ECG Anomaly Detection with LSTM-Autoencoder for Heartbeat Analysis

Abstract

Electrocardiogram (ECG) signals are central to cardiac health assessment but interpreting them accurately requires expertise. Traditional methods often lack interpretability, posing limitations in ECG signal analysis. In this study, we propose an LSTM-based autoencoder to detect anomalies in 1D ECG signals, taking inspiration from 2D image anomaly detection techniques. Our model addresses the limitations of conventional methods and enhances ECG signal analysis accuracy. Due to the scarcity of annotated ECG datasets, we adopt an unsupervised anomaly detection approach by harnessing LSTM-based autoencoders. This enables us to uncover underlying patterns in ECG signals, facilitating the detection of abnormal patterns without the need for explicit labeling. Experimental results demonstrate that our LSTM-based autoencoder produces competitive outcomes. This work underscores the potential of deep learning approaches to enhance both the accuracy and efficiency of anomaly hearbeat in ECG signals.