Noninvasive Electrocardiographic Imaging
Reconstruction of Epicardial Potentials, Electrograms, and Isochrones and Localization of Single and Multiple Electrocardiac Events
Background The goal of noninvasive electrocardiographic imaging (ECGI) is to determine electric activity of the heart by reconstructing maps of epicardial potentials, excitation times (isochrones), and electrograms from data measured on the body surface.
Methods and Results Local electrocardiac events were initiated by pacing a dog heart in a human torso–shaped tank. Body surface potential measurements (384 electrodes) were used to compute epicardial potentials noninvasively. The accuracy of reconstructed epicardial potentials was evaluated by direct comparison to measured ones (134 electrodes). Protocols included pacing from single sites and simultaneously from two sites with various intersite distances. Body surface potentials showed a single minimum for both single- and double-site pacing (intersite distances of 52, 35, and 17 mm). Noninvasively reconstructed epicardial electrograms, potentials, and iso-chrones closely approximated the measured ones. Single pacing sites were reconstructed to within ≤10 mm of their measured positions. Dual sites were located accurately and resolved for the above intersite distances. Regions of sparse and crowded isochrones, indicating spatial nonuniformities of epicardial activation spread, were also reconstructed.
Conclusions The study demonstrates that ECGI can reconstruct epicardial potentials, electrograms, and isochrones over the entire epicardial surface during the cardiac cycle. It can provide detailed information on local activation of the heart noninvasively. Its uses could include localization of cardiac electric events (eg, ectopic foci), characterization of nonuniformities of conduction, characterization of repolarization properties (eg, dispersion), and mapping of dynamically changing arrhythmias (eg, polymorphic VT) on a beat-by-beat basis.
The goal of noninvasive electrocardiographic imaging (ECGI) is to provide detailed spatial and temporal information on the electric activity of the heart from body-surface measurements. Conventionally, standard electrocardiography is used to obtain such information; however, it samples the body surface potential (BSP) only at a limited number of sites. From the ECG, the location of electric events in the heart can be inferred only in broad terms with very low resolution and a high degree of uncertainty. Moreover, one cannot mathematically reconstruct electric events in the heart from limited-lead ECG information because this requires the potential distribution over the entire torso (closed surface).
Studies demonstrate that some regional cardiac events can be discerned with multielectrode body surface potential mapping (BSPM).1 2 3 However, BSPM is limited in its ability to localize cardiac electric events because the potential at every point on the body surface is determined by the electric activity in the entire heart; it is also limited in its ability to resolve multiple electric events because the volume conductor between the heart and body surfaces smoothes the potential distribution.4 5 In contrast, epicardial potential maps reflect details of cardiac electric activity with high resolution, including the location of multiple myocardial events that cannot be resolved with BSPM.5 6 7 8 Because of the above limitations of BSPM, it has been the goal of many researchers to reconstruct cardiac electric events noninvasively from BSP data. Some have developed solutions in terms of equivalent intracardiac sources (for review, see Reference 99 ); others, directly in terms of epicardial isochrones.10 11 The approach taken in this work (as well as in our earlier work12 13 14 and in the work of others15 16 ) is to compute epicardial potentials noninvasively from the BSPM and to use these potentials to compute temporal electrograms and isochrones. The choice of epicardial potentials as the solution is supported by the established use of epicardial potential mapping as an important research and clinical tool and by its demonstrated ability to detect cardiac electric events with high resolution.17 In addition, the computed epicardial distribution is unique (only one solution exists),18 it can be more directly related to underlying physiological processes than equivalent sources can, and it can be evaluated (as done in this study) by direct comparison with measured epicardial maps.
It should be recognized that both epicardial potential mapping and excitation time (isochrone) mapping have specific advantages. Excitation time plots summarize the spread of excitation in one map. However, although the excitation times can be inferred from the sequence of potential maps, the reverse is not true, because extensive additional information is contained in the potential distribution. Moreover, in addition to excitation data, the potential distribution contains information on the repolarization and recovery processes as well. Potential maps also have the advantage of reflecting intramural activity before the activation front reaches the epicardium.8
One can envision many potential diagnostic and experimental uses for noninvasive ECGI. In the experimental setting, this would allow study of cardiac activation in the intact animal. This could potentially include studies of activation patterns during arrhythmias of various types, of effects of antiarrhythmic drugs and nonpharmacological (electric) interventions on such patterns, and of effects of neural activity on the cardiac activation and repolarization processes in the nonanesthetized animal. Potential uses in the clinical setting could include noninvasive localization of the origin (or origins) of ectopic activity, obtaining information on the activation sequence during a reentrant rhythm on a beat-by-beat basis, and evaluation of the efficacy of antiarrhythmic interventions (eg, drug, ablation) in a given patient.
Another important potential use of ECGI is in the area of screening and identifying patients with high vulnerability to arrhythmogenesis. Several ECG indices have been suggested as noninvasive measures of arrhythmia vulnerability. These include late potentials,19 T-wave alternans,20 21 and QT dispersion.22 For example, large QT dispersion in the 12-lead ECG has been suggested to reflect inhomogeneity of repolarization and to indicate increased vulnerability. Body surface maps of the spatial distribution of QT intervals show that zones of longest QT shift on the body surface during acute myocardial infarction.1 However, a fundamental limitation of ECG and BSPM is that they are measured on the body surface. Because the ECG provides a remote, integrated measure of complex cardiac events (it is a summation of contributions from all regions of the heart), it cannot be related to actual spatial differences and local properties of activation and repolarization in the heart itself. ECGI opens the possibility to reconstruct these properties on the epicardial surface noninvasively with the following advantages: (1) It is likely to enhance detection and provide better indices of vulnerability than the body surface ECG, which attenuates epicardial information and averages spatial nonuniformities; (2) it can provide a direct link to underlying mechanisms (eg, relate QT dispersion in the ECG to spatial dispersion of ventricular repolarization); and (3) it can relate ECG observations to localized processes in the heart and determine the location of the arrhythmogenic activity (eg, myocardial regions of prolonged activation-repolarization intervals that increase QT dispersion in the ECG). In this context, ECGI can be used to assess the effect of a pharmacological agent on the degree of epicardial dispersion, thereby evaluating its antiarrhythmic efficacy in a given patient.
Our previous work focused on the study of normal sinus rhythm (NSR) and tested the ability of ECGI to reconstruct important events (eg, right ventricular epicardial breakthrough) during NSR.12 14 During NSR, ventricular activation spreads rapidly along the endocardium with the aid of the Purkinje system and then propagates more slowly in a broad wave front from endocardium to epicardium.23 With activation from ectopic stimulation (unless subendocardial), there is no early involvement of the Purkinje system; the activation proceeds in all directions from the point of stimulation, with a faster velocity along fibers than across them. The faster activation along fibers is associated with stronger (dipolar) electric sources that dominate the pattern of the associated electric potential.2 24 Extensive research, both theoretical and experimental, has been conducted to understand the nature of this electric activity and its manifestation as potential distributions on the epicardial surface.8 24 25 26 It was found that early after pacing, a region of intense epicardial negativity forms superficial to the location of the pacing site. The epicardial potential distribution also contains one or two positive maxima that flank the region of negativity; a line drawn through the maxima regions and the center of the negative region has been shown to reflect the orientation of the myocardial fibers at the depth of intramural pacing. Similar patterns are observed in potential distributions measured on the endocardial surface during myocardial pacing.27 This pattern of the potentials is a direct consequence of the fiber orientation and can be explained by the oblique double layer model.2 24 26
The present study evaluates the ability of ECGI to provide information noninvasively on the electric activity of the heart. Specifically, it assesses the ability of this new functional imaging modality to reconstruct epicardial electric events. These include single or multiple sites of initial activity, epicardial potential patterns associated with single- or dual-site pacing, temporal epicardial electrograms in different regions of the heart, and epicardial activation sequences (including spatial nonuniformities of the epicardial spread of activation, ie, regions of sparse or crowded epicardial isochrones). Pacing sites in this study are all epicardial and are located both in the anterior and posterolateral regions of the heart. Single-site pacing provides a measure of the localization accuracy of ECGI; dual-site pacing (with varying intersite distances) provides a measure of its spatial resolution. In addition, dual-site pacing evaluates the ability of ECGI to reconstruct complex potential patterns, electrograms, and activation sequences when multiple, interactive wave fronts are present in the heart. The pacing protocols can also be viewed as simulating single or multiple ventricular ectopic foci. Focal sites of initiation of arrhythmogenic activity can result from abnormal automaticity, triggered activity, or microreentry.28 29 Because the focus is usually confined to a small region of the myocardium, it can be simulated by pacing the myocardium at a single site. Locating the ectopic focus (or multiple foci) is important for diagnosis and for guiding an interventional therapeutic procedure (eg, ablation).30
The following is a brief summary; details can be found in References 13, 14, 18, 31, and 3213 14 18 31 32 . ECG imaging requires solving the inverse problem of electrocardiography, and there are two major aspects to the computational methodology. The first entails the discretization of Laplace’s equation (using Green’s second theorem33 ) in the volume between the epicardial surface and the body surface; the boundary element method is used to accomplish this task.34 This discretization results in the following linear matrix relationship: where VE is the vector of epicardial potentials, VT is the vector of torso potentials, and A is the transfer matrix between heart and torso, which depends only on the geometry and the conductivities of the media in the volume. Equation 1 represents the forward problem of electrocardiography, that is, a computation of BSPs from potentials on the epicardium. The second aspect of the computational methodology is inverting Equation 1 to obtain an expression for the epicardial potentials in terms of the BSPs. The ill-posed nature35 of the inverse problem in electrocardiography (ie, its instability in the presence of noise) requires regularization of the solution. In the present study, Tikhonov zero-order regularization35 is used to stabilize the solution. It entails finding the epicardial solution, VE, that minimizes the following objective function: The first term in Equation 2 represents the least-squares solution of Equation 1. The second term in Equation 2 is a regularization term that imposes bounds on the amplitude of the solution. The regularization parameter, t, controls the degree of the imposed constraint. It provides a balance between the accuracy and stability of the solution, resulting in a close estimate of the epicardial potential distribution that is also stable. In this work, the regularization parameter, t, is found with the CRESO method15 and has been found to perform comparably to the “optimal” t that provides the minimum root mean square error between the computed VE and the measured VE.13 The CRESO t depends only on the vector VT and the matrix A. Computing the epicardial potentials VE, therefore, is completely noninvasive, requiring only the knowledge of the geometry (for the matrix A) and the torso electric potentials VT.
The inverse epicardial solutions were verified by use of a human-shaped torso tank similar to that used in our previous work13 14 31 but with many improvements (Fig 1A⇓). The tank, molded from the torso of a 10-year-old boy, was filled with an electrolytic solution and contained an isolated dog heart suspended in the proper human anatomic position. A second dog served to provide circulatory support for the isolated heart, with a modified Langendorff preparation that was stable for at least 4 to 5 hours. The electrode system of the tank for measuring electric potentials (Fig 1A⇓) consisted of 16 rows, each with 24 electrodes equally spaced on the body surface in the polar angle, θ. In addition to the 384 body surface electrodes, there were 918 electrodes along 384 rods that projected from the body surface into the volume toward the heart. All rods in the lower 6 rows were fixed; the rods in the upper 10 rows were fixed only in the cylindrical coordinates z (defining a given horizontal cross section) and θ (the polar angle). They were free to move in the radial direction and were pushed inward toward the heart once it was suspended in the tank (see Fig 1⇓). The rod tips were ≈1 cm from the surface of the epicardium. Potentials were also measured on a 64-electrode sock in direct contact with the epicardium. Fig 1B⇓ shows the rod-electrode configuration for one cross section of the tank. The innermost contour depicts the approximate heart boundary at this level. The next contour connects the rod-tip electrodes (“epicardial envelope”), and the outermost contour represents the body surface. The four asterisks on the inner contour depict the positions of the four pacing sites (see below). Fig 1C⇓ displays four overlapping views of the epicardium. Anatomic landmarks are displayed and identified beneath the plot. In addition, the four asterisks identify the pacing sites that are also displayed on the inner contour of Fig 1B⇓.
For the purposes of this work, the electric potentials measured on the torso provided the input to the inverse procedure, and the potentials measured at the rod tips, which constituted an effective envelope around the heart, provided the experimental verification of the inverse solution. The epicardial envelope was used instead of the epicardial sock because it provided more controlled conditions for evaluating the reconstruction procedure. The geometric positions of the rod-tip electrodes that composed the epicardial envelope were measured directly and in the same frame of reference as the torso electrodes. In contrast, the positions of the sock electrodes were known only relative to each other and in their own reference frame. Because of this inaccuracy, the epicardial envelope provided a more controlled environment in which to study and test the inverse methodology without additional uncertainty caused by geometric error. A potential disadvantage of using the epicardial envelope is the possible degradation of the potential patterns due to their distance from the epicardium. It was found (not shown here) that, in general, the potential patterns are not degraded: only potential amplitudes may be somewhat lower on the envelope compared with the sock. The data were acquired with a multiplexing system; 1366 electrodes divided among 8 banks (192 leads at a time). The data banks were time-aligned by use of 5 leads common to all banks. The signals were unipolar with reference to Wilson’s central terminal. Other characteristics were a sampling rate of 1 kHz (with sample and hold circuits), 12-bit resolution, and on-line amplification. All signals were gain-adjusted with calibration signals, and the baseline was subtracted.
Pacing was performed with subepicardial electrode pairs on four needles in the left ventricle and an additional electrode pair on the right atrium near the sinoatrial node. The four ventricular pacing sites were located along an imaginary line parallel to the atrioventricular groove and approximately halfway between it and the apex. Site 1 was located near the septum (see Fig 1B⇑ for a cross-sectional view and Fig 1C⇑ for a surface view); sites 2, 3, and 4 were at locations to the anatomic left of site 1, such that site 4 was located on the posterolateral left ventricle. Intersite distances between sites 1 and 2, between 2 and 3, and between 3 and 4 were ≈15, 20, and 17 mm, respectively. Each site was paced individually. Simultaneous dual pacing of sites 1 and 4, sites 1 and 3, and sites 3 and 4 tested the resolution of the inverse solution at intersite distances of ≈52, 35, and 17 mm, respectively. Because of the eccentric location of the heart within the torso, the various epicardial pacing sites were located at various distances from the chest wall. Measured along a line perpendicular to the epicardium, the distances from each of the four sites (1 to 4) to the chest wall were ≈19, 26, 82, and 99 mm, respectively. Ventricular pacing was done with current pulses of 2-ms duration and intensity just above threshold (generally 0.2 to 0.5 mA). Stimuli were delivered simultaneously to the ventricular pacing leads and to the right atrial pacing leads to prevent sinus beats from capturing the ventricles. The cycle length of pacing was ≈380 ms.
Fig 2⇓, modified from Taccardi et al,8 is a schematic of the basic epicardial patterns of potentials and currents associated with epicardial pacing. As discussed in the introduction, with epicardial pacing (central asterisk), a region of negativity forms (inside the ellipse in the figure). Because of the preferential activation along the fibers (shown as gray lines in the background), maxima (plus signs) form outside this negative region along the axis of the fibers. Notice that two corresponding minima (minus signs) form inside the negative region and that neither minimum coincides with the pacing site. This potential pattern is consistent with an equivalent source configuration of two opposite axial dipoles pointing from each minimum toward its corresponding maximum (see Fig 11⇓-8 and related discussion in Reference 22 ). On the basis of these properties, we determine the pacing site to be at the center of the region of negativity in both the measured and the inverse-computed maps. If we detect only one minimum in the region of negativity, it is because the epicardial potentials are determined with insufficient resolution to separate the two minima. In such cases, the center of the elliptical negative region is still used to identify the site of pacing. Ideally, the negative region is quasi-elliptical. In the experiments, this region approximates an ellipse, particularly during the early stages of propagation. We determine its center as the intersection point of the major and minor axes of an ellipse that best fits the negative region.
The torso potential data are displayed as contour maps in two views (anterior and posterior). The epicardial potential and isochrone data are shown as contour maps in four overlapping views (see Fig 1C⇑). A stylized epicardium very close in shape to the real epicardial rod-tip envelope was created for this display so that the epicardial potentials could be presented on a smooth surface. This prevented misinterpretation of the potential maps caused by small surface discontinuities when the real, nonsmooth epicardial (rod-tip) surface was projected onto two dimensions.
Isochrones for both measured and reconstructed epicardial data are computed by taking the time of epicardial activation at a given location as the time of maximum negative dV/dt of the temporal electrogram (“intrinsic deflection”) at that location. For the computed epicardial maps, first a potential map for each time frame is computed, then the time series of maps is organized by lead to provide temporal electrograms, and finally the intrinsic deflection for each lead is determined. In some cases, it is clear that the automated procedure for computing isochrones chooses the incorrect time of activation. This can happen when the activation time is found in a region of high-frequency noise or where there are discontinuities in the electrogram caused by the quasi-static inverse reconstruction of discrete maps in time. In these situations, the time of activation is corrected manually by taking information from the neighboring electrograms into account. Note that these isochrones are constructed on the epicardial envelope defined by the rod-tip electrodes and not directly on the epicardium. In this sense, they constitute “pseudoisochrones” on a surface close to the epicardium but not on the epicardium itself.
Single Pacing Sites
Fig 3⇓, plate 1 (top) shows the electric potentials for the first epicardial pacing site (site 1, see “Methods”), and its format is similar to that of Fig 3⇓, plates 2 and 3. Measured torso potentials (anterior and posterior views) are displayed on the left. The top and bottom rows on the right portion of the plate show the measured and inverse-reconstructed epicardial potential distributions, respectively. Each is displayed with the four overlapping views of Fig 1C⇑. All potentials are displayed as color contour plots; see figure legend for format details.
In this color plate, the first (1, most anterior) pacing site is stimulated, and an intense minimum (dark blue) is seen on the anterior view of the measured epicardial potentials (top row, right). In the torso potential distribution, there is also a single minimum anteriorly, demonstrating that the epicardial potentials are reflected on the body surface. With the torso potentials used to reconstruct the epicardial potentials noninvasively (bottom row, right), the intense minimum is reconstructed. In addition, the maxima flanking the minimum are seen in both the measured and the computed epicardial plots. The pacing site is marked with an asterisk. The reconstructed pacing site is located ≈7 mm from its measured location. The two maxima are reconstructed 20 and 28 mm, respectively, from their measured locations.
In Fig 4⇓, the most posterolateral pacing site (site 4) is shown. The format for this and subsequent black-and-white contour plots is described in the figure legend. In this figure (as in Fig 3⇑, plate 1), one minimum is also seen on the torso. This minimum, however, is shifted more posteriorly than that of Fig 3⇑, plate 1; the anterior maximum is more prominent here. In the measured epicardial potentials, one sees a single minimum and an asterisk in the center of an elliptical negative area that reflects the pacing site, with two flanking maxima. In the noninvasively computed epicardial potentials, the negative area contains two minima. These dual minima constitute the expected potential pattern as described in the discussion of Fig 2⇑. In our experiments, these minima are rarely detected, because the resolution (electrode density) is not sufficient. It is interesting that the minima are noninvasively reconstructed in their expected positions in the computed map despite their not being detected in the measured map. This can be explained by the fact that while the measured data are obtained with only 4 or 5 electrodes over the entire region of negativity, the computed potentials use data from the entire body surface. It is conceivable, then, that the computed potentials may have a better resolution than the measured potentials. The reconstructed pacing site (center of the ellipse, asterisk) is ≈4 mm from the position of the measured site. The positions of the reconstructed maxima are 0 and 28 mm from their respective measured locations. The error is reported as location errors of the noninvasively reconstructed maxima, minima, and pacing sites (relative to the measured ones) on the effective epicardial surface (composed of rod-tip electrodes). Because in the regions of these pacing sites the nodes (rod-tip electrodes) are spaced on the order of 15 mm apart, if the reconstructed extremum is shifted by only one electrode position, it will have an error of ≈15 mm; two electrode positions, ≈30 mm. Note that this error, reported on the effective epicardial surface, is larger than it would be on the epicardium itself because the effective epicardial surface is ≈1 cm outside of the epicardium, and distances on this surface are greater than their projections on the actual epicardium. For example, the rod-tip electrodes that measure the minima of pacing sites 3 and 4 are located ≈26 mm apart, but when the same rods are pushed in so that the rod tips touch the epicardium, these same rod-tip electrodes are ≈17 mm apart. One must therefore bear in mind that the actual epicardial error may be less than the reported error on the epicardial envelope.
Fig 5⇓ shows potentials for the two middle pacing sites (sites 2 and 3). The torso potentials are not shown, and only the epicardial views that contain relevant information (anterior and left views) are shown. For pacing site 2, the inverse-reconstructed pacing site is located 10 mm from its measured location, and the reconstructed maxima are located 0 and 27 mm from their measured locations. The center of the ellipse of site 3 is reconstructed to its exact measured location; the maxima are reconstructed 18 and 28 mm from their respective measured locations.
Fig 6⇓ shows a summary of the ability of ECGI to locate noninvasively the single pacing sites, positioned from anterior to posterolateral (again, only the anterior and left views are shown). In this figure, the epicardial mesh is shown instead of the contour plots, and only the pacing sites, identified by asterisks, are depicted. One can see that the progression of the pacing sites from anterior to posterolateral across the epicardium is reconstructed correctly in the noninvasively computed maps (bottom row; compare with measured sites in top row). The error in the location of the reconstructed pacing sites averages only 5 mm from their measured positions.
Multiple Pacing Sites
Fig 3⇑, plate 2, shows the potential distributions for two simultaneous pacing sites (1 and 4). The measured epicardial potential maps (top row, right) show the two intense negative areas that reflect the two sites of early epicardial activation. The torso potential distribution looks very much like the superposition of the torso potential distributions from the individual pacing sites (Fig 3⇑, plate 1, and Fig 4⇑), but notice that although the area of the minimum potential is broader in Fig 3⇑, plate 2, than in Fig 3⇑, plate 1, there is only one minimum. With the torso data used to reconstruct the epicardial potentials (bottom row, right), however, both epicardial minima are reconstructed as approximate negative ellipses, and the position errors of their center points relative to the measured ones are 7 and 4 mm, respectively.
Sites 1 and 4 are ≈52 mm apart on the heart; the distance between their reflected minima on the effective epicardial envelope is ≈65 mm. Sites 1 and 3 (anterior sites) are closer, at an intersite distance of ≈35 mm on the heart and 40 mm on the envelope (Fig 3⇑, plate 3). The torso potentials look almost identical to those of the single pacing site in Fig 3⇑, plate 1. The reconstructed epicardial potentials, however, resolve both minima distinctly and in their exact locations. A more challenging test of the method is the two posterolateral pacing sites (3 and 4), because they are located in a region of the heart that is farther from the chest wall, and also because the pacing sites are closer together (≈17 mm on the heart; 26 mm on the epicardial envelope). In Fig 7⇓, the noninvasively computed potentials reconstruct both posterolateral pacing sites distinctly. The error of position is ≈5 mm and 4 mm for sites 3 and 4, respectively. Again, the BSPs show only one minimum, failing to reflect these two simultaneous pacing sites.
Fig 8⇓ demonstrates the noninvasive reconstruction of temporal epicardial unipolar electrograms (format described in figure legend). In Fig 8A⇓, the usual four views of the epicardial surface are shown. Sample electrograms are displayed for sites close to (sites 1, 2, and 3), partially away from (4, 5, and 6), and far away from (7, 8, and 9) the pacing site. In panels B, C, and D, both measured and computed electrograms are displayed. Three main types of waveforms—monophasic negative (B), biphasic (C), and monophasic positive (D)—are reconstructed. Notice the close resemblance of the noninvasively reconstructed electrograms compared with the measured epicardial electrograms. The CCs for the plotted electrograms are printed in the figure and range from 0.810 to 0.998. Looking at the electrograms from the entire epicardial surface (not displayed), CC is >0.9 for 72% of all epicardial electrodes (54% with CC>0.95). For some outliers, the value of CC is poor, but CC is <0.5 in only 5% of the electrodes. A number of the computed electrograms show discontinuities of potential (“jagged” appearance) that do not exist in the measured electrograms. The reason for these discontinuities is that each time frame is computed with the quasi-static Tikhonov regularization scheme, in which each time frame is computed independent of all other time frames. In our procedure, we compute epicardial maps at discrete intervals of 1 ms. One would expect temporal discontinuities in the electrograms, because they are constructed from discrete maps without the application of a temporal smoothing procedure.14
Isochrones were computed for both the measured and the noninvasively reconstructed epicardial potentials. Fig 9⇓ shows the isochrones for site 1. Notice that the regions of earliest activation are reproduced in the computed isochrones, matching the measured ones. Notice also the spatial nonuniformities of isochrone density. In the measured isochrones, there are regions with apparent faster spread of epicardial activation (right, left, and posterior views) where the isochrones are sparse and regions with relatively slow activation spread where isochrones are crowded. For example, in the anterior view, there is a region of relatively slow spread between 47 and 69 ms, as evident by the crowding of isochrones in this area. Both the “fast” and “slow” areas are reproduced in the noninvasively computed isochrone map.
Figs 10⇓, 11⇓, and 12⇓ show the activation sequences (isochrones) associated with the double pacing sites of Fig 3⇑, plates 2 and 3, and Fig 7⇑, respectively. Notice that in all three figures, the noninvasively reconstructed patterns of activation (epicardial isochrones) closely resemble the measured ones. In all but Fig 12⇓, two distinct sites of earliest ventricular activation are reconstructed. This is true for sites that are 52 or 35 mm apart, anterior or posterolateral. Even in Fig 12⇓, in which the sites are posterolateral and only 17 mm apart, the first isochrone is elongated. This pattern implies a large region of earliest activation, which is consistent with the presence of more than a single pacing site.
The results of this study demonstrate that ECGI is capable of noninvasively reconstructing epicardial potentials, temporal electrograms, and isochrones from potentials measured on the body surface with good accuracy and resolution. The study evaluates the ability to identify, locate, and resolve single and multiple electric events, as simulated by ventricular pacing. With anterior and posterolateral pacing, BSPM is able to detect the existence of an event taking place anteriorly or posterolaterally. Conventional ECG would show this as well, and the fact that the QRS is wide could help identify the rhythm as ventricular. However, even the BSPM is unable to identify, with acceptable accuracy, the location of the electric event in the heart. Measured epicardial potentials reflect single or double pacing sites and identify their locations with good accuracy. The inverse computations reconstruct these locations noninvasively with an error of ≤10 mm. This result implies that the method is capable of reconstructing the location of arrhythmogenic foci with this accuracy. Further evaluation is needed in the setting of actual ventricular tachycardias and in the presence of structural heart disease.
BSPM is also limited in attempts to characterize the electric activity of the heart for more than a single event. When two sites are paced, even as far apart as 52 mm, the torso potentials show only one minimum (see Fig 3⇑, plates 2 and 3, in which the torso potential distributions for two pacing sites are very similar to that of plate 1 for one pacing site). With ECGI, however, sites as close as 17 mm apart on the heart are reconstructed well and the individual sites are resolved (Fig 3⇑, plate 2 and 3, and Fig 7⇑). This is true for the anterior minima, for which the pacing sites are 19 and 26 mm, respectively, from the chest wall, and also for the posterior minima, for which the pacing sites are as far as 82 and 99 mm from the chest wall. This has specific clinical relevance when more than one electric event is taking place simultaneously, especially if surgical or catheter ablation is indicated. It would then be imperative to know the number of regions of arrhythmogenic activity (eg, two preexcitation sites) and exactly where they are. In a more general sense, simultaneous pacing from two sites at various distances provides a measure of the ability of ECGI to resolve multiple electric events. It also demonstrates its capability to reconstruct epicardial potentials, electrograms, and isochrones for complex excitation patterns (ie, in the presence of multiple, interacting wave fronts) that can typically be present during cardiac activation.
In this study, we used dual pacing to generate complex patterns that result from more than a single wave front. However, further investigation is needed with patterns of activation that are even more complex. Studies by others36 using patient data showed that complex patterns were not reconstructed as accurately as simple patterns. However, our earlier studies12 14 of NSR, in which multiple activation fronts and breakthrough sites generate complex epicardial patterns, demonstrated that ECGI was able to reconstruct these potential distributions with good accuracy. The study mentioned above36 has two main sources of error: uncertainties in epicardial geometry and inconsistencies caused by the fact that measured and computed epicardial potentials were obtained in open and closed chest environments, respectively, and not simultaneously. Although these attempts to apply inverse methodology directly to patient data are promising, they also demonstrate the inherent difficulty of evaluating the reconstruction procedure in patients. These pioneering studies highlight the need for method testing in a more controlled environment. The possibility of direct evaluation under controlled conditions and known geometry is imperative for the success of this approach and for its development as a clinical and research tool. The studies in our laboratory were performed with an experimental setup that provides an ideal opportunity for evaluating the relationship between epicardial and torso potential distributions, electrograms, and isochrones in a controlled environment that closely resembles real life. It involves an isolated dog heart placed inside an exact replica of a human torso in the correct anatomic position. This setup, while approximating in vivo conditions, permits simultaneous recordings from the body surface and from the heart, thereby allowing direct comparison of the reconstructed epicardial potentials, electrograms, and isochrones with the measured ones without the above-described errors. We hope that studies in human subjects now in progress in our laboratory will further define the capabilities of ECGI, especially in the detection of more complex patterns.
In this study, the use of an epicardial envelope (rather than the epicardial sock; see “Experimental Methods” section) is beneficial because it contributes to the well-controlled conditions in terms of the geometry. In our laboratory, we have begun the initial stages of implementation and evaluation of ECGI in the clinical setting, in which we obtain the heart-torso geometry from computed tomography.37 The concept of using the epicardial envelope is attractive in clinical applications for the following reasons: (1) The potential distribution on an epicardial envelope is an accurate reflection of that on the epicardium; (2) the epicardial envelope is fixed throughout the cardiac cycle, free from the motion artifact introduced by contraction and filling of the heart; and (3) use of an epicardial envelope that encloses the heart at end diastole ensures that the volume between this envelope and the body surface is free of cardiac electric sources at all times, a necessary assumption of the inverse methodology.
All of the epicardial potential maxima are reconstructed with a worst location error of 28 mm, or to within 1 to 2 electrode positions. As stated in the introduction, the orientation of the measured epicardial potential patterns (relative orientation of maxima and minima) reflects the epicardial fiber orientation in the vicinity of the stimulation site. With this level of accuracy, the general orientation of the maxima and hence of the fibers in the vicinity of the pacing site can be determined noninvasively. This will be especially important in attempts to determine the intramural depth of stimulation because the myocardial fibers rotate in a counterclockwise direction from the epicardium to the endocardium. Therefore, we may expect that the orientation of the epicardial maxima with respect to the center of the negative area in the early time frames will reflect not only the orientation of the fibers per se but also the depth of the stimulation site8 (or of the arrhythmogenic focus in the clinical application).
ECGI reconstructs electric data on the epicardium. The ability to infer endocardial and septal activity from epicardial potentials is limited. In a parallel effort, we are developing similar methodology for reconstructing potentials, electrograms, and isochrones on the entire endocardial surface (including septal) from noncontact catheter measurements.27 38 Together, data on both the epicardial and endocardial surfaces of the heart will provide extensive information on myocardial electric activity.
Because it is sometimes difficult to examine a long series of electric potential maps, especially in a clinical setting where time is of the essence, an isochronal map can be helpful. Although much of the valuable information present in the potential distribution is not preserved in the isochronal map, significant information about the entire activation sequence can be seen at one glance. Moreover, for epicardial pacing, the patterns of the isochrones, like those of the potentials, are related to the local fiber orientation (anisotropy) in that they tend to be quasi-elliptical, with their major axis parallel to the fiber direction.8 The noninvasive isochronal maps correspond accurately to those computed directly from the measured potentials. The regions of earliest activation are located, and multiple pacing regions are resolved for sufficient intersite distance by the noninvasive reconstruction approach. In addition, spatial nonuniformities in the spread of epicardial activation (ie, regions of sparse and regions of crowded isochrones) are identified. This is an important property of the reconstructed isochrones because in pathological conditions, crowding of isochrones indicates a region of slow conduction. An area of slow conduction is a key component of the reentry circuit in reentrant arrhythmias,39 and its localization could provide important mechanistic and diagnostic information.
For small intersite distances (as in Fig 12⇑, 17 mm apart), the individual pacing sites are not resolved in the computed isochronal maps. Although the earliest isochrone is elongated, suggesting the presence of two pacing sites, it fails to provide their exact locations. In comparison, the computed potential distribution (Fig 7⇑, bottom) reconstructs two distinct minima, allowing one to locate the two pacing sites. The increased resolution of the potentials in comparison with the isochrones reflects a certain degree of uncertainty in determining the time of activation from individual unipolar electrograms. This limitation is manifest only close to the pacing sites, where high resolution is necessary. Isochrones provide the entire activation sequence at a glance, including regions of earliest activation and nonuniformities in the spread of activation. However, for the purpose of determining activation sites with high resolution, the early frames of the epicardial potential maps perform with greater accuracy.
The reconstruction of epicardial electrograms is important because it provides information on the temporal nature of activity in localized areas. An added benefit is that clinicians are accustomed to viewing and interpreting them. The results in Fig 8⇑ demonstrate, for the first time, the ability of ECGI and the inverse method to reconstruct electrograms of either monophasic or biphasic waveforms noninvasively and with great accuracy. It is also from these leads that the times of activation are computed, providing high accuracy for the reconstruction of noninvasive isochrone maps.
There are areas for improvement in this work. First, there is still some inaccuracy in the location of the extrema. Second, the time of earliest appearance of the epicardial negativity associated with the pacing site is, in general, several milliseconds later for the computed than for the measured potentials. This delay is not due to any capacitive, inductive, or propagation effects in the medium; it has been shown that such effects are negligible in the body volume conductor, which can be assumed to be purely resistive.40 Rather, it is most likely because the epicardial minimum is of very small magnitude at its earliest time, and its reflection on the torso is attenuated further, below the noise level. Temporal regularization14 incorporates the continuous nature of the cardiac activation process into the regularization procedure. It was shown to help in this type of situation so that events could be detected earlier. The applicability of this approach and its performance in terms of locating pacing sites will need to be explored in further studies. Temporal regularization would also be expected to improve the temporal discontinuities in the electrograms (Fig 8⇑) because it provides smoothing in time.
The tank-torso used in this study is homogeneous, and all forward and inverse computations are consistent with this property. This provides us with the opportunity to evaluate the inverse methodology without the added complexity introduced by the torso inhomogeneities. The formulation in terms of epicardial potentials used in this study takes into account all intracardiac inhomogeneities (eg, the intracavitary blood)18 ; however, external inhomogeneities (eg, the lungs) are not present in the tank-torso and are therefore not included in this mathematical formulation. The boundary element method can easily be expanded to include the effects of inhomogeneities as represented by secondary sources at interfaces (eg, the lung–skeletal muscle interface).2 Most theoretical and experimental studies suggest that the torso inhomogeneities affect the magnitudes of epicardial potentials but not their patterns5 31 41 42 or the sequence of epicardial activation (isochrones).43 Conversely, an inverse study that reconstructed epicardial isochrones directly11 suggests a more significant effect of the torso inhomogeneities on the accuracy of reconstruction. This issue deserves further consideration and will be addressed in future studies.
Finally, the magnitudes of the potentials tend to be lower in the computed than in the measured maps. This can be seen both in the contour maps of epicardial potentials and in the electrograms and is a natural consequence of the Tikhonov zero-order regularization scheme, which places constraints on the magnitudes of the potentials. This has minimal relevance in interpretation of the potential maps for localizing and resolving arrhythmogenic foci, including the effect of fiber orientation, since patterns (eg, positions of maxima and minima) and not amplitudes are considered. Similarly, isochronal maps are not affected, because determination of activation times involves the shape of the electrogram but not its amplitude.
This study concludes that ECGI is able to identify and locate, with good accuracy, single and multiple events by noninvasively computing the epicardial potential distribution from the BSPs. Specifically, this study demonstrates its ability to locate single and multiple pacing sites that simulate arrhythmogenic foci, suggesting its potential use in guiding interventional procedures (eg, ablation). From a broader perspective, the results of this study suggest a wide array of possibilities for potential clinical and experimental application, as follows. (1) The ability to map, simultaneously (during a single beat) and noninvasively, the electric activity of the heart opens the door for mapping of nonsustained or dynamically changing arrhythmias (eg, functional reentry circuits that continually change or drifting spiral waves44 ). Further research is needed to determine whether epicardial potential patterns and isochrones can identify a common pathway of multiple reentry circuits during polymorphic ventricular tachycardia and whether this information can be reconstructed by ECGI from body surface data. Such a common pathway constitutes a potential target for effective ablation of the tachycardia. (2) As demonstrated in this article, spatial nonuniformities of epicardial activation spread (eg, areas of sparse or crowded isochrones) can be located and mapped noninvasively with this technique. In pathological conditions, such nonuniformities might reflect nonuniformities of excitation spread that could play an important role in arrhythmogenesis. Our ability to recognize areas with variable density of isochrones in normal hearts sets the stage for further studies to determine whether slow conduction can also be detected during arrhythmic activity. Work in this direction has already begun in our laboratory. (3) The method could also be used to evaluate, in a given patient, the efficacy of antiarrhythmic drug therapy by noninvasively monitoring its effects on the spatial patterns of myocardial activation and recovery (eg, whether it increases or decreases dispersion of repolarization). (4) Recently, the need for a noninvasive method for identifying patients at risk of sudden death has been emphasized. Empirical indices from body-surface ECG (T-wave alternans,20 21 dispersion of QT intervals22 ) have been considered. The ability to “record” noninvasively from the heart itself will be very helpful in this regard by increasing detection sensitivity and by allowing one to relate these indices more directly to cardiac electric activity. For example, QT dispersion is determined on the basis of differences between individual leads on the body surface ECG. Since each surface lead reflects activity in the entire heart, this measure cannot be related to actual spatial heterogeneity of repolarization in the heart itself, which is a recognized arrhythmogenic property. With ECGI, one could obtain noninvasive information about the degree of spatial heterogeneity on the heart itself and locate the region that contributes to increased dispersion (and, therefore, to arrhythmogenesis). Similarly, alternans in the surface ECG could be related noninvasively to local beat-to-beat electric changes in the heart. (5) Another example of potential clinical use is increased specificity of differentiating between types of arrhythmias. Recently, ECG indices (QT interval changes in response to mexiletine or to an increase in heart rate) were shown to differentiate between two genetic types of the long-QT syndrome, LQT3 and LQT2, each requiring different therapy.45 46 A related study showed that different genotypes of the long-QT syndrome were associated with different phenotypic T-wave patterns on the ECG.47 In another study, two types of idiopathic ventricular tachycardia were differentiated in terms of initiation sites on the basis of BSP maps.48 It is highly probable that the specificity and ability to differentiate between arrhythmia types (especially important when the differential diagnosis is required for specific therapy) will be greatly enhanced by noninvasive analysis of reconstructed epicardial data rather than reliance on body surface data alone.
Like the clinical usefulness of a noninvasive electrophysiological imaging technique, one can envision its experimental potential. It could be used to study arrhythmias in the nonanesthetized, intact animal under physiological conditions. Moreover, it could provide a noninvasive tool for studying arrhythmogenesis in patients with chronic heart disease that develops and persists over the course of many years. In this situation, the mechanisms of arrhythmias and their properties might differ greatly from those in animal models.
Selected Abbreviations and Acronyms
|BSP||=||body surface potential|
|BSPM||=||body surface potential mapping|
|CRESO||=||composite residual and smoothing operator|
|NSR||=||normal sinus rhythm|
This study was supported by NIH-NHLBI grants HL-33343 (Dr Rudy) and HL-43276 (Dr Taccardi), American Heart Association National Center Grant-in-Aid 91006370 (Dr Rudy), and awards from the Nora Eccles Treadwell Foundation and the Richard A. and Nora Eccles Harrison Fund for Cardiovascular Research. Computer time was provided by Pittsburgh Supercomputing Center grant PSCA-21. We thank Yonild Vyhmeister, BS, for her assistance in the experiments and Robin Shaw, PhD, for his assistance in preparing figures.
- Received October 31, 1996.
- Revision received February 3, 1997.
- Accepted February 11, 1997.
- Copyright © 1997 by American Heart Association
Mirvis DM. Body Surface Electrocardiographic Mapping. Boston, Mass: Kluwer Academic Publishers; 1988.
Rudy Y. The electrocardiogram and its relationship to excitation of the heart. In: Sperelakis N, ed. Physiology and Pathophysiology of the Heart. Boston, Mass: Kluwer Academic Publishers; 1995:210-239.
De Ambroggi L, Musso E, Taccardi B. Body surface mapping. In: MacFarlane PW, Lawrie TDV, eds. Comprehensive Electrocardiography: Theory and Practice in Health and Disease. New York, NY: Pergamon Press: 1989;2:1015-1045.
Spach MS, Barr RC, Lanning CF, Tucek PC. Origin of body surface QRS and T wave potentials from epicardial potential distributions in the intact chimpanzee. Circulation. 1977;55:268-278.
Rudy Y, Plonsey R. A comparison of volume conductor and source geometry effects on body surface and epicardial potentials. Circ Res. 1980;46:283-293.
Arisi G, Macchi E, Baruffi S, Spaggiari S, Taccardi B. Potential fields on the ventricular surface of the exposed dog heart during normal excitation. Circ Res. 1983;52:706-715.
Spach MS, Barr RC. Ventricular intramural and epicardial potential distributions during ventricular activation and repolarization in the intact dog. Circ Res. 1975;37:243-257.
Taccardi B, Macchi E, Lux RL, Ershler PR, Spaggiari S, Baruffi S, Vyhmeister Y. Effect of myocardial fiber direction on epicardial potentials. Circulation. 1994;90:3076-3090.
Gulrajani RM, Savard P, Roberge FA. The inverse problem in electrocardiography: solutions in terms of equivalent sources. Crit Rev Biomed Eng. 1988;16:171-214.
Huiskamp G, van Oosterom A. The depolarization sequence of the human heart surface computed from measured body surface potentials. IEEE Trans Biomed Eng. 1988;35:1047-1058.
Huiskamp G, van Oosterom A. Tailored versus realistic geometry in the inverse problem of electrocardiography. IEEE Trans Biomed Eng. 1989;36:827-835.
Messinger Rapport BJ, Rudy Y. Noninvasive recovery of epicardial potentials in a realistic heart-torso geometry: normal sinus rhythm. Circ Res. 1990;66:1023-1039.
Messinger Rapport BJ, Rudy Y. Computational issues of importance to the inverse recovery of epicardial potentials in a realistic heart-torso geometry. Math Biosci. 1989;97:85-120. (Erratum, Math Biosci. 1990;99:141.)
Oster HS, Rudy Y. The use of temporal information in the regularization of the inverse problem of electrocardiography. IEEE Trans Biomed Eng. 1992;39:65-75.
Colli Franzone P, Guerri L, Tentoni S, Viganotti C, Baruffi S, Spaggiari S, Taccardi B. Mathematical procedure for solving the inverse problem of electrocardiography. Math Biosci. 1985;77:353-396.
Barr RC, Spach MS. Inverse calculation of QRS-T epicardial potentials from body surface potential distributions for normal and ectopic beats in the intact dog. Circ Res. 1978;42:661-675.
Shenasa M, Borggrefe M, Breithardt G. Cardiac Mapping. Mt Kisco, NY: Futura Publishing Co; 1993.
Rudy Y, Messinger Rapport BJ. The inverse problem in electrocardiography: solutions in terms of epicardial potentials. Crit Rev Biomed Eng. 1988;16:215-268.
Simson MB. Signal-averaged electrocardiography. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology: From Cell to Bedside. Philadelphia, Pa: WB Saunders Co; 1995:1038-1049.
Rosenbaum DS, Jackson LE, Smith JM, Garan H, Ruskin JN, Cohen RJ. Electrical alternans and vulnerability to ventricular arrhythmias. N Engl J Med. 1994;330:235-241.
Nearing BD, Huang AH, Verrier RL. Dynamic tracking of cardiac vulnerability by complex demodulation of the T wave. Science. 1991;252:437-440.
Surawicz B. Will QT dispersion play a role in clinical decision-making? J Cardiovasc Electrophysiol. 1996;7:777-784.
Durrer D, Van Dam RT, Freud GE, Janse MJ, Meijler FL, Arzbaecher RC. Total excitation of the isolated human heart. Circulation. 1970;41:899-912.
Colli Franzone P, Guerri L, Viganotti C, Macchi E, Baruffi S, Spaggiari S, Taccardi B. Potential fields generated by oblique dipole layers modeling excitation wavefronts in the anisotropic myocardium: comparison with potential fields elicited by paced dog hearts in a volume conductor. Circ Res. 1982;51:330-346.
Corbin LV, Scher AM. The canine heart as an electrocardiographic generator: dependence on cardiac cell orientation. Circ Res. 1977;41:58-67.
Colli Franzone P, Guerri L, Taccardi B. Potential distributions generated by point stimulation in a myocardial volume: simulation studies in a model of anisotropic ventricular muscle. J Cardiovasc Electrophysiol. 1993;4:438-458.
Khoury DS, Taccardi B, Lux RL, Ershler PR, Rudy Y. Reconstruction of endocardial potentials and activation sequences from intracavitary probe measurements: localization of pacing sites and effects of myocardial structure. Circulation. 1995;91:845-863.
Janse MJ, Wit AL. Electrophysiological mechanisms of ventricular arrhythmias resulting from myocardial ischemia and infarction. Physiol Rev. 1989;69:1049-1169.
Stevenson WG. Catheter mapping of ventricular tachycardia. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology: From Cell to Bedside. Philadelphia, Pa: WB Saunders Co; 1995:1093-1112.
SippensGroenewegen A, Spekhorst H, van Hemel NM, Kingma JH, Hauer RN, de Bakker JM, Grimbergen CA, Janse MJ, Dunning AJ. Localization of the site of origin of postinfarction ventricular tachycardia by endocardial pace mapping: body surface mapping compared with the 12-lead electrocardiogram. Circulation. 1993;88:2290-2306.
Messinger Rapport BJ, Rudy Y. Regularization of the inverse problem in electrocardiography: a model study. Math Biosci. 1988;89:79-118.
Rudy Y, Oster HS. The electrocardiographic inverse problem. Crit Rev Biomed Eng. 1992;20:25-45.
Jackson JD. Classical Electrodynamics. New York, NY: John Wiley & Sons; 1975.
Brebbia CA, Telles JCF, Wrobel LC. Boundary Element Techniques: Theory and Applications in Engineering. Berlin, Germany: Springer Verlag; 1984.
Tikhonov AN, Arsenin VY. Solutions of Ill-Posed Problems (trans from Russian). New York, NY: John Wiley & Sons; 1977.
Shahidi AV, Savard P, Nadeau R. Forward and inverse problems of electrocardiography: modeling and recovery of epicardial potentials in humans. IEEE Trans Biomed Eng. 1994;41:249-256.
Kaelber D, Haaga J, Rudy Y. Noninvasive in vivo determination of body surface and epicardial geometries for electrocardiographic imaging. Presented at the IEEE/EMBS 16th Annual International Conference; 1994; Baltimore, Md. Abstract.
Liu ZW, Jia P, Ershler PR, Taccardi B, Lux RL, Khoury DS, Rudy Y. Noncontact endocardial mapping: reconstruction of electrograms and isochrones from intracavitary probe potentials. J Cardiovasc Electrophysiol. 1997;8:415-431.
Ortiz J, Niwano S, Abe H, Rudy Y, Johnson NJ, Waldo AL. Mapping the conversion of atrial flutter to atrial fibrillation and atrial fibrillation to atrial flutter: insights into mechanisms. Circ Res. 1994;74:882-894.
Plonsey R. Bioelectric Phenomena. New York, NY: McGraw-Hill; 1969.
Messinger Rapport BJ, Rudy Y. The inverse problem in electrocardiography: a model study of the effects of geometry and conductivity parameters on the reconstruction of epicardial potentials. IEEE Trans Biomed Eng. 1986;33:667-676.
Gulrajani RM, Roberge FA, Mailloux GE. The forward problem of electrocardiography. In: MacFarlane PW, Veitch Lawrie TD, eds. Comprehensive Electrocardiology: Theory and Practice in Health and Disease. New York, NY: Pergamon Press; 1989;1:197-236.
Green LS, Taccardi B, Ershler PR, Lux RL. Epicardial potential mapping: effects of conducting media on isopotential and isochrone distributions. Circulation. 1991;84:2513-2521.
Davidenko JM, Pertsov AV, Salomonsz R, Baxter W, Jalife J. Stationary and drifting spiral waves of excitation in isolated cardiac muscle. Nature. 1992;355:349-351.
Schwartz PJ, Priori SG, Locati EH, Napolitano C, Cantu F, Towbin JA, Keating MT, Hammoude H, Brown AM, Chen LSK, Colatsky TJ. Long QT syndrome patients with mutations of the SCN5A and HERG genes have differential responses to Na+ channel blockade and to increases in heart rate: implications for gene-specific therapy. Circulation. 1995;92:3381-3386.
Rosen MR. Long QT syndrome patients with gene mutations. Circulation. 1995;92:3373-3375.
Moss AJ, Zareba W, Benhorin J, Locati EH, Hall WJ, Robinson JL, Schwartz PJ, Towbin JA, Vincent GM, Lehmann MH, Keating MT, MacCluer JW, Timothy KW. ECG T-wave patterns in genetically distinct forms of the hereditary long QT syndrome. Circulation. 1995;92:2929-2934.
Klug D, Ferracci A, Molin F, Dubuc M, Savard P, Kus T, Helie F, Cardinal R, Nadeau R. Body surface potential distributions during idiopathic ventricular tachycardia. Circulation. 1995;91:2002-2009.