Stanford AI Forecasts Health Risks Based on a Single Night’s Sleep Data
Researchers at Stanford Medicine have pioneered an artificial intelligence technology, SleepFM, that predicts potential health risks using data derived from a single night’s sleep. This innovative system scrutinizes intricate physiological signals from the brain, heart, and breathing patterns, unveiling hidden patterns indicative of health risks.
The AI model underwent training on a dataset of approximately 35,000 patients, aged between 2 and 96. The sleep studies were conducted at Stanford Sleep Medicine Center between 1999 and 2024. These polysomnography records were then paired with electronic health records spanning up to 25 years to ascertain if sleep data could foretell future diseases.
SleepFM showcased remarkable predictive accuracy for several conditions. The conditions and their respective C-index scores include:
- Parkinson’s disease (0.89)
- Dementia (0.85)
- Breast cancer (0.87)
- Prostate cancer (0.89)
- Heart attack (0.81)
The most pronounced results were observed for cancers, pregnancy complications, circulatory diseases, and mental health disorders, with prediction scores exceeding 0.8. The research findings were published in Nature Medicine on January 9, 2026.
Source: ScienceDaily