A graph-based cardiac arrhythmia classification methodology using one-lead ECG recordings.

Area of Science:

• Biomedical Engineering
• Computational Cardiology
• Machine Learning in Healthcare

Background:

• Cardiac arrhythmias are irregular heart rhythms that can lead to serious health complications.
• Accurate and timely diagnosis of arrhythmias is crucial for effective patient management.
• Single-lead electrocardiograms (ECGs) offer a portable and accessible method for cardiac monitoring.

Purpose of the Study:

• To develop and evaluate a novel graph-based methodology for classifying cardiac arrhythmia diseases using single-lead ECG signals.
• To compare the performance of different machine learning models in identifying various arrhythmia types.

Main Methods:

• A visibility graph technique was employed to transform time-series ECG signals into graph representations.
• Informative features were extracted from these graphs for subsequent classification.
• Three classifiers were investigated: graph convolutional neural network (GCN), multi-layer perceptron (MLP), and random forest (RF).
• The MIT-BIH arrhythmia database was used for training and validation.

Main Results:

• The multi-layer perceptron (MLP) model achieved the highest classification accuracy of 99.02%.
• The random forest (RF) model also demonstrated strong performance with an accuracy of 98.94%.
• The proposed graph-based approach proved effective for accurate arrhythmia classification.

Conclusions:

• The novel graph-based methodology offers a highly accurate approach for cardiac arrhythmia classification from single-lead ECGs.
• The multi-layer perceptron (MLP) is a highly effective classifier for this task, outperforming other models.
• This method holds promise for improving automated arrhythmia detection and diagnosis.