A patient-adaptive profiling scheme for ECG beat classification

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Clinical and telemedicine applications require automated electrocardiogram (ECG) signal processing and heart beat classification.
• Individual ECG morphologies vary, necessitating personalized cardiac profiling for accurate diagnosis and monitoring.
• Existing methods may lack the adaptability required for dynamic physiological conditions.

Purpose of the Study:

• To propose a patient-adaptive cardiac profiling scheme for automated ECG analysis.
• To develop a novel local ECG beat classifier for individual cardiac behavior profiling.
• To enable early warning flagging of abnormal cardiac behavior.

Main Methods:

• Utilized a wavelet-based mechanism for precise fiducial ECG point extraction.
• Implemented a novel local classifier for patient-specific normal cardiac behavior profiling.
• Employed a repetition-detection concept for adaptive profiling.

Main Results:

• Achieved 99.59% accuracy in beat detection.
• Demonstrated high classification accuracy of 97.42% for identifying cardiac abnormalities.
• Validated the technique on the MIT-BIH arrhythmia database.

Conclusions:

• The proposed patient-adaptive scheme effectively automates ECG analysis and cardiac profiling.
• The method accurately detects heartbeats and identifies abnormalities, crucial for arrhythmia diagnosis.
• This approach offers a valuable tool for early detection of abnormal cardiac behavior in clinical and remote monitoring.