Ensembled Feature based Multi-Label ECG Arrhythmia Classification.

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Arrhythmia is a primary cause of sudden cardiac death.
• Multi-label classification offers advantages over single-label approaches for detecting diverse cardiac conditions.
• Current deep learning models for multi-label arrhythmia classification face limitations like data dependency, overfitting, complexity, and poor interpretability.

Purpose of the Study:

• To propose a novel multi-label classification framework for electrocardiogram (ECG) analysis.
• To address the limitations of existing deep learning models in multi-label arrhythmia detection.
• To improve the accuracy and interpretability of classifying multiple concurrent cardiac arrhythmias.

Main Methods:

• Development of a multi-label classification framework incorporating ensembled hand-crafted features.
• Integration of the label powerset technique to preserve inter-class correlations.
• Validation using the CPSC 2018 ECG database.

Main Results:

• The proposed model achieved an accuracy of 88.52% and an F1-score of 88.45% for single-label classification.
• For multi-label classification, the model attained an accuracy of 88.11% and an F1-score of 86.32%.
• Demonstrated improved performance for specific cardiac conditions compared to state-of-the-art multi-label methods.

Conclusions:

• The proposed framework offers a robust alternative to deep learning models for multi-label arrhythmia classification.
• Accurate and early detection of multi-label ECG patterns, including atrial fibrillation and bundle branch blocks, can significantly reduce sudden cardiac death rates.
• The method enhances the classification of complex cardiac conditions, providing valuable clinical insights.