Why the study?
Reliable automatic anomaly detection is useful to support physicians reading ECG signals, prompting the design of a convolutional autoencoder system to assist in detecting disease-related anomalies.
Does a Convolutional Autoencoder (CAE)-based system improve anomaly detection in ECG signals compared to other state-of-the-art approaches?
Population
ECG signals from a simulated test set and a real test set
Comparison
CAE-based system vs other state-of-the-art ECG anomaly detection approaches
Key result
A Convolutional Autoencoder-based system for ECG anomaly detection achieved a ROC AUC of 99.75% on a real test set, outperforming other state-of-the-art approaches.
Authors
Loading...
May support ECG anomaly detection in decision support systems; leaves open clinical validation in prospective studies.
Does a Convolutional Autoencoder (CAE)-based system improve anomaly detection in ECG signals compared to other state-of-the-art approaches?
A novel Convolutional Autoencoder-based system demonstrates high accuracy (ROC AUC >97%) for automated anomaly detection in ECG signals.
Lomoio et al. (2023) studied ECG anomalies. Convolutional Autoencoder (CAE) based system vs. other state-of-the-art ECG anomaly detection approaches was evaluated on Anomaly detection performance (ROC AUC). A Convolutional Autoencoder-based system for ECG anomaly detection achieved a ROC AUC of 99.75% on a real test set, outperforming other state-of-the-art approaches.