Hi John,
What you could do is resample one of your images to an isotropic voxel size (e.g. if you had 0.5mm in-plane
resolution then go for a 0.5x0.5x0.5 mm voxel - which you can do with flirt and the -applyisoxfm call).
Then, you could register each of the other images to this image (with 6 DOF), take the outputs and average them
in fslmaths. It would probably be a good idea to try and apply the transformation matrices using applywarp
and with spline interpolation, to attempt to maximise the sharpness. You can also do this for the initial image
just using the identity matrix and the isotropic target volume (the result of the flirt call with -applyisoxfm).
So, to summarise, I would do:
1 - flirt with -applyisoxfm on one of the images to make an isotropic version
2 - applywarp with "--interp=spline" and the identity matrix (--premat=$FSLDIR/etc/flirtsch/ident.mat)
to make a "sharper" version of this isotropic image (again, using the same starting image)
3 - for each of the other two images, do a 6 DOF flirt registration to the result from step 2
4 - use applywarp with "--interp=spline" to resample the other two images, using the resulting
matrices from step 3.
5 - average the outputs of applywarp (all 3) using fslmaths
Hopefully this will give you a nice looking, higher quality, image.
All the best,
Mark
On 19 Apr 2012, at 04:04, John Smith wrote:
> Hi, I have 3 sets of CT slices of a single subject. One set is axial slices, another is coronal slices, and the other is sagittal slices. In each set, the slices are 4.5 mm thick. Is there any way to combine these sets to produce a single higher quality image?
>
> Thanks
> John
>
|