What is the clinical evidence from this study?

Study design: Observational. Population: Individuals undergoing standard 12-lead ECG (n=1132665). Intervention: ECG-MACE multi-task deep learning model vs. Single-task models and Framingham risk score. Primary outcome: Prediction of one-year first-ever major adverse cardiovascular events (heart failure, myocardial infarction, ischemic stroke, and all-cause mortality).

January 2, 2025Open Access

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

Q: What does this research mean for the field?

A novel multi-task deep learning model (ECG-MACE) utilizing solely standard 12-lead ECGs can predict one-year major adverse cardiovascular events with high accuracy (AUROCs of 0.90 for heart failure, 0.85 for myocardial infarction, 0.76 for ischemic stroke, and 0.89 for mortality) and outperforms the Framingham risk score for long-term prediction. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Key Result

The ECG-MACE model predicted one-year major adverse cardiovascular events with AUROCs of 0.90 for heart failure, 0.85 for myocardial infarction, 0.76 for ischemic stroke, and 0.89 for mortality.

Study Design

Type

Observational (n=1,132,665)

Multicenter

Yes

Structured PICO

Does a multi-task deep learning model utilizing standard 12-lead ECGs accurately predict 1-year first-ever major adverse cardiovascular events in adults without prior cardiovascular disease?

Population

1,078,629 adults (≥30 years old) without a history of heart failure, myocardial infarction, or ischemic stroke, and without pacemakers, contributing 2,821,889 standard 12-lead ECGs. Mean age ~60 years, ~50% female, based in Taiwan (Chang Gung Memorial Hospital). An external validation cohort included 54,036 adult patients (113,224 ECGs).

Intervention

ECG-MACE, a multi-task deep learning model (ResNet-18 neural network) utilizing standard 12-lead ECGs to predict four major cardiovascular events simultaneously.

Comparator

Framingham risk score (FRS) for long-term prediction, and single-task deep learning models for 1-year prediction.

Outcome

1-year first-ever major adverse cardiovascular events (MACE), specifically non-fatal myocardial infarction, non-fatal ischemic stroke, hospitalization-requiring heart failure, and all-cause mortality.hard clinical

A multi-task deep learning model utilizing standard 12-lead ECGs can accurately predict 1-year incident heart failure, myocardial infarction, and mortality, outperforming the Framingham risk score and offering a scalable tool for primary cardiovascular prevention.

Limitations

Ethnically confined to the Han-Chinese population, limiting generalizability to other ethnic groups
Hospital-based study predominantly comprising patients with pre-existing health conditions, potentially introducing bias
Inability to handle right-censored time-to-event data for patients with no subsequent medical activities
Inherent opacity of deep learning models presents a challenge in understanding the decision-making process

Abstract

Deep learning analysis of electrocardiography (ECG) may predict cardiovascular outcomes. We present a novel multi-task deep learning model, the ECG-MACE, which predicts the one-year first-ever major adverse cardiovascular events (MACE) using 2,821,889 standard 12-lead ECGs, including training (n = 984,895), validation (n = 422,061), and test (n = 1,414,933) sets, from Chang Gung Memorial Hospital database in Taiwan. Data from another independent medical center (n = 113,224) was retrieved for external validation. The model's performance achieves AUROCs of 0.90 for heart failure (HF), 0.85 for myocardial infarction (MI), 0.76 for ischemic stroke (IS), and 0.89 for mortality. Furthermore, it outperforms the Framingham risk score at 5-year MACEs and 10-year mortality prediction. Over 10-year follow-ups, the model-predicted-positive group exhibits significantly higher MACE incidences than the model-predicted-negative group (relative incidence ratio: HF: 15.28; MI: 7.87; IS: 4.74; mortality: 13.18). Using solely ECGs, ECG-MACE effectively predicts one-year events and exhibits long-term anticipation. It provides potential applications in preventive medicine.

Ask AI

Helpful

Bookmark

View Full Paper