P-Wave Spatial Variability in 12-Lead ECG Predicts Temporal Changes in Atrial Fibrillation

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Atrial fibrillation (AF) is linked to atrial remodeling.
• P-wave indices may predict AF risk.
• AF patients show greater P-wave morphology variability.

Purpose of the Study:

• Categorize AF patients using P-wave temporal variability.
• Classify AF patients via P-wave characteristics from standard ECGs.
• Investigate machine learning for AF stratification.

Main Methods:

• Unsupervised learning (k-means) for patient categorization.
• Feature extraction from 12-lead ECG P-waves.
• Supervised learning (SVM, Random Forest) for classification.
• Boruta algorithm and voting for feature selection.

Main Results:

• Two AF patient categories identified via VCG P-wave variability.
• Random Forest model achieved 80.2% accuracy.
• Temporal P-wave variation predictable from spatial ECG characteristics.

Conclusions:

• Machine learning can stratify AF patients using standard ECGs.
• This method offers a cost-effective approach for early risk identification.
• Larger datasets are needed for enhanced model validation.