A gait foundation model for prediction of multisystem health phenotypes from 3D skeletal motion.

Learned embeddings from 3D skeletal motion recorded by a single-depth camera predicted multisystem health phenotypes, improving age (r=0.69), BMI (r=0.91), and visceral adipose tissue area (r=0.83) in 3,328 adults.

• Why it matters: Gait is often treated as a symptom of specific conditions, but the field lacks a systemic biomarker that captures whole-body movement detail for broad health phenotyping.
• What they did: The study trained a gait foundation model on 5 motor tasks to generate joint-motion embeddings, then tested prediction across 18 body systems and 3,210 phenotypes.
• The result: Embeddings predicted 1,887 of 3,210 phenotypes and improved prediction in all 18 systems for males and 17 of 18 for females versus a baseline using age, BMI, visceral adipose tissue, and height.

AI summary of the abstract · may contain errors · not medical advice · read the paper.