Q: First time on list, sorry if I’ve missed similar threads before, but nothing popped up with any search. I’m hoping someone has a hint!

Issue: I’m trying to get stroke lesion segmentations into standard space to finally plug into VLSM (Bates et al) to correlate with behavioral measures. Problem is, I have non-standardized (clinical) scans to work with. Here’s my procedure:

1) Convert DICOMs into mnc format using UCLA debabeler. Seems to work.
2) Generate a transform (target.xfm) from target scan to standard scan using Register. Looks good enough.
3) Segment lesion in target brain using threshold fill in Display and “save labels as mnc” (lbl.mnc). Looks good.
4) Resample labels into standard space using:
mincresample -like standardbrain.mnc -transform target.xfm\\ lbl.mnc lbl_stnd.mnc
Reloading the transformed lbl_stnd.mnc on top of standardbrain.mnc in Display and the segmentation is WAY off, not even close.

Would anyone know what I’m doing wrong?
http://www.bic.mni.mcgill.ca/pipermail/minc-users/2008-May/002029.html

A:how did you do the registration with register. The register program computes the transform from volume 2 (the middle column) to volume 1 (the first column).

If you loaded the subject’s dataset into column one, and the target brain in volume 2, ie:

register subject_mri.mnc target mri.mnc

then your transform saved is backwards from what you want.

If this is the problem, then the following should solve it:

mincresample -like standardbrain.mnc -invert_transformation - 

transform target.xfm lbl.mnc lbl_stnd.mnc