Print

Print
Dear Bart,

On 8 Jun 2010, at 11:12, Bart Machilsen <[log in to unmask]> wrote:

Dear Dr. Litvak,

 

I just read your course slides on M/EEG preprocessing (which I downloaded from the SPM website), and I have a question about re-referencing. You write that for some purposes (source reconstruction and DCM) it is presently necessary to use average reference. Does this mean that re-referencing to the average of the mastoids is wrong for DCM, or that the software currently cannot handle this?


Yes, at present the forward computation for EEG assumes average reference. We've been expecting for a long time now that this will be relaxed but that will require some changes in the Fieldtrip code we are using and our collaborators haven't got to implementing these changes yet.

I am a bit confused, as in the SPM8 manual it is stated that “Re-referencing should mainly affect the analysis you do at the channel level and should have little impact on 3D source reconstruction and DCM as long as it is not too noisy” (p.106). I am planning to run a DCM analysis on my data (study on perceptual grouping), so I do not want to make too much errors...


That is a mistake. I'll fix it. At some stage we thought that the reference doesn't matter for source reconstruction and DCM but the we've realised that it does. I remember fixing that in the manual but it seems that there was another place where this was written. There are several other places in the manual which say quite clearly that you should convert to average reference.

 

On a related note: Is it possible to do source reconstruction and DCM without individually measuring electrode locations, i.e. by just assuming default locations for all sensors?


I think I've just answered this question on the SPM list a couple of days ago. The answer it can be done and many people do it but quantifying the error that this introduces is not simple. But in any case many EEG people especially ones using caps don't measure individual electrode locations.

Best,

Vladimir

 

Thanking you in advance,

 

bart machilsen

 

--------------------------------------------------

Bart Machilsen

Laboratory of Experimental Psychology

University of Leuven

Tiensestraat 102

3000 Leuven

Belgium

Tel: +32 (0)16/325885

Fax: +32 (0)16/326099

http://ppw.kuleuven.be/labexppsy