[FieldTrip] code change potentially relevant for MEG people

[jan-mathijs schoffelen]

Dear all,

Happy new year to all of you! I just made a change to the code which  
may be relevant to some. This change will be present in tonight's  
release version, and will be immediate for the in-house users at the  
Donders Centre.

If you are not doing MEG research you can stop here.
If you never use FieldTrip to compute leadfields for MEG data you can  
stop here.
If you never use the 'singleshell' method for your MEG leadfield  
computation, or if you don't know what that means, you can stop here.

There appeared to be a huge difference in magnitude of MEG-leadfields  
computed with FieldTrip using different algorithms. Particularly,  
there was a big discrepancy between the 'singleshell' method on the  
one hand, and the localspheres/singlesphere method on the other hand.  
It turned out to be the case that there was a scaling factor in the  
singleshell method, which (assuming the geometrical units were in cm)  
would yield a magnetic field/gradient in ft or ft/cm. At the moment  
FieldTrip does not everywhere explicitly impose particular physical  
units on the data and therefore also should not make particular  
assumptions. To make the three methods of leadfield computation more  
equivalent we changed the implicit unit-assuming code.

What does this all mean to you? Probably not an awful lot. Yet, you  
need to be aware that by scaling the leadfields, the magnitude of the  
inverse solution will probably also change (unless you use norm- 
normalized leadfields). It is not correct to compare analysis results  
using old-scale leadfields with analysis results using new-scale  
leadfields, if the respective analysis results have not been  
normalized in a meaningful way. In other words, comparing raw power is  
not correct, comparing neural activity indices, relative changes, t- 
statistics etc. is probably alright.

Sorry for the inconvenience,

Best wishes

Jan-Mathijs



Dr. J.M. (Jan-Mathijs) Schoffelen
Donders Institute for Brain, Cognition and Behaviour,
Centre for Cognitive Neuroimaging,
Radboud University Nijmegen, The Netherlands
J.Schoffelen at donders.ru.nl
Telephone: 0031-24-3614793