AI detects heart failure signs in sleep data

Researchers Zuzana Koscova and Gari Clifford from Emory University have developed an AI model that can identify early signs of heart failure from cardiac traces recorded during sleep screening. The model was described in the journal SLEEP on October 9, 2026.

Sleep data often goes unused

During polysomnography, the standard test for sleep disorders, electrocardiograms are recorded but typically not analyzed. These recordings are stored but rarely reviewed by medical professionals.

AI learns from large patient data

The AI was trained using data from 15,809 patients at Massachusetts General Hospital in Boston. It was later tested on 9,810 patients from Emory University Hospital and an additional 12,576 individuals, showing its ability to detect early signs of heart failure.

Algorithm focuses on specific sleep stages

The AI divides sleep into fifteen-second segments and links them to specific sleep stages. It places the most emphasis on the N1 stage, which occurs as a person transitions into sleep, when the nervous system changes most rapidly.

Risk increase linked to AI score

According to the study, a one-standard-deviation increase in the AI’s score was associated with a 69% higher risk of heart failure in one group of patients and a 133% increase in another, highlighting the potential predictive power of the model.

Atrial fibrillation also under study

The research also touches on the detection of atrial fibrillation, a type of irregular heartbeat, though the full implications of the AI’s ability to identify such conditions are still being explored.