Habenula Segmentation

---

Automated Human Habenula Segmentation Program

---

Requirements

• Python: version 2.7
• Python libraries: numpy, scipy, nibabel
• FSL

---

Input Files

• nifti1 is the only supported file format.
• AC-PC aligned T1w image
• T2w image that is registered to T1w image
• Habenula center positions:
  • a nifti1 file containing two markers - a voxel in the right habenula as 1 and a vxoel in the left habenula as 2
  • or, MNI152_0.7mm-to-ACPC warp file. (MNINonLinear/xfms/standard2acpc_dc.nii.gz if you used HCP pipeline)
• Optional: T1w/T2w which can be generated using fslmaths in FSL: fslmaths T1w -div T2w T1wDivT2w
• Note: The PreFreeSurfer in Human Connectome Project (HCP) Pipelines (https://github.com/Washington-University/Pipelines) is recommended for the AC-PC alignment and T2w-to-T1w registration.

---

How to Run

method 1: using command line

if you have habenula center position file, center.nii.gz

seg_hb.py -1 T1w.nii.gz -2 T2w.nii.gz  -c center.nii.gz [-m T1wDivT2w.nii.gz] [-o output.nii.gz]

if you have warp file, standard2acpc_dc.nii.gz

seg_hb.py -1 T1w.nii.gz -2 T2w.nii.gz -w standard2acpc_dc.nii.gz [-m T1wDivT2w.nii.gz] [-o output.nii.gz]

options

seg_hb.py -1 T1w.nii.gz -2 T2w.nii.gz [-m T1wDivT2w.nii.gz] [-o output.nii.gz] \
          [-c center.nii.gz] [-w warp_MNI152_0.7_to_ACPC] \
          [-t TEMPLATE_THRESHOLD] [-v VERBOSE] \
          [--min_volume MIN_VOLUME] [--t1min T1MIN] [--t1max T1MAX] \
          [--t2min T2MIN] [--t2max T2MAX] [--alpha ALPHA] \
          [--num_growing GROWTH]

---

method 2: using python script

import segment
segment.segment_hb(
        'T1w_filename',
        'T2w_filename',
        'myelin_filename',
        'hb_center_filename',
        'output_filename',
        min_volume = 80,
        t1min = None,
        t1max = None,
        t2min = 10,
        t2max = None,
        sig_factor = 0.9,
        growth = 5,
        verbose = False,
        return_volume=True,
        return_step_volume=False,
        return_center_of_mass=False,
        return_hb_index=False,
        return_hb_neighbor_index=False
        )

input:

• t1, t2, myelin: T1w, T2w, Myelin Nifti1 images
• hb center: mask Nifti1 file indicating Hb centers (1: right Hb, 2: left Hb)

output:

• output filename: output (segmentation) Nifti1 image file name

parameters:

• min_volume: minimum ROI volume (mm^3) for each Hb. Default 80.
• t1min, t1max, t2min, t2max: first thresold values for T1w, T2w.
• sig_factor: Factor of sigma for myelin threshold for Hb segmentation. Default 0.9. The higher factor, the smaller Hb.
• growth: Number of iteration for region growth method. 0 for disable it. Default 5.
• verbose: If True, Display details. Default False.

return:

• [color1 Hb volume, color2 Hb volume], [color1 Hb partial volume, color2 Hb partial volume], [list(left Hb index), list(right Hb index)]

Habenula Template

We release habenula templates: habenula_template directory.

• habenula_template_HCP_50_native_[left/right].nii.gz : habenula template in the MNI space generated from HCP 50 subjects' habenula segmentation in the native space.

---

Reference:

[1] Kim et al, Human habenula segmentation using myelin content. Neuroimage, 2016, 130 : 145-156 http://www.ncbi.nlm.nih.gov/pubmed/26826517

[2] Kim et al, Reproducibility of myelin content‐based human habenula segmentation at 3 Tesla. Human Brain Mapping, 2018, 39 : 3058-3071 https://doi.org/10.1002/hbm.24060