This is because DARTEL does not register the images with MNI space, but
instead transforms all the data to the average shape of all the individuals.
One of the the things on my list is to include an additional transform to put
all the pre-processed images into MNI space.
The simplest procedure for getting the data into MNI space would be that
described at
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0808&L=SPM&P=R16058
Generate a Template_6_sn.mat by using the Normalise button to do an spatial
normalisation of Template_6.nii,1 with the tpm/grey.nii image in the spm
distribution. The resulting Template_6_sn.mat will contain an affine
transform that could be used to bring the dartel warped images into MNI
space. The easiest thing to do then is to modify the headers of the images
that have been spatially normalised to the DARTEL average by pasting the
following text into MATLAB, and selecting the appropriate files:
% Select files
PN = spm_select(1,'.*_sn.mat','Select sn.mat file');
PI = spm_select(inf,'nifti','Select images');
% Determine affine transform from header
sn = load(deblank(PN));
M = sn.VG(1).mat/(sn.VF(1).mat*sn.Affine);
% Scaling by inverse of Jacobian determinant, so that
% total tissue volumes are preserved.
scale = 1/abs(det(M(1:3,1:3)));
% Pre-multiply existing headers by affine transform
for i=1:size(PI,1),
% Read header
Ni = nifti(deblank(PI(i,:)));
% Pre-multiply existing header by affine transform
Ni.mat = M*Ni.mat;
Ni.mat_intent='MNI152';
% Change the scalefactor. This is like doing a "modulation"
Ni.dat.scl_slope = Ni.dat.scl_slope*scale;
% Write the header
create(Ni);
end
Note that in subsequent analyses of the modified images, there will be a load
of warnings about QFORM0 being rounded. These can be ignored in SPM, as they
relate to a field that is unlikely to be used.
Best regards,
-John
On Thursday 23 October 2008 10:01, Saemann Philipp wrote:
> Hello,
>
> sorry to post very similar question again but we are stuck
> with a few questions on DARTEL here - though intersubject coregistration
> and statistical
> robustness are much improved compared with standard SPM5 VBM.
>
> (1) "DARTEL space" vs. "absolute" MNI space:
>
> We noted that the 6th (=last) template is systematically shifted in some
> areas we
> are very interested in (e. g . the hippocampus comes to lie 3-4 mm more
> superior than the original templates and the colin MNI single brain) - it
> brings a bit of a problem when using MSU or other MNI based localisation
> tools.
>
> Is there some trick to remain in MNI space and at the same preserve the
> good intersubject registration?
>
> (2) Can 1x1x1 mm3 images be written out form 1.5 x 1.5 x 1.5 mm3 flow
> fields, or does the process need to be started at the higher resolution
> from scratch?
>
> (3) Smoothing: We stepped down from 12 mm to 8 mm with DARTEL, which wrt
> some result aspects seems "ok", however, is there an objective way to get
> out of
> the arbitrariness in choosing the kernel size?
>
> Thanks very much for your help,
> best regards,
> Philipp
>
> Max Planck Institute of Psychiatry
> NMR Research Group
> Kraepelinstr. 2-10
> 80804 Munich
> Mail: [log in to unmask]
> Phone: 0049-89-30622-413
|