Gait Foundation Model Predicts Multi-System Biomarkers
Researchers led by Adam Gabet et al. (submitted 26 March 2026) develop a gait foundation model for 3D skeletal motion trained on 3,414 deeply phenotyped adults recorded via a depth camera across five motor tasks. Learned embeddings outperformed engineered features, predicting age (r = 0.69), BMI (r = 0.90), and visceral adipose tissue area (r = 0.82), and significantly predicted 1,980 of 3,210 phenotypic targets. Findings motivate translation to consumer-grade video and integration of gait as a scalable, passive vital sign.
Key Points
- 1Trained a gait foundation model on 3,414 adults using depth-camera 3D skeletal motion across five tasks
- 2Demonstrated embeddings predict age (r=0.69), BMI (r=0.90), VAT area (r=0.82), outperforming engineered features
- 3Enable passive, scalable multi-system health monitoring and phenotype prediction from consumer-grade video
Scoring Rationale
Strong multi-system validation and large cohort support high impact, limited by preprint status and translation needs.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
