<div dir="ltr">Dear Michael and James (cc Caterina),<div><br></div><div>Wow you guys are discussing an interesting topic (sorry if you are suffering from some trouble...)</div><div><br></div><div>Electrode co-registration is a tricky part and probably perfect fit cannot be desired with digitized locations (we use biosemi custom 256ch cap by the way). But you don't need to worry about the coregistration error too much. The channels will be projected to the scalp surface anyway, so heard I. According to Zeynep's paper (see Akalin Acar's papers...), most of the dipole error comes from other factors, such as unrealistic skull conductivity etc.</div><div><br></div><div>Caterina Piazza, who stayed with us for several months last year, is developing a solution to automatically judge whether 1 or 2 dipoles should be fit. In the upcoming meeting she'll present a poster. You may want to contact her.</div><div><a href="http://medicon2016.org/cfp.html">http://medicon2016.org/cfp.html</a><br></div><div><br></div><div>Caterina has infant data, and interestingly infant's EEG data shows much more bilateral dipoles than adult data. She found 15% r.v. threshold would exclude too many ICs. She came up with a scheme to justify that more than 15% r.v. was optimum. You may also want to ask her for this solution. </div><div><br></div><div>Makoto<br></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Sep 17, 2015 at 4:56 PM, Michael Boyle <span dir="ltr"><<a href="mailto:mrboyle@live.unc.edu" target="_blank">mrboyle@live.unc.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">Hey James,<div><br></div><div>Thanks for your reply! I haven't seen too many researchers actually using dipole fits to exclude components yet (just really method papers) so I feel a little in the dark on some key points:</div><div><br></div><div>1. Fitting a single dipole vs. symmetric dipoles. I have adopted the practice of fitting this model for all components and selecting the model that has the lower Bayesian Information Criterion (with the inputs to BIC being the residual variance and the number of parameters for the model (7 for the single dipole (6 dipole parameters plus the model noise term) and 10 for the 2-dipole model (since the second dipole only adds 3 degrees of freedom since the x-y-z coordinates for both dipoles must be symmetric))), but in reality the 2-dipole model will be the "ideal" (i.e. selected by lower BIC) model more often than it should when visually looking at component scalp maps. This is big because there are usually anywhere from a few to more than 10 components that would be excluded in single dipole model (due to either residual variance threshold or location outside the head) that don't get excluded if the 2-dipole model is used and vice versa.</div><div>2. Selecting the variance threshold for component rejection. In all my fits so far (this may be because of the poor coregistration of the EGI and BESA coordinate systems), 15% residual variance threshold is way too low. I would keep anywhere from 5% to 20% of components. I only get to the nominal 30% of components kept on average (I believe Makoto has said he experiences this) if I set my residual variance threshold as high as 35%.</div><div><br></div><div>Any thoughts on these issues?</div><div><br></div><div>Also, I'm not sure how similar the Biosemi 128 channel electrode locations are to the EGI 128 channel electrode locations, but I'll take a look at that transformation matrix and see if it gives anything better than what I got.</div><div><br></div><div>Best,</div><div>Michael</div></div><br><div class="gmail_quote"><div><div class="h5"><div dir="ltr">On Thu, Sep 17, 2015 at 6:52 PM James Jones-Rounds <<a href="mailto:jj324@cornell.edu" target="_blank">jj324@cornell.edu</a>> wrote:<br></div></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 class="h5"><div dir="ltr">Hi Michael,<div><br></div><div>I've had this problem in the past, too, using BioSemi's 128-channel system, which uses the ABCD montage (not 10-20). I had to do some manual warping, in the end, and I tried about a dozen set of parameters (x-, y-, and z-shift, pitch, roll and yaw, and radius changes in each dimension as well). For "pop_dipfit_settings", I use the standard MNI template brain image provided in EEGLAB's BEM folder, the BEM "standard_vol_SCCN.mat", and the "standard_alphabetic.elc" coordinate file, and these co-registration parameters:</div><div><br></div><div>'coord_transform', [0 -12 5 -0.1 0 -1.5708 97.88 92 97.8781] </div><div><br></div><div>This both looks qualitatively like the electrodes are all on the scalp, and out of the dozen or so parameters I tried, is the one that maximizes the number of components with residual variances under 25%, which I take as a general sign of accuracy and good-fit. Nevertheless, most of my dipoles are always stuffed a bit further into the brain than I think is accurate (assuming that the cortex is what is primarily responsible for the EEG signal in the first place). </div><div><br></div><div>I am soon going to be working on a project that has structural MR scans for each participant, which I hope will help improve the localization from the dipole-fitting process.</div><div><br></div><div>Hope that helps!</div><div><br>James<br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Sep 17, 2015 at 3:00 PM, <span dir="ltr"><<a href="mailto:eeglablist-request@sccn.ucsd.edu" target="_blank">eeglablist-request@sccn.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"><br>
<br>
1. Registering EGI 128 Channel GSN Electrode Locations to<br>
BESA/BEM Coordinates (Michael Boyle)<br>
2. Re: Processing Startle in EEGlab (Tara Miskovich)</blockquote></div></div></div></div><div dir="ltr"><div><div class="gmail_extra"><div class="gmail_quote"><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"><br>
<br><br>---------- Forwarded message ----------<br>From: Michael Boyle <<a href="mailto:mrboyle@live.unc.edu" target="_blank">mrboyle@live.unc.edu</a>><br>To: EEGLAB Discussion Mailing List <<a href="mailto:eeglablist@sccn.ucsd.edu" target="_blank">eeglablist@sccn.ucsd.edu</a>><br>Cc: <br>Date: Wed, 16 Sep 2015 22:36:47 +0000<br>Subject: [Eeglablist] Registering EGI 128 Channel GSN Electrode Locations to BESA/BEM Coordinates<br><div dir="ltr"><div><div><div>Hey all,<br><br></div>I have been trying to get a good registration between the sensor locations provided by EGI for their 128 channel geodesic sensor net and the default BESA and BEM coordinate systems that DIPFIT provides, but the results are quite unsatisfactory with the standard warping methods (no manual warping). I have tried using the automatic warp when aligning just the 10-20 sensor locations, the 10-10 sensor locations that EGI reports to be within 1cm of an EGI GSN electrode (<a href="ftp://ftp.egi.com/pub/documentation/technotes/HydroCelGSN_10-20.pdf" target="_blank">ftp://ftp.egi.com/pub/documentation/technotes/HydroCelGSN_10-20.pdf</a>), both with and without fiducials, and even just assessing the correspondence visually shows that the fits are quite poor. I was wondering if anyone has had success getting a good registration particularly with the BESA coordinates, since that is the dipole fitting method I've converged on. The most noticeable issue with my registrations is with the frontal electrodes in either model (the Fz's for example are clearly not well coregistered). Possibly as a result, I have had, in a few cases, DIPFIT place the well-fitting dipole for eye blink components slightly within the brain and thus not automatically exclude them.<br><br></div><div>I can send pictures of my best attempts at registration to anyone interested.<br><br></div>Thanks for your help!<br></div>Michael<br></div>
<br><br></blockquote></div></div></div></div><div dir="ltr"><div><div class="gmail_extra"><div class="gmail_quote"><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"></blockquote></div>-- <br><div><div dir="ltr"><div>James Jones-Rounds</div>Laboratory Manager<br>Human Development EEG and Psychophysiology (HEP) Laboratory,<div>Department of Human Development,<br>--------------------------------------------<br>Cornell University | Ithaca, NY<br></div><div><a href="tel:607-255-9883" value="+16072559883" target="_blank">607-255-9883</a></div><div><a href="mailto:eeg@cornell.edu" target="_blank">eeg@cornell.edu</a></div></div></div>
</div></div></div></div></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>_______________________________________________<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">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">eeglablist-request@sccn.ucsd.edu</a><br></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>