> 1. If the brain image contain non-brain area (i.e spines), is it
> important to crop it out before the registration?
It depends. If one image contains structures that the other doesn't,
then it can be a good idea to crop them first. If you are using
Coregister, then it sometimes helps to skull-strip first. If you are
using other SPM routines, then skull-stripping is often dealt with as
part of the model.
The best way to find out what works is to try it.
>
> 2. Is it possible to register (using SPM normalisation or aka
> non-linear regsitration) images with different array size to each
> other, is any caution required? or is there more danger of
> misregistration? I had a look at Chapter 5 Non-linear registration of
> the SPM book (Karl J Friston 2007), as I'm not good at math, I can't
> find the answer for this. Please help.
I would not suggest trying to warp one persons brain to another using
the Normalise button. In fact (for MRI anyway) I would probably suggest
not going anywhere near the normalise->estimate option.
Which options you use depend on what you hope to achieve. HDW (after
coregistration) probably works as well as anything in SPM for
londitudinal scans of the same subject. For inter-subject MR
registration, I would opt for Segment forllowed by Dartel. For PET, I
would suggest using the old Normalise button. Most bits of SPM can deal
with images of different sizes. Those that can't will give an error
message.
The SPM8 manual should tell you more, and has been written more from a
users perspective than the SPM and HBF books.
Best regards,
-John
--
John Ashburner <[log in to unmask]>
|