Elastic Statistical Shape Analysis of Biological Structures with Case Studies: A Tutorial

Area of Science:

• Computational Biology
• Biomedical Imaging
• Geometric Statistics

Background:

• Statistical shape analysis is crucial for understanding biological form and variation.
• Existing methods often lack invariance to object parameterization, limiting their applicability.
• A robust framework is needed for analyzing complex biological shapes, especially curves.

Purpose of the Study:

• To present a new statistical shape analysis framework for curves.
• To demonstrate its utility on diverse biological datasets.
• To provide tools for shape comparison, statistical modeling, and variability analysis.

Main Methods:

• Utilizes the square-root velocity function for shape representation.
• Employs an elastic metric for shape comparison and registration.
• Develops methods for computing shape means, covariances, and principal component analysis.
• Introduces Wrapped Gaussian models for shape space sampling.

Main Results:

• The framework achieves invariance to reparameterization, translation, rotation, and scale.
• Optimal registrations (point correspondences) are computed efficiently.
• Statistical analysis, including mean and covariance, is performed on various biological shapes.
• Case studies on leaf outlines, arteries, DTI tracts, tumors, and vertebrae demonstrate applicability.

Conclusions:

• The square-root velocity function framework offers a powerful approach for statistical shape analysis of curves.
• It provides robust tools for analyzing shape variability in biological data.
• The accompanying MATLAB package facilitates reproducible research in this domain.