Automated detection of myocardial infarction using ECG-based artificial intelligence models: a systematic review

Abstract

Clinical decision making in the emergency room needs to be fast and accurate, especially for myocardial infarction (MI) cases. The best way to address data-based decisions is through artificial intelligence techniques (AI), which haven’t been systematize for MI detection. Thereby, we performed a systematic review (PROSPERO: CRD42021229084). The literature search from Pubmed, Web of Science, Scopus, IEEE Xplore and Embase resulted in n = 48 included articles. 71% of those articles implemented deep-learning models, while the other 29% developed machine-learning models, from which Convolutional Neural Networks and Support Vector Machines were the most common architectures. Data pre-processing methods, ECG-derived features with their corresponding feature extraction techniques, dimensionality reduction and redundancy evaluation algorithms and classifier are discussed in the present work. Furthermore, public and private datasets are analyzed, and class balance is addressed. To the extent of our knowledge, the present work is one of the most comprehensive reviews that addressed systematically the characteristics of artificial intelligence algorithms for the detection of MI based on ECG information.