Unsupervised classification of ventricular extrasystoles using bounded clustering algorithms and morphology matching

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Ventricular extrasystoles (VEs) are irregular heartbeats originating outside the sinoatrial node, often linked to increased mortality post-myocardial infarction.
• Current VE screening relies on manual, time-intensive analysis of electrocardiogram (ECG) data, including heartbeat morphology and RR intervals.
• The need for automated, efficient VE classification methods is critical for timely diagnosis and patient management.

Purpose of the Study:

• To develop and validate a novel algorithm for the automatic classification of ventricular extrasystoles (VEs).
• To enable VE classification without requiring a training dataset or prior knowledge of the number of classes or specific heartbeat features.
• To improve the efficiency and accuracy of VE screening compared to traditional manual methods.

Main Methods:

• The algorithm employs bounded clustering, morphology matching, and RR interval analysis for automatic VE classification.
• It operates without a predefined training set, allowing for unsupervised learning and adaptation.
• The method analyzes heartbeat morphology and RR interval variations for accurate classification.

Main Results:

• The proposed algorithm successfully performs automatic classification of VEs.
• Validation studies confirm the algorithm's effectiveness in identifying and categorizing VEs.
• The system demonstrates robust performance without reliance on training data.

Conclusions:

• The novel algorithm offers an efficient and automated solution for ventricular extrasystole classification.
• This approach reduces the manual workload associated with VE screening, potentially improving patient outcomes.
• The algorithm's ability to classify VEs without a training set represents a significant advancement in automated cardiac rhythm analysis.