<html><head></head><body><div style="font-family:Helvetica Neue, Helvetica, Arial, sans-serif;font-size:small;"><div><div>Ramesh,</div></div><div>    Myself and hundres of others use automatic deletion of artifact that is objective and efficient.  There are three levels: 1- the use of the Hilbert transform to compute instantaneous time pont-by-time point Z scores with respect to a fully artifact free reference that has been scrutinized by numerous individuals and national institutions like NIH and FDA and 2- The selection of 10 seconds of an artifact free template that is then matched one to one or by adjustment to the remainder of the record to select artifact free samples and, 3- as a last resort manual selection of artifact free EEG based on expert judgement.   No matter what an expert needs to review all artifact free EEG samples to confirm that there is no artifact. </div><div><br></div><div>Because the artifact rejection is very fast one can repeat the automatic procedure and produce multiple artifact free selections and determine that they are essentially 0.95 the same and the clinical interpretation is unchanged with multiple independent selections.  One can do this in less than one minute with multiple selections of artifact free EEG and show the same clinical conclusions to a court.   Many users of Neuroguide do this in front of judges and attorneys in court trials and as a consequence QEEG is admitted in court in almost all states in the USA using this method.  In contrast ICA reconstruction is not accepted in court because the expert cannot repeat or replicate his reconstructed EEG and the opposing attorney prevails.</div><div><br></div><div>Here is a url to a You Tube that describes how multiple artifact free selections can be made showing very little change in phase differences and other metrics are the same independent of the selections: <a href="http://youtu.be/FQeHQ6u_nXg" rel="nofollow" target="_blank">http://youtu.be/FQeHQ6u_nXg</a></div><div><br></div><div>I also do not agree that all EEG contains artifact - artifact is produced by non-brain sources that obey the basic physics of external sources that have voltage gradients and magnitudes and shapes that are very different than EEG produced by sources inside of the skull.</div><div><br></div><div>Bob</div><div><br></div><div><br></div><div id="ydpfb50c349yahoo_quoted_7827098436" class="ydpfb50c349yahoo_quoted"><div style="font-family:'Helvetica Neue', Helvetica, Arial, sans-serif;font-size:13px;color:#26282a;"><div>On Wednesday, June 14, 2017, 7:31:18 PM EDT, Ramesh Srinivasan <r.srinivasan@uci.edu> wrote:</div><div><br></div><div><br></div><div><div id="ydpfb50c349yiv1980021979"><div>
    <p>Hi Bob <br clear="none">
    </p>
    <p>I certainly think your suggestion to "delete the one second of
      data" is preferable, if thats all there were to it.  But there are
      two additional things to consider - <br clear="none">
    </p>
    <p>(1) Many EEG studies are in populations, or use experimental
      paradigms, where inevitably there will be an
      eye-blink/eye-movement on a large fraction of trials.  <br clear="none">
    </p>
    <p>(2) I believe manual editing as you describe is also highly
      subjective, as is the selection of ICA components to remove.  We
      usually only remove the eye-movement component.  <br clear="none">
    </p>
    <p>In my opinion, there really is no such thing as artifact-free
      data, except for narrow-band signals like SSVEPs.   <br clear="none">
    </p>
    <p> ramesh<br clear="none">
    </p>
    <p> <br clear="none">
    </p>
    <div class="ydpfb50c349yiv1980021979yqt6888988912" id="ydpfb50c349yiv1980021979yqt43256"><div class="ydpfb50c349yiv1980021979moz-cite-prefix">On 06/14/2017 03:15 PM, Robert Thatcher
      wrote:<br clear="none">
    </div>
    <blockquote type="cite">
      </blockquote></div></div><div class="ydpfb50c349yiv1980021979yqt6888988912" id="ydpfb50c349yiv1980021979yqt83251"><div><div>
        <div>
          <div>Ramesh,</div>
        </div>
        <div>    Thank you for your post and I agree that artifact is
          broad-band and superimposed on many if not all of the EEG
          channels.  Reconstruction therefore will necessarily change
          relative phase which can be seen in the waves themselves and
          is accumulative in the average phase differences between
          channels.  As for your concern "<span>It's not obvious to me
            to prefer the original relative phase with the artifact
            components." I believe that you should have no concern
            because the original phase differences that are artifact
            free are real and produced by the underlying physiology and
            represent the summation of LFP due to synaptic rise times
            and synaptic integration times and conduction velocities
            between groups of neurons in networks of the brain.  The
            original phase differences must be preserved and not altered
            in any manner if one wants to study brain networks and
            dynamics.</span></div>
        <div><span><br clear="none">
          </span></div>
        <div><span>The simple solution is to not use ICA for artifact
            rejection and instead use algorhythms to delete the parts of
            the record that have artifact and retain the parts of the
            original record with no </span><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space">artifact.  Because
            of the stochastic and nonstationarity of the EEG as one
            increases the sample size then one converges toward the
            stable and reproducabe average of the instantaneous phase
            differences between channels that is not corrupted by
            artifact.  ICA reconstruction alters an entire 5 minute EEG
            recording even if there is ony a single 1 second of eye
            movement artifact.   Why not simply delete the one second
            artifact and then work with the remaining 4 minutes and 59
            seconds?</span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space"><br clear="none">
          </span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space">ICA is
            excellent for feature detection and can serve as "seeds" to
            guide further cross-spectral analyses only if the phase
            differences in the original recording are preserved.</span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space"><br clear="none">
          </span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space">Bob</span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space"><br clear="none">
          </span></div>
        <div><span class="ydpfb50c349yiv1980021979ydp5308b4ffApple-converted-space"><br clear="none">
          </span></div>
        <div><br clear="none">
        </div>
        <div><br clear="none">
        </div>
        <div class="ydpfb50c349yiv1980021979ydp293668c9yahoo_quoted" id="ydpfb50c349yiv1980021979ydp293668c9yahoo_quoted_7519025489">
          <div>
            <div>On Wednesday, June 14, 2017, 4:46:34 PM EDT, Ramesh
              Srinivasan <a shape="rect" class="ydpfb50c349yiv1980021979moz-txt-link-rfc2396E" href="mailto:r.srinivasan@uci.edu" rel="nofollow" target="_blank"><r.srinivasan@uci.edu></a> wrote:</div>
            <div><br clear="none">
            </div>
            <div><br clear="none">
            </div>
            <div>
              <div id="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460">
                <div>
                  <p>Hi All - <br clear="none">
                  </p>
                  <p>I think Bob is right that the relative phase will
                    be changed by deleting 1 or 2 artifact components.  
                    Any artifact is broad-band and hence has components
                    in each frequency bin.  When reconstructing the (in
                    this example, 19) channels, the relative phases will
                    change because some of the signal in each frequency
                    bin has been removed when using only 17 or 18
                    components. <br clear="none">
                  </p>
                  <p>The open question is whether the original relative
                    phase or the ICA-corrected relative phase is the
                    better estimate of the relative phase between the
                    populations that contributed to each electrode. 
                    It's not obvious to me to prefer the original
                    relative phase with the artifact components. <br clear="none">
                  </p>
                  <p>Part of the problem for me (and I do use EEGLAB's
                    ica) about identifying components as artifact in the
                    ICA is that I don't think they contain just the
                    artifact, they also contain some genuine brain
                    activity that we are removing.  This bothers me, but
                    I don't know a better solution.  Even the case of
                    the eye-movement artifact components is likely a
                    mixture. <br clear="none">
                  </p>
                  <p>I'd like to see this discussion move away from
                    algorithm to this harder question about artifact
                    removal. <br clear="none">
                  </p>
                  <p>ramesh <br clear="none">
                  </p>
                  <p><br clear="none">
                  </p>
                  <div class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460yqt3095400500" id="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460yqt85155">
                    <div class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460moz-cite-prefix">On
                      06/14/2017 10:43 AM, Robert Thatcher wrote:<br clear="none">
                    </div>
                    <blockquote type="cite"> </blockquote>
                  </div>
                </div>
                <div class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460yqt3095400500" id="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460yqt25440">
                  <div>
                    <div>
                      <div>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Iman,</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>    
                            Thank you for the information.  I could only
                            find a power point attachment of a
                            simulation in your post.  I did not find a
                            scientific publication where you compared
                            the phase differences changes between an
                            original EEG recording and a ICA
                            reconstruction after removing one or more
                            components.   Please re-send your study. 
                            Also please give the citation to any of your
                            publications or other’s publications where
                            phase differences were compared between the
                            original EEG recording and post ICA
                            reconstruction.  It will be interesting to
                            see if you found similar changes like in the
                            study by Montefusco-Siegmund et al or by
                            Georges Otte or even in the example pre vs
                            post data files that you can download from
                            the internet.  I am assuming that you have
                            downloaded the EEG data and then used a JTFA
                            like the Hilbert transform or even the FFT
                            cross-spectrum to prove to yourself that the
                            phase differences between the original and
                            the ICA reconstruction have not been
                            preserved.</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>As
                            for the mathematics concerning
                            reconstruction from a lower dimensional
                            matrix to a higher dimensional matrix where
                            there are no simple linear transforms I
                            refer you to Taken’s theorem where “<span>The
                              reconstruction preserves the properties of
                              the dynamical system that do not change
                              under smooth coordinate changes, but it
                              does not preserve the geometric shape of
                              structures in phase space.</span>”  Also,
                            in standard differential geometry math
                            courses the issue of lower dimensional
                            manifold mapping to higher dimensional
                            manifolds shows a loss of information in all
                            cases.  Also, commonsense operates here
                            where one tries to reconstruct 19 channels
                            of EEG using only 15 or 16 or 17 ICA
                            components hence a loss of information.</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Finally,
                            the brain is not a total chaotic organ.   As
                            demonstrated by many scientists (e.g.,
                            Nunez; Walter Freeman; Roberto-Pascual
                            Marqui; E. Roy John; Joel Lubar; etc)
                            coherence and phase differences are well
                            behaved and highly reproducible within and
                            between subjects.  Coherence and phase are
                            dependent on the number and strength of
                            connections between groups of neurons.  Here
                            is a URL to a study that tested Paul Nunez’s
                            two-compartmental model of Coherence and
                            Phase Differences and found that these
                            measures vary as a function of distance and
                            packing density:</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span><a shape="rect" href="http://www.appliedneuroscience.com/TWO-COMPARTMENTAL_MODEL_EEG_COHERENCE.pdf" rel="nofollow" target="_blank">http://www.appliedneuroscience.com/TWO-COMPARTMENTAL_MODEL_EEG_COHERENCE.pdf</a></span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Here
                            is a url to a study that used EEG LORETA
                            correlations to replicate Diffusion Tensor
                            Imaging measures of connectivity in the
                            brain:</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span><a shape="rect" href="http://www.appliedneuroscience.com/DTI-ThatcherHumanBrainmapping.pdf" rel="nofollow" target="_blank">http://www.appliedneuroscience.com/DTI-ThatcherHumanBrainmapping.pdf</a></span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Here
                            is a url to a study that measured phase lock
                            and phase shift duration from birth to about
                            16 years of age in 458 and where phase
                            differences were stable and well behaved:  <a shape="rect" class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460moz-txt-link-freetext" href="http://www.appliedneuroscience.com/PhaseresetDevelopment.pdf" rel="nofollow" target="_blank">http://www.appliedneuroscience.com/PhaseresetDevelopment.pdf</a></span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>If
                            you do a search of the National Library of
                            Medicine database (Pubmed) using the search
                            terms “EEG coherence” you will find 2,874
                            citations.  There is huge consistency in
                            this vast literature which would be
                            impossible if the brain was totally chaotic.</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Best
                            regards,</span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span> </span></p>
                        <p class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460ydp76ba64cbMsoNormal"><span>Robert</span></p>
                      </div>
                    </div>
                    <br clear="none">
                    <fieldset class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460mimeAttachmentHeader"></fieldset>
                    <br clear="none">
                    <pre>_______________________________________________
Eeglablist page: <a shape="rect" class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460moz-txt-link-freetext" href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" rel="nofollow" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a>
To unsubscribe, send an empty email to <a shape="rect" class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460moz-txt-link-abbreviated" href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" rel="nofollow" target="_blank">eeglablist-unsubscribe@sccn.ucsd.edu</a>
For digest mode, send an email with the subject "set digest mime" to <a shape="rect" class="ydpfb50c349yiv1980021979ydp293668c9yiv0275212460moz-txt-link-abbreviated" href="mailto:eeglablist-request@sccn.ucsd.edu" rel="nofollow" target="_blank">eeglablist-request@sccn.ucsd.edu</a></pre>
                    <br clear="none">
                  </div>
                </div>
              </div>
              <div class="ydpfb50c349yiv1980021979ydp293668c9yqt3095400500" id="ydpfb50c349yiv1980021979ydp293668c9yqt38584">_______________________________________________<br clear="none">
                Eeglablist page: <a shape="rect" href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" rel="nofollow" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br clear="none">
                To unsubscribe, send an empty email to <a shape="rect" href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" rel="nofollow" target="_blank">eeglablist-unsubscribe@sccn.ucsd.edu</a><br clear="none">
                For digest mode, send an email with the subject "set
                digest mime" to <a shape="rect" href="mailto:eeglablist-request@sccn.ucsd.edu" rel="nofollow" target="_blank">eeglablist-request@sccn.ucsd.edu</a></div>
            </div>
          </div>
        </div>
      </div>
    
    <br clear="none">
  </div></div></div></div></div></div></div></body></html>