Landmark-based Morphometrics#
Landmark-based morphometrics analyzes shape using discrete, biologically meaningful points called landmarks. This approach is implemented in ktch through Generalized Procrustes Analysis (GPA).
Landmarks#
A landmark is a point of correspondence on each object that matches between and within populations.
Configuration and Centroid Size#
A configuration is the complete set of landmarks for a single specimen, represented as a matrix of coordinates.
Centroid Size#
Centroid size is the standard measure of size in geometric morphometrics, defined as the square root of the sum of squared distances from each landmark to the centroid.
In ktch:
from ktch.landmark import centroid_size
cs = centroid_size(configurations)
Generalized Procrustes Analysis (GPA)#
GPA is the standard method for extracting shape information from landmark configurations. It removes variation due to:
Translation (position) - by centering configurations
Scale (size) - by normalizing to unit centroid size
Rotation (orientation) - by optimal rotation alignment
GPA Algorithm#
For a sample of configurations, GPA iteratively:
Center each configuration at the origin
Scale each configuration to unit centroid size
Rotate configurations to minimize distances to a reference
Compute the mean shape
Repeat until convergence
The result is a set of shape coordinates.
In ktch:
from ktch.landmark import GeneralizedProcrustesAnalysis
gpa = GeneralizedProcrustesAnalysis()
shapes = gpa.fit_transform(configurations)
Pre-shape Space and Shape Space#
After centering and scaling, configurations lie on a pre-shape space, a high-dimensional hypersphere.
After GPA removes orientation information, the specimens occupy Kendall’s shape space. The Procrustes distance between shapes corresponds to the great-circle distance on this space.
Tangent Space Approximation#
For practical analysis, data are projected onto a tangent space, which is a linear approximation at the mean shape. This enables standard multivariate statistics (PCA, regression, etc.).
Statistical Analysis of Shape#
Principal Component Analysis#
from sklearn.decomposition import PCA
pca = PCA()
pc_scores = pca.fit_transform(shapes)
Limitations#
Requires homologous landmarks across all specimens
Not suitable for structures lacking clear landmarks (see Semilandmarks for extending GPA to curves and surfaces)
See also
Semilandmarks for analyzing curves and surfaces using semilandmarks
What is Morphometrics? for comparison with harmonic methods
Generalized Procrustes analysis for practical examples
References#
Dryden, I.L., Mardia, K.V., 2016. Statistical Shape Analysis: With Applications in R, John Wiley & Sons. John Wiley & Sons.
Claude, J., 2008. Morphometrics with R, Springer Science & Business Media. Springer Science & Business Media. https://doi.org/10.1007/978-0-387-77789-4
Bookstein, F.L., 1997. Morphometric tools for landmark data: geometry and biology, Cambridge University Press. Cambridge University Press.