Comprehensive Uncertainty Quantification and Sensitivity Analysis for Cardiac Action Potential Models.

Area of Science:

• Computational biology
• Biophysics
• Mathematical modeling

Background:

• Computational models are crucial for healthcare predictions, necessitating rigorous evaluation via uncertainty quantification (UQ) and sensitivity analysis (SA).
• Current limitations hinder comprehensive UQ/SA for complex whole-heart models due to parameter variability and measurement challenges.
• Existing standards like ASME V&V40 highlight the need for robust model validation.

Purpose of the Study:

• To develop a feasible framework for comprehensive UQ/SA in cardiac cell models.
• To assess the robustness of a novel action potential (AP) model to parameter uncertainty.
• To identify key parameters influencing cardiac model behavior and potential "model failure".

Main Methods:

• Developed a moderately complex cardiac action potential (AP) model with 36 parameters.
• Prescribed input variability for all model parameters and used a hyper-parameter to scale uncertainty.
• Performed UQ and SA across various physiologically relevant model outputs, analyzing parameter influence and model behaviors.

Main Results:

• Demonstrated that the simulated AP is robust to low levels of parameter uncertainty, a novel finding for cardiac models.
• Observed diverse dynamics, including oscillatory behavior, at higher parameter uncertainty levels.
• Identified five specific parameters as highly influential on observed dynamics and potential model failure.

Conclusions:

• The developed framework enables comprehensive UQ/SA for cardiac cell models, addressing feasibility limitations.
• This approach facilitates the assessment of model robustness and the mitigation of "model failure" under uncertainty.
• The study represents a significant step towards reliable, uncertainty-aware computational tools for clinical decision-making.