Dear Peter,
>
> hello SPM-ers,
>
> I have problems normalizing my data. The volumes are realigned but the normalization
> goes wrong. Some of the scans in the normalization are cut, specially the scans in
> the begin and the scans in the end of the volumes. I have read that this can be
> normal, but when I display the normalized images with disp_analyze I get
> divide by zeros (which doesn't look right to me).
I honestly don't know what "disp_analyze is, but I would guess that as
long as you can see the images (not all voxels are zero) you should be
OK.
>
> If I run statistics on the normalized data, then I get the following errors:
>
> Warning: Log of zero
>
> Warning: Log of zero
>
> Warning: Log of zero
> ??? Index exceeds matrix dimensions.
>
> Error in ==> /rug103/home2/users/csg/csg739/spm96/spm_lambda.m
> On line 43 ==> if df <= length(LC2T); Lc2t = LC2T(df); end
>
> Error in ==> /rug103/home2/users/csg/csg739/spm96/spm_spm.m
> On line 315 ==> Lc2z = spm_lambda(df);
>
> Error in ==> /rug103/home2/users/csg/csg739/spm96/spm_fmri_spm_ui.m
> On line 509 ==> spm_spm(V,H,C,B,G,CONTRAST,ORIGIN,GX,HCBGnames,P,SIGMA,RT);
>
> ??? Error while evaluating callback string.
>
judging from the the error messages above your problem may not be
assosciated with the spatial normalisation as you have assumed. The
error messages indicate that you have negative degrees of freedom for
the design which you attempt to fit your data to, which in turn means
that you have defined more effects than you have independent
measurements.
>
> some info
> ---------
>
> The data is fmri
>
> The volumes consist of 6 scans, covering only part of the brain going from high
> in the forehead, over the ears to low in the back of the head.
>
Six times six = 36mm isn't really a great coverage in the axial
direction, and the spatial normalisation may have problems working out
the transforms in that direction.
> The structural scans were taken on the same place as the functionals.
>
I take it that you have done your structuals in the same was as the
functionals (6 slices of 6mm width). One point of doing structurals may
actually (in the case where you have incomplete coverage with your
functional data) be to have a scan on which to base the spatial
normalisation. This may be something you wish to consider in the
future.
> the resolution of the scans (both structural and functional are (256 x 256).
> the voxel-size is [0.89 0.89 6]
>
> the ac isn't on the scans
>
The spatial normalisation does not explicitly use the ac in any way, so
that should not be an issue.
> I did get results in spm for windows
>
Good for you. Possibly there are some differences in how you set up the
analysis.
> some more questions ?
> ---------------------
>
> Do i have to set the origin?
You don't have to set it, although it may be a good idea to do so. You say that you don't have the ac in the imaged volume, but possibly you could venture a guess as to where it is in relation to your volume.
> Is is neccessary to coregister? If so, what do I fill in as target, object (t1,T2 ?) ?
It may be necessary to coregister depending on exactly what you are
doing. If you are basing the spatial normalisation on the structurals
and applying the parameters to the functionals, then it may be a good
idea to coregister the functional data to your structural. Assuming you
want to use the structurals for spatial normalisation you should set
for your target image the appropriate modality for the structural
(probably best with T1) and for your object image you should set the
appropriate modality for that (T2).
> which of the default normalization options do you advice me to change ?
>
Given your poor coverage in the z-direction you may want too settle for
an affine registration (including zooms and shears, but excluding any
nonlinear componenets). You may even want to assume a value for the
z-axis zoom and not fit that value. You may work out a reasonable
assumption based on the x- and y-axes zooms and the data given in the
paper Ashburner et al. NeuroImage 1997; 6:344-352.
>
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|