AI reads health clues from how you move, predicting 1,887 of 3,210 traits in Nature Medicine study
coursiv_ · reddit · 2026-11-18
- Researchers trained an AI model on depth-camera recordings of 3,328 adults performing five movement tasks, learning coordinated knee/hip/torso motion patterns beyond simple walking speed.
- The model predicted 1,887 of 3,210 tested health traits, spanning sleep, metabolism, lipids, liver/kidney function, bone density and immunity.
- Movement data improved predictions over a baseline of age, BMI, visceral fat and height — meaning ordinary motion carries health information those measures miss.
- The study doesn't claim cameras can replace checkups; longitudinal tracking of individual movement change is the more useful next test.
More from Apps
- Microsoft's Perception security agent hits general availability with red/blue/green team loop — jennifermarsman · 2026-11-18
- Corenna MCP server builds signed Android/iOS apps from GitHub repos inside ChatGPT or Claude — Personal_You3422 · 2026-11-18
- Sales AI assistant Scout launches, claiming 10X reply rates on its own outbound — thisdudelikesAI · 2026-11-18
- Google AI Studio upgrades Android app building: HD emulator, Firestore sync, one-click Play publishing — gaganghotra_ · 2026-11-18
- Microsoft launches Copilot Workspace for team collaboration at Ignite — donasarkar · 2026-11-18
- Microsoft's Copilot IQ adds 350+ third-party business data connectors — testingcatalog · 2026-11-18