| Before you do statistical analysis, all images must be in register with
| each other. That's why you need to realign them. I took a look at the
| SPM preprocessing tutorial
| (http://www.mrc-cbu.cam.ac.uk/Imaging/preprocessing.html), and it says that
| SPM uses the mean image as the template all functional images are aligned to.
| I'm not quite sure why SPM creates a mean image for that- I'd say register
| all images to, for example, the first functional image. Perhaps an SPM
| expert would like to answer this?
Providing they are not too noisy, you can spatially normalise or coregister
using an individual image. If you do this, you should use one of your
unresliced images after realignment, rather than a resliced image. This
is because the masking at the edges of the realigned images can cause problems.
It is probably fractionally better to use the mean created by the realignment
procedure, but I suspect that this makes very little difference in most cases,
particularly for fMRI. Trilinear interpolation would be sufficiently accurate
to create this mean.
As an aside, the realignment for PET images is a two pass procedure. The
first pass aligns all images to the first in the series, then a mean is
created to which all the individual images are aligned again. This is not
done for fMRI alignment, purely in order to save time.
Best regards,
-John
|