pyclarcs 0.4.0

3-D anatomical surface analysis: symmetry, non-rigid registration, atlas construction and asymmetry mapping

🧠 pyclarcs

Python toolkit for the automated analysis of 3-D anatomical surfaces, with a focus on endocranial and bilateral structures.

clarcs is a command-line tool built around two core algorithms:

• Symmetry-plane estimation — Combès et al., CVPR 2008
• Non-rigid surface registration — EM-ICP with symmetric correspondences, TGD geodesic prior, and RKHS Wu-kernel M-step (Combès & Prima, CVIU 2019)

Combined into a full population-analysis pipeline: symmetry → alignment → non-rigid registration → atlas → asymmetry mapping.

---

Table of contents

• Installation
• Supported formats
• Commands
• Typical pipelines
  • Preprocessing and registration
  • Atlas construction
  • Asymmetry analysis on an atlas
• Python API
• Data — PaleoBRAIN dataset
• Registration benchmark
• Repository structure
• Scientific references
• Licence

---

Installation

pip install pyclarcs

Or from source:

git clone https://github.com/rdebroiz/pyclarcs
cd pyclarcs
pip install -e ".[dev]"

Dependencies (installed automatically):

Package	Role
numpy ≥ 1.21	Numerics
scipy ≥ 1.7	KD-tree, sparse linear algebra, graph shortest paths
vtk ≥ 9.0	Surface I/O and mesh processing
numba ≥ 0.57	JIT-compiled kernels (4× speedup on EM stages)
click ≥ 8.0	CLI framework

---

Supported formats

Format is inferred from the file extension.

Extension	Format	Read	Write
.vtk	VTK legacy PolyData (ASCII or binary)	✓	✓
.vtp	VTK XML PolyData	✓	✓
.vtu	VTK XML UnstructuredGrid (converted to PolyData)	✓	—
.ply	Stanford PLY	✓	✓
.stl	STereoLithography	✓	✓
.obj	Wavefront OBJ	✓	✓

Note — Asymmetry and deformation fields (VECTORS point data) require .vtk or .vtp.

---

Commands

Commands are grouped by purpose.

Preprocessing

Command	Description
clarcs reorient	Permute coordinate axes
clarcs recenter	Align symmetry plane to x = 0
clarcs centerofmass	Translate to match a reference's centre of mass
clarcs normalize	Uniformly scale + translate to match a reference
clarcs downsample	Decimate to a target vertex count or ratio

Symmetry analysis

Command	Description
clarcs symplane	Estimate the bilateral symmetry plane
clarcs mirror	Reflect a surface across its symmetry plane
clarcs asymmetry	Compute the pointwise asymmetry field

Non-rigid registration

Command	Description
clarcs nlregister	Non-rigid EM-ICP registration onto a reference

Atlas & population analysis

Command	Description
clarcs atlas	Build a mean shape atlas from a population
clarcs project-asym	Build atlas + project per-subject asymmetries onto it

---

Typical pipelines

Preprocessing and registration

# 1. Align symmetry plane to x = 0
clarcs recenter target.vtk target-rc.vtk --save-plane

# 2. Match size and centre of mass to the reference
clarcs normalize target-rc.vtk target-rcs.vtk --target reference.vtk

# 3. Non-rigid EM-ICP registration
clarcs nlregister target-rcs.vtk reference.vtk target-registered.vtk \
                  --deformation target-deformation.vtk

Atlas construction

# Build a mean shape atlas from a directory of preprocessed surfaces
clarcs atlas subjects/ atlas.vtk --save-registered

# Lower resolution first (useful for large surfaces)
clarcs downsample subjects/B01.ply subjects/B01-5k.ply --target-n 5000

Asymmetry analysis on an atlas

# All-in-one: build atlas, compute subject asymmetries, project onto atlas
clarcs project-asym subjects/ mean-asym.vtp --save-atlas atlas.vtp \
       --save-stats --save-individual

# Incremental (reuse pre-computed atlas and asymmetry fields)
clarcs asymmetry B01.ply B01-asym.vtk
clarcs project-asym subjects/ mean-asym.vtp \
       --atlas atlas.vtp \
       --registered-dir registered/ \
       --asymmetry-dir  asymmetry/

data/run_pipeline.py automates the preprocessing + registration sequence on the bundled test surfaces:

python data/generate_samples.py   # create test surfaces (once)
python data/run_pipeline.py results/

---

Python API

Symmetry plane

from pyclarcs.io import load_surface, save_plane_vtk
from pyclarcs.principal_axes import best_principal_axis_plane
from pyclarcs.coarse import coarse_symmetry
from pyclarcs.fine import em_icp_sym, em_icp_sym_corres

points, polygons = load_surface("surface.vtk")

plane = best_principal_axis_plane(points)
plane = coarse_symmetry(points, plane)
plane = em_icp_sym(points, plane)
plane = em_icp_sym_corres(points, plane)

plane.save("plane.pl")

Non-rigid registration

from pyclarcs.io import load_surface, load_surface_with_normals, save_surface
from pyclarcs.nonrigid import register, apply_deformation

mov_pts, mov_poly, mov_normals = load_surface_with_normals("target.vtk")
ref_pts, ref_poly, ref_normals = load_surface_with_normals("reference.vtk")

# Symmetric correspondences + TGD prior + RKHS M-step are all on by default
def_field = register(
    mov_pts, mov_normals,
    ref_pts, ref_normals,
    mov_poly, ref_poly,
)

warped = apply_deformation(mov_pts, def_field)
save_surface("registered.vtk", warped, mov_poly)

Atlas construction

from pyclarcs.io import load_surface_with_normals, save_surface
from pyclarcs.atlas import build_atlas

subjects = [load_surface_with_normals(p) for p in sorted(paths)]

atlas_pts, atlas_poly, registered = build_atlas(subjects, atlas_iter=3)
save_surface("atlas.vtk", atlas_pts, atlas_poly)

Asymmetry projection

from pyclarcs.io import load_vector_field
from pyclarcs.atlas import project_asymmetry_to_atlas
import numpy as np

# registered[i]: atlas vertices in subject i's space (from build_atlas)
# asym_fields[i]: asymmetry field at subject i's vertices (from clarcs asymmetry)
projected = project_asymmetry_to_atlas(registered, asym_fields, subject_pts)

mean_asym = np.stack(projected).mean(axis=0)   # (N, 3) mean field on atlas

---

Data — PaleoBRAIN dataset

75 brain surfaces (B01–B75) and 75 endocasts (E01–E75) from the PaleoBRAIN project (Balzeau A., 2025, doi:10.48579/PRO/KZMMLM).

# Download first 10 subjects (brains + endocasts), resample to 5 000 vertices
python data/download_paleobrain.py --n 10 --target-n 5000

# Brains only, all 75
python data/download_paleobrain.py --type brain

---

Registration benchmark

MNI pial endocranium (10 k vertices), synthetic deformation of 8 Gaussian bumps × 5 mm amplitude. RMS before: 2.28 mm.

Configuration	RMS after	Improvement
Baseline (Laplacian)	0.333 mm	85.4 %
+ Symmetric correspondences + TGD	0.267 mm	88.3 %
+ RKHS M-step (default)	0.190 mm	91.6 %

See docs/nlregister.md for full details and parameters.

---

Repository structure

pyclarcs/
├── pyproject.toml
├── README.md
├── docs/                           ← per-command documentation
│   ├── reorient.md
│   ├── symplane.md
│   ├── recenter.md
│   ├── centerofmass.md
│   ├── normalize.md
│   ├── mirror.md
│   ├── downsample.md
│   ├── nlregister.md
│   ├── atlas.md
│   ├── asymmetry.md
│   └── project-asym.md
├── data/
│   ├── generate_samples.py         ← generate synthetic + MNI test surfaces
│   ├── download_paleobrain.py      ← download PaleoBRAIN dataset
│   ├── benchmark_compare.py        ← compare clarcs / CPD / BCPD
│   └── run_pipeline.py             ← run the full pipeline end-to-end
└── src/pyclarcs/
    ├── _cli.py                     ← CLI entry-point (clarcs command)
    ├── symmetry.py                 ← SymmetryPlane class
    ├── principal_axes.py           ← inertia tensor initialisation
    ├── io.py                       ← multi-format surface I/O (VTK 9+)
    ├── coarse.py                   ← coarse ICP with trimmed estimator
    ├── fine.py                     ← EM-ICP annealing + doubly-stochastic
    ├── alignment.py                ← rigid transforms (recenter, rescale, …)
    ├── mesh.py                     ← adjacency, TGD, Wu kernel graph
    ├── nonrigid.py                 ← register(), _em_icp(), _build_hierarchy()
    ├── atlas.py                    ← build_atlas(), project_asymmetry_to_atlas()
    └── _numba_kernels.py           ← JIT-compiled inner loops (internal)

---

Scientific references

Combès B., Hennessy R., Waddington J., Roberts N., Prima S. Automatic symmetry plane estimation of bilateral objects in point clouds. IEEE CVPR 2008. Anchorage, United States.

Abadie A., Combès B., Haegelen C., Prima S. CLARCS, a C++ Library for Automated Registration and Comparison of Surfaces: Medical Applications. MeshMed @ MICCAI 2011. Toronto, Canada, pp. 117–126.

Combès B., Prima S. New algorithms for the deformable registration of brain images. Medical Image Analysis / CVIU 2019.

Balzeau A. (2025). Database of 75 endocasts and 75 brains obtained on the same sample of volunteers. doi:10.48579/PRO/KZMMLM

---

Licence

MIT