After running "recon-all -autorecon1" I saw that brain.mgz was missing some large areas in the cerebellum. I ran "recon-all -skullstrip -wsthreshold 35" and then checked my work with "tkmedit subjid brainmask.auto.mgz -aux T1.mgz" The skullstrip looked very good, so I went on to run autorecon2. brain.mgzwas never updated and the output of autorecon2 was horrendous. I am going to use mri_mask to generate a new brain.mgz using T1.mgz and brainmask.auto.mgz. Is this an appropriate solution to my problem? The document fstutorial/skullstriptutorial, which I used as my guide, indicates tht running recon-all with the -skullstrip flag will generate a new brain.mgz but this did not occur when I used the command. The autorecon help file correctly indicates that recon-all -skullstrip generates brainmask.auto.mgz and brainmask.mgz but lacks explanation of the command's various modes. It appears that the generation of brain.mgz is somehow lost between autorecon1 and autorecon2, at least in my experience.
FYI I am using freesurfer stable release 3.0.1 on a pentium 4 running Fedora Core 4
-Tom Bayer