Hi infant freesurfer gurus,

I need help with infant_recon_all.

I have existing third-party segmentations for six month old babies' T1w's.

I have relabeled them with label indices that bibsnet would output.

Then I tried running nibabies like so:

docker run --rm -i -u 1697:10 -v /usr/local/pkg/freesurfer7.3/.license:/opt/freesurfer/license.txt:ro -v /my_file_system/rawdata:/data:ro -v /my_file_system/derivatives/nibabies:/out -v /my_file_system/work/nibabies/sub-01:/work -v /my_file_system/derivatives/bibsnet:/deriv1 nipreps/nibabies:25.2.1 /data /out participant --participant-label 01 -s 6mo -w /work --verbose --derivatives /deriv1 --anat-only --fd-spike-threshold 0.2 --age-months 6 --surface-recon-method infantfs --cifti-output 91k --debug registration --project-goodvoxels --multi-step-reg --norm-csf --omp-nthreads 8 --mem 300000 --nprocs 8 --stop-on-first-crash --output-spaces fsLR MNIInfant:cohort-3 MNI152NLin6Asym:res-2 anat

It runs 30-60 minutes and seems to make some surfaces before bailing like so:

      writing spherical brain to /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/surf/lh.qsphere.nofix

      spherical transformation took 0.0770 hours

      FSRUNTIME@ mris_sphere  0.0770 hours 8 threads

      #VMPC# mris_sphere VmPeak  929720

      mris_sphere done

      /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/surf /work/nibabies_26_0_wf/single_subject_sub-01_ses-6mo_wf/infant_single_t1w_fit_wf/infantfs_surface_recon_wf/reconall 

      cp /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/surf/lh.qsphere.nofix lh.qsphere.nofix

      cp /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/surf/rh.qsphere.nofix rh.qsphere.nofix

      /work/nibabies_26_0_wf/single_subject_sub-01_ses-6mo_wf/infant_single_t1w_fit_wf/infantfs_surface_recon_wf/reconall 

      mri_or /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.right.41.mgz /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.left.2.mgz /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/wo

rk/aseg.16.mgz /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/wm.nonscaled.mgz

      processing 3 input files

      processing input volume 1 of 3: /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.right.41.mgz

      processing input volume 2 of 3: /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.left.2.mgz

      processing input volume 3 of 3: /out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.16.mgz

      error: mghRead(/out/sourcedata/freesurfer/sub-01_ses-6mo/work/sub-01_ses-6mo/work/aseg.16.mgz, -1): could not open file

      Segmentation fault (core dumped)

      

I suspect aseg.16.mgz is the Brain-Stem, and these third-party segmentations did not separate the brainstem into pons, medulla, and midbrain, so I left the brainstem unlabeled.  But after running into this issue, I tried relabeling the brainstem to label index 16 and ran nibabies again -- same error.

I have been looking at these web pages for possible work-arounds, but I could use some help, if you have any ideas:

https://github.com/freesurfer/freesurfer/blob/b4463b7610fb590aa48321dfcfaafda4197044c4/infant/infant_recon_all

https://github.com/sheremat/freesurfer/blob/master/mris_anatomical_stats/mris_anatomical_stats.c

Sorry for the long message and odd question, but as you know this timepoint is very challenging, and the segmentations are otherwise promising.

Thank you for any ideas you have for working around this issue,

Donna