← Azimuth Daily
Capturing data…
Capturing data…
SAT, 7 NOV · 88 ITEMS
"Gait foundation model predicts health phenotypes from 3D skeletal motion" — Nature Medicine · Science & Medicine
This paper presents a foundation model for gait analysis that uses data—captured by a depth camera while a person walks—to predict a wide range of across multiple body systems. Trained on over 3,400 deeply phenotyped adults, the model learns generalizable representations of walking patterns and can infer conditions such as cardiovascular, musculoskeletal, and neurological health from a short video. It represents a step toward non-invasive, scalable health screening using everyday movement.
The claim is based on a peer-reviewed study published in Nature Medicine (DOI: 10.1038/s41591-026-04724-9) and also available as a preprint on arXiv. As the primary source, the paper provides full details on the model architecture, training data, and validation results. The reported ability to predict multi-system health phenotypes is supported by the study's experiments, though independent replication would further strengthen confidence.
This is genuinely new research, published in 2026 (arXiv preprint March 2026, Nature Medicine publication likely around the same time).