Jana Svehlikova

Individualized model of torso surface for the inverse problem of electrocardiology.

PURPOSE We studied the implementation of a patient-specific torso model created without the use of magnetic resonance imaging in the inverse problem of electrocardiology. METHOD Three types of inhomogeneous numerical torso models were created, with different degrees of adjustment of the outer surface to patients, whereas the heart and lung models remained unchanged. The torso models were used in the inverse localization of small areas with repolarization changes from simulated difference integral QRST maps. The localization error (LE) was evaluated as the distance between the centers of the modeled and the inversely found area with repolarization changes. RESULTS The mean LE was 1.88 cm with the standard torso model. After adapting the torso shape, the mean LE was 1.83 cm, whereas after adapting both, the shape and electrode positions, the mean LE was 1.02 cm. CONCLUSION If torso imaging is not available, a torso model with adapted shape and electrode positions gives only slightly less accurate results.

Influence of Torso Model Complexity on the Noninvasive Localization of Ectopic Ventricular Activity

Abstract Location of premature ectopic ventricular activity was assessed noninvasively in five patients using integral body surface potential maps and inverse solution in terms of a single dipole. Precision of the inverse solution was studied using three different torso models: homogeneous torso model, inhomogeneous torso model including lungs and heart ventricles and inhomogeneous torso model including lungs, heart ventricles and atria, aorta and pulmonary artery. More stable results were obtained using the homogeneous model. However, in some patients the location of the resulting dipole representing the focus of ectopic activity was shifted between solutions using the homogeneous and inhomogeneous models. Comparison of solutions with inhomogeneous torso models did not show significantly different dispersions, but localization of the focus was better when a torso model including atria and arteries was used. The obtained results suggest that presented noninvasive localization of the ectopic focus can be used to shorten the time needed for successful ablation and to increase its success rate.

The effect of conduction velocity slowing in left ventricular midwall on the QRS complex morphology: A simulation study

UNLABELLED Midwall fibrosis is a frequent finding in different types of left ventricular hypertrophy. Fibrosis presents a local conduction block that can create a substrate for ventricular arrhythmias and lead to the continuous generation of reentry. Having also impact on the sequence of ventricular activation it can modify the shape of QRS complex. In this study we simulated the effects of slowed conduction velocity in the midwall in the left ventricle and in its anteroseptal region on the QRS morphology using a computer model. MATERIAL AND METHODS The model defines the geometry of cardiac ventricles analytically as parts of ellipsoids; the left ventricular wall is represented by five layers. The impulse propagation velocity was decreased by 50% in one and two midwall layers, respectively, in the whole left ventricle and in LV anterior region. The effects of slowed conduction velocity on the QRS complex of the 12-lead electrocardiogram are presented as 12-lead electrocardiograms and corresponding values of ECG criteria for left ventricular hypertrophy (ECG-LVH criteria): Gubner criterion, Sokolow-Lyon index (SLI) and Cornell voltage. RESULTS All simulated situations led to increased R wave amplitude in the lead I and of S wave in the lead III, showing a leftward shift of the electrical axis and increased values of ECG-LVH criteria based on limb leads alone or in combination with precordial leads (Gubner criterion, Cornell voltage). The slowed conduction velocity in the whole LV influenced the QRS complex voltage in precordial leads, having an impact on the SLI and Cornell voltage. The changes were pronounced if two layers were involved. CONCLUSION Using computer modeling we showed that the midwall slowing in conduction velocity modified the QRS complex morphology. The QRS complex changes were consistent with ECG-LVH criteria, i.e. QRS patterns usually interpreted as the effect of left ventricular hypertrophy (the increased left ventricular mass).

Influence of individual torso geometry on inverse solution to 2 dipoles.

BACKGROUND The purpose of this study was to observe the influence of variety in individual torso geometries on the results of inverse solution to 2 dipoles. METHODS The inverse solution to 2 dipoles was computed from the measured data on 8 patients using either standard torso with various shapes and sizes of the heart and lungs in it or using various outer torso geometries with the same inhomogeneities. The vertical position of the heart relative to the fourth intercostal level was kept constant in all models. The results were compared with the reference solution computed in standard torso. RESULTS The inverse solution was influenced in 4 of 8 cases by changes of torso geometry and only in 1 of 8 cases by changes of internal inhomogeneities. CONCLUSIONS The use of individual torso geometry with the knowledge of the true heart position is very important for correct inverse results.

Identification of Ischemic Lesions Based on Difference Integral Maps, Comparison of Several ECG Intervals

Identification of Ischemic Lesions Based on Difference Integral Maps, Comparison of Several ECG Intervals Ischemic changes in small areas of myocardium can be detected from difference integral maps computed from body surface potentials measured on the same subject in situations with and without manifestation of ischemia. The proposed method for their detection is the inverse solution with 2 dipoles. Surface potentials were recorded at rest and during stress on 10 patients and 3 healthy subjects. Difference integral maps were computed for 4 intervals of integration of electrocardiographic signal (QRST, QRSU, STT and STU) and their properties and applicability as input data for inverse identification of ischemic lesions were compared. The results showed that better (more reliable) inverse solutions can be obtained from difference integral maps computed either from QRST or from STT interval of integration. The average correlation between these maps was 97%. The use of the end of U wave instead of the end of T wave for interval of integration did not improve the results.

Noninvasive finding of local repolarization changes in the heart using dipole models and simplified torso geometry

Two inverse methods using dipole models for noninvasive assessment of local repolarization changes were investigated and compared in the simulation study. Lesions with changed repolarization were modeled by shortening of the action potential durations in ventricular regions typically influenced by occlusion of coronary arteries. Corresponding body surface potentials were computed using a multiple dipole model of the cardiac generator and an inhomogeneous torso model. Position of each lesion was then estimated by an inverse solution to a single dipole and to a group of five neighbouring dipoles. For both methods the lesion localization error was evaluated and its dependence on the lesion size and the noise in input data was studied. When no noise was present in the input data, the use of the inverse method to a group of dipoles instead of a single dipole resulted in an unsubstantial reduction of the mean localization error of small lesions from 0.6 to 0.5cm. For medium and especially for large lesions the mean localization errors decreased significantly from 1.1 to 0.6cm and from 2.3 to 1.0cm, respectively. The inverse solution to a group of five dipoles was more sensitive to noise. However, for large lesions it still gave better results than the solution to a single dipole if the signal to noise ratio was higher than 30dB.

QRS complex waveform indicators of ventricular activation slowing: Simulation studies

Diffuse or regional activation slowing in ventricular myocardium can result from different cardiac pathologies, such as left ventricular hypertrophy, ischemia or fibrosis. Altered ventricular activation sequence leads to deformations of the activation front and consequently to the changes in the QRS complex. Using a computer model we simulated the effect of slowed ventricular activation on the QRS waveform with a special interest in ECG changes which reproduce the ECG criteria of left ventricular hypertrophy (ECG-LVH). This paper describes results of a set of computer modeling experiments and discusses visual QRS patterns. Slowed ventricular activation in the whole left ventricle resulted in the prolongation of QRS duration, leftward shift of electrical axis, and increase in the QRS amplitude mainly in the precordial leads, having thus their main impact on simulated Sokolow-Lyon index and Cornell voltage. Slowed ventricular activation in the anteroseptal region resulted in a leftward shift of the electrical axis and increased values of ECG-LVH criteria seen in limb leads or in a combination with precordial leads (Gubner criterion, Cornell voltage). Transmural slowing and midwall slowing in two layers in the anteroseptal area led also to the QRS duration prolongation. Changes in QRS complex were more pronounced in the cases of transmural slowing as compared to the left ventricular midwall slowing. Using computer modeling, we showed that slowed ventricular activation is a potent determinant of QRS complex morphology and can mimic ECG patterns that are usually interpreted as the effect of left ventricular hypertrophy, i.e., increased left ventricular mass. These results contribute to understanding the variety of ECG finding documented in patients with LVH, considering not only anatomical enlargement but also the altered electrical properties of hypertrophied myocardium.

Geometrical constraint of sources in noninvasive localization of premature ventricular contractions

The inverse problem of electrocardiography for localization of a premature ventricular contraction (PVC) origin was solved and compared for three types of the equivalent cardiac electrical generator: transmembrane voltages, epicardial potentials, and dipoles. Instead of regularization methods usually used for the ill-posed inverse problems an assumption of a single point source representative of the heart generator was applied to the solution as a geometrical constraint. Body surface potential maps were simulated from eight modeled origins of the PVC in the heart model. Then the maps were corrupted by additional Gaussian noise with the signal-to-noise ratio (SNR) from 20 to 10dB and used as the input of the inverse solution. The inverse solution was computed from the first 30ms of the ventricular depolarization. It was assumed that during this period only a small part of the heart volume is activated thus it can be represented by a single point electrical source. Generally, the localization error was more dependent on the PVC origin position than on the type of the used heart generator. The most stable localization error between the inversely found results and the true PVC origin was not larger than 20mm for PVC origins located in the left ventricular wall and on the right ventricular anterior side. For such cases, the localization was robust to the noise up to SNR of 10dB for all studied types of the cardiac generator. For SNR 10dB the results became unstable mainly for the PVC origins in the septum and posterior right ventricle for the dipolar heart generator and for epicardial potentials defined on the pericardium when the range of the localization error increased up to 50mm. When the results for different electrical heart generators were considered altogether, the mean radius of the cloud of results did not exceed 20mm and the localization error of the cloud center was smaller than that obtained for a particular type of the cardiac generator. Combination of results from different models of a single point cardiac electrical generator can provide better information for the preliminary noninvasive localization of PVC than the use of one type of the generator.

First experience with PVC localization from clinical data

The inverse solution using a dipole as equivalent heart generator was used for assessment of the area of the premature ventricular contraction (PVC) starting point. In two patients with frequent PVC activity, body surface maps were recorded, and the patient specific torso model was created from whole torso CT scan. The position of the assumed ectopic beat origin was computed from the first 40 ms of activation when the activated area is small. The patients also underwent intracardial electrophysiological mapping when the true origin of the PVC beat was defined as the earliest activated point. The obtained inverse results were compared with that position. The inverse solution was computed for 10 chosen ectopic beats for each patient using both, homogeneous and inhomogeneous torso models. The inverse results for both cases were within the area of the PVC beat origin. The use of the more detailed inhomogeneous torso model instead of the homogeneous one did not demonstrate an improvement of localization.

Activation propagation in cardiac ventricles using homogeneous monodomain model and model based on cellular automaton

The activation propagation characteristics obtained when using homogeneous monodomain model (MM) of the cardiac ventricles and the model based on cellular automaton (CA) are compared in this study. The MM comprises the reaction — diffusion equation of the propagation and the modified FitzHugh-Nagumo equations of the electrical excitation of cardiac cells. This model was simulated in Comsol Multiphysics environment. Model based on CA was simulated in Matlab program environment. Realistic activation time of about 80 ms was obtained for the whole ventricles when activation was started in nine analytically defined points. Differences in activation times obtained from the numerical solutions using MM and CA models were less than ±10 ms.