Introduction: Cardiac simulations have become increasingly accurate at representing physiological processes, however such simulations often fail to capture the impact of parameter uncertainty in resulting predictions. Such omissions remain an impediment to clinical confidence in many cardiac simulations. Uncertainty quantification (UQ) is a set of numerical techniques that can capture variability of simulations based on model assumptions. While many UQ methods exist, their practical implementation can be challenging. Therefore, cardiac researchers need an accessible, unified, and flexible UQ framework that can be integrated with a variety of cardiac simulation problems. To that end, we created UncertainSCI.
Methods: UncertainSCI is a portable UQ framework that builds polynomial chaos (PC) expansions capable of modeling the forward stochastic error in simulations parameterized with random variables. UncertainSCI uses modern, non-intrusive methods that implement weighted approximate Fekete points, a strategy that parsimoniously explores parameter space. The result is an efficient, stable, and accurate PC emulator that can be quickly analyzed to compute output statistics. We have also created a simple python API to run UncertainSCI, minimizing the number of user inputs needed for guiding the UQ process.
Results: We have implemented UncertainSCI to (1) quantify the variability in boundary element method computed torso potentials in response to uncertainty in the heart position and orientation, (2) quantify the variability in the finite element method computed torso potentials in response to uncertainty in the conductivities of each biological tissue, and (3) quantify variability in the static biodomain simulation of ischemia with respect to uncertainties in the anisotropy ratios.
Conclusions: UncertainSCI utilizes modern mathematics and software engineering to make UQ accessible to cardiac researchers seeking to understand and improve their cardiac simulations. With the UQ support from UncertainSCI, emerging cardiac simulations can be better directed to the applications that will more likely be scientifically certain and clinically trustworthy.