Methods of modeling and classification of electrocardiograms

Abstract

The article thoroughly analyzes methods for modeling, processing, and classifying electrocardio signals based on the study of publications in the scientometric database Web of Science Core Collection for 2014-2024. 513 scientific papers reflecting the dynamics of the development of approaches to interpreting cardiac data were analyzed. Bibliometric analysis revealed an increase in publication activity with maximum indicators in 2019 (58 publications), 2022 (51 publications), and 2024 (54 publications), which confirms the relevance of the topic under study. Based on the keywords of the publications, a term cloud has been created, which demonstrates the prevalence of such concepts as electrocardio signal, deep learning, classification, signal processing, and heart rate variability. Methodological analysis has shown the evolution of approaches from traditional methods of digital signal processing to the introduction of innovative artificial intelligence technologies. In particular, the effectiveness of the use of convolutional neural networks for the classification of electrocardio signals that demonstrate an accuracy of more than 95% in diagnosing various types of arrhythmias has been investigated. The geographical distribution of publications showed the leadership of scientific institutions in the United States (19% of publications), India (12%), Germany (10%), and China (9%). The thematic direction is dominated by research in the field of engineering (51% of publications) and computer science (29%), which emphasizes the interdisciplinarity of the problem. The priority directions for the development of methods for analyzing ESN have been identified: the development of adaptive algorithms for filtering interference, the improvement of methods for extraction of diagnostically significant features, and the creation of intelligent classification systems based on hybrid architectures. Particular attention is paid to the synergy of machine learning methods with classical approaches to the analysis of ECS to increase the reliability of diagnosing cardiovascular pathologies.