Strain-Based Parameters for Infarct Localization: Evaluation via a Learning Algorithm on a Synthetic Database of Pathological Hearts

Abstract

Localization of infarcted regions is essential to determine the most appropriate treatment for patients with cardiac ischemia. Myocardial strain partially reflects the location of infarcted regions, which demonstrated potential use in clinical practice. However, strain patterns are complex and simple thresholding is not su cient to locate the infarcts. Besides, many strain-based parameters exist and their sensitivities to myocardial infarcts have not been directly investigated. In our study, we propose to evaluate nine strain-based parameters to locate infarcted regions. For this purpose, we designed a large database (n=200) of synthetic pathological finite-element heart models from 5 real healthy left ventricle geometries. The infarcts were incorporated with random location, shape and degree of severity. In addition, we used a state-of-theart learning algorithm to link deformation patterns and infarct location. Based on our evaluation, we propose to sort the strain-based parameters into three groups according to their performances in locating infarcts.

Population-Based Personalization of Geometric Models of Myocardial Infarction

“…This model requires setting the total number of iterations and the maximal radius of the spheres (2 parameters). -The second model generates an ellipsoid centered at a given endocardial point, and sets the infarcted region as the intersection between the ellipsoid and the myocardium [8]. It requires setting the short-and long-axes of the ellipsoid, which respectively lie on the radial and circumferential/longitudinal…”

“…Tissue-level geometrical models of the lesions have been proposed based on a regional prior about a given coronary territory [3][4][5] or even up to mimicking the wavefront phenomenon [6] for the lesion propagation around an existing coronary segmentation [7]. Here we rely on two very simple models whose output is rather straightforward to visualize and assess, in 2D, so that we can better focus on the relevance of the personalization process: one that iteratively models an infarct as the union of spheres of random size [4], and one that uses an ellipsoid centered on the endocardium to represent the infarct [8].…”

Domain Adaptation via Dimensionality Reduction for the Comparison of Cardiac Simulation Models

“…Correspondences between meshes were therefore obtained by registering left ventricular surface meshes using the currents representation and the large deformation diffeomorphic metric mapping framework, with the open-source software Deformetrica [14], and finally propagating the transformation to the left ventricular volumetric meshes. For both populations, data from each individual were finally interpolated at the cells and vertices of the template mesh using kernel ridge regression and the estimated mesh correspondences as in [13].…”

“…We are more interested in adapting the data from one database to another one, as pursued globally in [8,1]. A mathematically sound framework for Population Model #1 20 healthy cases processed as in [13]: -11 cases from cMAC-STACOM 2011 [12], -9 cases from our clinical collaborators. Left ventricular segmentations from CVI42 software 1 .…”