Inconsistent volume info

[EmmaRenauld]

Bug Report

Hi! I just build a singularity using the latest version (docker://deepmi/fastsurfer:latest), ran fastsurfer on a subject. Everything ran well, but then, when trying to view the results with:

    s=$SUBJECTS_DIR/my_subj
    freeview -v $s/mri/T1.mgz \
         -f $s/surf/lh.pial.T1:edgecolor=red $s/surf/lh.white:edgecolor=blue \
            $s/surf/rh.pial.T1:edgecolor=red $s/surf/rh.white:edgecolor=blue 

I get 4 times this warning:

WARNING: the source volume info is not consistent between the info contained
WARNING: in the surface data (0.696320,-10.885422,-0.250122) and that of the transform (0.696324,-10.885425,-0.250123).

It makes the additional option --ss file.png (save screenshot as a png) crash.

Description

Which functionality, script, interface or interaction does this bug concern?

run_fastsurfer.sh

Expected Behavior

Make sure the output is compatible with all other freesurfer tools.

Steps to Reproduce

1. Run freesurfer as in the documentation example:

singularity exec \
        --nv --no-mount home,cwd -e  \
        -B $outputs:$outputs -B $license:$license -B $link_path:$link_path \
        ./fastsurfer-gpu.sif \
        /fastsurfer/run_fastsurfer.sh \
                --fs_license $license \
                --t1 $link_file \
                --sid sub-${subj}_FastS --sd $outputs \
                --3T --threads 4 --vox_size 1 --device cuda

2. Run freeview (see command above)

Log Files / Screenshots

Will be shared privately upon request.

(If you can/want to share data privately with the FastSurfer team, you may use the following dropbox:
https://nextcloud.dzne.de/index.php/s/Z2qtHW8c7p3NSJ5)

Environment

Please describe how and which version of FastSurfer you are using. For this fill out the list below!

• Installation type: Apptainer

• FastSurfer Version: run_fastsurfer.sh --version all gives full version information, see also scripts/BUILD.log.

  • 2.5.4+cdfccea (stable)
  • 625a4b79defa6f2bf1c5d07777591643 /fastsurfer/checkpoints/CerebNet_axial_v1.0.0.pkl
  • 7d400765666150f14389764703f837f2 /fastsurfer/checkpoints/HypVINN_sagittal_v1.1.0.pkl
  • f28ab87df75492ad91f9ff5e49a2583c /fastsurfer/checkpoints/aparc_vinn_sagittal_v2.0.0.pkl
  • 8e9ebc9a1e0d84810c1a5d844b52c153 /fastsurfer/checkpoints/CerebNet_sagittal_v1.0.0.pkl
  • f16e78dd069eed2ec659674e0fdb697b /fastsurfer/checkpoints/aparc_vinn_axial_v2.0.0.pkl
  • 619578f4af014c573acd4f7eb8017ea2 /fastsurfer/checkpoints/CerebNet_coronal_v1.0.0.pkl
  • 97fb44467f12f8c673684a0b61169f31 /fastsurfer/checkpoints/FastSurferCC_localization_v1.0.0.pkl
  • 01b46b5aff0b5a718a410db769d353fc /fastsurfer/checkpoints/FastSurferCC_segmentation_v1.0.0.pkl
  • ab1339037fe72a96474918e3e43230ea /fastsurfer/checkpoints/HypVINN_axial_v1.1.0.pkl
  • fa699af3d60ca9ee06950e62ed04c288 /fastsurfer/checkpoints/aparc_vinn_coronal_v2.0.0.pkl
  • 55e30ae9c3e888bd3be3fccfbc7fbe04 /fastsurfer/checkpoints/HypVINN_coronal_v1.1.0.pkl

• OS: Linux Version / Windows / Mac: Linux, on a cluster (Alliance Canada)

Additional Context

Add any other context and comments about the problem here.

[dkuegler]

Hi,

so the differences between the volume infos are down to machine precision. I would really characterize this more as a freeview issue than a FastSurfer issue, nonetheless it needs fixing.

As a first step, I see that you are using T1.mgz as a reference image. I do not quite know, why you would want to use T1.mgz. In general, you would want to use orig.mgz, nu.mgz or norm.mgz. T1.mgz has additional "cleanup" operations performed inside that might now be what you want, for example filling of white matter. If you are after bias-field and intensity normalization, those are already part of the nu and norm images in FastSurfer.

So, please check, if the problem persists, if you use one of these three images.

Cheers

[EmmaRenauld]

Hi,

You're right, thanks. I tried with orig, nu and norm. The warning persists in all cases.

However, today, the --ss option did not crash anymore. I'm working on a cluster, so behaviour is probably more unstable than with a local installation. So this becomes a warning only issue, which is better.