tomoprint

Turn cryo-ET tomograms (.mrc) into clean, watertight, 3D-printable relief plates — "moon plate" style, where dense biological features (capsids, cells, viruses) read as craters carved by contrast through depth.

A tomogram is collapsed to a 2D heightmap (a single slice by default, or a slab / min·mean·max projection), conditioned (denoise, contrast, optional invert), and turned into a solid plate: a relief top surface + flat base + side walls, guaranteed manifold and watertight so it prints as-is. Drive it from a scriptable library + CLI, or the desktop GUI with live sliders and a 3D preview.

You can also crop to a region of interest — rectangle, ellipse, or freehand polygon (the plate takes that shape) — and optionally cut the plate into interlocking jigsaw pieces on export, so the print is a puzzle you assemble.

tomogram (Z,Y,X)  ──reduce──▶  heightmap (Y,X)  ──contrast/invert──▶  z = base + relief·v
                                                                              │
                                          watertight relief plate  ◀──mesh────┘
                                          (relief top + base + walls) ──▶ STL / OBJ / PLY

Install (micromamba)

Base Python 3.14 is too new for the VTK/Qt stack, so use a dedicated python=3.11 env:

micromamba create -y -n tomoprint -c conda-forge \
  python=3.11 \
  "numpy>=1.26,<2.3" "scipy>=1.13,<2" "mrcfile>=1.5,<2" "scikit-image>=0.24,<0.27" \
  "vtk>=9.3,<9.7" "pyvista>=0.44,<0.46" "pyvistaqt>=0.11,<0.12" "pyside6>=6.7,<6.10" \
  "pytest>=8" "pytest-cov>=5" "ruff>=0.6"
micromamba run -n tomoprint python -m pip install -e ".[dev]"

pip install -e also pulls the boolean stack (shapely, manifold3d, mapbox-earcut) used for shaped crops and the jigsaw cut, plus fast-simplification for mesh decimation.

The pure-core extra omits VTK/Qt (pip install -e ".[test]") if you only need the library + CLI.

Quickstart

# CLI: middle slice, 200 mm footprint, 6 mm relief, 2 mm base -> STL
micromamba run -n tomoprint tomoprint data/capsid_trim.mrc capsid.stl --report

# Min-intensity projection over +/-5 slices (emphasizes dense capsids), features popped out
micromamba run -n tomoprint tomoprint data/capsid_trim.mrc caps_min.stl \
  --mode min --half-thickness 5 --invert

# Large tomogram: bin x2 to ~256k triangles
micromamba run -n tomoprint tomoprint data/bot_left_trim.mrc bot.stl --bin 2 --report

# Crop to an elliptical region (round plate) centred on a feature
micromamba run -n tomoprint tomoprint data/capsid_trim.mrc disc.stl \
  --crop-shape ellipse --crop-cx 0.5 --crop-cy 0.45 --crop-w 0.6 --crop-h 0.6

# Cut the plate into a 5x4 interlocking jigsaw puzzle (one STL, each piece a separate body)
micromamba run -n tomoprint tomoprint data/capsid_trim.mrc puzzle.stl \
  --jigsaw --jigsaw-cols 5 --jigsaw-rows 4 --jigsaw-kerf 0.2

# Desktop GUI (interactive crop ROI + jigsaw overlay)
micromamba run -n tomoprint tomoprint-gui

# Library
from tomoprint import run_from_file, ReduceParams, GeometryParams
mesh, diag = run_from_file(
    "data/capsid_trim.mrc",
    reduce_p=ReduceParams(mode="slice", index=50),
    geo_p=GeometryParams(footprint_mm=200, relief_depth_mm=6, base_thickness_mm=2),
    out="capsid.stl",
)
print(diag)  # {'watertight': True, 'euler_number': 2, 'volume_mm3': ..., 'n_faces': ...}

Parameters

Group	Option	Default	Meaning
Source	--mode	slice	slice · slab_mean · min · mean · max
	--index	middle	slice / slab centre
	--half-thickness	0	± N slices for slab/projection
	--bin	1	integer downsample (simplify + denoise)
Contrast	--sigma	1.0	gaussian blur (px)
	--pclip-low/high	1 / 99	contrast percentile clip
	--invert	off	flip craters ↔ bumps
Relief	--footprint	200	longest plate side (mm)
	--relief	6	relief depth (mm)
	--base	2	solid base thickness (mm)
Crop	--crop-shape	off	rect · ellipse (polygon is GUI-only)
	--crop-cx/cy	0.5	region centre (0–1 of width/height)
	--crop-w/h	1.0	region size (0–1 of footprint)
Jigsaw	--jigsaw	off	cut the plate into puzzle pieces
	--jigsaw-cols/rows	4 / 3	puzzle grid
	--jigsaw-tab	0.22	tab size (fraction of cell edge)
	--jigsaw-kerf	0.2	gap between pieces (mm) for fit
Mesh	--taubin	0	Taubin smoothing iterations
	--decimate	off	keep-fraction (0–1)

Run tomoprint --help for the full list. The GUI exposes the same parameters as live sliders.

Development

micromamba run -n tomoprint pytest -m "not slow"   # fast unit tests (no VTK/Qt)
micromamba run -n tomoprint pytest -m slow          # real-MRC end-to-end (writes an STL)
micromamba run -n tomoprint ruff check src tests

Architecture: a pure algorithm layer (numpy/scipy/scikit-image/trimesh, no VTK/Qt) under src/tomoprint/, with cli.py and gui/ as thin frontends that call pipeline.py. See pyproject.toml extras [viz]/[gui]/[test]/[dev].