A Two-Stage Model Identification Method for Simulation of Electrical Wave Propagation in Heart Tissue

Abstract

Computer modeling and simulation is fast-moving towards clinical applications to advance cardiac diagnosis and aid treatment planning. As cardiac models become more popular and useful, the need to tailor these models for individual subjects has grown. Therefore, model identification becomes an important step of cardiac modeling. Cardiac model identification is a challenging task, particularly for simulation in high organizational scales such as tissue and organ scales, as the models are nonlinear and involve hidden ion channel gating variables. In this study, we proposed a two-stage model calibration algorithm to estimate the parameters of cardiac tissue model using membrane potential data. Specifically, an ensemble Kalman Filter (EnKF) is used in the first stage to track the membrane potentials and the hidden ion channel gating variables; the outputs from the EnKF are used as the initial values to calculate forward prediction in the second stage. The optimization algorithm minimizes the sum of squared errors in both stages through a coarse and a fine optimization. The proposed method is validated through multiple simulation designs, which shows good accuracy and efficiency.

Keywords

• Cardiac model
• ensemble Kalman Filter
• model calibration
• stochastic optimization