I would check the mri_em_register registrations of the step that immediately precedes mri_watershed (and produces talairach_with_skull.lta). We found that most of the variability and failures occurred in this step, and mri_watershed itself is a very robust brain extraction algorithm when it is given an accurate registration (my experience with BET is that it is probably not the best tool for getting very accurate brainmasks and sometimes takes away too much). I would recommend using the BET extracted image instead of the image with skull for the initial registration step, and then you can use mri_watershed on the original image and get good results every time. This is the type of strategy we use in the HCP Pipelines (except we use something more robust than BET for the initial mask).

Note that the "talairach" registration (done by the -talairach step with the talairach_avi tool) has nothing to do with the registration that precedes brain extraction. Perhaps it should is more robust and should be used, but the affine would have to be converted to the correct format.

Peace,

Matt.

On 3/6/13 3:53 PM, "Bruce Fischl" fischl@nmr.mgh.harvard.edu wrote:

you can use

mri_convert -rl ...

to reslice the output of bet so that it has the same geometry as the rest of our stream

cheers Bruce

On Wed, 6 Mar 2013, Winter, Warren wrote:

...

Hi again, We have had some suboptimal results from FreeSurfer's skullstripping algorithm, even after iteratively trying different watershed thresholds and after checking the quality of Talairach registration. The results are pretty good, but often leave some dura matter and bits of neck behind whose higher intensity values, I believe, may end up skewing recon-all's later estimation of the range of intensity values within grey matter, and consequently compromising the accuracy of pial surface estimation.

FSL's BET, however, seems to do a better job of extracting the brain, at least for our dataset. So, I was wondering if I might be able to skip over mri_watershed and feed autorecon2 and autorecon3 a brainmask.mgz file created using BET. I tried this procedure:

1. mri_convert <path/to/T1dcm> T1.nii.gz
2. bet T1.nii.gz BET_brainmask.nii.gz -A
3. mri_convert BET_brainmask.nii.gz BET_brainmask.mgz
4. recon-all -s <subjid> -autorecon1 -noskullstrip
5. mv BET_brainmask.nii.gz ./<subjid>/mri/brainmask.mgz
6. recon-all -s <subjid> -autorecon2 -autorecon3

And I got this error message, right after normalization2:

" #@# Mask BFS Tue Mar 5 22:52:05 EST 2013

/net/rc-fs-nfs/ifs/data/Shares/DMC-Sheridan2/projects/FOCUS/FreeSurfer/20 12_bet/mri

mri_mask -T 5 brain.mgz brainmask.mgz brain.finalsurfs.mgz

threshold mask volume at 5 ERROR: dimension mismatch between source and mask Linux rc-thunderball 3.2.21-mosix #1 SMP Mon Jul 2 08:55:48 EDT 2012 x86_64 x86_64 x86_64 GNU/Linux

recon-all -s 2012_bet exited with ERRORS at Tue Mar 5 22:52:06 EST 2013 "

I'm guessing that the dimensions of norm.mgz and aseg.mgz didn't match those of brainmask.mgz, resulting during normalization2 in a mismatch between brain.mgz and brainmask.mgz. Does that seem right, and is there a possible workaround?

Thank you!

Warren

-- Warren Winter Research Coordinator Boston Children's Hospital Sheridan Laboratory of Cognitive Neuroscience Division of Developmental Medicine 1 Autumn Street, AU 650 Boston, MA 02215 857-218-5224

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.