Analysis of Interpretable Machine Learning Classifiers for Myocardial Infarction Detection Using 12-Lead ECG Features

Abstract

Swift identification of cardiac events is crucial for enhancing patient outcomes. Electrocardiograms (ECGs) are noninvasive and can be quickly taken. Machine learning (ML) can then be used to assist in recognizing possible myocardial infarction. ECG recordings were extracted from a public database, and features related to heart rate variability and morphology were selected as descriptors for ML. The extracted data were preprocessed and classified using Waikato Environment for Knowledge Analysis. The Random Forest and Logistic Model Tree ML models were the best-performing models, achieving an accuracy of 87.40%. Other models, NaïveBayes, J48, and Reduced Error Pruning Tree, achieved accuracies of 86.86%, 81.23%, and 80.16%, respectively. Additionally, BayesNet and Logistic achieved accuracies of 86.33% and 84.99%, respectively. To validate the models internally, new data were taken from a completely new group of ECG recordings, and then the models were evaluated on each new internal independent dataset with unseen ECG data. BayesNet achieved an accuracy of 84.76%, and Random Forest managed to achieve 84.49% accuracy on the first internal independent dataset. These results show the power of ML in the rapid diagnosis of myocardial infarction and how it may aid in an emergency where quick  diagnosis and identification are crucial.

Received: 15 March 2026 | Revised: 3 August 2026 | Accepted: 20 August 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in PhysioNet at https://doi.org/10.13026/kfzx-aw45 (PTB-XL v1.0.3).

Author Contribution Statement

Chutian Chen: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization. Valentina L. Kouznetsova: Conceptualization, Writing – review & editing, Visualization, Supervision. Igor F. Tsigelny: Conceptualization, Methodology, Writing – review & editing, Supervision, Project administration.