Optimizing Network Simulation of Cardiac Electrical Dynamics.

Modeling and simulation play a critical role in cardiology research. Our recent study has discovered that the structural geometry of a heart can be effectively represented by a network. This, in turn, provides an opportunity to efficiently model and simulate cardiac dynamics using a sparse adjacency matrix. However, realizing the full potential of network simulation is highly dependent on optimization methodologies. The calibration of cardiac models involves substantial complexity. Cardiac electrical dynamics are not only chaotic and nonstationary but also computationally expensive, which poses significant challenges to traditional calibration methodologies. Thus, this paper presents a new statistical metamodeling framework for optimizing network simulation of cardiac electrical dynamics. First, a statistical surrogate is developed to predict the response of the computationally expensive simulation model under different experimental scenarios (i.e., parameter settings). Next, the uncertainty estimate of the statistical metamodel is leveraged to sequentially guide the selection of the next best parameter setting that yields the maximum expected improvement. As such, the optimal parameter setting for cardiac simulation can be efficiently identified through this iterative process. The proposed methodology is evaluated and validated through case studies on both 2D cardiac tissue and a whole heart. Experimental results show that the proposed statistical metamodeling approach efficiently calibrates network simulation of complex spatiotemporal dynamics.