Sensitivity of a data-assimilation system for reconstructing three-dimensional cardiac electrical dynamics.

Area of Science:

• Computational biology
• Biophysics
• Cardiovascular modeling

Background:

• Cardiac electrical modeling offers mechanistic insights but faces challenges with parameter uncertainty.
• Data assimilation combines computational models with experimental observations for improved state reconstruction.

Purpose of the Study:

• To extend data-assimilation techniques for reconstructing 3D cardiac electrical activity from surface voltage observations.
• To evaluate the impact of algorithmic and model parameters on reconstruction accuracy using synthetic data.

Main Methods:

• Utilized an ensemble Kalman filter for data assimilation.
• Reconstructed spatio-temporal dynamics of cardiac electrical states.
• Investigated sensitivity to algorithmic parameters and robustness to model errors.

Main Results:

• Achieved acceptable reconstruction error in many scenarios.
• Identified weakest performance in model-error conditions and extreme parameter regimes.
• Observed error increase after initial decrease when ensemble spread was reduced.

Conclusions:

• Data assimilation with ensemble Kalman filters shows promise for accurate cardiac state reconstruction.
• Ensemble spread management is crucial; increasing it may improve performance.
• Future work should explore additive inflation or multi-model ensembles for enhanced robustness.