| I am trying to perform a group comparison (20 normal controls +
| 48 AD patients) in the hope of comparing SPM with some dedicated
| software attached to our gamma camera. The images have
| all been registered to an inhouse template by the other software.
| Subsequently I have loaded the registered images into SPM99 and
| performed some Gaussian smoothing. The next thing required was to
| give the results some spatial meaning, so I wanted to
| perform an identical rigid body transformation to the SPM99 SPECT
| template on each image, then perform the group statistics
| The method I adopted was as follows:
| I created a mean image from both the controls and patients, then
| using the Coregistration facility in SPM99 with the Coregister Only
| option, coregistered the object (mean) image to the target (SPECT
| Template) image and consequently found the transformation
| matrix. I then applied this matrix to each and every image volume
| in my analysis..................I think?
When you do coregistration, it asks for object, target and other images.
To apply the estimated transformation from the object image to all the
images in the series, you would select them when it asked for other images.
You can make sure that the images are in register with the <Check Reg>
| As I'm not doing any reslicing, are the newly created transformed
| image sets simply replacing the original image data sets, or
| have I completely misinterpreted the purpose of coregistration?
The coregistration probably wont do a wonderful job of registering
the images to the SPM template, unless your in house template is the
same shape and size as those supplied with SPM. Coregistration estimates
a rigid body transformation (6 parameters), whereas more parameters would
be required for matching images of different shapes and sizes.
The results of the coregistration is a modified .mat file for the object
image and any "other" images you specified. Providing all the images
in a statistical analysis have the same .mat files (and dimensions),
then it should run OK. In this case, you probably don't need to do
| I just want to ensure that my original image volumes have all
| been transformed to the SPECT Template using the same
| transformation matrix throughout and to preserve both voxel size
| and counts.
Everything should be preserved, but the registration to the SPM templates
is probably only approximate.