Truace
Subscribe

Prediction of cardiovascular outcomes using electrocardiography from overnight polysomnography

Study objectives Polysomnography (PSG) is the gold standard for diagnosing sleep disorders, which are recognized as cardiovascular risk factors. Single-lead electrocardiograms (ECGs) are recorded during PSG but rarely analyzed, presenting an opportunity to leverage overnight ECGs for cardiovascular risk prediction. Materials and methods We applied a deep residual neural network with attention to single-lead ECGs from PSG, combined with sleep stage data, to predict 10-year risk of atrial fibrillation (AF), stroke…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1591· cite & verify
Prediction of cardiovascular outcomes using electrocardiography from overnight polysomnography
Sleep and Operational Stress by U.S. Naval Center for Combat and Operational Stress Control (NCCOSC). Public domain

In brief

Researchers took single-lead electrocardiograms that are recorded during overnight polysomnography but rarely analyzed, and combined them with sleep stage data in a deep residual neural network with attention. The model was originally trained to detect arrhythmias and was then finetuned on records from 15,809 Massachusetts General Hospital patients to estimate 10-year risk of several cardiovascular outcomes.

The work suggests that data already collected in routine sleep testing could be reused for cardiovascular risk screening without extra patient burden. The abstract excerpt gives no effect sizes or accuracy figures, and the outcomes came from health record codes, so how well the tool would perform in clinical decision-making remains to be shown.

Main points

  1. The model uses single-lead ECG from polysomnography together with sleep stage data.
  2. It predicts 10-year risk of atrial fibrillation, stroke, myocardial infarction, heart failure and death.
  3. It was finetuned on 15,809 Massachusetts General Hospital patients.
  4. External validation used Emory University Hospital (n = 9,810) and Beth Israel Deaconess Medical Center (n = 12,576) cohorts.

The gain

A neural network applied to single-lead ECGs already recorded during sleep studies, plus sleep stage data, was strongly associated with long-term risk of atrial fibrillation, stroke, heart attack, heart failure and death.

The rundown

Polysomnography is described as the gold standard for diagnosing sleep disorders, which are recognized cardiovascular risk factors. The ECG channel recorded during these studies is seldom analyzed, which the authors frame as an opportunity.

The network began as an arrhythmia detector and was adapted to long-term outcome prediction. Outcomes in the external cohorts were derived from electronic health records using ICD-9 and ICD-10 codes.

Cox proportional hazards models linked the network output to long-term risk. The authors conclude that integrating such predictions into polysomnography could improve detection and support personalized clinical decisions.

Sources

  1. Peer-reviewedSLEEP2026-10-09

The debate