Hi - here's a more detailed summary of how to fix your problem:
The problem is not related to padding or not padding - it's related to
putting the right constraints on the transform.
Your functional and structural scans are not very well calibrated - i.e.
you need different scaling in x and y to get a good fit (hence you need 9
or 12 parameters). However, because of the poor resolution in Z, the Z
scaling really needs to be constrained (hence for Z you need 7 DOF). Thus
none of the standard DOF settings apply to your data - you need a custom
pre-registration transformation to get excellent results.
The good news is that it's quite easy to find a suitable initialisation
transform (at least in this case!)
* run the default (12 parameter) registration
* examine the resulting transform - look for the resulting x-scaling and
y-scaling values; if the transformation is not doing too much then this is
approximately the top level value and the second value on the second line,
respectively (in your case 1.02 and 1.05).
* create a new initialisation matrix using these two values - put it in a
file called prescale:
1.02 0 0 0
0 1.05 0 0
0 0 1 0
0 0 0 1
* now run flirt with 7 DOF using this as the initialisation transform:
flirt -in Chen_RCrnrMth_1HzS104.folder.bet_meanimage \
-ref Anatomic.bet -omat dof7prescale.mat -out dof7prescale \
-init prescale -dof 7
Thanks,
Steve.
ps - note that this assumes that the Z scaling should be 1; this may not
always be correct, although it is in your case. If it's not, for data such
as this where the Z scaling is not well conditioned automatically, you may
need to try different values for Z scaling in the initialisation matrix.
|