Superficial White Matter

Citation

DeKraker, J., Cruces, R., & Hwang, Y. (2024). Superficial White Matter. Zenodo. https://doi.org/10.5281/zenodo.11510179

Description

Generates surfaces at various white matter depths (default 1, 2, and 3 milimiters). The depths are calculated based on the real world image resolution voxel size and transformed to milimiters.

Method

This is done by first computing a Laplace field over white matter (cortex to subcortex+ventricles), and then shifting an exiting white matter surface along that gradient. Stopping conditions are set by geodesic distance travelled.

White is the original wm surface, red, orange and yellow are depths 1mm, 2mm, and 3mm accordingly.

Installation

git clone https://github.com/jordandekraker/superficial-white-matter.git
pip install superficial-white-matter/

Usage with Freesurfer/Fastsurfer (example)

The code expects standard nifti orientation, in which the resolution is the diagonal of the header affine matrix. Running the inputs through fslreorient2std will ensure that everything is calculated correctly.

# This is the freesurfer/fastSurfer Subject directory
SUBJECTS_DIR=<path to surface subjects directory FreeSurfer/FastSurfer>

# Subject to process
SUBJECT=sub-01

# Output directory
OUT=<path to output directory>

# Convert segmentation to NIFTI
aparc_aseg=${OUT}/${SUBJECT}_aparc+aseg.nii.gz
mri_convert ${SUBJECTS_DIR}/${SUBJECT}/mri/aparc+aseg.mgz ${aparc_aseg}

# Reorient to standard
fslreorient2std ${aparc_aseg} ${aparc_aseg}

# 1. Calculate the Laplace field
python sWM/laplace_solver.py \
  ${aparc_aseg} \
  ${OUT}/${SUBJECT}_laplace-wm.nii.gz

# 2. Generate the surfaces for each hemisphere
for hemi in lh rh; do
  # White matter surface
  WM=${SUBJECTS_DIR}/${SUBJECT}/surf/${hemi}.white
  WM_gii=${OUT}/${SUBJECT}_hemi-${hemi}_label-white.surf.gii

  # Convert white matter to GIFTI
  mris_convert ${WM} ${WM_gii}

  # Calculate the SWM surfaces
  python sWM/surface_generator.py \
    "${WM_gii}" \
    ${OUT}/${SUBJECT}_laplace-wm.nii.gz \
    ${OUT}/${SUBJECT}_hemi-${hemi}_label-sWF_depth-
done

If you ran micapipe v0.2.0 or higher check the example script: example_usage.sh SWM is implemented in micapipe v0.2.3

`laplace_solver.py`

Solves Laplace equation over the domain of white matter.

Using grey matter as the source and ventricles as the sink.
Inputs are expected to be Free/FastSurfer aparc+aseg.mgz in .nii.gz format

Parameters
----------
NIFTI  :    str
            Parcellation file generated by Freesurfer/fastsurfre in nii.gz format (from mri/aparc+aseg.mgz).
NIFTI  :    str
            Output laplacian file path (nii.gz)

Returns
-------
NIFTI
    Laplacian image (nii.gz)

Usage
-----
laplace_solver.py aparc+aseg.nii.gz laplace-wm.nii.gz

`surface_generator.py`

Shifts a white matter surface inward along a Laplace field

Parameters
----------
GIFTI  :    str
            White matter surface in GIFTI format (surf.gii)
NIFTI  :    str
            laplacian image generated by laplace_solver.py
OUTPUT :    str
            path and name to the output surfaces
DEPTHS :    list [int | float] (OPTIONAL)
            DEFAULT=[1,2,3] List of depths to sample (in voxels)

Returns
-------
NIFTI
    a list of strings representing the header columns

Usage
-----
surface_generator.py hemi-L_label-white.surf.gii laplace-wm.nii.gz hemi-L_label-sWF_depth- 1,2,3