Regularization Techniques for ECG Imaging during Atrial Fibrillation: A Computational Study.

Area of Science:

• Biomedical Engineering
• Computational Electrophysiology
• Medical Imaging

Background:

• The inverse problem of electrocardiography (iECG) is crucial for understanding cardiac electrical activity but is typically studied under stable rhythms.
• The performance of iECG regularization techniques during dynamic conditions like atrial fibrillation (AF) remains underexplored.

Purpose of the Study:

• To assess various regularization techniques for estimating epicardial potentials, dominant frequency (DF), phase maps, and singularity point (SP) location during AF.
• To evaluate the robustness of these methods against noise and anatomical model imperfections.

Main Methods:

• Simulated body surface potentials (BSPs) using a realistic atria-torso model under sinus rhythm and two AF patterns.
• Applied 14 regularization techniques to noisy BSPs to estimate epicardial potentials.
• Computed DF, phase maps, and SP location from estimated potentials, using novel metrics for SP assessment.

Main Results:

• Bayes maximum-a-posteriori (MAP) method showed superior performance but requires prior information.
• Tikhonov-based methods demonstrated comparable performance to complex techniques in realistic AF scenarios.
• Dominant frequency and singularity point location estimation were robust to noise and pattern complexity.

Conclusions:

• Tikhonov regularization is a viable and effective approach for iECG during AF, particularly for DF and SP estimation.
• The proposed evaluation framework provides a robust benchmark for iECG algorithms and offers clinical insights.