Enabling forward uncertainty quantification and sensitivity analysis in cardiac electrophysiology by reduced order...

Area of Science:

• Computational Biology
• Biophysics
• Cardiovascular Research

Background:

• Cardiac electrophysiology modeling is crucial for understanding heart function.
• Uncertainty quantification (UQ) in these models is computationally intensive.
• Existing methods struggle with the high cost of full-order models.

Purpose of the Study:

• To develop a computationally efficient framework for forward UQ in cardiac electrophysiology.
• To integrate sensitivity analysis and uncertainty propagation.
• To reduce the computational burden of high-fidelity cardiac models.

Main Methods:

• Utilized the monodomain and Aliev-Panfilov models for cardiac electrical and ionic activity.
• Employed projection-based reduced-order models (ROMs) to decrease state-space dimensionality.
• Incorporated artificial neural network (ANN)-based models to account for ROM approximation errors.

Main Results:

• The proposed physics-based ROMs significantly reduced computational costs compared to full-order models.
• The framework successfully performed variance-based global sensitivity analysis and uncertainty propagation.
• ANNs enhanced the accuracy of the UQ pipeline by correcting ROM approximation errors.

Conclusions:

• The developed framework offers a computationally efficient solution for UQ in cardiac electrophysiology.
• Physics-based ROMs combined with ANNs provide a superior alternative to traditional regression-based emulators.
• This approach enables more accurate and efficient investigation of intra-subject variability in cardiac models.