Machine learning-based prediction of acute coronary syndrome using only the pre-hospital 12-lead electrocardiogram

Salah Al-Zaiti,Ervin Sejdić,Samir Saba,Lucas Besomi,Christian Martin-Gill,Richard Gregg,Stephanie Frisch,Ziad Faramand,Clifton Callaway,Zeineb Bouzid

doi:10.1038/s41467-020-17804-2

Abstract

Prompt identification of acute coronary syndrome is a challenge in clinical practice. The 12-lead electrocardiogram (ECG) is readily available during initial patient evaluation, but current rule-based interpretation approaches lack sufficient accuracy. Here we report machine learning-based methods for the prediction of underlying acute myocardial ischemia in patients with chest pain. Using 554 temporal-spatial features of the 12-lead ECG, we train and test multiple classifiers on two independent prospective patient cohorts (n = 1244). While maintaining higher negative predictive value, our final fusion model achieves 52% gain in sensitivity compared to commercial interpretation software and 37% gain in sensitivity compared to experienced clinicians. Such an ultra-early, ECG-based clinical decision support tool, when combined with the judgment of trained emergency personnel, would help to improve clinical outcomes and reduce unnecessary costs in patients with chest pain.

Highlights

Prompt identification of acute coronary syndrome is a challenge in clinical practice
This problem results from the low sensitivity of the electrocardiogram (ECG) and initial clinical data to predict the presence of ongoing acute myocardial ischemia in those with acute coronary syndrome (ACS)
Using different machine learning (ML)-based classifiers trained and tested on separate prospective cohorts, we arrived at a generalizable model that outperforms both commercial interpretation software as well as experienced clinicians

Summary

Introduction

Prompt identification of acute coronary syndrome is a challenge in clinical practice. More than 50–75% of the seven million patients with chest pain are admitted to the hospital because the initial clinical evaluation is not sufficient to rule in or rule out ACS This problem results from the low sensitivity of the electrocardiogram (ECG) and initial clinical data to predict the presence of ongoing acute myocardial ischemia in those with ACS. Other studies focused only on the identification of patients with STEMI18, while others used single-lead ECGs or individual heart beats for algorithm development All of these limitations diminished the clinical utility for predicting acute ischemia in the general, nonselected chest pain populations seen at ED settings. A key feature of our ML method is that it utilizes traditional ECG features, but it takes advantage of novel temporal–spatial features of the 12-lead

Objectives

Methods

Results

Conclusion