Hypothalamus Sub-Segmentation (HypVINN)

The hypothalamus module (hypothal): a deep learning network (HypVINN) that segments the hypothalamus into its subunits together with adjacent structures, and computes summary statistics for them. It can use a T2-weighted image in addition to the T1-weighted one.

What it computes

  • A sub-segmentation of the hypothalamus and adjacent structures, including the third ventricle, the mammillary bodies, the fornix and the optic tracts (hypothalamus.HypVINN.nii.gz), and a mask of it.

  • Summary statistics for these structures (hypothalamus.HypVINN.stats), based on the bias-field corrected T1-weighted image.

  • With a T2-weighted image: the bias-field corrected T2 image and its registration to the T1 image.

  • With --qc_snap: a snapshot of the segmentation for visual quality control.

The output files are listed in the output files overview.

What it needs

  • A T1-weighted image is highly recommended, see the requirements to input images. Voxel sizes down to 0.7 mm are supported natively (smaller voxels are experimental).

  • Optionally a T2-weighted image of the same subject. FastSurfer registers it to the T1 image with FreeSurfer tools, so a T2 image also needs a FreeSurfer license.

  • Bias-field corrected images, which the segmentation computes for you. With --no_biasfield, the module expects images that were corrected beforehand.

The module adds a few minutes on a GPU to the segmentation.

Options

The module runs by default, as part of the segmentation. The options of run_fastsurfer.sh for it:

  • --no_hypothal: skip this module.

  • --t2 <t2_path>: also use a T2-weighted image.

  • --reg_mode <coreg|robust|none>: how the T2 image is registered to the T1 image: coreg (default, FreeSurfer’s mri_coreg), robust (mri_robust_register), or none if both images are already co-registered.

  • --qc_snap: create quality control snapshots in qc_snapshots.

Usage:

run_fastsurfer.sh --sd <subjects_dir> --sid <subject_id> --t1 <t1_path> \
    --t2 <t2_path> --fs_license <freesurfer_license_path>

All options are described in the run_fastsurfer.sh reference. To run the network script directly, for example on an existing FastSurfer subject that was processed without this module, see HypVINN in the command reference.

References

If you use the hypothalamus sub-segmentation in your research, please cite:

  • Estrada S, Kuegler D, Bahrami E, Xu P, Mousa D, Breteler MMB, Aziz NA, Reuter M. FastSurfer-HypVINN: Automated sub-segmentation of the hypothalamus and adjacent structures on high-resolutional brain MRI. Imaging Neuroscience 1 (2023), 1–32. doi:10.1162/imag_a_00034