A Deep Learning Method for ECG Heartbeat Classification Based on Gaussian Features

Researchers have developed a deep learning method for electrocardiogram (ECG) heartbeat classification that leverages Gaussian features to improve the precision of arrhythmia detection. The study addresses the limitations of traditional methods in terms of feature extraction precision and discrimination capability.

The methodology begins with the identification of the R-peak using a visibility graph algorithm. Once these segments are isolated, the researchers fit five Gaussian components to the segmented signal. This process yields sixteen highly discriminative features that serve as the input for the classification model.

To process these features, the study develops an AMH–BiLSTM architecture. This model integrates convolutional neural networks, bidirectional long short-term memory networks, and multi-head attention mechanisms. The architecture is completed with two fully connected layers that perform the final classification task.

The researchers evaluated their method using the MIT–BIH Arrhythmia Database.

Over 10 independent random seeds, the proposed method achieved a mean classification accuracy of 97.93%. The precision was recorded at 98.10%, while the recall stood at 97.92%. The F1-score reached 97.99%. Additionally, the model demonstrated a mean AUC of 99.90%.

When compared to a baseline model under the same experimental setup, the proposed method showed improvements in classification performance. The authors also provided reference comparisons with recent representative methods, noting that their approach demonstrates competitive performance under comparable evaluation protocols.

Subject of Research: Deep learning method for ECG heartbeat classification using Gaussian features

Article Title: A deep learning method for ECG heartbeat classification based on Gaussian features

Article References: Li, Y., Liu, J., Shu, L., & Li, Y. (2026). A deep learning method for ECG heartbeat classification based on Gaussian features. Scientific Reports. https://doi.org/10.1038/s41598-026-76659-7

Image Credits: AI Generated

DOI: 10.1038/s41598-026-76659-7

Keywords: ECG, Deep Learning, Arrhythmia, Gaussian Features, BiLSTM, MIT-BIH Database, Cardiovascular, Signal Processing, Classification, SMOTE, deep, learning