A Method for Context-Based Adaptive QRS Clustering in Real Time.

Area of Science:

• Biomedical Engineering
• Cardiovascular Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Continuous electrocardiogram (ECG) monitoring is crucial for diagnosing cardiac arrhythmias.
• Fast detection of alterations in normal heart conduction patterns from ECG signals remains a challenge.
• Existing methods often struggle with real-time processing and adapting to evolving QRS morphologies.

Purpose of the Study:

• To present a novel real-time method for adaptive clustering of QRS complexes from multilead ECG signals.
• To identify and characterize the set of QRS morphologies present during long-term ECG monitoring.
• To provide a tool for fast localization of deviations from normal cardiac conduction patterns.

Main Methods:

• Sequential processing of QRS complexes using adaptive clustering based on temporal context.
• Dynamic cluster representation via evolving templates that adapt to QRS morphology changes.
• Implementation of rules for cluster creation, merging, and removal, including noise detection.
• Utilized derivative dynamic time warping to address beat-to-beat misalignment.

Main Results:

• Achieved high validation purity rates: 98.56% on the MIT-BIH Arrhythmia Database and 99.56% on the AHA ECG Database.
• Demonstrated superior performance compared to previous offline QRS morphology analysis solutions.
• Successfully met real-time processing requirements for continuous ECG monitoring applications.

Conclusions:

• The proposed real-time adaptive clustering method effectively identifies diverse QRS morphologies in long-term ECG recordings.
• This approach offers a significant advancement in the rapid detection of cardiac conduction abnormalities.
• The method provides a robust and efficient tool for enhancing the diagnosis of arrhythmias.