Multi-object spring level sets (MUSCLE).

Area of Science:

• Medical image analysis
• Computational geometry
• Computer-aided diagnosis

Background:

• Accurate geometric modeling of multi-object systems is crucial for medical imaging tasks.
• Existing methods often struggle with maintaining object integrity during complex deformations.
• Bridging the gap between registration and segmentation requires a versatile data representation.

Purpose of the Study:

• To introduce a new data structure for geometrically modeling multi-objects with deformable and fluid-like behaviors.
• To enable seamless integration of deformable registration and active contour segmentation.
• To demonstrate the utility of this data structure in segmenting human brain structures.

Main Methods:

• The proposed data structure integrates a label mask, distance field, and springls (disconnected triangles).
• It offers sub-voxel precision, parametric control, automatic re-meshing, and correspondence tracking.
• Guarantees topological correctness, preventing self-intersections, air-gaps, and overlaps.

Main Results:

• Demonstrated the application of existing registration algorithms and active contour segmentation to the new data structure.
• Successfully segmented 74 cortical and subcortical structures in the human brain.
• The model exhibited elastic and fluid-like properties, facilitating inter-task interpretability.

Conclusions:

• The novel data structure provides a robust and precise geometric model for multi-object systems.
• It enhances interpretability between deformable registration and active contour segmentation tasks.
• This approach shows significant potential for accurate anatomical structure segmentation in neuroimaging.