A patient-adapting heartbeat classifier using ECG morphology and heartbeat interval features

Area of Science:

• Biomedical Engineering
• Medical Informatics

Background:

• Automated analysis of electrocardiograms (ECG) is crucial for diagnosing cardiac conditions.
• Accurate heartbeat classification is essential for effective arrhythmia monitoring.

Purpose of the Study:

• To develop and evaluate an adaptive system for automatic ECG heartbeat classification.
• To improve the performance of automated arrhythmia detection systems.

Main Methods:

• An adaptive system using a global classifier and a patient-specific local classifier was developed.
• The system adapts by training the local classifier with expert-corrected ECG beats.
• The adapted classifier combines global and local classifications for improved beat annotation.

Main Results:

• The performance of the adaptive classifier increases with the amount of local training data.
• Significant performance boosts were achieved with minimal adaptation, even with single-class data.
• The adaptive system demonstrated improved automated heartbeat classification accuracy.

Conclusions:

• Patient-specific adaptation significantly enhances automated ECG analysis.
• The developed system offers a highly beneficial improvement for automatic arrhythmia monitoring.
• This approach represents a notable advancement over existing automated heartbeat classification systems.