Hello!

I think we have found the issue that is causing the recons to fail with out of memory. If you see the below 2 brains you can see the WM mask going out of bounds. Not sure what could have caused this (perhaps network latency during the editing process?)

What we are looking to see is if there is a way to reliably fix this issue and submit another recon? And any leads as to how we can avoid such situation to occue in the future?

Thanks in advance!

--Sowmya

From: Sambamoorthy, Sowmya
Sent: Thursday, September 18, 2025 1:06 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Cc: Kidwai, Ambereen <Amber.Kidwai@utdallas.edu>
Subject: RE: Recon failing: normal vector of length zero at vertex NNNNN with N faces

Ok, thank you.

From: Nolan, Jackson <JNOLAN5@mgh.harvard.edu>
Sent: Thursday, September 18, 2025 10:57 AM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Cc: Kidwai, Ambereen <Amber.Kidwai@utdallas.edu>
Subject: [Freesurfer] Re: Recon failing: normal vector of length zero at vertex NNNNN with N faces

I don’t see anything in the attached log that looks like a smoking gun.

It is still possible that the process ran out of memory and was killed by the system, but it is not caught in the log file. You can check system logs to help determine if this is the case.

From: Sambamoorthy, Sowmya <Sowmya.Sambamoorthy@UTDallas.edu>
Date: Wednesday, September 17, 2025 at 4:24 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Cc: Kidwai, Ambereen <Amber.Kidwai@utdallas.edu>
Subject: [Freesurfer] Re: Recon failing: normal vector of length zero at vertex NNNNN with N faces

External Email - Use Caution

Thank you for your response, Jackson. That log did run out of resources and could not allocate memory. This is another one that does not have the “cannot allocate memory” error

*************************************************************

INFO: assuming .mgz format

writing corrected surface to 'orig.premesh'

7.3.2

before topology correction, eno=-96 (nv=411988, nf=818040, ne=1230124, g=49)

using quasi-homeomorphic spherical map to tessellate cortical surface...

Correction of the Topology

Finding true center and radius of Spherical Surface...done

Surface centered at (0,0,0) with radius 100.0 in 36 iterations

marking ambiguous vertices...

590641 ambiguous faces found in tessellation

segmenting defects...

17 defects found, arbitrating ambiguous regions...

analyzing neighboring defects...

17 defects to be corrected

0 vertices coincident

reading input surface /cvl/rugg/ruggdata/UTD/EXPT2/FMR/fmri102/freesurfer/ru/fmri102_054/surf/rh.qsphere.nofix...

reading brain volume from brain...

reading wm segmentation from wm...

Reading original properties of orig.nofix

Reading vertex positions of inflated.nofix

Computing Initial Surface Statistics

-face loglikelihood: -0.1953 (-0.0977)

-vertex loglikelihood: -0.2060 (-0.1030)

-normal dot loglikelihood: -2.1836 (-2.1836)

-quad curv loglikelihood: -1.7030 (-0.8515)

Total Loglikelihood : -4.2879

CORRECTING DEFECT 0 (vertices=302859, convex hull=16466, v0=0)

PIDs (2457437 2457506) completed and logs appended.

Linux cvlruggcompute2.utdallas.edu 4.18.0-553.16.1.el8_10.x86_64 #1 SMP Thu Aug 8 17:47:08 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

recon-all -s fmri102_054 exited with ERRORS at Wed Sep 10 20:17:48 CDT 2025

To report a problem, see http://surfer.nmr.mgh.harvard.edu/fswiki/BugReporting

From: Nolan, Jackson <JNOLAN5@mgh.harvard.edu>
Sent: Wednesday, September 17, 2025 3:14 PM
To: Freesurfer support list <freesurfer@nmr.mgh.harvard.edu>
Cc: Kidwai, Ambereen <Amber.Kidwai@utdallas.edu>
Subject: [Freesurfer] Re: Recon failing: normal vector of length zero at vertex NNNNN with N faces

Hello Sowmya,

From the errors at the bottom of the screenshot, it looks like the processes are failing due to insufficient memory. (“error: Cannot allocate memory”)

It also looks like the recons are being run via the ‘reconbatchjobs’ utility, but I might be incorrect about that.

I would suggest running the recons on a machine with more RAM, and/or running the jobs in sequence rather than parallel, so the jobs are not competing for resources.

Best,

Jackson

From: Sambamoorthy, Sowmya <Sowmya.Sambamoorthy@UTDallas.edu>
Date: Wednesday, September 17, 2025 at 3:38 PM
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Cc: Kidwai, Ambereen <Amber.Kidwai@utdallas.edu>
Subject: [Freesurfer] Recon failing: normal vector of length zero at vertex NNNNN with N faces

External Email - Use Caution

Hello!

Need help getting past these errors (attached recon log) for reference. Many of my brains are throwing errors like below:

“normal vector of length zero at vertex 101440 with 6 faces

first tangent vector of length zero at vertex 101440”

Last few lines of recon.log below:

Any help appreciated!

Thank you in advance for your support.

--Sowmya