AI Analysis of Sleep Data Predicts Disease Risk – New Study Reveals Key Insights

Researchers at Stanford Medicine have developed an artificial intelligence model that can analyze a single night’s sleep data to predict a person’s risk of developing over 100 different health conditions.

The model, named SleepFM, was trained on nearly 600,000 hours of polysomnography data collected from 65,000 participants. Polysomnography records multiple physiological signals during sleep, including brain activity, heart activity, respiratory patterns, eye movements and leg movements.

According to Emmanuel Mignot, MD, PhD, the Craig Reynolds Professor in Sleep Medicine and co-senior author of the study, sleep studies generate a wealth of physiological data that is currently underutilized. “We record an amazing number of signals when we study sleep,” said Mignot. “It’s a kind of general physiology that we study for eight hours in a subject who’s completely captive. It’s very data rich.”

James Zou, PhD, associate professor of biomedical data science and co-senior author, noted that sleep has been relatively overlooked in AI research compared to fields like pathology or cardiology, despite its importance to overall health. “From an AI perspective, sleep is relatively understudied,” said Zou.

The researchers emphasized that SleepFM represents the first use of artificial intelligence to analyze such large-scale sleep data for disease risk prediction. By leveraging advances in AI, the model can extract meaningful patterns from the complex, multimodal data gathered during polysomnography that were previously difficult to interpret at scale.