<div dir="ltr">Dear Kelly,<div><br></div><div><div>> I've already gone through half the subjects and rejected artifactual components (I'm rejecting about 3-10 components per file - this may be too many?</div><div><br></div><div>No. I usually see only 10-20 good ICs, regardless of the number of channels. I have a big plan to investigate why it is so, but that's another story.</div><div><br></div><div>> I'm careful to only reject things that are clearly blinks, cardiac artifacts, saccades, or what looks like single bad channels or bad EMG). <a href="https://georgetown.box.com/s/78olhxp6wx5k0x3q8u02nhypsd1cgfnc" target="_blank">This screenshot</a> shows another subject with a few components pulled up - #3 which seems to be clearly brain, and #27, which would be an EMG artifact that I might reject along with the clear blink/cardiac components (1,2,7,9,14) </div><div><br></div><div>I would not reject IC7, depending on its PSD plot. If it shows 1/f curve (EEGLAB's plot is linear in both axes, so 1/f should be smooth exponential curvature), don't reject it.</div><div><br></div><div>> If this still looks poor, I can re-run everything through ICA and try the 1.5Hz filter, reducing the number of components, and subtracting the extra channel (FYI, when referencing, I did use the method where I add a dummy channel, re-ref, then remove the dummy channel - so maybe this is unnecessary?). </div></div><div><br></div><div>I would say your decomposition is clean to very clean.</div><div>If you doubt if what you get could be suboptimal, go ahead and run 1.5Hz high-pass to make comparison (when people run into a dilemma in preprocessing 'Oh, should I do A or B?' they <i>stop. </i>This is a bad habit of qualitative thinking's fault.<i> </i>The best answer is always try both and evaluate the difference quantitatively. If the difference seems trivial, it means choosing whichever is fine). In the end you have to choose either one of which, but it is free to run the analysis to see the changes. I can't predict what will come out here, because high cut-off highpass can sometimes substantially improve ICA in general.</div><div><br></div><div>Makoto</div><br><div class="gmail_quote"><div dir="ltr">On Fri, Oct 26, 2018 at 7:47 AM Kelly Michaelis <<a href="mailto:kcmichaelis@gmail.com">kcmichaelis@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr">Hi Arno, Marius and Makoto,<div><br></div><div>Thank you all for the helpful advice! </div><div><br></div><div>To answer Arno's question about other ICA algorithms, I tried runica, and I found AMICA produced better results. I did also try re-running it with just removing the channels and not interpolating before ICA, and found that it didn't really make a difference. </div><div><br></div><div>As for the amount of data - I have two recording sessions per subject, and the sessions are split into two .set files (I reapplied saline to lower impedances halfway through the session). Each .set file is submitted to ICA independently. Each file is about 30 minutes of data, but once that gets epoched to 1sec epochs before ICA, it's trimmed considerably, so that before ICA I end up with (data = 128x250x2052), depending on the # of epochs removed. So perhaps this is part of why things don't look as pretty as they could? <br></div><div><br></div><div>My data are fairly noisy - lots of EMG and non-repetitive artifacts - but this being my first major EEG study I didn't have much to compare my ICA decompositions to except tutorials online. After hearing from you all, perhaps what I'm getting isn't all that abnormal for 128 channels and somewhat noisy data. Following Makoto's advice, I have been checking the component scroll data for each file, and it looks fine. </div><div><br></div><div>I've already gone through half the subjects and rejected artifactual components (I'm rejecting about 3-10 components per file - this may be too many? I'm careful to only reject things that are clearly blinks, cardiac artifacts, saccades, or what looks like single bad channels or bad EMG). <a href="https://georgetown.box.com/s/78olhxp6wx5k0x3q8u02nhypsd1cgfnc" target="_blank">This screenshot</a> shows another subject with a few components pulled up - #3 which seems to be clearly brain, and #27, which would be an EMG artifact that I might reject along with the clear blink/cardiac components (1,2,7,9,14) </div><div><br></div><div>If this still looks poor, I can re-run everything through ICA and try the 1.5Hz filter, reducing the number of components, and subtracting the extra channel (FYI, when referencing, I did use the method where I add a dummy channel, re-ref, then remove the dummy channel - so maybe this is unnecessary?). </div><div><br></div><div>I'm curious to know what you all think. Thank you!</div><div><br></div><div>Kelly</div><div><br></div><div><br></div><div><br></div><div><br></div></div></div></div><br><div class="gmail_quote"><div dir="ltr">On Thu, Oct 25, 2018 at 2:47 PM Makoto Miyakoshi <<a href="mailto:mmiyakoshi@ucsd.edu" target="_blank">mmiyakoshi@ucsd.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr">Dear Kelly,<div><br></div><div>I like Marius's advice. Try 1.5Hz high-pass. Also, your data look fine with me.</div><div><br></div><div>See the pictures shown here. If you screw up in copying weight matrices, you'll NEVER know it by looking at scalp maps. You MUST check EEG.icaact which is ICA activation on the scroll data.</div><div><a href="https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Tips_for_copying_ICA_results_to_other_datasets_.2806.2F26.2F2018_updated.29" target="_blank">https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Tips_for_copying_ICA_results_to_other_datasets_.2806.2F26.2F2018_updated.29</a><br></div><div><br></div><div><span style="font-family:arial,helvetica,sans-serif">> I got one suggestion to reduce the number of components down to something like 64, but this article by Fiorenzo, Delorme, Makeig recommends against that. </span><br></div><div><br></div><div>I have been very worried that Fiorenzo's message was widely misunderstood. He did not say using PCA is unconditionally bad. Instead, his message was that do not use PCA to save time. PCA is a lossy compression, so you do lose information. Also, usually up to 20 principal components explains >95% of data, but most of the variance here is artifacts (eye blinks etc). Although 'keeping 95% variance' is a widely-used engineering standard practice, it does not hold for EEG.</div><div><br></div><div>Using PCA is TOTALLY VALID if you know you need to reduce data dimension. There is no evidence, for example, that channel rejection is better than PAC dimension reduction.</div><div><br></div><div>Makoto</div><div><br></div><div><br></div></div><br><div class="gmail_quote"><div dir="ltr">On Tue, Oct 16, 2018 at 8:18 AM Kelly Michaelis <<a href="mailto:kcmichaelis@gmail.com" target="_blank">kcmichaelis@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr"><font face="arial, helvetica, sans-serif">Hi everyone,</font><div><font face="arial, helvetica, sans-serif"><br></font></div><div><font face="arial, helvetica, sans-serif">I'm wondering if anyone can help shed some light on why I'm getting such poor ICA decomposition and what to do about it. I've tried a number of pipelines and methods, and each one is about this bad (The link below has pictures of the scalp maps from two files below). I'm using a 128 channel EGI system. Here is my pipeline:</font></div><div><font face="arial, helvetica, sans-serif"><br></font></div><div><font face="arial, helvetica, sans-serif">1. Import, low pass filter at 40Hz, resample to 250Hz, high pass filter at 1Hz</font></div><div><font face="arial, helvetica, sans-serif">2. Remove bad channels and interpolate, then re-reference to average ref</font></div><div><font face="arial, helvetica, sans-serif">3. Epoch to 1s epochs, remove bad epochs using joint probability</font></div><div><font face="arial, helvetica, sans-serif">4. run AMICA using PCA keep to reduce components to 128-#chans interpolated</font></div><div><font face="arial, helvetica, sans-serif">5. Load raw data, filter same as above, resample, remove bad chans, interpolate, re-reference</font></div><div><font face="arial, helvetica, sans-serif">6. Apply ICA weights to continuous, pre-processed data</font></div><div><font face="arial, helvetica, sans-serif">7. Do component rejection</font></div><div><font face="arial, helvetica, sans-serif"><br></font></div><div><font face="arial, helvetica, sans-serif">What am I missing? Does anyone see any glaring errors here? My data are a bit on the noisy side, and while I do capture things like blinks and cardiac artifacts pretty clearly, I get the artifacts loading on a lot of components, and I'm not getting many clear brain components. I got one suggestion to reduce the number of components down to something like 64, but this article by Fiorenzo, Delorme, Makeig recommends against that. </font></div><div><font face="arial, helvetica, sans-serif"><br></font></div><div><font face="arial, helvetica, sans-serif">Any ideas?</font></div><div><font face="arial, helvetica, sans-serif"><br></font></div><div><font face="arial, helvetica, sans-serif">Thanks,</font></div><div><font face="arial, helvetica, sans-serif">Kelly</font></div><div><br></div><div>Scalp maps:</div><div><a href="https://georgetown.box.com/s/1dv1n5fhv1uqgn1qc59lmssnh1387sud" target="_blank">https://georgetown.box.com/s/1dv1n5fhv1uqgn1qc59lmssnh1387sud</a><br></div></div></div><br><div class="gmail_quote"><div dir="ltr">On Thu, Oct 11, 2018 at 11:10 AM Kelly Michaelis <<a href="mailto:kcmichaelis@gmail.com" target="_blank">kcmichaelis@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div dir="ltr">Hi everyone,<div><br></div><div>I'm wondering if anyone can help shed some light on why I'm getting such poor ICA decomposition and what to do about it. I've tried a number of pipelines and methods, and each one is about this bad (I've attached pictures of the scalp maps from two files below). I'm using a 128 channel EGI system. Here is my pipeline:</div><div><br></div><div>1. Import, low pass filter at 40Hz, resample to 250Hz, high pass filter at 1Hz</div><div>2. Remove bad channels and interpolate, then re-reference to average ref</div><div>3. Epoch to 1s epochs, remove bad epochs using joint probability</div><div>4. run AMICA using PCA keep to reduce components to 128-#chans interpolated</div><div>5. Load raw data, filter same as above, resample, remove bad chans, interpolate, re-reference</div><div>6. Apply ICA weights to continuous, pre-processed data</div><div>7. Do component rejection</div><div><br></div><div>What am I missing? Does anyone see any glaring errors here? My data are a bit on the noisy side, and while I do capture things like blinks and cardiac artifacts pretty clearly, I get the artifacts loading on a lot of components, and I'm not getting many clear brain components. I got one suggestion to reduce the number of components down to something like 64, but this article by Fiorenzo, Delorme, Makeig recommends against that. </div><div><br></div><div>Any ideas?</div><div><br></div><div>Thanks,</div><div>Kelly</div><div><br></div><div><br></div></div></div>
</blockquote></div>
_______________________________________________<br>
Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" rel="noreferrer" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
To unsubscribe, send an empty email to <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu" target="_blank">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
For digest mode, send an email with the subject "set digest mime" to <a href="mailto:eeglablist-request@sccn.ucsd.edu" target="_blank">eeglablist-request@sccn.ucsd.edu</a></blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr" class="m_8095431202922221246m_-4832123330088500424m_-6297412784448039492m_-5573628341717825067m_-1818756473059668117gmail_signature" data-smartmail="gmail_signature"><div dir="ltr">Makoto Miyakoshi<br>Swartz Center for Computational Neuroscience<br>Institute for Neural Computation, University of California San Diego<br></div></div></div>
</blockquote></div>
</blockquote></div><br clear="all"><div><br></div>-- <br><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr">Makoto Miyakoshi<br>Swartz Center for Computational Neuroscience<br>Institute for Neural Computation, University of California San Diego<br></div></div></div>