phenotastic 0.5.1

3D plant phenotyping.

Documentation

3D plant phenotyping package for segmentation of early flower organs (primordia) from shoot apical meristems in 3D images.

Features

• 3D Image Contouring: Morphological active contour methods for extracting surfaces from 3D image stacks

• Mesh Processing: Smoothing, remeshing, and repair operations for 3D meshes

• Domain Segmentation: Curvature-based segmentation of meshes into regions (domains)

• Pipeline System: Configurable recipe-style pipelines for reproducible workflows

Installation

uv pip install phenotastic

Or install from source:

git clone https://github.com/supersubscript/phenotastic.git
cd phenotastic
uv pip install -e ".[dev]"

Quick Start

Using the Python API

from phenotastic import PhenoMesh, Pipeline, load_preset
import pyvista as pv

# Load a mesh
polydata = pv.read("my_mesh.vtk")
mesh = PhenoMesh(polydata)

# Process with the default pipeline
pipeline = load_preset()
result = pipeline.run(mesh)

# Access results
print(f"Mesh has {result.mesh.n_points} points")
print(f"Found {len(result.domains.unique())} domains")

Using the CLI

# Run with default pipeline
phenotastic run image.tif --output results/

# Run with custom config
phenotastic run image.tif --config my_pipeline.yaml

# Generate a config template
phenotastic init-config my_pipeline.yaml

# List available operations
phenotastic list-operations

# List available presets
phenotastic list-presets

# Validate configuration
phenotastic validate my_pipeline.yaml

# View a mesh interactively
phenotastic view mesh.vtk --scalars curvature

Pipeline Configuration

Phenotastic uses a recipe-style YAML configuration for defining pipelines. Each step specifies an operation name and optional parameters.

Example Configuration

steps:
  # Create mesh from contour
  - name: create_mesh
    params:
      step_size: 1

  # Smoothing
  - name: smooth
    params:
      iterations: 100
      relaxation_factor: 0.01

  # Remesh to regularize faces
  - name: remesh
    params:
      n_clusters: 10000

  # More smoothing
  - name: smooth
    params:
      iterations: 50

  # Domain segmentation
  - name: compute_curvature
    params:
      curvature_type: mean

  - name: segment_domains

  - name: merge_small
    params:
      threshold: 50

  - name: extract_domaindata

Default Pipeline

Phenotastic provides a default pipeline that includes the full workflow from 3D image to domain analysis. The default pipeline is automatically used when calling load_preset() without arguments or when running the CLI.

Available Operations

Image/Contour Operations

• contour: Generate binary contour from 3D image using morphological active contours

• create_mesh: Create mesh from contour using marching cubes

• create_cellular_mesh: Create mesh from segmented image (one mesh per cell)

Mesh Processing Operations

• smooth: Laplacian smoothing

• smooth_taubin: Taubin smoothing (less shrinkage than Laplacian)

• smooth_boundary: Smooth only boundary edges

• remesh: Regularize faces using ACVD algorithm

• decimate: Reduce mesh complexity by removing faces

• subdivide: Increase mesh resolution by subdividing faces

• repair_holes: Fill small holes in the mesh

• repair: Full mesh repair using MeshFix

• make_manifold: Remove non-manifold edges

• filter_curvature: Remove vertices outside curvature threshold range

• remove_normals: Remove vertices based on normal angle

• remove_bridges: Remove triangles where all vertices are on the boundary

• remove_tongues: Remove tongue-like artifacts

• extract_largest: Keep only the largest connected component

• clean: Remove degenerate cells

• triangulate: Convert all faces to triangles

• compute_normals: Compute surface normals

• flip_normals: Flip all surface normals

• correct_normal_orientation: Correct normal orientation relative to an axis

• rotate: Rotate mesh around an axis

• clip: Clip mesh with a plane

• erode: Erode mesh by removing boundary points

• ecft: ExtractLargest, Clean, FillHoles, Triangulate (combined operation)

Domain Operations

• compute_curvature: Compute mesh curvature (mean, gaussian, minimum, maximum)

• filter_scalars: Apply filter to curvature field (median, mean, minmax, maxmin)

• segment_domains: Create domains via steepest ascent on curvature field

• merge_angles: Merge domains within angular threshold from meristem

• merge_distance: Merge domains within spatial distance threshold

• merge_small: Merge small domains to their largest neighbor

• merge_engulfing: Merge domains mostly encircled by a neighbor

• merge_disconnected: Connect domains isolated from meristem

• merge_depth: Merge domains with similar depth values

• define_meristem: Identify the meristem domain

• extract_domaindata: Extract geometric measurements for each domain

PhenoMesh Class

PhenoMesh extends PyVista’s PolyData class, adding convenient methods for 3D plant phenotyping workflows. It can be used anywhere a PolyData is expected.

from phenotastic import PhenoMesh
import pyvista as pv

# Create from PyVista mesh
mesh = PhenoMesh(pv.Sphere())

# PhenoMesh is a PolyData
isinstance(mesh, pv.PolyData)  # True

# Process
mesh = mesh.smooth(iterations=100)
mesh = mesh.remesh(n_clusters=5000)
curvature = mesh.compute_curvature(curvature_type="mean")

# Visualize
mesh.plot(scalars=curvature, cmap="coolwarm")

# Convert to plain PyVista PolyData if needed
polydata = mesh.to_polydata()

Development

# Install development dependencies
uv sync --group dev

# Run tests
uv run pytest

# Type checking
uv run ty check

# Linting
uv run ruff check src/phenotastic/

# Pre-commit hooks
uv run pre-commit run --all-files

Citation

If you use Phenotastic in your research, please cite:

Åhl, H., Zhang, Y., & Jönsson, H. (2022). High-throughput 3D phenotyping of plant shoot apical meristems from tissue-resolution data. Frontiers in Plant Science, 13, 827147.

BibTeX:

@article{aahl2022high,
  title={High-throughput 3d phenotyping of plant shoot apical meristems from tissue-resolution data},
  author={{\AA}hl, Henrik and Zhang, Yi and J{\"o}nsson, Henrik},
  journal={Frontiers in Plant Science},
  volume={13},
  pages={827147},
  year={2022},
  publisher={Frontiers Media SA}
}

License

GNU General Public License v3

Author

Henrik Ahl ([e-mail])