Hi,
Sorry for the lack of reply but it didn't sound like you were
interested in an FSL solution from your previous email.
I think that your previous problems may have stemmed from
using mm coordinates (as pointflirt needs them in a special,
FSL-specific convention). Try using voxel coordinates in pointflirt
(with the --vox flag) and this will hopefully give you a matrix
that applies correctly to images. Note that the relationship
between voxel and mm coordinates is not trivial when dealing
with flirt matrices, so it may not be possible to check whether
your matrix is doing the right thing easily in matlab, but you
can use img2imgcoord to check how well the transformation
is mapping the coordinates. Sorry this is tricky but this tool is
rarely used so is not well documented I'm afraid.
Also, I'm not even sure if you need to be using pointflirt.
In your first email you have already decided for yourself that
you need to map landmark points, but in my experience it is
possible to register images without resorting to this. Have you
tried using mutualinfo as the cost and doing it without brain
extraction? Or possibly with a cost function weighting image
defined on the template image? If you cannot get good
results with pointflirt then feel free to upload some example
data to:
http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
and send us the reference number. It would also be useful
to know what kind of approach you used in AFNI in order to
get your previous result.
All the best,
Mark
On 22 Nov 2010, at 21:40, Uenal wrote:
> Hi.. I sorted this out but using AFNi... I would be glad if some could help me out to do this with FSL instead of AFNI...
> Thanks,
> ..
> U.
>
|