Regional Capture of Fibrillating Ventricular Myocardium
Evidence of an Excitable Gap
Abstract Previous investigations have suggested that during ventricular fibrillation (VF) pacing stimuli are incapable of evoking propagated ventricular activations. To determine whether regional myocardial capture could be achieved during rapid pacing in VF, extracellular unipolar potentials were sampled (2 kHz) and recorded from 506 Ag-AgCl electrodes arranged in a rectangular grid (22×23, 1.12-mm spacing) embedded in a plaque overlying two pacing electrodes in the epicardium of the anterobasal right ventricle in pentobarbital-anesthetized pigs (25 to 30 kg, n=6). During separate episodes of electrically induced VF, two bursts of 40 monophasic stimuli (10 mA, 2-millisecond duration) were asynchronously applied to the stimulating electrodes in either a bipolar, unipolar anodal, or unipolar cathodal mode. Evidence of regional capture was provided by (1) animating the first temporal derivative of the extracellular potentials, (2) analyzing interbeat interval patterns, and (3) employing the Karhunen-Loeve decomposition method to quantify the repetitiveness of spatiotemporal patterns of activation. Regional capture of ventricular myocardium during VF was observed when pacing stimuli fell late in the local myocardial activation interval and when the pacing cycle length was 80% to 115% of the mean subplaque activation cycle length. When myocardial activations became phase locked to the pacing stimuli, repeatable spatiotemporal patterns of activation followed each stimulus. Poincaré sections at the plaque border revealed that during VF prior to pacing, interbeat intervals were irregular but were driven by pacing to stable fixed values at times corresponding to our qualitative declaration of regional capture. A similar correspondence was demonstrated between the time of capture, defined by direct observation of the activation patterns, and a rise in the power contained in the first two spatial modes of a Karhunen-Loeve decomposition. These data demonstrate that appropriately timed stimuli produce regional capture of fibrillating right ventricular myocardium in the pig and support the existence of an excitable gap during VF in this model.
A recently proposed therapeutic strategy designed to exert control over cardiac tachyarrhythmias relies on the finding that small, appropriately timed exogenous perturbations to a chaotic system induce order and periodicity.1 For this strategy to be practically realized, however, local electrical stimulation during fibrillation must predictably influence myocardial activation dynamics over a significant spatial region. This phenomenon has been observed in the canine left atrium2 3 but not in the ventricles. Based on inferences from intracellular and extracellular recordings,4 5 6 7 8 some9 10 but not all11 12 investigators have asserted that fibrillating ventricular myocardium is immediately reactivated upon recovering excitability. These previous findings9 10 suggest that the intervals of activation during ventricular fibrillation (VF) represent the shortest possible refractory periods to circulating wave fronts, so that no fully excitable gap is present. Yet recent optical,11 extracellular,12 and microelectrode13 recordings in addition to theoretical predictions14 15 suggest that a fully excitable gap (time between recovery of excitability and next activation) or a partially excitable gap (time during relative refractory period during which a suprathreshold stimulus is capable of producing a propagated action potential) may be present during established VF.
Therefore, we reasoned that because a partially excitable gap is necessary for wave front propagation during VF, certain forms of electrical stimulation should be capable of altering patterns of activation by producing new wave fronts that propagate away from the stimulation site and interact with intrinsic wave fronts. Thus, the purpose of this study was to use high-resolution cardiac mapping techniques to determine whether critically-timed suprathreshold electrical stimuli locally capture ventricular myocardium during electrically-induced fibrillation.
Materials and Methods
In six pigs (25 to 35 kg) of either sex, we used a 528-channel cardiac mapping system and custom computer programs to acquire mapping data and then reconstruct spatiotemporal patterns of epicardial activation before, during, and after periodic electrical stimulation of fibrillating ventricular myocardium. Animal care adhered to the guiding principles accepted by the National Institutes of Health. Each animal was tranquilized with acepromazine (1.1 mg/kg) and ketamine (22 mg/kg). Subsequently, pentobarbital was given intravenously to induce (30 mg/kg) and maintain (0.05 mg/kg per minute) a deep plane of anesthesia. Central venous access was obtained through the left jugular vein for delivery of maintenance fluids and drugs. Volume-controlled ventilation was achieved using a cuffed endotracheal tube and a Harvard respirator. Partial pressures of O2 and CO2 as well as arterial blood pH, Ca2+, K+, Na+, and HCO3− were determined hourly. Normal physiologic levels of electrolytes and blood gases were maintained by adjusting ventilation parameters and infusing sodium bicarbonate, calcium chloride, or potassium chloride in appropriate concentrations. The surface electrocardiogram (lead II) and femoral arterial blood pressure signals were simultaneously displayed on a monitor. The heart was suspended in a pericardial cradle via a median sternotomy. Two spherical electrodes (0.8-mm diameter) for myocardial stimulation, separated by 5 mm, were positioned on the surface of the heart through subepicardial needle tracks on the right ventricular outflow tract such that the Ag-AgCl exposed tips were flush with the epicardium. The diameter of the insulated wire leading from the stimulating electrodes to the stimulator was 0.3 mm. The plaque was positioned so that the pacing electrodes were approximately centered beneath it. Pledgeted sutures at each corner maintained contact between a 506-electrode mapping plaque in a grid pattern (22×23, 1.12-mm center-to-center interelectrode spacing) and the underlying right ventricular myocardium, as previously described.16 This right ventricular location was chosen because it offered favorable exposure for plaque attachment. The electrograms were recorded from the exposed tips of Ag wires (125 μm) embedded in the plaque. Titanium defibrillation electrodes were positioned in the superior vena cava (coil surface area, 8 cm2) and on the epicardium of the left ventricular apex (wire mesh patch diameter, 2 cm).
The 506 epicardial unipolar electrograms, referenced to the left leg, were band-pass filtered between 0.5 and 500 Hz, sampled at 2 kHz (14-bit resolution, 9.766-μV step size), and recorded on videotape.17 18 Subplaque stimuli were generated by a programmable constant current stimulator controlled by a computer.19 Trains of 2-millisecond rectangular monophasic stimuli (n=40) were applied in bipolar, unipolar anodal, and unipolar cathodal modes to the stimulating electrodes beneath the plaque. For unipolar modes, the return electrode was clipped to the diaphragm. This train was long enough to provide a large variety of coupling intervals between intrinsic VF activations and the stimuli. A range of pacing cycle lengths was chosen (100 to 140 milliseconds) to span the intrinsic activation rate during VF, 120 to 125 milliseconds.20 A 10-mA stimulus strength, which was at least 10 times the diastolic threshold, was used for all stimuli in all animals in an attempt to excite partially refractory tissue close to the electrodes.
Bipolar pacing stimulus thresholds were determined to the nearest 0.1 mA in each animal at a basic drive cycle length of 300 milliseconds by increasing stimulus strength until the heart was captured. VF was induced by application of alternating current (60 Hz) for 1 to 2 seconds between a pacing electrode (stainless steel wire, 2-mm loop) sutured to the epicardium of the right ventricular apex and an indifferent electrode off the heart. Two trains of 40 subplaque stimuli with the same cycle length were applied, the first 2 seconds after VF induction and the second 2 to 3 seconds after the first train ended. Twenty-seven combinations of the nine stimulation cycle lengths (100 to 140 milliseconds, 5-millisecond steps) and the three stimulation modes were tested in random order in each animal. After each combination was tested, VF was terminated via electric countershock (single-capacitor biphasic waveform, 6 milliseconds first phase/6 milliseconds second phase, 4- to 6-A peak, Ventritex HVS-02), and the procedure was repeated until all cycle length/stimulation mode combinations had been performed. A recovery period of at least 5 minutes was observed between fibrillation/defibrillation episodes. After the study, the animals were killed by electrically inducing VF, and markers were placed to document macroscopic plaque location and orientation. The plaque was removed and the heart was harvested, weighed, and fixed in formalin.
To determine whether pacing captured the fibrillating tissue, we visualized animations of the first temporal derivative (5-point dV/dt) of the extracellular potentials on a scientific workstation (Sun Microsystems, Inc) displayed as a 22×23 array of elements, each of which represented 1 of the 506 electrogram recordings. When dV/dt became more negative than −0.5 V/s, that element was shaded gray. Any dV/dt value between −1.1 and −0.5 V/s was displayed as a level of gray, each level corresponding to a 0.1-V/s gradation, with darker levels of gray indicating more negative values of dV/dt. All dV/dt values greater than −0.5 V/s were displayed as white. Our analyses were performed on the first temporal derivative of the extracellular electrograms, because this method removes baseline drift and better represents local myocardial activity.21 Animations were displayed approximately 40 times slower than real time to represent visually activation fronts moving across the surface of the heart beneath the plaque. This method of analysis avoids the difficulties of detection of discrete activation times during VF and the ambiguities associated with the grouping of beats required for construction of isochron maps.22
Determination of Myocardial Capture
We used three methods to determine that the tissue within the mapped region was captured by the pacing stimuli during VF. First, a qualitative visual criterion was established for the spatiotemporal animations of dV/dt. To satisfy this criterion, contiguous activation fronts produced by the stimulus were required to propagate to at least one of the borders of the mapped region before arrival of the next stimulus. Recorded data before and after application of every stimulus were viewed in this way to rapidly screen data for evidence of myocardial capture during VF. These visual techniques, while compelling, may not be viewed by some as quantitative proof that rapid stimulation during VF produces regional capture. Therefore, two independent quantitative methods were used to validate our findings based on visual assessments of the animated epicardial activation dynamics.
The first quantitative method compared the means and variances of the interbeat intervals during VF for three categories of behavior: (1) free-running VF without subplaque stimulation, (2) subplaque stimulation during VF when the animation provided no clear evidence of phase locking to the stimuli, and (3) 1:1 phase locking between stimulus artifact and local electrograms. This procedure is equivalent to assessing the location and dispersion of the Poincaré return map23 relating each activation interval (In) to the previous activation interval (In−1). Poincaré maps are commonly used in dynamic system analysis to represent the relationship between successive states. For the case we present, the states are interbeat intervals. These maps convey information about regularity of the interval separating the local activations at some fixed point in space. Interbeat intervals were obtained by taking a time difference between successive local activations. Activation times were automatically assigned to the maximum dV/dt within a 25-millisecond window encompassing a rapid downslope in the electrogram signal from the same four recording sites in each case. These sites were at the center of a quadrant defining the mapped region. Activation times inadvertently assigned to stimulus artifacts were deleted manually.
We reasoned that during episodes of regional capture, the mean and dispersion of myocardial activation intervals should be different than during episodes when the stimuli did not capture the myocardium. All bipolar stimulation episodes that contained sustained capture during VF were analyzed in this way. For each episode, time series from the same four electrodes at the quadrant centers were grouped into three subseries corresponding to each behavior type identified by our qualitative visual criteria. The mean interbeat interval (I) and its SD (ςI) were calculated for each subseries. All activation intervals were considered in this analysis for episodes having at least five consecutive cycles of regional capture and for episodes having at least five consecutive cycles for which regional capture was not observed.
We employed a second method to quantitatively demonstrate that rapid pacing locally captures fibrillating myocardium and thereby controls activation sequences beneath the mapping plaque. The Karhunen-Loeve (K-L) decomposition method allowed us to isolate and study spatial patterns of myocardial activation, as previously described.16 The K-L decomposition represents observed data as a weighted sum of orthogonal spatial modes. The data obtained from the mapping system in this study consist of samples (M) from each electrode in an array of electrodes (N). A large, M×N data matrix (X) can be formed whose rows (r⃗I) are “snapshots” of the state of the array at each instant in time. Each snapshot can be exactly reconstructed from the weighted sum of N orthogonal spatial patterns, v⃗: Each snapshot can also be approximated using fewer than N modes: where P<N. K-L modes are the optimal spatial patterns for such a representation because a given number of K-L modes approximates the original data better, in terms of mean squared error, than does the same number of any other modes (eg, Fourier modes).24 K-L modes are determined, for each data set, as the eigenvectors of a spatial autocovariance matrix.24 Spatiotemporal behavior that is repetitive and organized over its whole extent requires only a few modes to account for most of its variance, or “energy.” Therefore, in our analysis the fraction of total energy contributed by the first two K-L modes [C(2)] was chosen as a measure of the degree of order in activation patterns recorded under the plaque. C(2) was computed (matlab version 4.2a, The MathWorks, Inc) as a running average over a 0.5-second window that was moved through the data file at 0.25-second intervals. Stimulus artifacts were removed from the data stream. The window duration chosen for our analyses allowed several cycles to contribute to each C(2) value. A substantially narrower window might have impeded our ability to resolve changes in dominant spatial patterns, whereas a wider window might have attenuated changes in spatial patterns occurring over a few hundred milliseconds.
The hearts from the six animals weighed 142±18 g (mean±SD). The bipolar stimulation pacing stimulus threshold was 0.6±0.2 mA. Therefore, the 10-mA stimulus strength that was used in all experiments was 18±5 times the pacing threshold at a drive cycle length of 300 milliseconds.
Animation of Cardiac Activation Dynamics
The activation fronts during VF were commonly fractionated, as a result of slowed conduction and conduction block; fronts frequently turned and collided with other wavelets with only moderate spatiotemporal repeatability (Fig 1A⇓ and 1B⇓). When pacing stimuli were applied during episodes of electrically induced VF, most of the pulses had no observable effect on the tissue, presumably because the tissue was refractory (Fig 1C⇓). However, when a pacing stimulus occurred with a long coupling interval after the previous VF activation at the pacing site, stimulation gave rise to activation fronts that propagated away from the pacing site (Fig 1D⇓). If the tissue adjacent to the pacing site then remained inactivated until the next stimulus was applied, the next stimulus also generated an activation front that was 1:1 phase locked with the stimulus artifact. In general, as the number of phase-locked beats increased, the area of captured myocardium increased. However, when the pacing cycle length was too long, intrinsic VF activation fronts activated the tissue at the pacing site before the next stimulus was applied, periodicity of activation was lost, and subsequent stimuli did not capture the tissue.
When pacing stimuli evoked 1:1 phase locking over all or even relatively small (≈20%) portions of the mapped region, striking differences in activation patterns were observed when compared with either VF without pacing or VF with pacing that did not evoke phase-locked beats (Fig 1⇑). We observed 1:1 phase locking of the extracellular electrograms to the stimulus artifact in every animal. However, the spatial extent of capture was not the same for all episodes within animals or between animals. Sustained myocardial capture occurred only when the pacing cycle length was 80% to 115% of the local intrinsic activation rate, similar to the coupling intervals for capture of myocardium during atrial fibrillation3 and in functionally defined circuits of rapid tachyarrhythmias in infarcted dogs.25
Table 1⇓ summarizes the data derived from the animations of epicardial activation patterns from all animals. Note that sustained phase locking (≥5 consecutive beats) occurred for each stimulation mode and was observed during 9±4 (mean±SD) pacing bursts in each animal. Once sustained capture was established, it was either lost by premature activation at the stimulation site or continued until the stimulation burst ended. For bipolar stimulation, this “loss of capture” during the burst of 40 stimuli occurred in 9 of 25 cases, whereas it occurred in 5 of 13 and 6 of 14 episodes during which the stimuli were anodal and cathodal, respectively. In the remaining sustained-capture episodes, phase locking was maintained until the pacing burst ended. In two bipolar stimulation episodes (120- and 125-millisecond cycle lengths) and in one unipolar cathodal stimulation episode (110-millisecond cycle length), all 40 stimuli produced 1:1 phase locking lasting 4.8, 5.0, and 4.4 seconds, respectively. Capture was significantly more frequent (P<.05 by χ2 test) during the second pacing burst (18 of 25) than during the first (7 of 25) for bipolar stimulation. This trend was also observed for unipolar cathodal and anodal stimulation. To determine the local timing criteria immediately preceding sustained capture during VF, the coupling interval between the first phase-locked beat and the previous activation interval was manually extracted at a point ≤4 mm from the site of stimulation for each of the episodes displaying sustained phase locking (see Table 1⇓).
Whereas the spatiotemporal animation of activation dynamics within the mapped region provided compelling evidence that pacing stimuli produced regional myocardial capture during VF, we sought to verify this direct observation by using two quantitative methods: Poincaré mapping and the K-L decomposition.
Fig 2A⇓ shows a time series of interbeat intervals produced by propagated activation wave fronts at one recording site near the border of the mapped region ≈1 cm from the stimulation site. Fig 2B⇓ shows the Poincaré map of data from Fig 2A⇓. The tightly clustered intervals (solid circles) at 115 milliseconds on the line of identity correspond to beats during which animations of cardiac activation dynamics revealed capture of the entire mapped region. Episodes during which the entire mapped region was captured appear similar to the frames shown in Fig 1D⇑. In general, when pacing stimuli produced regional capture, the difference between myocardial activation cycle length and the pacing cycle length was lower than when the pacing stimuli did not capture fibrillating ventricular myocardium. This observation is supported by the statistically significant differences in ςI during episodes in which rapid subplaque stimulation produced regional capture compared with ςI during free-running VF (Table 2⇓). These data provide confirmation of our qualitative inference that pacing stimuli capture the tissue beneath the plaque.
Data from a representative K-L decomposition calculation are shown in Fig 3⇓, which plots fractional energy contained in the first two K-L modes [C(2)]. The time at which pacing was begun is indicated by the change from solidcircles to open squares. At 4.5 seconds (solidcircles), we declared capture of the tissue within the mapped region using a qualitative visual criterion described earlier. For the case shown in Fig 3⇓, a marked rise in C(2) was temporally coincident with changes in the spatial patterns that we directly observed from the animations. An increase in C(2) is associated with higher spatial order. Therefore, the K-L decomposition also independently validated our visual interpretation of the data. Similar analyses were performed on all episodes containing sustained phase locking, with consistent results; increased C(2) coincided with times during which the applied stimuli apparently captured the fibrillating myocardium and thereby produced repeatable subplaque activation patterns (see Table 3⇓).
Fig 4⇓ shows both frames from animation of dV/dt and individual unipolar electrogram recordings (2-second segment) from five sites along a diagonal of the mapped region during stimulation of fibrillating myocardium. Wave fronts of activation are produced by the stimulus and propagate away from the stimulation site near the center of the mapped region. The patterns of activation were repeatable and demonstrate sustained regional capture of myocardium during VF.
There are several possible explanations for the observation that a region of ventricular myocardium >5 cm2 may be captured by rapid periodic pacing during VF. First, the stimulus strength (10 mA) was about 15 to 25 times the diastolic pacing threshold. This suprathreshold stimulus could have formed a virtual cathode26 that directly excited a sufficient volume of myocardium so that the activation front formed at the border of this directly excited region was large enough to excite the adjacent tissue and hence propagate. Second, repetitive stimulation (40 pulses) at a fixed cycle length may have caused additional local shortening of the refractory period to an extent that subsequent stimuli “pried” their way into the tissue, even though a fully excitable gap was not present during VF.9 However, we occasionally observed capture by the first stimulus of the train, suggesting that sometimes a fully excitable gap might be present before the first stimulus. Even if a fully excitable gap were not present, the stimulus strength used in this study was probably sufficient to stimulate tissue in a partially excitable gap. Third, pigs have longer intrinsic activation intervals during VF (about 125 milliseconds) than some other animals, eg, dogs (75 to 85 milliseconds).27 It is not known how much of this difference is caused by a greater excitable gap and how much is caused by a longer refractory period in pigs.
The animation method we used to rapidly analyze data was an important part of this study. Had a more traditional means of analysis been chosen that involved manual definition of activation times at each sensing site, analysis of all data would have been nearly impossible, because definition of nearly 10 million activation times would have been required before construction and subsequent interpretation of over 19 000 isochron maps. Thus, the animation method described here is extremely useful and efficient and allows quantitative characterization of patterns of activation on the ventricular epicardium.28
Regional capture occurred more frequently during the second stimulation burst (about 10 seconds after VF was induced) than during the first burst, which was applied about 3 seconds after VF was induced. It is possible that not only the duration of VF before myocardial stimulation but also the previous application of stimuli (first stimulation burst) may influence regional capture.
Our findings in the intact, working heart confirm the theoretical prediction of Winfree14 that an excitable gap is a necessary electrophysiological characteristic of self-perpetuating wave fronts of activation observed in fibrillating ventricular myocardium.16 28 However, the results from this study do not conclusively demonstrate whether the excitable gap during VF in this model is fully excitable or partially excitable.
Bruce KenKnight was a visiting scientist from the Therapies Research Department of Cardiac Pacemakers, Inc, at the time this work was performed. This work was supported in part by the National Institutes of Health research grants HL-28429, HL-44066, and HL-42760; National Science Foundation Engineering Research Center grant CDR-8622201; and by Cardiac Pacemakers, Inc. We wish to acknowledge the expert technical support of S.B. Melnick and R.G. Walker.
Reprint requests to Bruce H. KenKnight, Cardiac Rhythm Management Laboratory, UAB Station, Volker Hall, Room G78A, Birmingham, AL 35294-0019. E-mail email@example.com.
- Received October 24, 1994.
- Accepted July 5, 1995.
- © 1995 American Heart Association, Inc.
Garfinkel A, Spano ML, Ditto WL, Weiss JN. Controlling cardiac chaos. Science. 1992;257:1230-1235.
Allessie M, Kirchhof C, Scheffer GJ, Chorro F, Brugada J. Regional control of atrial fibrillation by rapid pacing in conscious dogs. Circulation. 1991;84:1689-1697.
Kirchhof C, Chorro F, Scheffer GJ, Brugada J, Konings K, Zetelaki Z, Allessie M. Regional entrainment of atrial fibrillation studied by high-resolution mapping in open-chest dogs. Circulation. 1993;88:736-749.
Sano T, Tsuchihashi H, Shimamoto T. Ventricular fibrillation studied by the microelectrode method. Circ Res. 1958;6:41-46.
Downar E, Janse MJ, Durrer D. The effect of acute coronary artery occlusion on subepicardial transmembrane potentials in the intact porcine heart. Circulation. 1977;56:217-224.
Akiyama T. Intracellular recording of in situ ventricular cells during ventricular fibrillation. Am J Physiol. 1981;240:H465-H471.
Swartz JF, Jones JL, Fletcher RD. Characterization of ventricular fibrillation based on monophasic action potential morphology in the human heart. Circulation. 1993;87:1907-1914.
Opthof T, Misier ARR, Coronel R, Vermeulen JT, Verberne HJ, Frank RGJ, Moulijn AC, van Capelle FJL, Janse MJ. Dispersion of refractoriness in canine ventricular myocardium: effects of sympathetic stimulation. Circ Res. 1991;68:1204-1215.
Lammers WJEP, Allessie MA, Rensma PL, Schalij MJ. The use of fibrillation cycle length to determine spatial dispersion in electrophysiologic properties used to characterize the underlying mechanism of fibrillation. N Trends Arrhythmia. 1986;2:109-112.
Jalife J, Davidenko JM. Spiral waves as a mechanism of reentrant excitation in isolated cardiac muscle. In: Shenasa M, Borggrefe M, Breithardt G, eds. Cardiac Mapping. Mount Kisco, NY: Futura Publishing Co Inc; 1993:607-623.
Cha Y-M, Birgersdotter-Green U, Wolf PL, Peters BB, Chen P-S. The mechanism of termination of reentrant activity in ventricular fibrillation. Circ Res. 1994;74:495-506.
Zhou X, Wolf PD, Rollins DL, Afework Y, Smith WM, Ideker RE. Effects of monophasic and biphasic shocks on action potentials during ventricular fibrillation in dogs. Circ Res. 1993;73:325-334.
Winfree AT. The electrical thresholds of ventricular myocardium. J Cardiovasc Electrophysiol. 1990;1:393-410.
Winfree AT. How does ventricular tachycardia decay into ventricular fibrillation? In: Shenasa M, Borggrefe M, Breithardt G, eds. Cardiac Mapping. Mount Kisco, NY: Futura Publishing Co Inc; 1993:655-680.
Wolf PD, Rollins DL, Blitchington TF, Ideker RE, Smith WM. Design for a 512 channel cardiac mapping system. In: Mikulecky DC, Clarke AM, eds. Biomedical Engineering: Opening New Doors. Proc Annu Biomed Eng Soc. New York, NY: New York University Press; 1990:5-13.
Smith WM, Wolf PD, Simpson E, Danieley ND, Ideker RE. Mapping ventricular fibrillation and defibrillation. In: Shenasa M, Borggrefe M, Breithardt G, eds. Cardiac Mapping. Mount Kisco, NY: Futura Publishing Co Inc; 1993:251-260.
Rollins DL, Wolf PD, Ideker RE, Smith WM. A programmable cardiac stimulator. Proc Comput Cardiol. 1992:507-510.
Johnson EE, Idriss SF, Cabo C, Melnick SB, Smith WM, Ideker RE. Evidence that organization increases during the first minute of ventricular fibrillation in pigs mapped with closely spaced electrodes. J Am Coll Cardiol. 1992;19:90A. Abstract.
Berbari EJ, Lander P, Scherlag BJ, Lazzara R, Geselowitz DB. Ambiguities of epicardial mapping. J Electrocardiol. 1992;24:16-20.
Glass L, Mackey MC. From Clocks to Chaos: The Rhythms of Life. Princeton, NJ: Princeton University Press; 1988:1248.
Waldecker B, Coromilas J, Saltman AE, Dillon SM, Wit AL. Overdrive stimulation of functional reentrant circuits causing ventricular tachycardia in the infarcted canine heart: resetting and entrainment. Circulation. 1993;87:1286-1305.
Wikswo JP, Wisialowski TA, Altemeier WA, Balser JR, Kopelman HA, Roden DM. Virtual cathode effects during stimulation of cardiac muscle: two-dimensional in vivo experiment. Circ Res. 1991;68:513-530.
Zhou X, Guse P, Wolf PD, Rollins DL, Smith WM, Ideker RE. Existence of both fast and slow channel activity during the early stage of ventricular fibrillation. Circ Res. 1992;70:773-786.