Hi, John. Thank you so much for the reply.
We resliced first, then realigned because these data were acquired in interleaved fashion, which I've been led to believe should be pre-processed in a 1.) reslice and 2.) realign order.
We could reprocess the data (though not too happy to do this, but alas), but now I worry that realign and then reslice on interleaved data will cause other more significant problems. Or am I deluded by fMRI "lore" about this pre-processing order reversal for interleaved acquisition?
Regards,
Jeff
---- John Ashburner <[log in to unmask]> wrote:
> The problem isn't because of the voxel sizes, but is instead because the
> images are in different orientations. Did you reslice then realign, or did
> you realign and then reslice? If you realign without reslicing (or reslice
> before realigning) then the headers will indicate that the images are all in
> different orientations. Try this:
>
> P = spm_get(Inf,'*.img','Select the images in the analysis');
> VY = spm_vol(P);
> R = zeros(numel(VY),12);
> for i=1:numel(VY),
> R(i,:) = spm_imatrix(VY(i).mat);
> end
>
> figure
> plot(R)
>
>
> A few functions in SPM need all the data to be aligned so that a voxel in one
> image exactly corresponds to the same voxel in another image. These
> functions do not reslice the data on-the-fly (because it is likely to lead to
> problems to do with the boundary condidions used for the smoothing). Because
> there is no on-the-fly reslicing, the only option is to crash out, reporting
> that the images are not aligned voxel-for-voxel.
>
> Best regards,
> -John
>
> On Thursday 01 November 2007 01:45, you wrote:
> > John,
> >
> > We have manual ROIs in native voxel space that I'm attempting to use as
> > masks in Marsbar on smoothed, non-normalized functional data (3.75 x 3.75 x
> > 3.80 mm voxels). However, I can't get out scripts to estimate the models
> > and perform the contrasts due to a orientation/voxel size error message.
> > The data were acquired in an interleaved fashion, so were resliced and then
> > realigned. Each subjects' three runs have the same bounding box and voxel
> > dimensions, so I'm uncertain why I keep getting the error message.
> >
> > Error using ==> spm_fmri_spm_ui
> > images do not all have same orientation & voxel size
> >
> > I attempted the plot visualization using the brief script code:
> >
> > P = spm_get(Inf,'*.img','Select the images in the analysis');
> > VY = spm_vol(P);
> > dims = cat(1,VY.dim);
> > figure
> > plot(dims(:,1:3));
> >
> > I receive the attached plot graph, which appears to suggest no outliers
> > over the course of the three runs. Not certain what's going on. I've also
> > attached a single functional image if it would help ferret out the problem.
> > I did notice that the voxel and world space were not the same and that
> > there are no .mat files for normalized data. So, is it a problem with SPM
> > not liking the presence of files with associated .mat files when attempting
> > to construct and estimate a model? Our manual ROIs were constructed to
> > match the functional world space, not voxel space.
|