Artificial intelligence-enabled electrocardiogram for mortality and cardiovascular risk estimation: a model...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Informatics

Background:

• Artificial intelligence (AI)-enabled electrocardiography (ECG) shows promise for predicting future disease and mortality.
• Current AI-ECG models lack individual patient actionability, explainability, and biological plausibility.
• The AI-ECG risk estimator (AIRE) platform was developed to address these limitations.

Purpose of the Study:

• To develop an AI-ECG platform (AIRE) that provides actionable, explainable, and biologically plausible risk predictions.
• To predict not only the risk of mortality but also the time-to-mortality using a single ECG.
• To validate the AIRE platform across diverse, transnational patient cohorts.

Main Methods:

• The AIRE platform was developed using deep learning and a discrete-time survival model on a large secondary care dataset (1,163,401 ECGs from 189,539 patients).
• AIRE was validated in five diverse, transnational cohorts, including volunteers, primary care, and secondary care patients.
• Phenome-wide and genome-wide association studies were conducted to identify biological pathways associated with predicted risk.

Main Results:

• AIRE accurately predicts all-cause mortality (C-index 0.775 in development, 0.638-0.773 in validation).
• AIRE also predicts future ventricular arrhythmia, atherosclerotic cardiovascular disease, and heart failure with high accuracy.
• Identified biological pathways include cardiac structure/function changes and genes linked to aging and metabolic syndrome.

Conclusions:

• AIRE is an actionable, explainable, and biologically plausible AI-ECG risk estimation platform.
• The platform has potential for worldwide clinical use in various contexts for short-term and long-term risk estimation.
• AIRE represents a significant advancement in AI-driven cardiovascular risk prediction.