SleepFM: One Night of Sleep Predicts 130 Diseases

Researchers from Stanford Medicine introduced SleepFM—a foundation multimodal AI model capable of predicting the risk of approximately 130 diseases based on data from a single night of polysomnography. The publication, widely covered on January 9, 2026, details how the algorithm analyzes patterns of brain activity, breathing, and heart rate, revealing hidden correlations inaccessible to doctors.

SleepFM successfully forecasts risks of dementia, cardiovascular diseases, and even certain types of cancer. Unlike narrowly specialized models, SleepFM is trained on a massive dataset (14,000+ recordings) and functions as a universal diagnostic tool. This is a significant step from reactive to predictive medicine, where wearable devices could become the first detectors of serious pathologies.

Source: Stanford Medicine