomy. We
compared the ECoG electrical potentials recorded from
the dura to those
recorded from the scalp and there was a 20 to 50 times
higher amplitude but the
phase differences were essentially the same because they
are produced by
summated synaptic potentials and global cortico-cortical
connectivity in the
white matter. </span></span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span class="yiv3996281630ydp71640538apple-converted-space"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);"> </span></span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">You stated: “In
your case, the average<span class="yiv3996281630ydp71640538apple-converted-space"> </span></span>reference
is far from zero because 19 10-20 channels, which are
located<span class="yiv3996281630ydp71640538apple-converted-space"> </span><span style="font-size:12.0pt;line-height:107%;color:#26282A;"><br clear="none">
<span style="">on the top half of the head sphere, can hardly
sum up to zero.”</span></span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);"> </span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">This is correct and as
Desmedt
and others point out even 128 channel recordings do not
produce a zero
electrical field. This would require
placing electrodes on the bottom of the brain or skull or
like Don Tucker jokes
a valid average reference requires “Recording from the
other side of the moon”. This is one of the reasons
that Desmedt and
others argue against the use of an average reference and
are in favor of a
single common reference.</span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);"> </span></p>
<p class="yiv3996281630ydp71640538MsoNormal"><span style="font-size:12pt;line-height:107%;color:rgb(38, 40, 42);">Robert<br clear="none">
</span></p>
</div>
<div><br clear="none">
</div>
<div><br clear="none">
</div>
<div class="yiv3996281630yahoo_quoted" id="yiv3996281630yahoo_quoted_8665233753">
<div style="font-family:'Helvetica Neue', Helvetica, Arial, sans-serif;font-size:13px;color:#26282a;">
<div>On Saturday, June 24, 2017, 6:52:55 AM EDT, Stefan
Debener <a rel="nofollow" shape="rect" class="yiv3996281630moz-txt-link-rfc2396E" ymailto="mailto:stefan.debener@uni-oldenburg.de" target="_blank" href="mailto:stefan.debener@uni-oldenburg.de"><stefan.debener@uni-oldenburg.de></a> wrote:</div>
<div><br clear="none">
</div>
<div><br clear="none">
</div>
<div>
<div dir="ltr">Dear Robert,<br clear="none">
<br clear="none">
Ok, I guess I have to give up on you. Of course you
ascribe the <br clear="none">
"original" times series some magic, and this is one of
several flaws in <br clear="none">
your reasoning. To cite your own published paper:<br clear="none">
<br clear="none">
" PROBLEMS WITH RE-MONTAGING AND DISTORTIONS OF THE
ORIGINAL TIME SERIES <br clear="none">
If the original EEG/event-related potential (ERP) time
series is <br clear="none">
transformed into a second time series by using the
average reference <br clear="none">
then the original phase differences from three electrode
locations may <br clear="none">
be scrambled and lost. For example, with an average
reference the entire <br clear="none">
surface of the brain is not measured, thus the averaging
does not create <br clear="none">
a true zero potential at each instant of time."
(citation taken from <br clear="none">
<a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf">http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf</a>).<br clear="none">
<br clear="none">
Your phrase "distortion of the original time series"
does make sense <br clear="none">
only if you believe that the original time series
represents somehow a <br clear="none">
gold standard, something special (something "magic",
forgive my poor use <br clear="none">
of English) that is magically close to the contributions
from brain <br clear="none">
generators to the surface-recorded signal. In my view
this is not <br clear="none">
justified, for instance because it disregards the fact
that recording <br clear="none">
settings will influence how the data will be recorded,
that is, they <br clear="none">
determine phase and amplitude! I argued before that
attributing <br clear="none">
something special to the "original" time series is
highly misleading, <br clear="none">
the "original" time series is not closer to the brain
signal, in <br clear="none">
contrast it may be pretty far away from it, not only
because phase of <br clear="none">
mixed brain generators does not make much sense, but
also becasue of all <br clear="none">
the artifactual influences not accounted for. Now if
you really believe <br clear="none">
that one particular reference scheme is appropriate in
bringing the <br clear="none">
original data close to the brain while all others are
rubbish, then I <br clear="none">
wish you good luck in trying to convince the community.
I predict that <br clear="none">
nobody will take you serious. The reference discussion
has been going on <br clear="none">
for ages, and there are good reasons to change the
reference (online or <br clear="none">
offline) depending on what the purpose of the
study/analysis is.<br clear="none">
<br clear="none">
I argue that any post-recording signal processing that
changes the <br clear="none">
morphology of a time series will change the phase values
as well, NOT <br clear="none">
just the average reference, and NOT just ICA. In your
case, the average <br clear="none">
reference is far from zero because 19 10-20 channels,
which are located <br clear="none">
on the top half of the head sphere, can hardly sum up to
zero. Only a <br clear="none">
full equidistant spatial coverage of the head sphere
would make the <br clear="none">
spatial average approximate zero potential. Check this
paper for <br clear="none">
in-depth discussion on the average reference and the
bias it introduces: <br clear="none">
<a rel="nofollow" shape="rect" target="_blank" href="https://www.ncbi.nlm.nih.gov/pubmed/10402104">https://www.ncbi.nlm.nih.gov/pubmed/10402104</a>).
If you don't like the <br clear="none">
average reference, fair enough, then simply take any
other reference, <br clear="none">
but you will observe the same "distortion" on phase
values. By the way, <br clear="none">
non-invasive EEG does not measure signals from the
brain, there is a <br clear="none">
skull in between and a couple other layers, all with
different <br clear="none">
conductivies...just another reason to be aware of the
inverse problem <br clear="none">
and trat signals originally recorded from the scalp with
great care.<br clear="none">
<br clear="none">
The second flaw in your reasoning is that you believe
there are <br clear="none">
"artifact-free" intervals. The report I included showed
(far from <br clear="none">
perfect, but clearly evident!) heart-electrical
activity. Recording <br clear="none">
parameters strongly determine whether such "EKG"
contributions show up <br clear="none">
in ICA decompositions or not (such as sub-10/20 spatial
coverage). Now, <br clear="none">
only because you don't see heart-electrical activity in
your "original" <br clear="none">
19-ch 10/20 recordings, can you seriously claim that the
heart of your <br clear="none">
participant was not beating during recording? Of course
not! All it says <br clear="none">
is that the influcene may be stronger or weaker
represented in your <br clear="none">
recordings (depending on individual differences, and,
again, recording <br clear="none">
parameters). With your philosophy, you gonna miss it if
it does not jump <br clear="none">
into your eyes.<br clear="none">
<br clear="none">
Anyway, I hope others join the discussion, I am giving
up.<br clear="none">
<br clear="none">
Best,<br clear="none">
Stefan<br clear="none">
<br clear="none">
<br clear="none">
Am 23.06.17 um 19:16 schrieb Robert Thatcher:<br clear="none">
><br clear="none">
> Dear Stefan,<br clear="none">
><br clear="none">
> Thank you for your dilligence and dedication to
this important <br clear="none">
> issue. I am pleased that you are in agreement with
myself and <br clear="none">
> Georges and Gert and scienific publications that
"a spatial filter <br clear="none">
> operation such as ICA or other, the phase
differences may indeed be <br clear="none">
> different." I would add that we have not yet found
an instance where <br clear="none">
> ICA reconstruction did not alter phase differences
between channel and <br clear="none">
> there I would change the word "may" and phrase to
"are indeed different".<br clear="none">
><br clear="none">
> You sated: "As a toy example I include the common
average reference. "<br clear="none">
><br clear="none">
> In Neuroguide we do not allow one to compute phase
diffferences or <br clear="none">
> coherence when using a common reference. We found
with simulation and <br clear="none">
> real data that the common reference mixes the
phases differences <br clear="none">
> between all channels and itself distorts phase
differences and often <br clear="none">
> in strange ways, for example, if one or two
channels has a suddent <br clear="none">
> large amplitude alpha or theta rhythm then the
phase differences <br clear="none">
> between channels that do not have a alpha or beta
or theta rhythm are <br clear="none">
> altered. If one uses a single common reference
then if an alpha <br clear="none">
> rhythm appeas for example in O1/2 then there is not
change in phase <br clear="none">
> differences in channels where there is no alpha.
We also compared two <br clear="none">
> sine waves at different phase differences, e.g., 30
deg, 60 deg, 90 <br clear="none">
> deg, etc and mixed different amounts of white noise
in one of the <br clear="none">
> channels we found linear reductions in coherence
(i.e., the phase <br clear="none">
> stability over time) as a function of the SNR and
the mean phase <br clear="none">
> differences were stable until the noise was too
high and when <br clear="none">
> coherencer was near zero. In contrast, uses the
averge refence and <br clear="none">
> repeats the same signal and noise test with
different phase <br clear="none">
> differences then the mean phase difference is
quickly lost and there <br clear="none">
> is no valid measure of coherence. Here is a url to
a publication <br clear="none">
> that discusses this topic: <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf">http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf</a><br clear="none">
><br clear="none">
> You stated: "Your claim that ICA has somehow
corrupted the data such <br clear="none">
> that previously super reliable clinical effects all
over a sudden <br clear="none">
> vanished is not convincing either."<br clear="none">
><br clear="none">
> I never said this. There is no ICA reconstruction
in NeuroGuide and <br clear="none">
> the over 3,000 users of Neuroguide have not
complained about coherence <br clear="none">
> or phase and they always obtain repeatable measures
when the retest <br clear="none">
> patients. It was only the WinEEG users that are
worried about ICA <br clear="none">
> and they never said that they got "super reliable
clinical effects". <br clear="none">
> They just noticed that coherence using the WinEEG
after ICA <br clear="none">
> reconstruction was totally different than when they
use NxLink or SKIL <br clear="none">
> or Neurorep or Neuroguide or Brindx or other
software, etc.<br clear="none">
><br clear="none">
> You stated: "Now which phase values are valid,
those obtained by one <br clear="none">
> particular reference scheme or those by another? In
my view they are <br clear="none">
> both arbitraty"<br clear="none">
><br clear="none">
> See my earlier reply regarding average references
and the Laplacian <br clear="none">
> reference in regard to the accurate and
reproducible measures of phase <br clear="none">
> differences as opposed to using a single common
reference. I am <br clear="none">
> recopying the link here: <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf%20">http://www.appliedneuroscience.com/Coh_phasediff&phase_resetinEEG-ERP.pdf
</a><br clear="none">
> here is another review on this topic: <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Brain%20Connectivity-A%20Tutorial.pdf">http://www.appliedneuroscience.com/Brain%20Connectivity-A%20Tutorial.pdf</a><br clear="none">
><br clear="none">
> You stated: “there is no such magically clean raw
brainsignal <br clear="none">
> available in the first place!”<br clear="none">
><br clear="none">
> No one says that “magic” is involved in the EEG.
However, the physics <br clear="none">
> of EEG is involved and during a 5 minute to 20
minute EEG recording <br clear="none">
> there is plenty of artifact free data. There are
over 100,000 QEEG <br clear="none">
> publications in the National Library of Medicine
with high effect <br clear="none">
> sizes and high test retest reliability and highly
reproducible <br clear="none">
> findings. My concern and the concern of many
others is that ICA <br clear="none">
> reconstruction alters phase differences in an
entire EEG recording <br clear="none">
> even if there are only a few instances of
artifact. The reason that <br clear="none">
> QEEG has been so successful and so widely used
since the late 1950s is <br clear="none">
> because people have been successful in avoiding or
deleting artifact <br clear="none">
> and selecting multiple artifact free parts of the
record and achieving <br clear="none">
> 0.95 or higher test retest reliability. Over
reaction to the <br clear="none">
> presence of small amounts of artifact is not a
justification for <br clear="none">
> altering the electricity of the brain including
network dynamics such <br clear="none">
> as average synaptic rise times and conduction
velocities and couplings <br clear="none">
> between groups of neurons.<br clear="none">
><br clear="none">
> You stated: “Artifacts not accounted for adulterate
EEG phase values”<br clear="none">
><br clear="none">
> We agree on this but this is not what the
discussion is about. The <br clear="none">
> concerns of many is that the phase differences in
the artifact free <br clear="none">
> sections are altered. Rarely if ever do
clinicians/scientists bother <br clear="none">
> computing the phase differences during an eye
movement artifact. <br clear="none">
> Usually phase differences are zero and this physics
fact plus the <br clear="none">
> electrical gradients from the eyes is how many
people detect eye <br clear="none">
> movement artifact and then omit the artifact from
analyses without <br clear="none">
> using ICA. The main focus in QEEG is the parts of
the recording <br clear="none">
> where there is no artifact and the electrical
potentials are generated <br clear="none">
> by the brain inside the skull.<br clear="none">
><br clear="none">
> Also, thank you for your use of the Hilbert
transform, this is only of <br clear="none">
> the tools that we use and it allows one to evaluate
phase differences <br clear="none">
> in every individual time sample in the artifact
free sections and <br clear="none">
> prove that each and every one of the phase
differences for all <br clear="none">
> channel combinations is altered by ICA
reconstruction. You may be <br clear="none">
> interested that we use the Hilbert transform to
measure phase shift <br clear="none">
> and phase lock duration that have a high
correlation with Autism and <br clear="none">
> intelligence in short distance connections and with
use the Hilbert <br clear="none">
> transform to measure the magnitude of information
flow (phase slope <br clear="none">
> index) and intelligence. Here are some hyperlinks
to these studies: <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Intelligence-phase_reset_Nature.pdf">http://www.appliedneuroscience.com/Intelligence-phase_reset_Nature.pdf</a><br clear="none">
><br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Autism%20Thatcher%20et%20al.pdf">http://www.appliedneuroscience.com/Autism%20Thatcher%20et%20al.pdf</a><br clear="none">
><br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Default_Network_LORETA_Phase_Reset-Thatcher_et_al.pdf">http://www.appliedneuroscience.com/Default_Network_LORETA_Phase_Reset-Thatcher_et_al.pdf</a><br clear="none">
><br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Intelligence%20&%20information%20flow-Thatcher%20et%20al%202016.pdf">http://www.appliedneuroscience.com/Intelligence%20&%20information%20flow-Thatcher%20et%20al%202016.pdf</a><br clear="none">
><br clear="none">
> We are also using the Hilbert transforms for
cross-frequency network <br clear="none">
> dynamics including phase-amplitude coupling. I do
not believe that <br clear="none">
> we would have discovered these important network
correlations if we <br clear="none">
> had used ICA reconstruction.<br clear="none">
><br clear="none">
> Robert<br clear="none">
><br clear="none">
><br clear="none">
> On Friday, June 23, 2017, 9:53:01 AM EDT, Stefan
Debener <br clear="none">
> <<a rel="nofollow" shape="rect" ymailto="mailto:stefan.debener@uni-oldenburg.de" target="_blank" href="mailto:stefan.debener@uni-oldenburg.de">stefan.debener@uni-oldenburg.de</a>>
wrote:<br clear="none">
><br clear="none">
><br clear="none">
> Dear Robert,<br clear="none">
><br clear="none">
> I have expanded my illustration and now consider
the phase differences<br clear="none">
> between two channels, slides 13 to 16 of the
updated pdf:<br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="https://www.dropbox.com/s/e70qhf91dgc5anu/Thatcher_summary_2.pdf?dl=0">https://www.dropbox.com/s/e70qhf91dgc5anu/Thatcher_summary_2.pdf?dl=0</a><br clear="none">
><br clear="none">
> Note that phase values were derived by the Hilbert
transform of the<br clear="none">
> bandpass filtered signal, as explained by W Freeman
here:<br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.scholarpedia.org/article/Hilbert_transform_for_brain_waves">http://www.scholarpedia.org/article/Hilbert_transform_for_brain_waves</a><br clear="none">
><br clear="none">
> More details on the particular implementation I
used are here:<br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="https://de.mathworks.com/help/signal/ref/hilbert.html">https://de.mathworks.com/help/signal/ref/hilbert.html</a><br clear="none">
><br clear="none">
> If you measure phase differences between two
channels, consider the<br clear="none">
> result as your gold standard, and then apply a
spatial filter operation<br clear="none">
> such as ICA or other, the phase differences may
indeed be different. I<br clear="none">
> assume any spatial filter (that effectively
spatially filters the data)<br clear="none">
> changes phase values and phase difference values.
As a toy example I<br clear="none">
> include the common average reference. If you apply
a common average<br clear="none">
> reference to the raw data, then bandpass filter as
before, and compare<br clear="none">
> the phase difference values to your "gold
standard", then the phase<br clear="none">
> differences will change as well. Now which phase
values are valid, those<br clear="none">
> obtained by one particular reference scheme or
those by another? In my<br clear="none">
> view they are both arbitraty, since recording
settings as well as<br clear="none">
> preprocessing steps may have a strong impact on the
actually measured<br clear="none">
> phase. There is no reason to assume that a change
in phase, or in phase<br clear="none">
> differences, "adulterates" a magically clean phase
signal obtained from<br clear="none">
> the raw data - simply because there is no such
magically clean raw brain<br clear="none">
> signal available in the first place!<br clear="none">
><br clear="none">
> Your claim that ICA has somehow corrupted the data
such that previously<br clear="none">
> super reliable clinical effects all over a sudden
vanished is not<br clear="none">
> convincing either. Artifacts such as eye blinks and
lateral eye<br clear="none">
> movements are very common, I hope you can agree at
least here. Now, keep<br clear="none">
> in mind that they contribute fixed spatial
patterns - as long as the<br clear="none">
> electrodes cap does not shift during acquisition
the projections of the<br clear="none">
> sources of those artifacts do not change. My
illustrations above show<br clear="none">
> very clearly how artifacs indeed adulterate phase
values, just as Arnos<br clear="none">
> illustrations do! Now, if you disregard artifactual
influences you may<br clear="none">
> end up with highly reliable connectivity effects -
but they tell you<br clear="none">
> very little about brain function! Even more
troubling, if you compare<br clear="none">
> two individuals EEGs (say, one "healthy", one
"abnormal"), then a<br clear="none">
> different amount of artifacts in the data, if not
carefully taken care<br clear="none">
> of during preprocessing, will produce spurious
results that are falsely<br clear="none">
> attributed to differences in brain function.
Actually, given that many<br clear="none">
> artifacts often contribute much more variance to
that raw signals than<br clear="none">
> (reasonably well validated) brain signals, such as
fronto-midline theta,<br clear="none">
> this is actually very likely! So, what we learn is
that:<br clear="none">
><br clear="none">
> Artifacts not accounted for adulterate EEG phase
values<br clear="none">
><br clear="none">
> Best,<br clear="none">
><br clear="none">
> Stefan<br clear="none">
><br clear="none">
><br clear="none">
><br clear="none">
> Am 22.06.17 um 20:30 schrieb Robert Thatcher:<br clear="none">
> > Dear Stefan,<br clear="none">
> > The attachment did not contain any measures
of phase differences<br clear="none">
> > between channels. It is very difficult to
visually see differences<br clear="none">
> > in phase differences. One must use the
cross-spectrum to calculate<br clear="none">
> > phase differences and compare phase
differences in degrees. Phase<br clear="none">
> > difference varies from -180 to 180 degrees and
one must look at the<br clear="none">
> > numbers. Below is a url to the two power
points that also show<br clear="none">
> > visually similar EEG tracings but also
computed the instantaneous<br clear="none">
> > phase differences using the Hilbert transform
(complex demodulation).<br clear="none">
> > Four identical time points were selected and
they demonstrated<br clear="none">
> > totally different phase differences with
respect to the O1 channel and<br clear="none">
> > the other 18 channels. No matter what
reference channel one selects<br clear="none">
> > and no matter what identical time points one
selects there are always<br clear="none">
> > large differences in the phase difference
between channels in all<br clear="none">
> > frequency bands. I also computed the average
phase difference in the<br clear="none">
> > artifact free parts of the record and the
averages were statistically<br clear="none">
> > significantly different at P < 0.0001 and
the same for the FFT.<br clear="none">
> ><br clear="none">
> > Proof of phase difference adulteration is in
the power points. I am<br clear="none">
> > again copying the hyperlink here:<br clear="none">
> ><br clear="none">
> ><br clear="none">
> > <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Phase_Diff-Original_&_Delorme-Post-ICA-4_time_points.zip">http://www.appliedneuroscience.com/Phase_Diff-Original_&_Delorme-Post-ICA-4_time_points.zip</a><br clear="none">
> ><br clear="none">
> ><br clear="none">
> > This cannot be explained by a low quality ICA
reconstruction because<br clear="none">
> > the ICA reconstruction was conducted by Arnu
using EEGLab software.<br clear="none">
> ><br clear="none">
> > Robert<br clear="none">
> > On Thursday, June 22, 2017, 2:00:19 PM EDT,
Stefan Debener<br clear="none">
> > <<a rel="nofollow" shape="rect" ymailto="mailto:stefan.debener@uni-oldenburg.de" target="_blank" href="mailto:stefan.debener@uni-oldenburg.de">stefan.debener@uni-oldenburg.de</a>
<br clear="none">
> <mailto:<a rel="nofollow" shape="rect" ymailto="mailto:stefan.debener@uni-oldenburg.de" target="_blank" href="mailto:stefan.debener@uni-oldenburg.de">stefan.debener@uni-oldenburg.de</a>>>
wrote:<br clear="none">
> ><br clear="none">
> ><br clear="none">
> > Dear Robert,<br clear="none">
> ><br clear="none">
> > I looked up some own data and find absolutely
no evidence in favour of<br clear="none">
> > your ICA phase adulteration claim, see the
attached pdf report. I guess<br clear="none">
> > you simply used a poor ICA implementation,
and/or a poor component<br clear="none">
> > selection. The attached example is in full
accordance with Arnos reply,<br clear="none">
> > with the difference that I zoom into a clearly
visibile alpha<br clear="none">
> > oscillation, to have a reference brain signal.
The example shows no<br clear="none">
> > evidence that occipital alpha phase is biased
by ICA eye blink<br clear="none">
> > correction. This is a very typical example and
based on a quick and<br clear="none">
> > dirty ICA decomposition, nothing fancy, to
keep this demo simple. Better<br clear="none">
> > preprocessing and component selection would
easily further improve the<br clear="none">
> > signal quality.<br clear="none">
> ><br clear="none">
> > Best,<br clear="none">
> ><br clear="none">
> > Stefan<br clear="none">
> ><br clear="none">
> ><br clear="none">
> ><br clear="none">
> > Am 20.06.17 um 19:53 schrieb Robert Thatcher:<br clear="none">
> > ><br clear="none">
> > > Dear Arno,<br clear="none">
> > ><br clear="none">
> > > 1)*On Phase Differences in the Original
vs the Delorme ICA<br clear="none">
> > > Reconstruction: *We can agree or disagree
about whether or not some<br clear="none">
> > > small eye movement artifact was in the
hand selection that I did. But<br clear="none">
> > > that misses the main point here. That is
the ICA reconstruction<br clear="none">
> > > alters each and every data point in the
entire record including all<br clear="none">
> > > artifact free portions no matter what one
selects. For example, the<br clear="none">
> > > record is 6 minutes and 51 seconds = 411
seconds. The Mitsar sample<br clear="none">
> > > rate was 250 samples per second = 102,750
data samples. Phase<br clear="none">
> > > difference for each frequency band for
each and every one of the<br clear="none">
> > > 102,750 data samples has been altered by
your own ICA reconstruction<br clear="none">
> > > in the EDF file that you emailed to me.
Unless you were to sit next to<br clear="none">
> > > me or if we do a Team Viewer it is not
possible for me to demonstrate<br clear="none">
> > > this for all of the data points and then
create a power point for all<br clear="none">
> > > of these data samples. However, I can
show some exemplars, for<br clear="none">
> > > example, I have created two figures at 4
different time points (1 sec;<br clear="none">
> > > 2:27 sec; 42 sec & 5:49 sec) that you
can download. You can extract<br clear="none">
> > > each screen capture and expand them so
that you can see that the exact<br clear="none">
> > > same time points were selected and the
Hilbert transform JTFA for the<br clear="none">
> > > 4 time points resulted in different phase
differences in all channel<br clear="none">
> > > combinations with respect to O1 for all
frequencies. The same is true<br clear="none">
> > > no matter which channel is selected to
compute the phase differences<br clear="none">
> > > in degrees. The same is true also if one
computes averages of the<br clear="none">
> > > instantaneous phase differences or if one
uses the FFT. Here is the<br clear="none">
> > > download URL:<br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > <br clear="none">
> <a rel="nofollow" shape="rect" target="_blank" href="http://www.appliedneuroscience.com/Phase_Diff-Original_&_Delorme-Post-ICA-4_time_points.zip">http://www.appliedneuroscience.com/Phase_Diff-Original_&_Delorme-Post-ICA-4_time_points.zip</a><br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > > 2)*On the WinEEG ICA Reconstruction: *I
agree that having access to<br clear="none">
> > > ICA components themselves and the
topography is critical in<br clear="none">
> > > understanding exactly what the WinEEG
software did. Unfortunately, I<br clear="none">
> > > personally do not have access to the
WinEEG software.<br clear="none">
> > > Clinician/Scientists in Australia use the
WinEEG software and they<br clear="none">
> > > were the ones that expressed concern
about phase difference distortion<br clear="none">
> > > at a workshop in Adelaide and gave me the
original and the WinEEG ICA<br clear="none">
> > > eye movement corrected files in EDF
format. They explained that they<br clear="none">
> > > removed only one ICA component for eye
movement before they<br clear="none">
> > > reconstructed a new time series. At
first, I was impressed because<br clear="none">
> > > the eye movements were absent in the
reconstructed time series. I<br clear="none">
> > > then was able to use JTFA (Hilbert
transform) to compare the two edf<br clear="none">
> > > files and discovered that all of the
phase differences for all<br clear="none">
> > > channels for all frequencies had been
altered by the ICA<br clear="none">
> > > reconstruction including artifact free
periods. I could demonstrate<br clear="none">
> > > this by individual time comparisons or
averages of instantaneous phase<br clear="none">
> > > differences or by the FFT. A user of
WinEEG explained that they do<br clear="none">
> > > not throw away the original raw digital
data, however I was told that<br clear="none">
> > > they believe that the ICA reconstructed
times series is artifact free<br clear="none">
> > > and therefore they compute means and
standard deviations for their<br clear="none">
> > > normative database using the ICA
reconstructed data and not the hand<br clear="none">
> > > edited or artifact deleted original data
samples like other commercial<br clear="none">
> > > companies do. Your ICA reconstructed
time series is actually less<br clear="none">
> > > different than the original phase
difference in comparison to the<br clear="none">
> > > WinEEG ICA. Nonetheless, both your ICA
reconstruction and the WinEEG<br clear="none">
> > > reconstructions are significantly
different than the original <br clear="none">
> recording.<br clear="none">
> > ><br clear="none">
> > > Best regards,<br clear="none">
> > ><br clear="none">
> > > Robert<br clear="none">
> > ><br clear="none">
> > > Cp���<br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > > On Tuesday, June 20, 2017, 1:12:41 AM
EDT, Arnaud Delorme<br clear="none">
> > > <<a rel="nofollow" shape="rect" ymailto="mailto:arno@ucsd.edu" target="_blank" href="mailto:arno@ucsd.edu">arno@ucsd.edu</a>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:arno@ucsd.edu" target="_blank" href="mailto:arno@ucsd.edu">arno@ucsd.edu</a>>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:arno@ucsd.edu" target="_blank" href="mailto:arno@ucsd.edu">arno@ucsd.edu</a>
<br clear="none">
> <mailto:<a rel="nofollow" shape="rect" ymailto="mailto:arno@ucsd.edu" target="_blank" href="mailto:arno@ucsd.edu">arno@ucsd.edu</a>>>>
wrote:<br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > > Dear Robert,<br clear="none">
> > ><br clear="none">
> > > 1) *On my ICA decomposition analysis on
your data.* You have selected<br clear="none">
> > > a subset of the file where there is 1
minute and 41 second data of eye<br clear="none">
> > > free data. I was only able to select 40
seconds in the same file, and<br clear="none">
> > > I also showed that even in this short
file, there was some residual<br clear="none">
> > > eye movements. Jason and Stefan agreed
with me. This is the reason why<br clear="none">
> > > ICA components power spectrum over
frontal channels (and frontal<br clear="none">
> > > channels only) was affected below 10 Hz
frequency band in my data<br clear="none">
> > > analysis. So on my ICA decomposition, our
disagreement comes from the<br clear="none">
> > > interpretation. You feel that the power
we remove at low frequency in<br clear="none">
> > > frontal channel is not eye movement. In
an attempt to convince you, I<br clear="none">
> > > have picked up a clean region from your
EDF dataset, and did some<br clear="none">
> > > dipole localization at this latency. We
see that in the clean data,<br clear="none">
> > > the best dipolar fit (with 2 symmetrical
dipoles) ends up near the eye<br clear="none">
> > > balls with a residual variance of 6.9%.
Hopefully this convinces you<br clear="none">
> > > that your data is not free of eye
movement artifacts. If you are<br clear="none">
> > > willing to take a step further you might
contemplate the idea that ICA<br clear="none">
> > > can remove this residual spurious
activity.<br clear="none">
> > ><br clear="none">
> > > 2) *On the WinEEG ICA decomposition
analysis.* It is critical for us<br clear="none">
> > > to see the scalp topography (and if
possible continuous activity) of<br clear="none">
> > > the components the people at the
Australia workshop selected. Without<br clear="none">
> > > this, it is not possible for us to
comment on the cleaned data. I<br clear="none">
> > > agree with you that there was some phase
distortion in alpha (visible<br clear="none">
> > > directly in the raw data in the first
email you sent) and that this<br clear="none">
> > > should not be the case. However, without
seing the ICA decomposition,<br clear="none">
> > > it is not possible for us to conclude as
to wether people selected the<br clear="none">
> > > wrong ICA components or if the ICA
decomposition implemented in this<br clear="none">
> > > software is buggy (ICA is not a simple
algorithm and it is sensitive<br clear="none">
> > > to numerical imprecision and a lot of
other parameters - a suboptimal<br clear="none">
> > > implementation could easily explain the
WinEEG results). Also, you<br clear="none">
> > > seem to imply that the WinEEG people were
running ICA on their data<br clear="none">
> > > then throwing away the raw data (which is
why their ICA biased<br clear="none">
> > > neurofeedback database is useless for
practical purposes). Is that<br clear="none">
> > > correct? One should never throw away the
raw data. If they did throw<br clear="none">
> > > away the raw data, it is an indication
that the WinEEG are not<br clear="none">
> > > rigorous in their approach and therefore
might not have implemented<br clear="none">
> > > ICA in an optimal way. If it is not the
case, one may easily<br clear="none">
> > > reconstruct the database of measures with
or without ICA decomposition<br clear="none">
> > > (assuming ICA is done right which does
not seem to be the case) then<br clear="none">
> > > assess data measure distoritions (power,
phase index, etc…) in a<br clear="none">
> > > statistical fashion.<br clear="none">
> > ><br clear="none">
> > > Best wishes,<br clear="none">
> > ><br clear="none">
> > > Arno<br clear="none">
> > ><br clear="none">
> > > <a rel="nofollow" shape="rect" target="_blank" href="http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf%20">http://sccn.ucsd.edu/~arno/download/clean_edf_file_analysis2.pdf
</a><br clear="none">
> <<a rel="nofollow" shape="rect" target="_blank" href="http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf%20">http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf%20</a>><br clear="none">
> > <<a rel="nofollow" shape="rect" target="_blank" href="http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf%20">http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf%20</a>><br clear="none">
> > > <<a rel="nofollow" shape="rect" target="_blank" href="http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf">http://sccn.ucsd.edu/%7Earno/download/clean_edf_file_analysis2.pdf</a>><br clear="none">
> > ><br clear="none">
> > >> On Jun 18, 2017, at 11:44 AM, Robert
Thatcher<br clear="none">
> > <<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>> <br clear="none">
> <mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>>>
<div class="yiv3996281630yqt6084624675" id="yiv3996281630yqtfd48737"><br clear="none">
> > >> <mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>>
<br clear="none">
> <mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>
<mailto:<a rel="nofollow" shape="rect" ymailto="mailto:rwthatcher2@yahoo.com" target="_blank" href="mailto:rwthatcher2@yahoo.com">rwthatcher2@yahoo.com</a>>>>>
wrote:<br clear="none">
><br clear="none">
> ><br clear="none">
> > >><br clear="none">
> > >> <Pre-ICA-Hand Artifact free
selections.edf><br clear="none">
> ><br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > ><br clear="none">
> > > Dieser Nachrichteninhalt wird auf
Anfrage komplett heruntergeladen.<br clear="none">
> ><br clear="none">
> ><br clear="none">
><br clear="none">
<br clear="none">
</div>
</div>
</div>
</div>
</div>
</div>
</blockquote>
<br clear="none">
</div>
</div></div></html></html></div></div></div></div></div></body></html>