[MINC-users] nlpfit minctracc

[Marc BOUFFARD]

Hello,

Using the following command on two separate computers (yorick and a linux
machine)

nlpfit -init_xfm lin.xfm source.mnc target.mnc nlin.xfm

and then using the resulting grid file to compute the determinant of the
jacobian using mincblob:

mincblob -det grid.mnc det.mnc


yields differing det.mnc files.  This is not caused by mincblob since this
command gives the same result given the same grid file.  Therefore the
differing det.mnc files are caused by differing grid files.  In
particular, on the linux machine there are many artifacts outside the
brain not present in the files obtained from yorick.  Inside the brain the
patterns are similar although the values are not exactly the same (within
about 15%).

The templates and files used on both computers are exactly the same.  The
only difference I could find was minctracc.  On yorick:

yorick 157% minctracc -version
The program <minctracc> was built from:
Package MNI AutoReg, version 0.98k, compiled by swmgr at yorick
(mips-sgi-irix6.5) on 2003-11-26 at 16:46:16

and on the linux machine:

oliver% minctracc -version
The program <minctracc> was built from:
Package mni_autoreg 0.99.2, compiled by rotor at romeo (i686-pc-linux-gnu) on
Mon May  8 23:16:34 EDT 2006


What between the two versions could cause the difference between the grid
files and consequently the det.mnc files?  The fact that most differences
are outside the brain and on the edges suggests a masking difference but I
have not defined a mask in either case.

Marc