Comparing semi-landmarking approaches for analyzing three-dimensional cranial morphology.

Area of Science:

• Biomedical imaging
• Geometric morphometrics
• Computational anatomy

Background:

• Three-dimensional (3D) imaging generates complex data requiring detailed shape descriptions.
• Semi-landmarks enhance shape information but pose challenges in 3D placement regarding time, cost, and artifacts.
• Dense surface sampling is crucial for accurate 3D shape analysis.

Purpose of the Study:

• To implement and compare three strategies for dense semi-landmark sampling on 3D surfaces.
• To evaluate the shape information captured by each dense sampling strategy.
• To assess the robustness of these methods against noise and data variability.

Main Methods:

• Implemented three dense sampling strategies: patch, patch-thin-plate spline (TPS), and pseudo-landmark sampling.
• Applied methods to great ape skulls (three species).
• Evaluated shape information by estimating transforms between individuals and a population average template using mean root squared error.

Main Results:

• Landmark sets from all methods yielded shape estimations comparable or superior to manual landmarks alone.
• The patch method showed sensitivity to noise and missing data, producing outliers.
• Patch-TPS and pseudo-landmarking demonstrated robust performance with noisy and variable data.

Conclusions:

• All tested landmarking strategies effectively increased shape information density compared to manual landmarks.
• Patch-TPS and pseudo-landmark sampling provide reliable and robust methods for dense 3D surface sampling.
• Trade-offs exist between point correspondence, spacing, coverage, repeatability, and computational cost.