An improved method to detect arrhythmia using ensemble learning-based model in multi lead electrocardiogram (ECG)

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Arrhythmia, a condition of irregular heart rhythm, poses significant health risks.
• Current single-lead electrocardiogram (ECG) methods for arrhythmia detection lack sufficient sensitivity and specificity.
• Accurate and early detection of arrhythmias is critical for timely and effective patient treatment.

Purpose of the Study:

• To develop and evaluate an improved ensemble learning approach for enhanced arrhythmia detection using multi-lead ECG data.
• To introduce a novel feature extraction technique utilizing a sliding window of 5 R-peaks for improved signal analysis.
• To compare the performance of the proposed Fine Tuned Boosting (FTBO) model against other ensemble methods like bagging and stacking.

Main Methods:

• Implementation of a Fine Tuned Boosting (FTBO) ensemble learning model for multi-class arrhythmia detection.
• Development of a new feature extraction method based on a 5 R-peak sliding window applied to multi-lead ECG signals.
• Comparative analysis of FTBO with bagging and stacking models, including parameter tuning, using the MIT-BIH arrhythmia database.

Main Results:

• The proposed FTBO model demonstrated high sensitivity, specificity, and accuracy across multiple arrhythmia types.
• Achieved 100% sensitivity and specificity for Atrial Fibrillation (AF) detection.
• Attained 99% sensitivity and specificity for Premature Ventricular Contraction (PVC) and nearly 96% for Atrial Premature Contraction (PAC) detection.

Conclusions:

• The developed Fine Tuned Boosting (FTBO) model offers a significant advancement in arrhythmia detection accuracy using multi-lead ECG data.
• The novel 5 R-peak sliding window feature extraction technique improves the model's ability to identify complex cardiac irregularities.
• This approach shows substantial potential for early and reliable diagnosis of life-threatening cardiac arrhythmias.