Inverse estimation of cardiac activation times via gradient-based optimization

Area of Science:

• Computational biology
• Cardiac electrophysiology
• Medical imaging

Background:

• Patient-specific computational cardiac models integrate multiscale data for heart disease research.
• Cardiac inverse modeling offers a promising approach for personalizing these models.
• Emerging inverse modeling techniques require feasibility assessments.

Purpose of the Study:

• To numerically investigate the identifiability of initial activation sequences using a partial differential equation-constrained optimal control approach.
• To assess the feasibility of personalizing cardiac models from synthetic extracellular potential observations.
• To evaluate the robustness of the method against noise and varying simulation parameters.

Main Methods:

• Utilized a partial differential equation-constrained optimal control framework.
• Employed the bidomain approximation and 2D representations of cardiac tissue.
• Generated synthetic extracellular potential data with varying noise levels and cell membrane kinetics models.

Main Results:

• Successfully recovered activation times and durations of stimuli from noisy synthetic data.
• Demonstrated that electrocardiogram-relevant sampling frequencies (1 kHz) are sufficient.
• Showed that coarser spatial resolutions than standard can be effectively used.
• Confirmed the optimization's effectiveness even when synthetic data used different cell kinetics models.

Conclusions:

• The presented optimal control approach shows potential for identifying cardiac activation sequences from clinical data.
• This method can serve as an extension to existing cardiac imaging techniques for personalized medicine.
• The findings support the clinical application of inverse modeling for guiding cardiac therapies.