This is why I keep repeating the statement about never changing the flip
settings part way through an analysis. The images were originally considered
left-handed, so the .mat files (if you are still using old software) or
headers (if you are using the new stuff - SPM5 onwards) of the realigned
images will encode this right-handed orientation.
I'm guessing that you are still using an old SPM version, in which case the
first image of the series won't have a .mat file (because the positional
information can be encoded by the origin and voxel sizes of the ANALYZE
.hdr). After you change the flip settings, this image will then be
considered as stored in a left-handed way.
If you are still using old software, then you can just delete all the .mat
files and start again.
Unless you want to use some old toolboxes, then I would strongly suggest
switching to SPM5 or later, and using e.g. MRIcron (rather than MRIcro) to do
the conversion. SPM itself is also able to convert many forms of DICOM.
Best regards,
-John
On Wednesday 12 November 2008 20:59, Michael Bannert wrote:
> dear spm community,
>
> after running our preprocessing scripts on a few of our subjects
> we've found out that our spm default.analyze.flip value was set to 1,
> which is not what we wanted because our raw analyze images had
> already been flipped to the neurological orientation in the
> conversion step from dicom to analyze. this resulted in our images
> being flipped twice. we realized this when there was a mismatch
> between the coordinates reported by spm and mricro.
>
> so now we decided to do the preprocessing with the flip value set to
> 0. no big deal, i thought. the problem is that we get a strange
> warning that goes like this:
>
> "Principal matrix logarithm is not defined for A with nonpositive
> real eigenvalues. A non-principal matrix logarithm is returned."
>
> this warning is issued by funm.m which is called from logm.m in
> spm_uw_estimate.m
>
> what can we do about this warning? what does it mean?
>
> since both funm.m and logm.m are matlab functions i figured that this
> might be a version problem (i run spm within matlab R2006b) but even
> when temporarily replacing these functions with their equivalents
> from matlab R2007a, i get the same warnings.
>
> i read that there's a known conflict in spm_uw_esimate between a
> variable name used and function name from a matlab toolbox and i also
> used the fix attached in this mail ( http://www.jiscmail.ac.uk/cgi-
> bin/wa.exe?A2=ind03&L=SPM&P=R501185 ) but that didn't help either.
>
> any suggestions are greatly appreciated!
>
> best regards,
> michael
|