Autoencoders for ECG Anomaly Detection: A Survey
Journal of Computing in Education, Sports and Health · 2025
DOI: 10.69996/jcesh.2025004 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Computing in Education, Sports and Health, 2025. Verfasst von Shreevats Dhyani und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Anomaly detection in Electrocardiogram signals is very important for the identification of abnormal heart conditions such as arrhythmias and ischemia. Recently, Autoencoders, especially Variational Autoencoders (VAEs) and Convolutional Autoencoders (CAEs), have been used as effective tools for this purpose. These unsupervised machine learning models can learn efficient representations of normal cardiac activity, detecting deviations that signal abnormalities. Autoencoders are trained on normal ECG signals, and anomalies are flagged when the reconstruction error exceeds a predefined threshold. Our experiments demonstrated strong performance with the model achieving 97% accuracy, 0.93 precision, and 1.00 recall for anomaly detection. The F1-scores were 0.96 for anomalies and 0.97 for normal signals, confirming the effectiveness of Autoencoders compared to traditional methods. The paper reviews Autoencoder performance across various ECG datasets and compares it to other anomaly detection techniques. This includes challenges such as data imbalance, noise, and the requirement for high-quality labelled datasets. The paper also indicates that hyperparameter tuning and feature engineering are among the critical factors affecting model performance. Finally, future directions are emphasized, such as the integration of Autoencoders with techniques such as transfer learning and hybrid models for accuracy and possible real-time clinical applications.
Zitieren
APA
Dhyani, S., & Butola, R. (2025). Autoencoders for ECG Anomaly Detection: A Survey. Journal of Computing in Education, Sports and Health. https://doi.org/10.69996/jcesh.2025004
Vancouver
Dhyani S, Butola R. Autoencoders for ECG Anomaly Detection: A Survey. Journal of Computing in Education, Sports and Health. 2025. doi:10.69996/jcesh.2025004
BibTeX
@article{Dhyani2025,
author = {Dhyani, Shreevats and Butola, Ravi},
title = {Autoencoders for ECG Anomaly Detection: A Survey},
journal = {Journal of Computing in Education, Sports and Health},
year = {2025},
doi = {10.69996/jcesh.2025004}
}
EUVIMED