Hi.
I apologize for reposting my question, but I am afraid it got lost in a forked
message thread.
I am trying to compare effect of brain extraction (BET) and more importantly
RF bias field correction (FAST) on accuracy of rigid body registration of my
MR images.
I would like to compare cost function (normalized mutual information) values
if I apply different transformation matrices to input image. I think that I
can get the cost function value from flirt using "-verbose x" command line
option. Example of output is as follows and the value is above "^^^^".
7 : -1.1338 :: 0.852393 0.00369968 0.0853298 62.6364 69.734 -21.1884
^^^^^^
I get transformations by registering reference and processed input images.
Since I can not compare the last value of cost functions from processed input
images, I want to apply the transformation matrix (-applyxfm) to un-processed
input image. However, using both "-verbose x" and "-applyxfm" options I am
not getting cost function value.
I tried to use "-init my.mat" option and modified schedule file with 1mm scale
and hoped to get the value with "measurecost", but it doesn't work.
I have been able to perfectly register horriblly looking interventional MR
images using FSL, while other registration tools (Analyze, ITK, SPM) failed
badly. The neurosurgeons like my results a lot. I would be grateful for any
help on how to get the cost function value.
Cheers,
Martin
|