Thank you for that pointer in the right direction. I wonder if I could ask
a few other questions.
I am trying to register segmented T1 GM maps in a subjects before and after
development of a brain lesion to one another. I have done this by careful
betting, followed by applying FAST. In trying to register the resulting GM
images I have first tried to use FLIRT (with -inmasking of pathology) to
derive a transformation matrix, before using FNIRT (again with inmasking).
My questions are :
1) Looking at the VBM scripts, in fslvbm_3_proc am I right in thinking
native GM images are registered to a group specific template using just
FNIRT - if so why is there no need to use FLIRT first ?
2) When I try and use FLIRT (several options tried including affine/rigid
body models, weighting in mask, costfunction=normcorr, interp=trilinear) on
my GM images I get strange results - images flipped, and sheared/zoomed. I
get much better results using FLIRT with a command line that incorporates
the VBM grey matter config file, but also with an inmask as above, and a
native dilated subject refmask. I would be grateful for any advice here -
eg is it critical to derive a FLIRT matrix before using FNIRT ?
3) Would it be better to derive the FLIRT matrix from betted, unsegmented
images, and then apply this to GM images ?