Mayo Clinic researchers trained a deep convolutional neural network on over 11,000 ECG records to detect OSA, achieving an overall AUC of 0.80 and higher performance in women (AUC 0.82) versus men (AUC 0.73).

An AI-powered ECG algorithm from Mayo Clinic can detect obstructive sleep apnea from standard 12-lead ECGs, offering faster, cheaper, and more accessible screening with a particular strength in identifying OSA in women.

The approach envisions more at-home diagnostics, remote monitoring, and automated testing, accompanied by careful attention to regulatory, bias, and consent considerations to ensure equitable deployment.

The AI-based method is designed to be faster, more cost-effective, and more accessible, aiming to broaden diagnostic reach and efficiency.

The public health impact could be substantial by addressing underdiagnosis—especially in women—while reducing cardiovascular and cognitive risks associated with untreated OSA, in a condition global prevalence cited at over 936 million adults.

This development aligns with a broader trend toward explainable, scalable AI diagnostics that leverage routinely collected data to improve early detection and patient outcomes, potentially reshaping sleep medicine workflows.

Future work will refine accuracy across diverse populations, integrate with electronic health records and wearables, and expand AI-ECG applications from screening to treatment monitoring, all while navigating regulatory, privacy, and bias concerns.

ECG-based screening offers a non-invasive, scalable alternative to polysomnography, potentially enabling broader use especially in primary care and remote settings and helping close gender gaps in diagnosis.

Beyond diagnosis, the AI framework could evaluate how different OSA treatments impact cardiovascular risk, informing more personalized treatment decisions.

While equity benefits are noted, there are cautions about potential over-diagnosis and false positives, underscoring the need for transparent validation and human-in-the-loop oversight.

Published in JACC: Advances, this work positions AI-assisted ECG analysis as a transformative tool in sleep medicine and cardiology, with implications for personalized cardiovascular risk management in OSA patients.

Overall effort focuses on closing gender gaps in OSA diagnosis by prioritizing women, addressing historic underdiagnosis due to differing symptom presentations.