SleepFM Predicts Over 100 Diseases From Sleep
Stanford Medicine researchers publish SleepFM on January 6 in Nature Medicine, an AI foundation model trained on 585,000 hours of polysomnography from 65,000 participants to predict risks for over 100 conditions. Linked to longitudinal records for 35,000 patients (1999–2024), SleepFM identified 130 diseases with concordance index above 0.75, including Parkinson’s, breast and prostate cancer, dementia, and cardiovascular events.
Key Points
- 1Learns physiological signals from single-night polysomnography to predict 130 diseases with C-index above 0.75.
- 2Uses leave-one-out contrastive learning to harmonize multimodal channels and reconstruct missing physiological signals.
- 3Enables time-to-event risk stratification for clinicians and researchers, informing screening and preventive interventions.
Scoring Rationale
High novelty and robust peer-reviewed validation, with limited immediate clinical deployment due to integration and explainability challenges.
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
