Current-Based 4D Shape Analysis for the Mechanical Personalization of Heart Models

Abstract

Abstract. Patient-specific models of the heart may lead to better understanding of cardiovascular diseases and better planning of therapy. A machine-learning approach to the personalization of an electro-mechanical model of the heart, from the kinematics of the endo-and epicardium, is presented in this paper. We use 4D mathematical currents to encapsulate information about the shape and deformation of the heart. The method is largely insensitive to initialization and does not require on-line simulation of the cardiac function. In this work, we demonstrate the performance of our approach for the joint estimation of three parameters on one heart geometry. We manage to retrieve parameters such that the model matches the 4D observations with an accuracy below the voxel size, in less than three minutes of computation.

A Meshfree Representation for Cardiac Medical Image Computing

“…In the past two decades, though there is a rapidly growth in external model constraints, relatively little attention has been paid to representation and computation strategy for cardiac image analysis [6] – [11] . Over decades, finite element methods (FEMs), and, to a lesser extent, boundary element methods (BEMs), are the most commonly computational strategies for cardiac image analysis because meshing of analysis domains among discrete points has become the standard numerical representation.…”

A self-taught artificial agent for multi-physics computational model personalization

“…A wide variety of manual and (semi-)automatic model parameter estimation approaches have been explored, including Aguado-Sierra et al (2010; Augenstein et al (2005); Chabiniok et al (2012); Delingette et al (2012); Itu et al (2014); Konukoglu et al (2011); Le Folgoc et al (2013); Marchesseau et al (2013); Neumann et al (2014a,b); Prakosa et al (2013); Schmid et al (2006); Seegerer et al (2015); Sermesant et al (2009); Wallman et al (2014); Wang et al (2009); Wong et al (2015); Xi et al (2013); Zettinig et al (2014). Most methods aim to iteratively reduce the misfit between model output and measurements using optimization algorithms, for instance variational (Delingette et al, 2012) or filtering (Marchesseau et al, 2013) approaches.…”