I'm trying to use Flirt from FSL to get talairach.xfm. Here are the steps I took: 1) recon-all -autorecon1 -notalairach -subjid $subject 2) mri_convert $SUBJECTS_DIR/$subject/mri/nu.mgz $SUBJECTS_DIR/$subject/mri/nu.img 3) bet $SUBJECTS_DIR/$subject/mri/nu $SUBJECTS_DIR/$subject/mri/nu_brain 4) fsl_rigid_register -i $SUBJECTS_DIR/$subject/mri/nu_brain.img -r $FREESURFER_HOME/average/mni305_brain.cor.mgz -o $SUBJECTS_DIR/$subject/mri/nu_reged.img -dof 12 -cost corratio -xfmmat $SUBJECTS_DIR/$subject/mri/transforms/talairach.xfm
The mni305_brain.cor.mgz is the skull-striped mni305.cor.mgz from average folder. The registered images from flirt is fine. However, when I checked the transform with the tkregister2: tkregister2 --mgz --s $subject --fstal It seems it's a bad transforms. I switched the input and ref volumes in step 4. The result is improved, but still not good. I tried to use orig.mgz instead of nu.mgz, the result is the same. Anything I missed?
Thanks,
Wayne