Assessment of regularization techniques for electrocardiographic imaging.

Area of Science:

• Biomedical Engineering
• Computational Electrophysiology
• Medical Imaging

Background:

• Electrocardiographic imaging (ECGI) reconstructs epicardial potentials from torso ECGs.
• The ECGI inverse problem is ill-posed, requiring regularization for stable solutions.
• Existing regularization methods lack rigorous comparative analysis for clinical applicability.

Purpose of the Study:

• To systematically compare various regularization techniques for solving the ECGI inverse problem.
• To identify optimal regularization methods for accurate epicardial potential reconstruction.
• To evaluate performance across idealized and realistic torso models.

Main Methods:

• Tested 13 regularization techniques including Tikhonov and total variation algorithms.
• Utilized a unified simulation framework with idealized dipole models.
• Employed an electrolytic human torso tank with a canine heart model for in-vivo validation.

Main Results:

• Non-quadratic (total variation) and Tikhonov regularizations (1st and 2nd order) demonstrated superior performance.
• These methods yielded comparable average reconstruction errors.
• Outperformed other tested regularization techniques in accuracy.

Conclusions:

• Total variation and Tikhonov regularization are effective methods for ECGI.
• These techniques offer promising solutions for clinically relevant epicardial potential reconstruction.
• Further research should focus on these superior methods for clinical translation.