A machine learning approach relating 3D body scans to body composition in humans

Area of Science:

• Nutrition science
• Obesity research
• Biometrics

Background:

• Quantifying the relationship between body fat and anthropometry is a long-standing challenge.
• Traditional anthropometric measurements like BMI have limitations in accurately representing body composition.
• 3D body shape imaging offers a rapid and precise method for collecting extensive anthropometric data.

Purpose of the Study:

• To derive homogenous body shape archetypes from a large dataset of 3D scanned anthropometry.
• To pair these body shape archetypes with accurate body composition data.
• To develop a machine learning approach for linking 3D body shape to health outcomes.

Main Methods:

• Utilized principal components and cluster analysis to identify four body shape archetypes from 3D scans of US Army recruits.
• Paired these archetypes with dual-energy X-ray absorptiometry body composition data from a separate participant group.
• Applied a machine learning model to connect large-scale 3D anthropometric data to detailed body composition measurements.

Main Results:

• Successfully derived four distinct body shape archetypes from a large, diverse population.
• Established a method to link 3D body shape data to precise body composition measures.
• Demonstrated the feasibility of using machine learning to bridge large anthropometric datasets with smaller, detailed health outcome datasets.

Conclusions:

• 3D body scanning technology enables rapid, comprehensive anthropometric data collection.
• The developed machine learning approach effectively pairs body shape archetypes with body composition.
• This methodology can be extended to analyze relationships between 3D body shape and various health outcomes in large populations.