Researchers develop gait foundation model to predict 3,210 health phenotypes
A new gait foundation model was trained on 3,328 adults who performed five motor tasks recorded with a single‑depth camera. The model learns embeddings that capture how all body joints move together over time, rather than relying on engineered gait features such as cadence or stride length.
Key points
- Model trained on 3,328 adults using a single‑depth camera and five motor tasks
- Embeddings outperformed engineered features for age (r=0.69), BMI (r=0.91) and visceral adipose tissue area (r=0.83)
- Embeddings predicted 1,887 of 3,210 phenotypes and improved predictions across 18 systems
The embeddings outperformed engineered features for predicting age (Pearsonʼs r = 0.69), body mass index (r = 0.91) and visceral adipose tissue area (r = 0.83). Across 3,210 tested phenotypes, the embeddings significantly predicted 1,887, improving prediction in all 18 body systems for males and 17 of 18 for females, and enhancing clinical diagnosis and medication use predictions.
The study suggests gait can serve as an independent, multisystem biosignal. Integrating gait analysis into routine clinical assessment could provide a scalable, passive vital sign for early detection of diverse health conditions.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Baseten releases Serverless Loops for open-model fine-tuning with per-token billing · 1 src
- Insilico Medicine releases open-source Longevity Grok AI system with 30 agents · 1 src
- Ro launches research lab to analyze patient data for clinical insights · 1 src
- Open ASR Leaderboard adds Egyptian and Nigerian English accents · 1 src
- Researchers link AI agent power to human Goal attainment in formal models · 1 src
Comments
via GitHub Discussions