Artificial Intelligence-Enabled Electrocardiography to Screen Patients with Dilated Cardiomyopathy

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Undiagnosed dilated cardiomyopathy (DC) poses risks including sudden cardiac death.
• Current screening methods like echocardiography are costly and labor-intensive.
• Standard electrocardiography (ECG) lacks sensitivity and specificity for DC detection.

Purpose of the Study:

• To evaluate the diagnostic performance of an artificial intelligence-enabled electrocardiography (AI-ECG) algorithm for detecting reduced left ventricular ejection fraction (LVEF) in DC patients.
• To assess the applicability of AI-ECG as a screening tool for DC.

Main Methods:

• A cohort of 421 DC patients and 16,025 controls with 12-lead ECGs and echocardiography-measured LVEF were analyzed.
• An AI algorithm was developed and tested for its diagnostic performance in identifying LVEF ≤45% (indicating DC).
• Sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) were calculated at varying DC prevalence rates (1% and 5%).

Main Results:

• The AI-ECG model achieved an area under the curve (AUC) of 0.955 for detecting LVEF ≤45%.
• The algorithm demonstrated high sensitivity (98.8%) and specificity (44.8%).
• At 1% DC prevalence, NPV was 100% and PPV was 1.8%; at 5% prevalence, NPV was 99.9% and PPV was 8.6%.

Conclusions:

• AI-ECG exhibits high sensitivity and NPV for detecting DC.
• AI-ECG presents a simple, cost-effective screening tool for DC.
• This technology has potential implications for screening first-degree relatives of DC patients.