Model-based approach to the localization of infarction

Abstract

A model-based approach to noninvasively determine the location and size of the infarction scar is proposed, that in addition helps to estimate the risk of arrhythmias.The approach is based on the optimization of an electrophysiological heart model with an introduced infarction scar to fit the multichannel ECG measured on the surface of the patient's thorax. This model delivers the distributions of transmembrane voltages (TMV) within the ventricles during a single heart cycle.The forward problem of electrocardiography is solved in order to obtain the simulated ECG of the patient. This ECG is compared with the measured one, the difference is used as the criterion for optimization of model parameters, which include the site and size of infarction scar. IntroductionMortality of patients in the immediate postinfarction period, as well as during the first year after myocardial infarction, is most often due to sudden death from ventricular fibrillation [1]. Thus the development of a cardiac model, which would allow to estimate the probability of arrhythmia for a specific patient, must include a correct implementation of ischemia and/or infarction.The aim of this work is to build a model of the patient's heart that contains infarction scar. It is optimized until the simulated multichannel ECG is as near as possible to the measured one. It is assumed that the location and size of infarction obtained in this manner corresponds to the real one. Taking part in the PhysioNet Challenge 2007 [2] should prove or disprove this assumption.Within the scope of the Challenge, 4 data sets were proposed for patients suffering from different infarctions. Each of the data sets contained anatomical information about the heart and the locations of 352 electrodes on the surface of the patient's thorax. Corresponding 352-channel ECGs were also provided. The first two data sets were considered as training and contained the information about the parameters of infarction scar. The main task of the Challenge consisted in the estimation of location and size of infarction scar for the latter two patients based on the pro- Figure 1. Model of the heart used in this work. Atria, ventricles and excitation conduction system are shown. vided data. Methods Cellular automatonA cellular automaton based model of the heart is employed in this work. The anatomy of the heart has been taken from the Visible Male Project [3]. The cardiac geometry is defined on a regular mesh with the resolution of 1 × 1 × 1 mm 3 . This model contains ventricles and atria, sinus and AV-nodes. An excitation conduction system is generated as a tree-shaped structure, with two branches going from the AV-node towards the apex of both ventricles, afterwards giving rise to multiple branches of Purkinje fibers connected to the ventricular endocardium through a large amount of junctions (see figure 1).The cellular automaton is implemented as follows. If an excitation appears in some voxel, it is transferred to all the neighboring voxels containing excitable tissue. The velocity o...

Noninvasive Transmural Electrophysiological Imaging Based on Minimization of Total-Variation Functional

“…Table II lists quantification comparison with the gold standard. As shown, the ischemic centers are correctly identified in both patients and the accuracy is comparable to the best result available [5], [49], [50]. In particular, in case 2 our method shows much higher accuracy in localizing the two separated ischemic regions with segment overlap %, while the best available SO reported in the literature was 33.33.…”

Noninvasive Imaging of Electrophysiological Substrates in Post Myocardial Infarction

“…Table 2 compares the current results with our previous MI evaluation which solely depends on volumetric TMP estimates [7], as well as existent results of case 3 and 4 obtained using case 1 and 2 for training. In brief, Dawoud et al solved the problem by epicardial potential imaging from BSP [5], Farina et al estimated the site and size of spherical infarct models [6], and Mneimneh et al produced the best results using simple ECG signal analysis [13]. In both cases, our results are comparable to the best results and substantially improved over the other two reconstruction results.…”

“…Electrophysiological reconstruction on heart surfaces assumes the substrate to be homogeneous and defines its extent only on heart surfaces [5]. Using a spherical infarct model, [6] estimated its location and size from BSP by deterministic optimization. The predefined infarct shape, however, does not allow flexible data-driven descriptions of intricate 3D infarct structures.…”

An a posteriori error estimator for model adaptivity in electrocardiology

“…The geometry, the space discretization and the fibers direction used are the same as in the previous test case. In this case the conductivity of the tissue is non-homogeneous, the extracellular conductivities being amplified, while the intracellular conductivities are reduced, as suggested in [10].…”