AI model predicts cardiovascular disease risk using sleep data

Researchers have developed an artificial intelligence model capable of predicting the risk of cardiovascular disease and death up to ten years in advance by analyzing heart activity during sleep. The study, published in the journal ‘Sleep’, utilizes electrocardiogram (ECG) data recorded during polysomnography—a standard test used to diagnose sleep disorders like insomnia and sleep apnea.

The deep learning model was trained on a dataset of over 15,000 patients from Massachusetts General Hospital and subsequently validated using datasets from Emory University Hospital and Beth Israel Deaconess Medical Center. The findings indicate that the AI can effectively categorize individuals into long-term cardiovascular risk groups.

According to the research, higher risk scores identified by the model were associated with significantly increased probabilities of developing atrial fibrillation, stroke, heart failure, and all-cause mortality. Notably, the model maintained its predictive accuracy for heart attacks and irregular heart rhythms even after adjusting for conventional risk factors such as age, sex, BMI, diabetes, and hypertension.