Kelli,
In the newest dev version (Jan28 or Feb 10), certain defaults changed, which appear possibly not for the better. Could you try re-running recon-all on the failed subject, using flags which I'll describe in a second?
Attached is a new recon-all. To install, type this from where you have saved the attachment:
gunzip recon-all.gz chmod a+x recon-all cp recon-all $FREESURFER_HOME/bin/
Then, we'd first like you to try adding more iterations to the nu_correct step, as possibly your subject has large non-uniformities. Type the following (where <subject> is the name):
recon-all -s <subject> -autorecon1 -clean -nuiterations 4
This increases the iterations of the 'nu_correct' step from the default of 2 up to 4. See if the resulting skull strip is any better.
Note: the -clean flag will delete your control points! In fact, for now, please don't use control points! (not just yet anyway)
If the skull-strip still fails, try:
recon-all -s <subject> -autorecon1 -clean -no-wsatlas
The -no-wsatlas flag bypasses using an atlas during the skull-strip step (mri_watershed). We've had mixed results with using this atlas (mostly for the better), but its possible that it is the source of your troubles.
Lastly, you can try both flags:
recon-all -s <subject> -autorecon1 -clean -nuiterations 4 -no-wsatlas
If even this fails, then we may want to have a look at the subject files.
Thank you in advance for your patience!
Nick (and Xiao)
On Fri, 2006-02-10 at 16:49 -0500, dominick@bu.edu wrote:
Hi,
I had been having a lot of trouble with the skull stripping in the newest dev version (it was taking off the top half of the head and leaving head and skull on the bottom half). I was able to fix this by redoing the intensity normalization step from autorecon1 using control points. My question is - In the brains where I did not have a problem with skull stripping, but after autorecon1 I have white matter intensity values below 110, should I redo intensity normalization with control points on them as well?
Thanks!
Kelli
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer