Post-processing techniques for making reliable measurements from curve-skeletons

Area of Science:

• Biomedical Engineering
• Medical Imaging Analysis
• Computational Biology

Background:

• Interconnected 3-D biological networks are common.
• Curve-skeletons quantify network geometry (e.g., path lengths, tortuosities, thicknesses).
• Standard curve-skeletons are sensitive to small surface features and imaging noise.

Purpose of the Study:

• To develop robust post-processing techniques for curve-skeletons.
• To reduce sensitivity to small-scale surface features in geometric measurements.
• To introduce a more reliable measure of cross-sectional dimension.

Main Methods:

• Developed new post-processing techniques for curve-skeletons.
• Utilized a minimal sphere-network representation for object sampling.
• Defined a novel 'modal radius' for cross-sectional dimension measurement.
• Achieved sub-voxel accuracy in measurements.

Main Results:

• Measurements of lengths and thicknesses are less sensitive to surface noise.
• The modal radius is more robust than the internal radius for cross-sectional dimension.
• The techniques effectively quantify geometric parameters of complex networks.
• Demonstrated application on trabecular bone and tumor vascular networks.

Conclusions:

• The new techniques provide more reliable quantification of 3-D biological network geometry.
• These methods improve the robustness of curve-skeleton analysis, particularly with noisy imaging data.
• The modal radius offers a superior single-measure metric for local structure size.