Registration-based workflow for shape study

Area of Science:

• Computational Biology
• Geometric Morphometrics
• Bioinformatics

Background:

• Quantitative shape analysis is hindered by challenges in aligning diverse anatomical structures.
• Existing computational registration methods in morphometrics have limitations.
• Automated 3D registration is crucial for advancing morphology analysis.

Purpose of the Study:

• To present an improved two-step automated 3D registration pipeline for morphometrics.
• To enhance the accuracy and efficiency of shape variation analysis.
• To overcome limitations of previous automated registration techniques.

Main Methods:

• Developed an updated version of Automated 3D Geometric Morphometrics (auto3dgm) with improved installation, interface, and efficiency.
• Implemented a Surface Analysis, Mapping, and Segmentation (SAMS) module for informed registration.
• Tested the pipeline on diverse anatomical structures for alignment quality and efficiency.

Main Results:

• The updated auto3dgm demonstrated faster processing and higher alignment efficiency with fewer pseudolandmarks.
• SAMS-based registration generated biologically homologous feature points, resolving auto3dgm pseudolandmark issues.
• The combined pipeline offers more accurate 3D registrations compared to auto3dgm alone.

Conclusions:

• The presented automated 3D registration pipeline significantly enhances the practicality of morphometric analysis.
• This approach provides more accurate registrations, enabling advanced machine learning applications in morphology.
• The improved methods facilitate more effective quantitative analysis of shape variation across species.