Stanford Develops SleepFM Predicting 130 Conditions
Stanford researchers describe SleepFM, a foundation model that analyzes polysomnography to predict future risk for 130 health conditions, reporting results in Nature Medicine. Trained on nearly 600,000 sleep-hours from 65,000 participants and using leave-one-out contrastive learning, it achieved C-index values above 0.8 for cancers, circulatory and neurological conditions, and mortality. The study suggests potential use with wearables for early risk screening, though cohorts were PSG-referred.
Key Points
- 1Trains foundation model on ~600,000 sleep-hours from 65,000 patients to predict 130 conditions.
- 2Demonstrates high prognostic power (C-index >0.8) for cancers, cardiovascular, neurological, and mortality risks.
- 3Enables potential clinical and wearable-based early risk screening, guiding preventive monitoring and interventions.
Scoring Rationale
High novelty and broad clinical relevance from peer-reviewed Nature Medicine study, tempered by limited generalizability from PSG-referred cohorts.
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
