Researchers have developed a new AI model capable of predicting an individual’s risk of developing more than 100 health conditions using sleep data, highlighting the growing role of sleep as a window into overall health.
The model, called SleepFM, was developed by an international team of researchers, including scientists from Stanford University in the United States. It was trained on nearly 600,000 hours of sleep data collected from around 65,000 participants, making it one of the largest datasets used for sleep-focused AI research to date. The findings have been published in the journal Nature Medicine.
SleepFM is designed as a foundation model, meaning it can learn general patterns from large-scale sleep data and apply them across multiple health-related tasks. In initial evaluations, the model demonstrated strong performance on established sleep analysis benchmarks, including identifying different stages of sleep and assessing the severity of sleep apnoea.
Beyond traditional sleep metrics, the researchers found that the model could extract deeper physiological signals from sleep patterns that correlate with a wide range of diseases. These include cardiovascular, metabolic, neurological, and respiratory conditions, suggesting that sleep data may serve as an early indicator of disease risk long before clinical symptoms appear.
The development of SleepFM underscores the potential of AI-driven analysis of wearable and clinical sleep data to support early risk stratification, preventive healthcare, and personalised medicine. As access to large-scale sleep data continues to expand through wearables and home monitoring devices, such models could play an important role in integrating sleep health into routine disease prediction and management strategies.
