source positions - brain regions

[Robert Oostenveld]

hi Nathan

On 15 Jun 2006, at 11:42, Nathan Weisz wrote:
> however now i'm a little bit stuck. from here, what is the best
> method of finding out which dipoles correspond (very roughly) to
> which brain regions, so i can extract specifically those to look at
> the sourcewaveforms? unfortunatley no mri.
> I am aware of the sourceplot function, but was uncapable of
> producing a meaningful output.

So far it seems ok. The headmodel is in cm, the electrodes as well (I
presume, otherwise it would have given warnings and/or errors), so
the source locations are also expressed in cm, since they are
expresed in the same coordinate system. I suggest that you have a
look at http://www2.ru.nl/fcdonders/fieldtrip/doku.php?
id=fieldtrip:documentation:frequently_asked_questions#how_are_the_differ
ent_head_and_mri_coordinate_systems_defined to get a better
understanding in general.

What we normally do in such a case is to use an indivudual headmodel,
and what we thus subsequently would do is to take an anatomical MRI
of the same subject, in which the MRI is aligned to the same head/
model coordinate system (using the ears and nose fiducials). Since
you do not have a MRI, you could take a standard MRI (i.e. the MNI
template in single_subj_T1 from SPM). Some time ago I (or Arno)
posted a homogenous transformation matrix "besa2spm" on the eeglab
mailing list (*). That homogenous transformation matrix transforms
from besa coordinates for a spherical model to MNI coordinates.

besa2spm = [
     0.0101   -0.9400         0    0.5588
     1.1889    0.0080    0.0530  -18.0041
    -0.0005   -0.0000    1.1268    1.8045
          0         0         0    1.0000 ];

spm2besa = inv(besa2spm);

The transformation matrix is for the "besa" spherical model in
EEGLAB, which has different sphere radii (85mm outer sphere) than you
have used. So you should add a scaling from the besa radii to yours.

your2besa = eye(4);
your2besa(1,1) = 8.5/12;
your2besa(2,2) = 8.5/12;
your2besa(3,3) = 8.5/12;

then

your2spm = besa2spm * your2besa
spm2your = inv(your2spm);

It is most convenient to apply the inverse transformation to the MRI
corodinate system. I.e. do
   mri = read_fcdc_mri('single_subj_T1.mnc')
then mri.transform describes the voxels in SPM/MNI coordinates, so
   mri.transform = spm2your * mri.transform;
changes the anatomical MRI into your headmodel coordinates.

Subsequently you can interpolate the functional values onto the
anaomtical voxel locations with SOURCEINTERPOLATE, and plot the
functional values overlaid over the anatomical with SOURCEPLLOT or
SLICEINTERP.

best regards,
Robert

PS (*) the besa2spm coordinate transformation matrix was constructed
using ELECTORDENORMAMALIZE, which warps electrodes. In this case the
spherical 1020 electrodes were warped to the corresponding electrodes
in the MNI template. The homogenous transformation matrix describes
the warp/deformation and can be used as well for the dipoles inside
the head.