ECG beat classification using a cost sensitive classifier

Area of Science:

• Biomedical Engineering
• Cardiology
• Machine Learning

Background:

• Accurate electrocardiogram (ECG) beat classification is crucial for diagnosing cardiac arrhythmias.
• Existing methods may struggle with ambiguous beats or varying costs of misclassification and rejection.

Purpose of the Study:

• To develop and evaluate a novel ECG beat classification system incorporating Support Vector Machines (SVMs) with a dynamic rejection mechanism.
• To improve classification accuracy and efficiency by intelligently handling uncertain beats.

Main Methods:

• ECG signal preprocessing followed by QRS complex detection and segmentation.
• Feature extraction including frequency information, RR intervals, QRS morphology, and AC power coefficients.
• Classification using Support Vector Machines (SVMs) with a decision rule employing dynamic reject thresholds based on misclassification and rejection costs.

Main Results:

• The proposed system demonstrated significant performance enhancement on the MIT-BIH arrhythmia database.
• Average accuracy reached 97.2% without rejection and 98.8% with minimal classification cost.
• The dynamic rejection strategy effectively improved the overall classification performance.

Conclusions:

• The developed SVM-based ECG beat classification system with dynamic rejection offers a robust and accurate method for arrhythmia detection.
• The system's ability to adapt reject thresholds based on cost functions enhances its practical utility in clinical settings.