Artificial Intelligence in ECG interpretation - review article

Abstract

Introduction and purpose

Developments in medical and information technology are leading to improvements in the quality of medical care. This paper aims to show how the development of artificial intelligence can affect more effective interpretation of ECG, thus contributing to greater efficiency for clinicians, and to describe the limitations and potential for further development of artificial intelligence in ECG interpretation.

Material and methods

A review paper using articles describing the application of artificial intelligence in ECG interpretation from 1997 to 2024. The search terms used to find publications were “artificial intelligence”, “ecg”, “deep learning”, “machine learning”, “neural networks”, “arrhythmia”, “left ventricular dysfunction” and “cardiomyopathy” using the Pubmed and Google Scholar databases.

Results

The present work has shown that artificial intelligence can be applied in the interpretation of ECGs for the following heart-related conditions: left ventricular dysfunction, atrial and ventricular arrhythmias, prediction of cardiovascular events, cardiomyopathy, valvular defect, monitoring of sleep quality, diagnosis of severe depression, presence of myocardial infarction, detection of the risk of transient myocardial ischemia, diagnosis of chronic maternal and fetal stress, and even determination of the age and sex of the subject. The biggest limitations of artificial intelligence are the need to verify the diagnosis made by the algorithm in order to detect possible errors. In addition, the use of personal data to train an artificial intelligence algorithm to diagnose specific medical conditions can be controversial, which can interfere with data protection rules. These data may contribute to a better understanding of artificial intelligence in ECG interpretation and its wider use in daily medical care practice.