Corpus Callosum Pipeline¶
A deep learning-based pipeline for automated segmentation, analysis, and shape analysis of the corpus callosum in brain MRI scans. Also segments the fornix, localizes the anterior and posterior commissure (AC and PC) and standardizes the orientation of the brain.
Overview¶
This pipeline combines localization and segmentation deep learning models to:
Detect AC (Anterior Commissure) and PC (Posterior Commissure) points
Extract and align midplane slices
Segment the corpus callosum
Perform advanced morphometry for corpus callosum, including subdivision, thickness analysis, and various shape metrics
Generate visualizations and measurements
The output files are described here. The structure of the JSON files describing corpus callosum measures is documented below. Advanced options, like custom subdivision schemes and quality control are described in the FastSurfer-CC documentation.
JSON Output Structure¶
The pipeline generates two main JSON files with detailed measurements and analysis results:
stats/callosum.CC.midslice.json (Middle Slice Analysis)¶
This file contains measurements from the middle sagittal slice and includes:
Shape Measurements (single values):¶
total_area: Total corpus callosum area (mm²)total_perimeter: Total perimeter length (mm)circularity: Shape circularity measure (4π × area / perimeter²)cc_index: Corpus callosum shape index (length/width ratio)midline_length: Length along the corpus callosum midline (mm)curvature: Average curve of the midline (degrees), measured by angle between its sub-segmentscurvature_body: Average curve of the center 65% of the midline (degrees), measured by angle between its sub-segments
Subdivisions¶
areas: Areas of CC using an improved Hofer-Frahm sub-division method (mm²). This gives more consistent sub-segments while preserving the original ratios.curvature_subsegments: Average curve in the CC subsegments (see ‘curvature’)
Thickness Analysis:¶
thickness: Average corpus callosum thickness (mm)thickness_profile: Thickness profile (mm) of the corpus callosum slice (100 thickness values by default, listed from anterior to posterior CC ends)
Volume Measurements:¶
cc_num_voxel: Segmentation-based (masks) CC voxel count within a 5mm slab around the midsagittal plane (partial voxels at the edges are weighted to achieve exactly 5mm width). Multiply byvoxel_volumeto get the volume in mm³.cc_volume: Surface-based (contour) CC volume estimate in mm³, computed from the CC contours across all valid slices assuming 5mm slab width. Only reliable whencc_num_failed_slicesis 0.nullif fewer than 2 contour slices processed successfully.
Anatomical Landmarks:¶
All anatomical landmarks are given image voxel coordinates (LIA orientation)
ac_center: Anterior commissure coordinates in original image space (orig.mgz)pc_center: Posterior commissure coordinates in original image space (orig.mgz)ac_center_oriented_volume: AC coordinates in standardized space (orient_volume.lta)pc_center_oriented_volume: PC coordinates in standardized space (orient_volume.lta)ac_center_upright: AC coordinates in upright space (cc_up.lta)pc_center_upright: PC coordinates in upright space (cc_up.lta)
stats/callosum.CC.all_slices.json (Multi-Slice Analysis)¶
This file contains comprehensive per-slice analysis when using --slice_selection all:
Global Parameters:¶
slices_in_segmentation: Total number of slices in the segmentation volumevoxel_size: Voxel dimensions [x, y, z] in mmvoxel_volume: Volume of a single voxel in mm³cc_num_failed_slices: Number of slices for which surface processing failedsubdivision_method: Method used for anatomical subdivisionnum_thickness_points: Number of points used for thickness estimationsubdivision_ratios: Subdivision fractions used for regional analysiscontour_smoothing: Gaussian sigma used for contour smoothingslice_selection: Slice selection mode used
Per-Slice Data (slices array):¶
Each slice entry contains the shape measurements, thickness analysis and sub-divisions as described above.