Hi freesurfers,
To summarize the forwarded thread, some months ago I was experiencing segmentation faults at the mri_segment step in the processing stream, and the fix turned out to be running that particular caselist with v5. My lab is still using 4.2.0 in general, however, since we would have to rerun a couple hundred cases with the new version; thus, we probably won't do an update until a future major version release, like v6 or something.
So my question now is, we're experiencing this same problem with a couple scans in our general data set (my guess is that this is due to a relatively high, i.e. >10,000, number of "sparsely connected voxels"?). We want these scans' FreeSurfer data to be compatible with our other FreeSurfer data, so I was wondering if there is a solution to this problem that doesn't involve running these with v5. Is there any hope for us in that respect?
Thanks, Diandra
-------- Original Message -------- Subject: Re: [Freesurfer] Segmentation fault during mri_segment Date: Fri, 30 Jul 2010 08:19:29 -0400 (EDT) From: Bruce Fischl fischl@nmr.mgh.harvard.edu To: Diandra Lucia diandra@bwh.harvard.edu References: AANLkTimMnZL_OcAvEs-wlMwckBaREviS97cwRqd_7_Dm@mail.gmail.com Pine.LNX.4.62.1007212054060.9035@gate.nmr.mgh.harvard.edu AANLkTil_Tzs_-6y0xbGE3sDjxTMfdxMAFDmdx-P8bJRp@mail.gmail.com Pine.LNX.4.62.1007231724570.30598@gate.nmr.mgh.harvard.edu AANLkTimVoVhDKe0K+3XyO5QpTJ-oncnCLcMh_w=Bx9Ho@mail.gmail.com
Hi Diandra,
you can read the release notes when it's out and decide. There are a bunch of changes, and it depends on what you care about
cheers Bruce On Thu, 29 Jul 2010, Diandra Lucia wrote:
Hi Bruce,
Sorry for the delay -- yeah, I guess I can do that for these cases, as this is actually a collaborator's data. We have such a large dataset internally, though, that rerunning all of our cases would be a pain... But this set is pretty small, so we can run them all on v5. (How improved is v5? Would it be worth considering rerunning >100 cases?)
Thanks, Diandra
On Fri, Jul 23, 2010 at 5:25 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Diandra, that one ran through fine for me. Do you want to wait a couple of weeks for V5 to be out and try it?
Bruce On Thu, 22 Jul 2010, Diandra Lucia wrote:
Hi Bruce,
I sent you a sample case via https://www.nmr.mgh.harvard.edu/facility/filedrop/index.html; please let me know if you need anything else. I just sent it to your address; let me know if I need to share it with others as well.
Thanks so much, Diandra
On Wed, Jul 21, 2010 at 8:54 PM, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
Hi Diandra,
can you send us one of these datasets so we can see if we can track it down?
thanks Bruce On Wed, 21 Jul 2010, Diandra Lucia wrote:
Hi again,
As we process more cases, we've run into four cases that have had the same problem. Each has segfaulted during mri_segment during autorecon2; here is one example:
#@# Mask BFS Mon Jul 19 23:38:35 EDT 2010 /home/pnlstaff/freesurfer/subjects/case2403/mri
mri_mask -T 5 brain.mgz brainmask.mgz brain.finalsurfs.mgz
threshold mask volume at 5 Writing masked volume to brain.finalsurfs.mgz...done. #-------------------------------------------- #@# WM Segmentation Mon Jul 19 23:38:41 EDT 2010
mri_segment brain.mgz wm.seg.mgz
doing initial intensity segmentation... using local statistics to label ambiguous voxels... computing class statistics for intensity windows... WM (105.0): 105.7 +- 6.1 [80.0 --> 125.0] GM (72.0) : 71.1 +- 8.6 [30.0 --> 96.0] setting bottom of white matter range to 79.6 setting top of gray matter range to 88.2 doing initial intensity segmentation... using local statistics to label ambiguous voxels... using local geometry to label remaining ambiguous voxels...
reclassifying voxels using Gaussian border classifier...
removing voxels with positive offset direction... smoothing T1 volume with sigma = 0.250 removing 1-dimensional structures... 12443 sparsely connected voxels removed... thickening thin strands.... Segmentation fault Linux compute-2-7.local 2.6.9-34.ELsmp #1 SMP Fri Feb 24 16:56:28 EST 2006 x86_64 x86_64 x86_64 GNU/Linux
recon-all exited with ERRORS at Mon Jul 19 23:40:24 EDT 2010
They all segfaulted at exactly the same place. These were run with freesurfer-Linux-rh9-stable-pub-v4.2.0 (without the new version of the mri_ca_normalize you sent me recently). Please let me know if you need anything else (data, etc) to help troubleshoot this.
Thanks, Diandra