Velocity-Based Cardiac Contractility Personalization with Derivative-Free Optimization

Abstract

Cardiac contractility personalization from medical images is a major step for biophysical models to impact clinical practice. Existing gradient-based optimization approaches show promising results of identifying the maximum contractility from images, but the contraction and relaxation rates are not accounted for. A main reason is the limited choice of objective functions when their gradients are required. For complicated cardiac models, analytical evaluation of the gradient is very difficult if not impossible, and finite difference approximation may introduce numerical difficulties and is computationally expensive. We remove such limits by using derivative-free optimization, and propose a velocity-based objective function on identifying the maximum contraction, contraction rate, and relaxation rate simultaneously with intact model complexity. Experiments on synthetic data show that the parameters are better identified using the velocity-based optimization than the position-based one. Experiments on clinical data show that the framework can obtain personalized contractility consistent to the physiologies of the patients.