Dear Jesper,
a quick take on fnirt:
Assume we estimate nonlinear warp coefficients using fnirt with the --aff starting guess. Now this transform shall be applied to another image using applywarp. It seems like this should be done NOT using --premat with the previous *.mat of the starting guess of using --aff in fnirt, right? (That is the question;)
Hope all is well,
cheers-
Andreas
PS: But what if I want to just estimate the warps from the affine and combine the linear and nonlinear registrations? (2nd question;)
|