Summary

seg2mesh is a toolkit for processing segmentatation label images, extracting isocontours from label images as triangular meshes, and processing these meshes with various downstream tasks.

For label image processing, operations such as:

• Resampling

  • To adjust image resolution and/or apply transforms
  • To merge labels in different coordinate systems or resolutions

• Morphological Opening and Closing

• Island Removal

• Binary Pruning (To remove spurious label regions)

• A multistep procedure to make labels contiguous

are provided.

For mesh processing, operations such as:

• Taubin Smoothing
• Uniform Remeshing with the Approximated Discreted Centroidal Voronoi Diagram algorithm
• Adaptive Remeshing with mmg3d using a metric automatically assigned based on local curvature

are provided.

seg2mesh also enables characterization of geometric changes between processed and unprocessed meshes including:

Distance metrics:

• Hausdorff distance
• Mean symmetric surface distance
• Root mean square distance

Classification metrics:

• Dice coefficient
• Intersection over union
• Accuracy

Getting Started

Please refer to the seg2mesh Documentation for detailed instructions on installation, example usage, and API documentation auto-generated via Sphinx.