[FieldTrip] Quality of beamforming on MNI template vs single-subject MRI

"Jörn M. Horschig" jm.horschig at donders.ru.nl
Thu Apr 4 14:50:33 CEST 2013

Dear Andrea,

> Questions:
> Does anyone know why ?

I am not sure as you provide minimalistic insight into what you are 
exactly doing. If you use a single subject grid, the actual grid points 
for which you estimate source activity will be slightly different than 
when you base this on a template, so that might be one reason. Also you 
are probably doing monte carlo statistics, for which p-values will vary 
anyway for different runs (sidenote: check the newly implemented p-value 
confidecen interval, if they are overlapping, no need to worry). 
However, since you do not say how much the p-values differ, how many 
permutations you are running, what other parameters you choose etc., 
it's hard to judge whether this is something to worry about.

> Is the use of the MNI template the only (and best) way to allow 
> statistical analysis across subjects in the source space ? 
No, but I think the answer most suited to what you actually want to ask 
is yes.
No, because you can use any other template,you can define any T1 as a 
template, e.g. of one of your subjects. Yes, however, because using the 
MNI template you know what coordinate (roughly) corresponds to what 
anatomical region. With a different template you would need to find a 
reliable way to determine an anatomical label.
There is also a way to do everything on subject-specific space and only 
after source reconstruction warp back to the MNI template. This used to 
be the way before I started working on all this, so I am not exactly 
sure how this works. The only thing I know is that we recommend the way 
that you took.
Another way would be to use a localizer task to anatomically localize 
the region of interest per subject, beam only activity of that region 
and run statistics only on that one region. Then there is no need to go 
to some template space.

Hope this helps :)


On 4/4/2013 2:18 PM, Andrea Brovelli wrote:
>  Dear all,
> I used DICS and compared the quality of the beamforming using two 
> different source spaces (3D grids):
> 1) MNI template space as described in 
> http://fieldtrip.fcdonders.nl/example/create_single-subject_grids_in_individual_head_space_that_are_all_aligned_in_mni_space
> 2) Single subject 3D grid using anatomical MRI
> The number of sources in the two models is similar (approx. 3500 
> sources). However, the results using the MNI template are less 
> significant. That is, the statistical analysis testing a significant 
> increase in power post-stimulus with respect to baseline give higher 
> p-values (in some region of interest).
> Questions:
> Does anyone know why ?
> Is the use of the MNI template the only (and best) way to allow 
> statistical analysis across subjects in the source space ?
> thanks a lot
> Andrea

Jörn M. Horschig
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Neuronal Oscillations Group
FieldTrip Development Team

P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

E-Mail: jm.horschig at donders.ru.nl
Tel:    +31-(0)24-36-68493
Web: http://www.ru.nl/donders

Visiting address:
Trigon, room 2.30
Kapittelweg 29
NL-6525 EN Nijmegen
The Netherlands

More information about the fieldtrip mailing list