A nonlinear trimmed moving averaging-based system with its application to real-time QRS beat classification

Area of Science:

• Biomedical Engineering
• Cardiovascular Signal Processing
• Machine Learning in Healthcare

Background:

• Arrhythmia detection is crucial for diagnosing heart conditions.
• Real-time analysis of electrocardiogram (ECG) signals is essential for immediate patient monitoring.
• Accurate classification of QRS complexes, particularly abnormal beats, is a key challenge in ECG analysis.

Purpose of the Study:

• To develop and evaluate a real-time QRS beat classification system.
• To identify abnormal heartbeats of ventricular origin using a novel nonlinear approach.
• To enable immediate beat diagnosis in parallel with QRS detection.

Main Methods:

• Implementation of a real-time QRS beat classification system.
• Utilizing a nonlinear trimmed moving average filter for beat analysis.
• Parallel processing with a real-time QRS detector for immediate diagnosis.
• Evaluation using ECG recordings from the MIT-BIH arrhythmia database.

Main Results:

• The proposed system achieves a beat classification rate exceeding 99.5%.
• The nonlinear trimmed moving average filter effectively identifies ventricular abnormal beats.
• The system operates in real-time, providing immediate diagnostic capabilities.

Conclusions:

• The developed real-time QRS beat classification system demonstrates high accuracy and efficiency.
• The nonlinear filter approach is effective for identifying ventricular arrhythmias.
• This system has the potential to significantly improve real-time cardiac monitoring and diagnosis.