Deciphering atrial repolarization morphology: A spline interpolation framework for atrial arrhythmia diagnosis

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Atrial repolarization (Ta wave) characterization is challenging due to its low amplitude and overlap with the QRS complex.
• Existing methods struggle to isolate and analyze the Ta wave effectively.

Purpose of the Study:

• To visualize the Ta wave within the QRS complex using a spline interpolation framework.
• To evaluate the utility of Ta wave features for improved atrial arrhythmia classification.

Main Methods:

• ECG data from Sinus Tachycardia (SiT) and Atrial Tachycardia (AT) patients were analyzed.
• Spline interpolation models were employed to synthesize the hidden Ta wave from PR and ST segments.
• Validation using Atrio-Ventricular Block (AVB) ECGs identified the optimal interpolation model.
• Machine learning models, including a stacked ensemble, were trained using P, Ta, and P-Ta wave features.

Main Results:

• The clamped cubic & B-spline model achieved the best performance (SSIM 0.7, power spectrum difference 1.33%) in interpolating the Ta wave.
• Derived Ta wave features (dispersion, area, peak location, etc.) were extracted.
• A stacked ensemble model achieved 99% classification accuracy and 0.99 F1 score for atrial arrhythmia detection.

Conclusions:

• The Ta wave, when characterized using spline interpolation, offers valuable features for classifying atrial arrhythmias.
• Integrating Ta wave features with traditional P wave analysis enhances diagnostic capabilities.
• The proposed interpolation method is adaptable for various clinical applications.