A multitask deep learning model utilizing electrocardiograms for major cardiovascular adverse events prediction.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Electrocardiography (ECG) analysis using deep learning shows promise for predicting cardiovascular outcomes.
• Cardiovascular diseases remain a leading cause of mortality worldwide, necessitating advanced predictive tools.

Purpose of the Study:

• To develop and validate a novel multi-task deep learning model, ECG-MACE, for predicting one-year major adverse cardiovascular events (MACE).
• To assess the model's performance against established risk scores and evaluate its long-term predictive capabilities.

Main Methods:

• A multi-task deep learning model (ECG-MACE) was trained and validated on a large dataset of 2,821,889 standard 12-lead ECGs from a Taiwanese hospital.
• External validation was performed using data from an independent medical center (113,224 ECGs).
• Model performance was evaluated using Area Under the Receiver Operating Characteristic curves (AUROCs) for various cardiovascular events and mortality.

Main Results:

• The ECG-MACE model achieved high AUROCs: 0.90 for heart failure (HF), 0.85 for myocardial infarction (MI), 0.76 for ischemic stroke (IS), and 0.89 for mortality.
• The model outperformed the Framingham risk score in predicting 5-year MACE and 10-year mortality.
• Over 10-year follow-ups, individuals predicted positive for MACE by the model showed significantly higher event incidences compared to those predicted negative.

Conclusions:

• The ECG-MACE model effectively predicts one-year cardiovascular events and demonstrates long-term predictive value using only ECG data.
• This AI-driven approach holds significant potential for applications in preventive cardiology and early disease detection.