Improving electrocardiographic imaging solutions: A comprehensive study on regularization parameter selection in...

Area of Science:

• Biomedical Engineering
• Computational Electrophysiology
• Medical Imaging

Background:

• Electrocardiographic imaging (ECGI) requires precise regularization parameter (λ) selection for accurate inverse electrogram reconstruction.
• Quantifying the impact of signal noise and geometric uncertainties on ECGI regularization is critical.
• Existing methods lack robust strategies for identifying sub-optimal λ due to these uncertainties.

Purpose of the Study:

• To introduce a novel method for selecting the regularization parameter (λ) in ECGI using Tikhonov regularization and L-curve optimization.
• To specifically address the influence of electrical noise in body surface potential map (BSPM) signals and cardiac mesh inaccuracies on λ selection.
• To develop a more reliable ECGI solution by accounting for signal and geometry uncertainties.

Main Methods:

• Utilized 19 atrial simulations with varying rhythms and substrate complexities, incorporating white Gaussian noise (40 dB to -3 dB).
• Introduced cardiac mesh displacements (1-3 cm) to simulate geometrical uncertainties and analyzed their effect on the L-curve.
• Quantified the regularization parameter, maximum curvature, and the L-curve's β angle, validating findings with real patient BSPM data.

Main Results:

• Maximum L-curve curvature inversely correlated with signal-to-noise ratio and positioning errors.
• The L-curve's β angle directly correlated with electrical noise and was unaffected by geometrical errors.
• The proposed λ adjustment based on the β angle yielded more reliable ECGI solutions compared to traditional corner-based methods.

Conclusions:

• Adjusting λ using the β angle provides superior ECGI solutions over methods relying solely on the L-curve's corner.
• ECGI activation map information is preserved under uncertainty when the regularization parameter is appropriately selected.
• The proposed regularization parameter selection criteria enhance the accuracy and reliability of ECGI solutions.