Revolutionizing Risk Prediction: The Power of AI-Driven Sleep Analysis

The AI model SleepFM predicts risk for more than 100 health conditions from a single night of sleep recordings, offering a scalable, noninvasive risk signal with immediate relevance for screening and triage workflows. Its scope includes oncologic, neuropsychiatric, and cardiometabolic domains—broadening detection pathways clinicians can leverage.

This approach differs from traditional risk tools by using sleep physiology rather than laboratory values or genomic profiles. It synthesizes multisystem risk across diverse endpoints instead of optimizing for a single disease, adding a parallel data stream that augments existing clinical risk models.

Core physiologic recordings are standard sleep channels: EEG, ECG, EMG, pulse waveform/oximetry, and airflow from polysomnography and many ambulatory monitors. These inputs produce features across sleep architecture, heart-rate variability, respiratory events, and movement that the model maps to disease probabilities. Such signals act as digital biomarkers: heart-rate variability and nocturnal arrhythmia patterns from ECG correlate with cardiometabolic risk, while sleep fragmentation and reduced slow-wave activity on EEG associate with cognitive decline.