In my experience, the FreeSurfer's brain extraction works better than the BET. Although, now with an option of iterative BET in the latest FSL, the outcomes are much better than before. In cases where brain extraction failed in my analysis, I found that it was the bias field correction which didn't do the best job, though the images looked ok. In these cases, even playing with watershed parameters didn't provide statisfactory results. So I tried FLS's bias correction followed by the watershed, and this always helped.

Martin

On Thu, 13 Dec 2007, Bruce Fischl wrote:

Hi Simon,

yes, you would need to skip the watershed step (it is our skull stripping), or better yet, replace it with bet, since at some points we need the skull on e.g. to do TIV estimation. Do you find our skullstripping is failing? In our hands the current version is extremely robust - working on 99% of the brains we test.

cheers, Bruce

On Thu, 13 Dec 2007, Simon Rushton wrote:

...

I've been having problems with a few brains (errors during processing). I found advice on the web about running BET (the FSL brain extraction tool) prior to running recon-all

http://lcni.uoregon.edu/mediawiki/index.php/SOFT:FSL_FreeSurfer

I've tried this and that has helped. (Searching the mail archive I see that at least one other person on this list appears to be doing the same.) My question is, is this a sensible thing to do? I ask because looking at the output I see that at least one of the early processing stages appears to expect a non-stripped brain:

mri_watershed -brain_atlas /Applications/freesurfer/average/ RB_all_withskull_2007-08-08.gca transforms/talairach_with_skull.lta T1.mgz brainmask.auto.mgz

Does this suggest that I either should not try the BET trick or that if I do that I should make some modifications to the recon-all scripts?

simon

---

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

---

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

I've found it varies. A couple of brains that didn't go through with the FS brain extraction tool went through with BET. Maybe recon-all could include an option to try BET? Obviously have the most reliable tool as default but allow the user to easily test an alternative tool before starting to dig deep into other things?

simon

On 14 Dec 2007, at 09:40, Martin Kavec wrote:

In my experience, the FreeSurfer's brain extraction works better than the BET. Although, now with an option of iterative BET in the latest FSL, the outcomes are much better than before. In cases where brain extraction failed in my analysis, I found that it was the bias field correction which didn't do the best job, though the images looked ok. In these cases, even playing with watershed parameters didn't provide statisfactory results. So I tried FLS's bias correction followed by the watershed, and this always helped.

Martin

On Thu, 13 Dec 2007, Bruce Fischl wrote:

...

Hi Simon,

yes, you would need to skip the watershed step (it is our skull stripping), or better yet, replace it with bet, since at some points we need the skull on e.g. to do TIV estimation. Do you find our skullstripping is failing? In our hands the current version is extremely robust - working on 99% of the brains we test.

cheers, Bruce

On Thu, 13 Dec 2007, Simon Rushton wrote:

...

I've been having problems with a few brains (errors during processing). I found advice on the web about running BET (the FSL brain extraction tool) prior to running recon-all http://lcni.uoregon.edu/mediawiki/index.php/SOFT:FSL_FreeSurfer I've tried this and that has helped. (Searching the mail archive I see that at least one other person on this list appears to be doing the same.) My question is, is this a sensible thing to do? I ask because looking at the output I see that at least one of the early processing stages appears to expect a non-stripped brain: mri_watershed -brain_atlas /Applications/freesurfer/average/ RB_all_withskull_2007-08-08.gca transforms/ talairach_with_skull.lta T1.mgz brainmask.auto.mgz Does this suggest that I either should not try the BET trick or that if I do that I should make some modifications to the recon- all scripts? simon _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

---

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

yes, with the current version it seems to almost never fail.

On Fri, 14 Dec 2007, Nick Schmansky wrote:

Simon or Martin,

If you have a chance, can you try the -multistrip option of recon-all with a subject that you know fails the default fs skullstrip? That option launches eight instances of mri_watershed all with different parameter combos, and the best one is selected. If all of those fail, can you send us the failed subject data? We are always looking for test cases.

The command is: recon-all -s <subj> -multistrip

Nick

On Fri, 2007-12-14 at 14:47 +0000, Simon Rushton wrote:

...

I've found it varies. A couple of brains that didn't go through with the FS brain extraction tool went through with BET. Maybe recon-all could include an option to try BET? Obviously have the most reliable tool as default but allow the user to easily test an alternative tool before starting to dig deep into other things?

simon

On 14 Dec 2007, at 09:40, Martin Kavec wrote:

...

In my experience, the FreeSurfer's brain extraction works better than the BET. Although, now with an option of iterative BET in the latest FSL, the outcomes are much better than before. In cases where brain extraction failed in my analysis, I found that it was the bias field correction which didn't do the best job, though the images looked ok. In these cases, even playing with watershed parameters didn't provide statisfactory results. So I tried FLS's bias correction followed by the watershed, and this always helped.

Martin

On Thu, 13 Dec 2007, Bruce Fischl wrote:

...

Hi Simon,

yes, you would need to skip the watershed step (it is our skull stripping), or better yet, replace it with bet, since at some points we need the skull on e.g. to do TIV estimation. Do you find our skullstripping is failing? In our hands the current version is extremely robust - working on 99% of the brains we test.

cheers, Bruce

On Thu, 13 Dec 2007, Simon Rushton wrote:

...

I've been having problems with a few brains (errors during processing). I found advice on the web about running BET (the FSL brain extraction tool) prior to running recon-all http://lcni.uoregon.edu/mediawiki/index.php/SOFT:FSL_FreeSurfer I've tried this and that has helped. (Searching the mail archive I see that at least one other person on this list appears to be doing the same.) My question is, is this a sensible thing to do? I ask because looking at the output I see that at least one of the early processing stages appears to expect a non-stripped brain: mri_watershed -brain_atlas /Applications/freesurfer/average/ RB_all_withskull_2007-08-08.gca transforms/ talairach_with_skull.lta T1.mgz brainmask.auto.mgz Does this suggest that I either should not try the BET trick or that if I do that I should make some modifications to the recon- all scripts? simon _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

---

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

---

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

---

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