Digital Stethoscope-Derived Single-Lead Electrocardiogram and Artificial Intelligence to Detect Low Ejection...

Area of Science:

• Cardiology
• Medical Devices
• Artificial Intelligence in Healthcare

Background:

• Preoperative echocardiography is crucial but challenging in resource-limited settings.
• Artificial intelligence (AI) is emerging for diagnosing low ejection fraction (EF) from electrocardiograms (ECGs).

Purpose of the Study:

• To evaluate the diagnostic accuracy of an AI-powered electronic stethoscope for identifying reduced left ventricular ejection fraction (LVEF ≤40%) in preoperative patients.
• To assess the utility of this tool in optimizing resource allocation for preoperative screening.

Main Methods:

• A multicenter, prospective, observational diagnostic accuracy study involving 981 preoperative patients across 10 hospitals in India.
• Single-lead ECGs and heart sounds were collected using the Eko CORE 500 electronic stethoscope.
• An AI algorithm analyzed recordings to categorize LVEF as ≤40% or >40%, validated against echocardiography.

Main Results:

• The final cohort included 867 patients; 138 had LVEF ≤40%.
• The algorithm achieved an AUC-ROC of 0.73 (95% CI, 0.67-0.78) for detecting LVEF ≤40%.
• Sensitivity was 60.1% and specificity was 81.3%, with better performance in noncardiac surgery patients.

Conclusions:

• The Eko digital stethoscope demonstrates potential as a clinical decision-making tool for preoperative assessment.
• Its ability to detect reduced EF, particularly in noncardiac surgery, suggests utility as a rule-out test.
• Consideration for use in specific patient populations and resource-limited settings is warranted.