Author + information
- Received February 22, 2001
- Revision received September 5, 2001
- Accepted September 7, 2001
- Published online January 2, 2002.
- Gregory P Walcott, MD∗,* (, )
- G.Neal Kay, MD, FACC∗,
- Vance J Plumb, MD, FACC∗,
- William M Smith, PhD∗,†,
- Jack M Rogers, PhD†,
- Andrew E Epstein, MD, FACC∗ and
- Raymond E Ideker, MD, PhD, FACC∗,†
- ↵*Reprint requests and correspondence:
Dr. Gregory P. Walcott, B140 Volker Hall, 1670 University Blvd., University of Alabama at Birmingham, Birmingham, Alabama 35294, USA.
Objectives This study was designed to characterize the organization of ventricular fibrillation (VF) on the endocardium of humans.
Background Most proposed mechanisms for the maintenance of VF postulate the propagation of a number of activation wave fronts that reenter to maintain the arrhythmia. We tested the hypothesis that, in patients undergoing internal cardioverter-defibrillator implantation, VF consists primarily of a few large wave fronts on the endocardium.
Methods Electrograms were recorded from a 36-electrode catheter in the left ventricle of 16 patients during VF. Activation times were chosen for a 2-s epoch for each fibrillation episode, and a two-dimensional Kolmogorov-Smirnov test was performed to determine if activation occurred randomly along the catheter over that time interval. The maximum cross-correlation was found for all possible pairs of electrodes on the catheter, and these values were plotted relative to the distance between the two electrodes. An exponential curve was then fit to the data, and a length constant was determined. Activation times were grouped into wave fronts along the catheter, and the lengths of the wave fronts were estimated.
Results The Kolmogorov-Smirnov test showed that activation was not random along the catheter in any of the patients studied. The correlation length determined was 9 ± 2 cm. The number of wave fronts recorded by the catheter was 9.2 ± 2.9 wave fronts/s. The length of the pathway of each wave front along the catheter was 6.5 ± 4.5 cm.
Conclusions Ventricular fibrillation is well organized on the endocardial surface of humans, consisting primarily of a few large wave fronts on the order of 6 to 9 cm.
The final common arrhythmia in a majority of patients suffering sudden cardiac death is a hemodynamically unstable tachyarrhythmia, either ventricular fibrillation (VF) or ventricular tachycardia, that eventually degenerates into VF (1). Ventricular fibrillation is a difficult rhythm to study in patients and animal models because of its complexity and hemodynamic instability.
In 1887, MacWilliam characterized VF as a state in which “the ventricular muscle is thrown into a state of irregular arrhythmic contraction” (2). Wiggers used high-speed cinematography to describe four stages of progressive disorganization through which VF transitions from initiation to cardiac quiescence (3). More recently, multi-channel computerized electrical and optical cardiac mapping have been used to record electrograms during VF (4–10), and quantitative measures have been developed to describe VF. Correlation and coherence length have been used to estimate the spatial organization of VF in animals (11,12). Rogers et al. developed a series of metrics to describe VF recorded from 504 electrodes on the epicardium of pigs. These metrics include wave front size, the number of wave front fragmentations and collisions that occur per unit time, the conduction velocity of the wave fronts and the repeatability of wave fronts over time (13). Gray et al. (14–16)have recently shown the applicability of spiral-wave theory to VF. Almost all of these studies reported that VF is organized with activation fronts following pathways that are centimeters in size.
All of these studies, though, were performed in animals, where the heart size is smaller and the activation rate of VF is faster than in humans. Although several studies have characterized VF in humans from body-surface electrodes (17–19), single extracellular electrodes (20)or monophasic action potential electrodes (21)and epicardial multi-site mapping in Langendorff-perfused explanted hearts (22), little has been done to map VF from multiple sites simultaneously in intact humans.
In this study, we tested the hypothesis that endocardial activation during VF in humans consists primarily of a few large wave fronts by simultaneously recording over a wide expanse of the left ventricle (LV). We analyzed these recordings to determine whether activation occurred randomly or in an organized pattern. We estimated the extent of spatial organization during VF using the correlation length. Finally, we grouped activations to estimate the size and number of wave fronts recorded along the catheter. From these data we estimated the size and number of wave fronts in the human heart during VF.
This protocol was approved by the University of Alabama at Birmingham Internal Review Board for human studies, and written informed consent was obtained from each participant. Patients who were undergoing implantation of an internal cardioverter-defibrillator and who had no contraindications to having an electrophysiology catheter placed in the LV were eligible to participate in this study. Contraindications included a thrombus in the LV, a mechanical aortic or mitral valve, peripheral vascular disease including previous aorto-bifemoral bypass or femoral-distal bypass surgery, or failure to obtain informed consent.
A 36-pole catheter with 1-mm platinum bands spaced 4 mm apart (ElectroCatheter Corp., Rahway, New Jersey) was advanced from a femoral artery and across the aortic valve. It was positioned in the LV so that pole 1 (most distal) was in the antero-basal segment of the LV free wall, poles 16–22 were at the LV apex and pole 36 (most proximal) was in the postero-basal segment of the interventricular septum (Fig. 1). A standard decapolar or hexapolar catheter was placed in the coronary sinus. This catheter was advanced as far as possible into the coronary sinus to record electrograms from the base of the LV free wall.
In the first eight patients, electrograms were recorded from every other pole of the 36-pole LV catheter and the coronary sinus catheter using a 32-channel mapping system (Pruka Engineering, Dallas, Texas). Unipolar electrograms were recorded at a sampling rate of 1 kHz at a gain of 500 and band-pass filtered from 5 Hz to 500 Hz. Electrograms were saved on an optical disk for off-line analysis.
In the last eight patients, a custom-designed 144-channel mapping system was used to record unipolar electrograms from all 36 poles of the LV catheter and all poles of the coronary sinus catheter. Electrograms were recorded at a gain of 100 and band-passed filtered from 5 Hz to 500 Hz.
Electrograms were recorded during the VF that was induced during normal defibrillation testing, performed as part of the defibrillator implantation procedure. Ventricular fibrillation was induced by burst pacing and was allowed to continue for 8 to 10 s before a defibrillation test shock was delivered. The data analyzed in this paper were recorded in the 2 s just before the delivery of the test shock. Either one or two episodes of VF were analyzed from each patient.
Test of activation randomness along length of catheter
Activation times were determined using a computerized algorithm and were manually edited. The maximum downslope of the electrogram was chosen as the local activation time (23). The activation rate during VF was calculated for each patient by averaging the R-R interval for each electrogram over the 2 s of data analyzed. A Kolmogorov-Smirnov test of randomness was performed on the activation times (24). Briefly, this test determines the probability that data in a two-dimensional plot are consistent with a uniform random distribution. In this case, the two dimensions of data were electrode number and activation time. The data for each VF episode were considered random if the p value determined for that episode was >0.05.
Correlation length determination
After the mean value of the electrogram amplitude was subtracted from each point of the electrogram (to remove any DC offset), cross-correlations were calculated between every pair of electrograms with lags from −2 s to 2 s. Cross-correlations were normalized so that the auto-correlation for each electrode was 1. Then the maximum correlation between each pair of electrodes was plotted as a function of distance between the two electrodes. Using non-linear regression, an exponential curve of the equation Max(R) = e−d/λ, where Max(R) is the maximum correlation between each pair of electrodes and d is the distance between the two electrodes, was fit to the data points to determine the length constant, λ, the distance at which the correlation falls to 0.37 (11). Analysis was performed on a Sun Sparc station 20 (Sun Microsystems, Sunnyvale, California) using MATLAB software (The Mathworks Inc., Natick, Massachusetts).
Estimation of the size and number of wave fronts
The activations determined as previously described were grouped into wave fronts. Activations occurring within 40 ms of each other on adjacent electrodes were grouped into a single wave front. A minimum of three adjacent electrodes meeting this criterion was required to define a wave front. The length of the pathway of a wave front along the catheter was defined as the number of electrodes grouped into the wave front multiplied by the inter-electrode spacing (0.4 cm).
Sixteen patients, age 59 ± 14 years, (mean ± standard deviation) were studied. Patient characteristics are shown in Table 1. Electrode 1 of the catheter was always in an antero-basal position, and electrode 36 of the catheter was always in a postero-basal position. An example of recorded electrograms from the 36-pole catheter is shown in Figure 2. Surface lead II is shown in the first trace. The subsequent traces are electrograms from the LV catheter.
The activation interval during VF was 218 ± 36 ms. Figure 3Ashows the activation times from a second patient. For comparison, Figure 3Bshows a plot with the same number of data points as in panel A, but plotted in a uniform random distribution. For each VF episode tested, the distribution of activation over the catheter for the 2 s analyzed was not consistent with a uniform random distribution, p < 0.05 for each episode.
Figure 4shows electrograms from a third patient that demonstrate highly organized activations on the endocardium, while the body surface lead II recording appears disorganized. Figure 5is a plot of maximum cross-correlation between pairs of electrodes on the catheter as a function of distance between the electrodes. The length constant averaged 9 ± 2 cm across the group of patients. The total length of the line of electrodes on the catheter was 14 cm. Because the catheter spanned a large portion of the endocardial surface, albeit only in one plane, these data suggest that there are only a few wave fronts on the endocardium at any given time during VF in humans.
The average number of wave fronts that were recorded by the catheter was 9.2 ± 2.9 wave fronts/s. The size of the wave fronts measured along the catheter was 6.5 ± 4.5 cm. An example of the wave front isolation is shown in Figure 6.
The main findings from this study are threefold. First, the activation pattern during human VF is not random. Second, the correlation length along the catheter during VF is long, 9 ± 2 cm. Third, activations could be grouped into wave fronts with a mean length of 6.5 ± 4.5 cm along the catheter. All of these results suggest that 8 to 10 s after electrical induction, VF in humans consists primarily of a few large wave fronts on the endocardium at any given time.
Over the past 125 years, several electrophysiologic mechanisms have been proposed for the maintenance of VF (25). Most proposed mechanisms postulate the propagation of a number of activation wave fronts that reenter to maintain the arrhythmia. Classically, VF is thought to consist of multiple disorganized wavelets that follow constantly changing reentrant pathways. Although some results suggest that the number of wandering wavelets is large, others suggest that during the first minute of VF the number of wave fronts is small, perhaps only one (14–16). This conclusion is supported by several pieces of evidence. First, on the basis of cinematographic studies, Wiggers stated (26)that soon after its onset VF cannot be described adequately as asynchronous contraction of individual myocardial fibers, but rather that coordination and synchrony are present over large sections of myocardium (26). These large sections progressively decrease in size and increase in number over the first minutes of VF. Second, Garrey et al. (27)showed that a critical mass of myocardium, at least one fourth of the ventricular muscle in dogs, must be present for VF to persist. Third, frequency analysis of humans and dogs demonstrated a power spectrum with a well-defined peak and higher harmonics (28), suggesting some degree of organized activation during VF. Recent animal data suggest that one or two primary wave fronts located in the regions with the fastest activation rates may drive the rest of the heart and that these one or two activating regions, not the entire myocardium, are responsible for the maintenance of fibrillation (29,30).
The Kolmogorov-Smirnov test is a non-parametric test that determines whether a distribution of points is uniform across space. The results of the Kolmogorov-Smirnov test here showed that the activation times recorded during VF on the 36-pole catheter did not occur randomly but rather in an organized fashion in all patients. The simplest explanation for this organization is that individual wave fronts are larger than the 4-mm inter-electrode spacing along the catheter. These single wave fronts “sweep” along the catheter, activating multiple electrodes in sequence. In contrast, if activation wave fronts were smaller than 4 mm, then individual wave fronts would activate single electrodes, giving rise to a random activation pattern.
In order to better define the size of activation wave fronts during VF, we used a spatial correlation length measure. Correlation length is a widely used measure of spatial organization. Longer correlation lengths are associated with larger structures and more organized behavior. Correlation length has been used to help understand a wide variety of complex systems (31–33). The correlation length has recently been proposed as an estimate of the size of the reentrant pathway for both atrial fibrillation (34,35)and VF (11).
The correlation length measured here in humans is much longer than that in pigs, which was 4 to 10 mm (11). There are several possible reasons for this difference. First, the activation rate during VF is almost twice as fast in pigs as it is in humans. Second, the animal hearts were normal, whereas a majority of the patients had some form of structural heart disease. Third, and probably most important, we mapped the endocardial surface, whereas the epicardial surface of the heart was mapped in pigs. Worley et al. (36)recorded electrograms from electrodes along a plunge needle that was positioned transmurally in dog hearts. They showed that after 1 to 2 min of VF the electrograms recorded from the endocardium were sharper and less complex than those recorded from the epicardium. Future studies need to be performed to better understand this difference.
Wave front isolation showed that the average size of a wave front was 6.5 ± 4.5 cm. Taken together with the correlation length, we can estimate the number of wave fronts on the endocardial surface at any given time. If it is assumed that all parts of the endocardium behave similarly during VF, simple scaling arguments can be used to estimate the global complexity of fibrillation. If wave fronts on the endocardium are on average 6.5 to 9 cm (the range given by our two measures of wave front size, the correlation length and wave front isolation), then each wave front may cover an area of 42 to 81 cm2(6.5 cm2to 9 cm2) (11). If we assume an endocardial area of 154 cm2(π∗(length of the catheter/2)2), then we would estimate that two to four areas of activation should exist on the endocardium at any given time. Though a rough calculation, these numbers suggest that VF on the endocardium of humans consists primarily of a few relatively large wave fronts at any given time.
The major limitation of this study is that the recordings were made only from the endocardium and from a single line of electrodes. Ventricular fibrillation is likely to be more complex than the view that this catheter provides. The heart is a three-dimensional structure, and the wave fronts of VF can propagate in all three directions (37). As multipolar three-dimensional catheters become available and/or epicardial plaque mapping of patients undergoing open-heart surgery is performed (38), it may be possible to map fibrillation more completely and come to more complete conclusions about the size and number of wave fronts that support the arrhythmia, as well as to describe how those wave fronts interact to perpetuate the arrhythmia.
We did not differentiate between VF and polymorphic tachycardia. The differentiation between these two rhythms is usually based on rate. We did not see the number or size of wave fronts change with the activation rate; therefore, there is no reason to think that our results are skewed by episodes of polymorphic tachycardia.
Understanding the patterns of activation during VF may lead to more effective therapeutic options for halting arrhythmias. If there are many small wave fronts in the ventricle during fibrillation, then there is little possibility of “synchronizing” a shock to some feature of the fibrillation in order to lower the defibrillation threshold. If there are only a few large wave fronts during VF, then there is a greater possibility of “synchronizing” a defibrillation shock.
It may also be possible to use pacing-sized stimuli to help lower the defibrillation threshold in humans. KenKnight et al. (39)have shown that there is an excitable gap during VF in pigs on the epicardium and that 5 to 20 cm2can be “captured” by pacing during the arrhythmia (40). Our data suggest that there are fewer wave fronts on the endocardium during VF in humans than there are on the epicardium of pigs. Therefore, it may be easier to stimulate and capture the ventricle during fibrillation in humans, and it may be possible to stimulate a larger region of the heart with each pacing pulse. Controlling activation in a region of the heart during fibrillation, especially in a region that contains the earliest sites of global activation after failed defibrillation shocks, may help lower the defibrillation threshold.
Ventricular fibrillation in patients who have suffered cardiac arrest or sustained VF is characterized by a few large wave fronts of activation.
☆ Supported in part by NIH grant HL 66256 and a research grant from Guidant Corp., St. Paul, Minnesota.
- left ventricle, left ventricular
- ventricular fibrillation
- Received February 22, 2001.
- Revision received September 5, 2001.
- Accepted September 7, 2001.
- American College of Cardiology
- Fye W.B.
- Johnson E.E.,
- Idriss S.F.,
- Melnick S.B.,
- Wolf P.D.,
- Smith W.M.,
- Ideker R.E.
- KenKnight B.H.,
- Windecker S.,
- Chattipakorn N.,
- et al.
- Witkowski F.X.,
- Penkoske P.A.
- Lee J.J.,
- Kamjoo K.,
- Hough D.,
- et al.
- Wit A.L.,
- Hoffman B.F.,
- Cranefield P.F.
- Gray R.A.,
- Jalife J.,
- Panfilov A.V.,
- et al.
- Gray R.A.,
- Pertsov A.M.,
- Jalife J.
- Stewart A.J.,
- Allen J.D.,
- Adgey A.A.J.
- Swartz J.F.,
- Jones J.L.,
- Fletcher R.D.
- Wu T.-J.,
- Ong J.J.C.,
- Hwang C.,
- et al.
- Fasano G.,
- Franceschini A.
- Garrey W.E.
- Herbschleb J.N.,
- Heethaar R.M.,
- Tweel L.,
- Meijler R.L.
- Chen J.,
- Mandapati R.,
- Berenfeld O.,
- Skanes A.C.,
- Jalife J.
- Zaitsev A.V.,
- Berenfeld O.,
- Mironov S.F.,
- Jalife J.,
- Pertsov A.M.
- Hohenberg P.C.,
- Shraiman B.I.
- Chen P.-S.,
- Wolf P.D.,
- Dixon E.G.,
- et al.
- Walcott G.P.,
- Holman W.L.,
- Rogers J.M.,
- et al.
- KenKnight B.H.,
- Bayly P.V.,
- Gerstle R.J.,
- et al.