SurfaceNets for Multi-Label Segmentations with Preservation of Sharp Boundaries.

Area of Science:

• Medical image analysis
• Computer graphics
• Computational geometry

Background:

• Generating accurate 3D surface meshes from segmented medical images with multiple material types is challenging.
• Existing methods may struggle with preserving topological features and sharp boundaries between different tissue types.

Purpose of the Study:

• To extend the 3D SurfaceNets algorithm for generating high-quality triangle meshes from multi-label 3D medical image data.
• To ensure preservation of topology and sharp material boundaries while guaranteeing user-specified accuracy.
• To provide an efficient and interactive tool for exploring the trade-off between surface accuracy and smoothness.

Main Methods:

• Extension of the 3D SurfaceNets algorithm to handle multi-label volumes.
• Development of a method to preserve sharp boundaries between segmented materials.
• Implementation of techniques for interactive control over accuracy and surface smoothing.
• Provision of open-source C++ library and a Qt/OpenGL-based application for experimentation.

Main Results:

• Generation of smooth, high-quality triangle meshes from multi-label 3D medical images.
• Preservation of original topology and sharp interfaces between different material labels.
• Achieved user-specified accuracy with interactive control over surface fairing.
• Demonstrated efficiency suitable for interactive exploration of parameters.

Conclusions:

• The extended 3D SurfaceNets algorithm effectively generates high-quality, topologically consistent surface meshes from multi-label medical images.
• The method offers a valuable tool for 3D rendering, tetrahedralization, and analysis of complex medical datasets.
• The open-source implementation facilitates further research and application in medical imaging.