Geometric morphometrics approach for classifying children's nutritional status on out of sample data

Area of Science:

• Geometric Morphometrics
• Biomedical Imaging
• Anthropometry

Background:

• Traditional alignment-based classification methods in geometric morphometrics struggle with classifying individuals outside the original study sample.
• This limitation is particularly relevant for infant and child nutritional assessment using body shape images, where new individuals are frequently encountered.
• Classification rules derived from a reference sample require sample-dependent preprocessing (e.g., Procrustes analysis, allometric regression) for out-of-sample application.

Purpose of the Study:

• To develop methods for obtaining shape coordinates of new individuals for classification.
• To analyze the impact of different template configurations on the registration of out-of-sample raw coordinates.
• To improve the accuracy of children's nutritional status evaluation using arm shape analysis from photographs.

Main Methods:

• Proposing novel approaches to generate shape coordinates for individuals not included in the initial study.
• Investigating the use of various template configurations for registering raw coordinate data from new individuals.
• Analyzing sample characteristics and collinearity among shape variables for optimized classification.

Main Results:

• The study presents methods for processing out-of-sample data in geometric morphometrics.
• Different template configurations were evaluated for their effect on registration accuracy.
• Understanding sample characteristics and variable collinearity is shown to be critical for classification performance.

Conclusions:

• The proposed methods facilitate the classification of new individuals in geometric morphometrics, especially for nutritional assessment.
• Effective template selection and understanding of shape variable relationships are key for accurate classification of children's nutritional status.
• The findings support the development of tools like the SAM Photo Diagnosis App© for offline, updated nutritional screening.