> We are finally dragging ourselves into the 21st century and moving all
> analyses to SPM2. A problem which emerges (and to which we couldn't find
> the answer in the archives) is how to deal with all our SPM99-specific PET
> ligand templates.
>
> Our "raw" images are Analyze, L on R, +x voxel size, and get displayed in
> SPM2 as L on L, -x. Creating a template from scratch in SPM2 leads to
> images which are Analyze, L on L, -x. SPM2 templates (*.mnc) get displayed
> in SPM2 as L on L, +x - so far, so consistent.
>
> SPM99 templates get displayed in SPM2 as L on R, -x; this situation would
> lead to the wrong sides being matched during normalisation. This is due to
> the x flipping during their construction in SPM99. We therefore want to
> convert our SPM99 templates to something compatible with SPM2.
>
> 1. Are we correct in thinking that flipping the SPM99 templates in x will
> be sufficient? This leads to them being displayed as L on L, -x.
You can simply do a left-right flip of your template images. This is because
they are purely axial with the x co-ordinate of the origin in the centre.
>
> 2. Ideally, we would like to have the (flipped, spm-2-ified '99 templates)
> in *mnc format and displayed as L on L, +x, like the supplied MNI/ICBM
> SPM2 *.mnc templates, but haven't had much luck with such conversions; for
> example Andrew Janke's ana2mnc from www.cmr.uq.edu.au/~rotor/software/ did
> not find the spm origin and we ended up with a template without origin and
> L on L, -x. Mark, I think you were working with Andrew some time ago on an
> improved version that would recognise the origin - any progress on that? Or
> are there any other tried & tested solutions?
A few extra arguments would need to be passed in order to put voxel-sizes,
origins etc into the MINC file. These links may give you some ideas:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0307&L=spm&P=R25187&I=-1
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0306&L=spm&P=R44552&I=-1
Use the Check-Reg button with your flipped SPM99 template versus the MINC
version to make sure the origin is OK.
>
> 3. A related question - some of our tools for SPM99 were written with 1x1x1
> mm voxels, thereby ending up with a 182x218x182 matrix. They were fully
> compatible with the 2x2x2 mm, 91x109x91 matrix MNI/ICBM templates, and
> could, for example, just be resliced to 2x2x2 if need be without problems.
> Due to concern about unequivocality of the origin, SPM2 writes 1x1x1mm
> images into a 181x217x181 matrix. Clearly only one slice has been lost in
> every dimension, leading to loss of symmetry - but which slice has been
> lost (inf or sup, right or left, ant or post)? Am I correct in thinking
> that, in order to avoid misplacement by one mm in each dimension, I have to
> renormalise the images used for the tool construction using SPM2?
The origin in SPM99 could be out by half a voxel anyway (because of rounding
to the closest value). In SPM2, the code was written to ensure that the
origin fell at the centre of a voxel. For example, consider a counding box
from -21:21, with a voxel size of 2mm. In SPM99, this would be at voxels:
bb = [-21 21];
vx = 2;
bb(1):vx:bb(2)
In SPM2, the voxels would be such that a zero is included:
bb = round(bb/vx)*vx
bb(1):vx:bb(2)
Best regards,
-John
|