<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Arial; font-size: 12pt; color: #000000'><span>Hi Marc,<br><br>this comment is not related to your code example below, because I did not understand it. The Knösche (2002) paper suggests to use a minimum norm solution as a plausible, but temporary model to carry the information from the original sensor position to the new positions. If you use a plausible spatial arrangement of the MEG sensors and the source space together with a low regularization you might easily achieve even smaller RV values than 2% for the transformation original-to-original. For all other transformations, you need to keep in mind, that you always extrapolate the magnetic field distribution from the position & orientation of your measurement towards the position & orientation you finally wish to have. Nevertheless, it works as long as the differences in position & orientation are not too big (see the Knösche paper for some tests). The critical examples are those in which you try to get closer to the sensor via the head position correction. In these cases, you actually ask for an improvement of your signals which were unfortunately measured suboptimally. Due to the larger distance between sensors and brain, your data most likely suffers from a complete loss of impact of some of the more distant sources - it is therefore impossible to get this information back by whatsoever mathematical method.<br><br>We have used this method on a regular basis from about 2000 on until 2006 to correct for different head positions between measurement blocks whenever the signal-to-noise ratio in single blocks was too low to get stable inverse solutions from the single block data already. In the latter case, there is no need to use this method as the computation of the inverse solution for each single block is an simpler, alternative option.<br><br>best wishes,<br>Burkhard<br><br><br><br><span name="x"></span>--<br>Dr. Burkhard Maess<br>Max Planck Institute for Human Cognitive and Brain Sciences<br>Stephanstr. 1a, P.O. Box 500355, D-04303 Leipzig<br>Aussenstelle Bennewitz, phone/fax: +49(3425)8875-2526/-2511 mail: maess 'at' cbs.mpg.de, http://www.cbs.mpg.de<span name="x"></span><br></span><br><hr id="zwchr"><b>From: </b>"Marc Recasens" <recasensmarc@gmail.com><br><b>To: </b>fieldtrip@science.ru.nl<br><b>Sent: </b>Wednesday, 30 May, 2012 11:47:51 AM<br><b>Subject: </b>[FieldTrip] megrealign across runs/sessions<br><br>Hi everyone,<div><br></div><div>I'm considering the possibility to append the MEG data (CTF-275) from 3 different runs (recorded within the same day but with different headpositions in the dewar) into one single dataset. That is, combine my datasets in the sensor-space.</div>
<div>I've been reading about the possibility to use the ft_megrealign function in order reconstruct the magnetic fields onto a standard gradiometer location,</div><div>However, in the literature this is mainly used to average data across subjects rather than across runs.</div>
<div><br></div><div>I did a test using the following code:</div><div><i>cfg= [];</i></div><div><div><i>cfg.template{1} = run1.grad;</i></div><div><i>cfg.template{2} = run2.grad;</i></div><div><i>cfg.template{3} = run3.grad;</i></div>
<div><i><br></i></div><div><i>cfg.vol = vol; % single shell headmodel computed from individual MRI</i></div><div><i>cfg.inwardshift = 3;</i></div><div><i>cfg.verify = 'yes';</i></div><div><i>cfg.feedback = 'yes';</i></div>
<div><i>[interp1] = ft_megrealign(cfg, run1); % trial-based data</i></div></div><div><div><i>[interp1] = ft_megrealign(cfg, run1);</i></div><div>[interp1] = ft_megrealign(cfg, run1);</div><br class="Apple-interchange-newline">
</div><div>acording to the results (I show the highest RV), the difference between the original and the realigned data seem really small (which I assume it's good)</div><div><div><i>original -> template RV 2.22 %</i></div>
<div><i>original -> original RV 2.11 %</i></div><div><i>original -> template -> original RV 2.14 %</i></div></div><div><br></div><div><br></div><div>I'm wondering whether anyone has experience in using ft_megrealign across runs/sessions and can recommend it (any advise is welcomed).<br>
</div><div>According to Knosche (2002), the method seems good but I'd like to know whether someone has test it in real-life situations (especially taking into account the head position differences in the z axis)</div>
<div><br></div><div>Can affect the accuracy of the subsequent source reconstruction?</div><div><br></div><div><div><br></div>-- <br>Marc Recasens</div><div>PhD Student<br>Universitat de Barcelona<br>Tel.: +34 639 24 15 98<br>
<br>
</div>
<br>_______________________________________________<br>fieldtrip mailing list<br>fieldtrip@science.ru.nl<br>http://mailman.science.ru.nl/mailman/listinfo/fieldtrip</div></body></html>