Robust variational segmentation of 3D bone CT data with thin cartilage interfaces.

Area of Science:

• Medical Imaging
• Computational Biology
• Biomedical Engineering

Background:

• Accurate segmentation of 3D bone CT data is crucial for biomechanical analysis.
• Existing methods struggle with thin cartilage interfaces, leading to segmentation errors.
• Robust bone segmentation is essential for downstream predictive simulations.

Purpose of the Study:

• To develop a novel two-stage variational approach for accurate 3D bone CT segmentation.
• To specifically address and overcome challenges posed by thin cartilage interfaces.
• To enable automated and integrated bone segmentation for predictive simulations.

Main Methods:

• A two-stage variational segmentation strategy was employed.
• The first stage utilized a flux-augmented Chan-Vese model for initial segmentation.
• The second stage incorporated a phase-field fracture-inspired model to refine segmentation topology and eliminate spurious bridges.

Main Results:

• The proposed method achieved robust segmentation of 3D bone CT data, including challenging thin cartilage regions.
• Accurate segmentation topology was obtained, enabling reliable identification of individual bone objects.
• Validation was successful on femur and vertebra bone datasets, demonstrating effectiveness in hip, intervertebral disc, and synovial joint areas.

Conclusions:

• The developed two-stage variational approach offers a significant advancement in 3D bone CT segmentation.
• The methodology effectively handles thin cartilage interfaces, improving segmentation accuracy and topology.
• The approach holds strong potential for full automation and seamless integration with finite element-based predictive bone simulations.