<div dir="ltr">Dear Mori,<div><br></div><div>> Are my interpretation correct?<br><div class="gmail_extra"><br></div><div class="gmail_extra">No. See below.</div><div class="gmail_extra"><br></div><div class="gmail_extra">> Liue's singular value decomposition approach removes all the gradient artefacts completely before running ICA on data. So nothing will remain for ICA to detect and separate.<br></div><div class="gmail_extra"><br></div><div class="gmail_extra">His SVD-based method works <i>better</i> than the simple template subtraction in my opinion, but no method can do this job perfectly due to data non-stationarity. After all, data quality depends on how cooperative subjects are.</div><div class="gmail_extra"><br></div><div class="gmail_extra">> Also, gradient artefacts are not similar to blink and muscle artefacts which can be separated by ICA.<br></div><div class="gmail_extra"><br></div><div class="gmail_extra">Depending on your TR and the number of slices, it can make a peak at the theta, alpha, and beta ranges. ECG is harder to remove, and it tends to overlap theta.</div><div class="gmail_extra"><br></div><div class="gmail_extra">So it's pretty challenging.</div><div class="gmail_extra">If you haven't seen Paul Sajda's nice invention of EEG cap for fMRI-EEG recording, you should check it. In his talk he said this hardware-level solution is very important for good data quality.</div><div class="gmail_extra"><br></div><div class="gmail_extra">Makoto</div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Nov 20, 2014 at 10:05 AM, mori larin <span dir="ltr"><<a href="mailto:morilarin88@gmail.com" target="_blank">morilarin88@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Dear Makoto,<div><br></div><div>Thank you for your response and comments, I appreciate it.</div><div>So from what I understand, Liue's singular value decomposition approach removes all the gradient artefacts completely before running ICA on data. So nothing will remain for ICA to detect and separate. Also, gradient artefacts are not similar to blink and muscle artefacts which can be separated by ICA.</div><div>Are my interpretation correct?</div><div><br></div><div>Best regards,</div><div>Mori</div><div class=""><div class="h5"><div><br>On Sunday, 16 November 2014, Makoto Miyakoshi <<a href="mailto:mmiyakoshi@ucsd.edu" target="_blank">mmiyakoshi@ucsd.edu</a>> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">> In your experience, does the remaining effect of gradient artefact appear in a single component (after <span style="font-size:13px;font-family:arial,sans-serif">singular value decomposition approach and </span>running ICA ) similar to other types of artefacts like muscle and blink components?<br><div class="gmail_extra"><br></div><div class="gmail_extra">Theoretically no, due to non-stationality. Practically, it depends, but generally it does not work.</div><div class="gmail_extra"><br></div><div class="gmail_extra">> If yes, what are the characteristics of the gradient component (in time, topography and frequential domains)? <br></div><div class="gmail_extra"><br></div><div class="gmail_extra">ERP looks like the subtracted GA template.</div><div class="gmail_extra"><br></div><div class="gmail_extra">Makoto</div><div class="gmail_extra"><br><div class="gmail_quote">On Sun, Nov 16, 2014 at 9:09 AM, mori larin <span dir="ltr"><<a>morilarin88@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Dear Makoto,<div><br></div><div>In your experience, does the remaining effect of gradient artefact appear in a single component (after <span style="font-family:arial,sans-serif;font-size:13px">singular value decomposition approach and </span>running ICA ) similar to other types of artefacts like muscle and blink components? If yes, what are the characteristics of the gradient component (in time, topography and frequential domains)? </div><div><br></div><div>Best regards,</div><div>Mori</div><div><br></div></div><div><div><div class="gmail_extra"><br><div class="gmail_quote">On 30 June 2014 18:48, Makoto Miyakoshi <span dir="ltr"><<a>mmiyakoshi@ucsd.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Dear Morin,<div><br></div><div>Generally speaking...</div><div><span><div><br></div><div>> Could it be any number and which criteria needs to be considered? <br></div><div><br></div></span><div>Isn't it better to choose the one so that the original sampling rate is an integral multiple of the one chosen?</div><span>
<div><br></div><div>> should I have to remove gradient artefacts before running ICA and then trying to find the remaining effect of gradient artefact in ICA components?<br></div><div><br></div></span><div>Yes definitely. It is because grandient artifact has very high amplitude. Actually I recommend you try it yourself to see what happens. For artifact subtraction I liked Liu's singular value decomposition approach (NeuroImage 2012) because it does not smear out artifacts.</div>
<div><br></div><div>Makoto</div><div><br></div><div><div class="gmail_extra"><div class="gmail_quote"><div><div>On Mon, Jun 30, 2014 at 8:04 AM, mori larin <span dir="ltr"><<a>morilarin88@gmail.com</a>></span> wrote:<br>
</div></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div><div><div dir="ltr">Dear list,<div><br></div><div>I am using EEG data and I have two questions:</div>
<div>1) I am not sure about sampling rate. The EEG data was recorded at 5000 Hz and I have to down sample it for further work. I used 256 Hz and I do not know is it correct or not. How should we select the re-sampling rate? Could it be any number and which criteria needs to be considered? </div>
<div><br></div><div>2) For the EEG data which is recorded simultaneously with fMRI data, in order to remove gradient and BCG artefacts automatically from the data using ICA , should I have to remove gradient artefacts before running ICA and then trying to find the remaining effect of gradient artefact in ICA components? (and what are the methods to remove it) or I have to run ICA on the contaminated data directly? The latter I think I have to expect more components associated to gradient artefacts because the amplitude of the gradient artefacts are larger than brain signals. </div>
<div><br></div><div>I really appreciate it if you could help me,</div><div><br></div><div>Regards,</div><div>Morin</div></div>
<br></div></div>_______________________________________________<br>
Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
To unsubscribe, send an empty email to <a>eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
For digest mode, send an email with the subject "set digest mime" to <a>eeglablist-request@sccn.ucsd.edu</a><span><font color="#888888"><br></font></span></blockquote></div><span><font color="#888888"><br><br clear="all">
<div><br></div>
-- <br><div dir="ltr">Makoto Miyakoshi<br>Swartz Center for Computational Neuroscience<br>Institute for Neural Computation, University of California San Diego<br></div>
</font></span></div></div></div></div>
</blockquote></div><br></div>
</div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br><div><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></div>
</blockquote></div>
</div></div></blockquote></div><br><br clear="all"><div><br></div>-- <br><div class="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></div></div>