I think the best you can do would be to re-normalise the images that have
been flipped during the spatial normalisation, making sure that SPM is
picking up the correct flip setting (no warning messages about
spm_flip_analyze_images).
The problem could have been due to running functions without first running
spm_defaults. This function is not run automatically when Matlab starts. It
is usually run when the spm user interface is started up, so if functions are
called without first starting spm, then it may get the flip wrong (and a
warning sent to the screen).
Another possibility may be that some preprocessed data came from someone
else, who has their flip set differently.
Another explanation would be that it could be a wierd Matlab thing that I
have seen before. Something to do with it not recognising global
variables, but it is quite intermittent. What version of Matlab are you
using? I can't remember which Matlab version I saw this error with. I
think it was either 6.0 or 6.1, on a Sun workstation.
Best regards,
-John
> I am trying to re-estimate a previously specified
> model in SPM2. I saved the different versions of
> SPM.mat soon after
> a) fmri models - design
> b) fmri models - data
> and also after estimation and contrasts etc.
>
> When i try to do a RFX analysis taking contrats images
> from these, i got an error ... images do not
> all have same orientation & voxel size
>
> It was because the spm_defaults was somehow not loaded
> automatically when matlab starts and in our lab the
> defaults.analyze.flip is set to 0. So when configuring
> the design matrix, i used to get an array of error
> message for almost every file ... Warning: cant get
> default analyze orientation -assuming flipped
> and sometimes i explicitly loaded the spm_defaults, so
> these were in different orientation.
>
> Apparently it is the spm_flip_analyze_images.m that
> determines the default orientation and it is called by
> spm_vol_ana.m to fill in the VY variable.
> I tried to reconfigure the design matrices by setting
> SPM.xY.VY=[]; and SPM=spm-fmri_spm_ui(SPM); and re-
> estimated the design matrix SPM=spm_spm(SPM); and
> reevaluated the contrasts. But this did not help as
> some analyses had voxel dimensions of -2x2x2 while
> others had 2x2x2.
>
> Another point is that for both sets, the preprocessed
> data i.e. swa*.img had voxel dimensions of 2x2x2.
>
> Is there a way i can do the re estimation of the
> previously specified models.
|