A Systematic Review of AI Techniques for ECG Analysis in Acute Coronary Syndrome Detection

Abstract

Electrocardiography (ECG) is an essential cardiovascular medicine tool, offering non-invasive information regarding cardiac electrical activity to diagnose many conditions. ECG interpretation, though, can be intricate, laborious, and variable across human interpreters. Artificial intelligence (AI), such as machine learning (ML) and deep learning (DL), has become a game-changer with the potential to enhance the precision, efficiency, and availability of ECG analysis. This systematic review synthesizes the evidence on the diagnostic performance of AI models in ECG interpretation for acute coronary syndrome (ACS), comparing AI’s effectiveness with that of human experts and traditional methods. The methodology followed established systematic review guidelines, such as PRISMA, with comprehensive searches across databases like PubMed and Scopus. Included studies utilized high-level ML/DL architectures, e.g., CNNs, and were risk of bias assessed using tools such as QUADAS-2. Diagnostic accuracy measures, e.g., sensitivity, specificity, and Area Under the Curve (AUC), were estimated through bivariate random-effects models. Results show that AI models exhibit high diagnostic accuracy in ACS conditions. For Acute Coronary Syndromes (ACS), AI models recorded AUROCs of as much as 0.997, outperforming human experts in speed and accuracy. Even with these numbers, issues persist, such as AI-associated misdiagnosis, performance variability, and data-related issues like quality and imbalance. Patient privacy and algorithmic bias ethical issues also present challenges. The practical implementation of AI in clinical practice relies on continuous improvement, verification, and coordination among AI vendors, clinicians, and policymakers to overcome the technical and practical challenges.