Hi, We have 15 out of 16 data sets with Final Objective Function value 0.1. We also find some deviations from the talairach (or MNI) volume when we visually inspect the images (fixed vs. movable) in tkregister2. But we are unsure how large the deviations must be for it to be a problem, so we rely on the Final Objective Function value. When it comes to the intensity normalization, that's even harder to inspect visually, so we don't really know how to check that (unless there are really huge effects). But when we use the brain.mgz image as input for the talairach, the Final Objective Value improves, that's why we've started using that. So we need to know how to proceed from there. (We use the "recommended reconstruction," so we check the talairach and skullstrip after autorecon1.)
Thanks! -- yours, Lars M. Rimol
depends why you needed to improve the talairach. If it's just for reporting purposes and the filling/intensity normalization all worked fine then you don't need to.
cheers, Bruce
Hi, After having redone mri_convert with an edited brain.mgz as input file,
in
order to improve the talairach transform, and after having changed the
file
names (brain.xfm -> talairach.xfm), should we then go to autorecon2? (or
do
we need to re-run some more scripts from autorecon1?)
-- yours, Lars M. Rimol
Hi Lars,
You should be able to just move on with -autorecon2 as normal. The only need to re-do anything would be if your poor talairach caused some other step to be wrong (i.e., the skull strip failed or the normalization failed because the talairach was wrong). If those are fine, then continue with -autorecon2.
If you ever need to re-run -autorecon1 for these subjects just be sure to use the flag -notalairach so that your manually corrected talairach won't be overwritten.
Jenni
On Fri, 9 Dec 2005, Lars M. Rimol wrote:
Hi, We have 15 out of 16 data sets with Final Objective Function value 0.1. We also find some deviations from the talairach (or MNI) volume when we visually inspect the images (fixed vs. movable) in tkregister2. But we are unsure how large the deviations must be for it to be a problem, so we rely on the Final Objective Function value. When it comes to the intensity normalization, that's even harder to inspect visually, so we don't really know how to check that (unless there are really huge effects). But when we use the brain.mgz image as input for the talairach, the Final Objective Value improves, that's why we've started using that. So we need to know how to proceed from there. (We use the "recommended reconstruction," so we check the talairach and skullstrip after autorecon1.)
Thanks!
yours, Lars M. Rimol
depends why you needed to improve the talairach. If it's just for reporting purposes and the filling/intensity normalization all worked fine then you don't need to.
cheers, Bruce
Hi, After having redone mri_convert with an edited brain.mgz as input file,
in
order to improve the talairach transform, and after having changed the
file
names (brain.xfm -> talairach.xfm), should we then go to autorecon2? (or
do
we need to re-run some more scripts from autorecon1?)
-- yours, Lars M. Rimol
Hi Lars,
definitely trust your eyes more than the objective function! If the xforms are reasonable everything should work fine, but if you care about talairach averaging or coords for reporting you should adjust them.
cheers,
Bruce
On Fri, 9 Dec 2005, Lars M. Rimol wrote:
Hi, We have 15 out of 16 data sets with Final Objective Function value 0.1. We also find some deviations from the talairach (or MNI) volume when we visually inspect the images (fixed vs. movable) in tkregister2. But we are unsure how large the deviations must be for it to be a problem, so we rely on the Final Objective Function value. When it comes to the intensity normalization, that's even harder to inspect visually, so we don't really know how to check that (unless there are really huge effects). But when we use the brain.mgz image as input for the talairach, the Final Objective Value improves, that's why we've started using that. So we need to know how to proceed from there. (We use the "recommended reconstruction," so we check the talairach and skullstrip after autorecon1.)
Thanks!
yours, Lars M. Rimol
depends why you needed to improve the talairach. If it's just for reporting purposes and the filling/intensity normalization all worked fine then you don't need to.
cheers, Bruce
Hi, After having redone mri_convert with an edited brain.mgz as input file,
in
order to improve the talairach transform, and after having changed the
file
names (brain.xfm -> talairach.xfm), should we then go to autorecon2? (or
do
we need to re-run some more scripts from autorecon1?)
-- yours, Lars M. Rimol
freesurfer@nmr.mgh.harvard.edu