INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

IN 35-50% OF PATIENTS who succumb to congestive heart failure (CHF), the mode of death is sudden and unexpected (22, 32). The majority of these sudden deaths are likely due to ventricular fibrillation (VF). Recent clinical trials showing improved survival using internal defibrillators in high-risk patients with CHF (30, 48) underscore the importance of attaining a clear understanding of the mechanisms of initiation and maintenance of VF, some of which remain elusive. Although mapping studies of ventricular activation during VF in animals have enhanced our understanding of many basic properties of this arrhythmia (19), equivalent data in animal models of chronic CHF are lacking. Consequently, relatively little is known about how VF in dilated hearts with left ventricular (LV) dysfunction may differ from VF in normal hearts.

Studies of VF activation patterns in normal hearts have demonstrated relatively large, distinct reentrant wavefronts during early-onset VF (25). Interaction between wavefronts at a vulnerable period may be responsible for the production of new wavefronts and subsequent perpetuation of VF. Quantification of these wavefronts and identification of possible patterns in the sequence of activation may be crucial to our understanding of onset, maintenance, and subsequent therapy of VF in CHF. Despite evidence for a dominant role of the LV endocardial surface in the maintenance of VF (54), previous mapping studies of VF have utilized recordings from limited sections of myocardium. The study reported herein was designed to compare full-surface endocardial mapping data of VF in dogs with CHF to VF in dogs with normal hearts. Because ischemic heart disease is the most common cause of CHF, we used a model of myocardial infarction followed by rapid ventricular pacing to induce congestive heart failure (15).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

This study was approved by the Animal Studies Subcommittee of the Minneapolis Veterans Administration Medical Center, conformed to the Federal Animal Welfare Act administered by the United States Department of Agriculture, followed the guidelines of the National Research Council's Guide for Care and Use of Laboratory Animals, and was performed in laboratories approved by the American Association for Accreditation of Laboratory Animal Care.

Two groups of dogs were studied: six normal dogs and seven dogs with CHF induced by ischemic injury followed by rapid ventricular pacing. Normal dogs were anesthetized with the use of intravenous pentobarbital sodium induction (25 mg/kg) with halothane maintenance. A pacing catheter was placed in the right ventricular apex via a femoral vein, and a noncontact, multielectrode catheter (EnSite; Endocardial Solutions, St. Paul, MN) was introduced via a femoral artery, advanced retrograde across the aortic valve, and positioned in the mid-LV. After baseline recordings were made in sinus rhythm, VF was induced by either brief application of current from a 9-V battery or by rapid pacing delivered via the right ventricular pacing electrode. Continuous recordings were made for at least 5 min.

CHF dogs had myocardial infarction induced under surgical anesthesia with the use of a coronary angioplasty catheter. The balloon was positioned in the midvessel (or slightly distal) of either the circumflex or the anterior descending coronary artery, inflated for 90 min, and then removed. A unipolar pacing lead was positioned under fluoroscopic guidance in the right ventricular apex and tunneled to a modified high-rate pacemaker generator (Medtronic, Minneapolis, MN) in a subcutaneous pocket in the neck. After a minimum 2-wk recovery period, the pacemaker was programmed to 250 beats/min, and the animals were paced continuously for 3 wk. At the end of 3 wk, the pacemaker was deactivated, the animals were again anesthetized, and the multi-electrode catheter was positioned in the LV as described above. Baseline recordings were made, VF was induced, and endocardial activation maps were recorded continuously for at least 5 min.

The catheter used in this study has an array of 64 electrodes positioned around a balloon that, when inflated within the cardiac chamber, provides an ellipsoid 1.8 cm in diameter and 4.6 cm long. These electrodes are connected to a computer system through a breakout box that also accommodates signals from three separate ring electrodes attached to the catheter shaft, a 12-lead surface electrocardiogram, and up to 32 standard electrophysiology electrodes. The electrode signals are amplified and digitized in an interface unit and fed into a computer workstation that can display individual electrograms as well as a real-time dynamic display of endocardial activation patterns projected onto a three-dimensional model of the LV.

The model of the LV is created by "tracing" the endocardium with a standard electrophysiology (EP) catheter. Fluoroscopic guidance is used to manipulate the tip of the EP catheter into as many positions as possible within the ventricle. The position of the EP catheter tip is continuously monitored relative to the balloon electrode array over a nominal 2-min time period by use of a 50-µA 5.68-kHz current delivered through the tip electrode while the potential on the array electrodes is sensed. Because the coordinates of the array electrodes are predetermined, the position of the roving electrode can be determined by use of a point-source model (United States patent 5662108). A smooth spline fit of the most distal locations of the EP catheter tip are used to reconstruct a model of the LV (14).

Chamber geometry, known locations of the array electrodes, and passive noncontact measured endocardial potentials at each electrode on the array are combined by use of a boundary element inverse method to solve LaPlace's equation and reconstruct potentials for the endocardial surface. This system computes 3,360 endocardial (virtual) potentials, with the virtual potential at each site refreshed at a rate of 1,200/s. The potentials are color coded by amplitude and displayed on the three-dimensional model of the LV to produce a high-resolution, full-surface real-time electroanatomic map of endocardial activation. The potentials can also be translated into virtual electrograms by displaying potential amplitude at each site versus time.

The computerized mapping technique used in this study has previously been validated in the LV during normal sinus rhythm by Gornick et al. (14) and in the right atrium during sinus rhythm, atrial flutter, and atrial fibrillation by Kadish et al. (21). To insure the technique is useful in VF, we compared virtual electrograms with contact electrograms in three normal anesthetized dogs during VF. A multielectrode catheter (Crista Cath; Cordis Webster, Diamond Bar, CA) was placed in the LV together with the EnSite system, and unipolar endocardial electrograms were compared with their corresponding virtual electrograms for each of 16 separate sites. After induction of VF, the endocardial electrograms from the tip electrode of the multielectrode catheter were recorded for 3 min, and then the 15 other endocardial electrodes were recorded sequentially for a minimum of 15 s each. The individual recordings from each site were analyzed in 2-s segments by use of cross correlation to compare the endocardial electrograms with the corresponding virtual electrograms. Because the sampling rate was 1,200 Hz, this produced 2,400 x-y pairs per sample segment. The correlation coefficients for each sample segment were averaged to get a mean correlation coefficient for each of the 16 individual sites. All 16 sites were then averaged to provide a grand mean, felt to be a fair representation of how well virtual electrograms match directly recorded endocardial electrograms. This grand mean, including all three dogs, was 0.724 ± 0.155, with a median of 0.723. If only those sites within 34 mm of the center of the EnSite catheter are included, the grand mean is 0.781 ± 0.141, with a median of 0.777. The three dogs used here for system validation were not included in the group of normal dogs used for comparison to CHF dogs.

In both the normal and CHF dogs, 2 s of VF were analyzed at two different times, the first beginning 10 s after the onset of VF (Wiggers' stage II) and the second 5 min later. The fundamental frequency components of the VF waveforms were calculated with the use of fast Fourier transform (FFT) over a range of 2-20 Hz for selected surface leads and virtual endocardial electrograms. The peak frequency and centroid frequency of the power spectrum were determined for each FFT at the time points of interest. Centroid frequency represents that frequency for which the area under the power curve for all frequencies above that point would equal the area under the power curve for all frequencies below that point (4). It is defined as
for i = 1 to n, where f is frequency and p is power.

To quantify the number of distinct wavefronts present at any time on the endocardial surface of the LV, color potentials projected onto the electroanatomic map of the LV were set such that white represented a small range of amplitude representing the most negative potentials present, and a rainbow spectrum represented progressively more positive potentials. The scale was adjusted for each animal during sinus rhythm to provide the most optimal view of normal depolarization. This generally resulted in the white color representing approximately the most negative 15% of the voltage range for each dog. The scale was not adjusted subsequently during VF. This process is analogous to producing a color contour map of a mountain range, with the colors representing voltage amplitude instead of mountain altitude. The number of simultaneous noncontiguous wavefronts present at any given time was determined by counting the number of distinct, separate white areas present on the endocardial surface. These noncontiguous wavefronts were manually counted every 10 ms for a 2-s period 10 s after the onset of VF and again at 5 min. FFT analysis of the number of noncontiguous wavefronts present over time was performed to provide an estimate of the dominant frequency with which individual wavefronts emerged and dissipated, or the "turnover" rate of VF.

Enlargement of the LV and decreased LV function in response to the infarction and rapid pacing in the CHF dogs were documented by obtaining echocardiograms before intervention and again at the terminal study after 3 wk of rapid pacing. Echocardiograms were performed in the right lateral position, using an HP Sonos 1000 phased array two-dimensional imaging system (Hewlett-Packard, Andover, MA). For the preterminal study, the echocardiograms were obtained after the pacemaker had been turned off. Internal diameters of the LV in systole and diastole (LVID_s and LVID_d, respectively) were measured from the septum to posterior wall in the short-axis view at the level of the chordae tendineae. Ejection fraction was estimated with the use of planimetry to calculate ventricular end-systolic and end-diastolic areas in the short-axis view at the level of the chordae tendineae and by applying the single-plane ellipse formula to transpose to systolic and diastolic volumes. Postmortem examinations of the hearts of the CHF dogs quantified the extent of myocardial infarction induced by the intracoronary 90-min balloon inflation. The LV was trimmed and cut into transverse sections at 5 mm-intervals and stained with tetrazolium blue, and the area of scar was obtained by planimetry. The percent area for each slice (except base and apex) was averaged, and the average was multiplied by the weight of the slice to get the amount of scar in each slice. These were then summed and divided by the total LV weight to get the percentage of ventricle that is scar.

Data are presented as means ± SD, and differences were considered significant for P < 0.05. Student's unpaired t-test was used for comparisons between groups (normal vs. CHF), and Student's paired t-test was used for within-group comparisons. When multiple comparisons were made within a group, i.e., in comparing several endocardial electrograms to surface electrograms for each animal, analysis was done using analysis of variance for repeated measures on the same elements.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

LV function. The dogs in the CHF group developed significant LV enlargement and dysfunction in response to the intervention. Baseline LV short-axis LVID_d before any intervention in the seven CHF dogs was 2.8 ± 0.2 cm. This increased to 4.3 ± 0.6 cm at the final study, i.e., after induction of a myocardial infarction followed by 3 wk of rapid ventricular pacing (P < 0.005). Ejection fraction decreased from 68.8 ± 13.3 to 22.1 ± 4.1% (P < 0.001) as a result of the infarction and rapid pacing. LV damage induced by the 90-min coronary balloon occlusion resulted in scar tissue ranging from 0 to 1.5% of the total LV myocardium in the seven CHF dogs, with a mean of 0.6 ± 0.6%. The single dog with no infarction on pathological examination had ventricular tachycardia during angioplasty balloon inflation, requiring premature reperfusion before completion of 90 min of occlusion. Echocardiograms were not performed in the six normal dogs in which VF was induced without prior intervention.

Full-surface endocardial mapping. The real-time color-coded display of virtual potentials provided a dynamic view of endocardial events during ventricular activation that cannot be fully represented with static images; however, Figs. 1 and 2 present snapshot photos of several relevant time points. Ventricular activation in a normal dog during sinus rhythm is seen in Fig. 1. As noted in MATERIALS AND METHODS, the color coding is scaled such that white is maximum negative voltage. Early septal breakthrough in normal sinus rhythm can be seen in Fig. 1A at the onset of the QRS. Figure 1, B-D, shows subsequent spread of depolarization throughout the endocardium at 6-ms intervals. Figure 2 shows a representative time sequence from the same normal animal in VF. In Fig. 2, the panels are at 30-ms intervals. Note that several wavefronts are evident at once and that the number of noncontiguous wavefronts varies with time.

View larger version (123K): [in this window] [in a new window]   Fig. 1.   Ventricular activation maps color coded by amplitude at the onset of the QRS (A) and subsequently at 6-ms intervals (B-D) for a normal dog in sinus rhythm. Each map is oriented as if viewer is looking into the left ventricular (LV) cavity through the base of the heart with the aortic and mitral valves removed and the multielectrode array balloon catheter positioned in the LV cavity. Septal breakout can be seen in A, with white color representing peak negative voltage. B-D: depolarization front spreads throughout the ventricle.

View larger version (145K): [in this window] [in a new window]   Fig. 2.   Serial ventricular activation maps color coded by amplitude in the same normal dog as in Fig. 1 but at 10 s after onset of ventricular fibrillation (VF; A) and subsequently at 30-ms intervals (B-D).

View larger version (22K): [in this window] [in a new window]   Fig. 3.   Representative examples of a computer-generated endocardial electrogram and simultaneous surface lead I electrocardiogram, each from 2 different animals during VF. Both are clearly in VF as evidenced by lead II, but the endocardial electrogram in A is highly organized, whereas the endocardial electrogram in B is more irregular, consistent with the surface lead.

Surface leads versus endocardial electrograms. The occasional disparity between the appearance of the virtual endocardial electrograms and the surface ECG raised the possibility that the basic characteristics of the two waveforms (surface vs. endocardial) might differ. We therefore performed FFT on two surface ECG leads (II and V2) and four representative endocardial virtual electrograms (at midventricle on the septum, anterior, lateral, and posterior walls) over the same 2-s time period beginning 10 s after the onset of VF. An example for a typical case demonstrating the results for both a surface and virtual endocardial lead is presented in Fig. 4.

View larger version (12K): [in this window] [in a new window]   Fig. 4.   Example of the power spectrum obtained by fast Fourier transform (FFT) of surface lead II and a virtual electrogram from a normal dog 10 s after the onset of VF.

View larger version (12K): [in this window] [in a new window]   Fig. 5.   Normalized power spectra combined by summation for all congestive heart failure (CHF) dogs, obtained by FFT of surface lead II and a virtual endocardial electrogram.

Effects of CHF on frequency spectra. To determine whether development of CHF alters the fundamental frequency characteristics of fibrillation waveforms, the same leads as noted above were compared in CHF dogs and normal dogs. During early-onset VF (10 s), neither the peak nor centroid frequency differed in CHF compared with normal dogs for any of the two surface or four virtual sites (i.e., all P values > 0.05 by use of Student's unpaired t-test).

Quantitative assessment of ventricular wavefronts. The full-surface real-time three-dimensional computerized display of endocardial potentials allowed us to observe the emergence, dispersion, and dissipation of wavefronts over time. We quantitated the number of active wavefronts by counting the number of distinct noncontiguous wavefronts present at 10-ms intervals for all dogs over a 2-s period at 10 s after the onset of VF and again 5 min later, with no interim therapy other than continuing ventilator support. The results are presented in Table 1. The average number of wavefronts present after 10 s of VF was decreased after 5 min of VF for every dog in the normal group, and this decrease was statistically significant for P < 0.05. These data are more completely presented in the histogram of Fig. 6, which shows the downward shift in the distribution of number of wavefronts present at any given time as VF progresses from 10 s to 5 min. The average number of wavefronts present at any time decreased from 10 s to 5 min in the CHF dogs as well, but this change did not reach statistical significance (P = 0.18). Most interestingly, the number of wavefronts present in the CHF dogs at the onset of VF (10 s) was significantly less (P < 0.005) than in the normal dogs (Table 1 and Fig. 7). The CHF dogs appear at the onset of fibrillation to be somewhat similar to the normal dogs after the latter have been fibrillating for 5 min.

View larger version (14K): [in this window] [in a new window]   Fig. 6.   Histogram of no. of distinct noncontiguous wavefronts present every 10 ms for a 2-s period at 10 s after onset of VF (solid bars) and again after 5 min of sustained VF (open bars). The distribution shifts toward fewer wavefronts present at any given time at 5 min of VF compared with early onset (10 s).

View larger version (15K): [in this window] [in a new window]   Fig. 7.   Histogram of no. of distinct noncontiguous wavefronts present every 10 ms for a 2-s period at 10 s after onset of VF for all normal (NL) dogs (solid bars) compared with all CHF dogs (open bars). The distribution shifts toward fewer wavefronts active at any given time in CHF dogs compared with normals.

Fibrillation turnover. As the quantitative data were being tabulated, it became evident that there was a general pattern to the way in which the number of wavefronts appeared and disappeared. This prompted us to apply FFT to the number of wavefronts present over time to obtain an assessment of the turnover rate of VF, i.e., the frequency with which wavefronts emerge and dissipate. The raw data (obtained every 10 ms for a 2-s period beginning 10 s after the onset of VF) for a representative animal are shown in Fig. 8, along with the FFT analysis on that series of numbers. The peak frequency for all normal dogs 10 s after the onset of VF was 9.0 ± 1.3 Hz, and this was unchanged 5 min later (7.9 ± 3.4 Hz, P = 0.26). Centroid frequency was also unchanged from 10 s (10.0 ± 0.3 Hz) to 5 min (10.2 ± 0.6 Hz, P = 0.99). Similar results were seen in the CHF dogs, where peak frequency at 10 s (8.0 ± 1.6 Hz) remained unchanged at 5 min (9.0 ± 2.5 Hz, P = 0.37), and centroid frequency was also unchanged (9.6 ± 1.3 to 9.6 ± 0.7 Hz, P = 0.99). At 10 s after the onset of VF, there was no significant difference between normal and CHF dogs for either peak frequency or centroid frequency. Thus the turnover rate for VF wavefronts does not change after 5 min of sustained fibrillation and is not different in normal ventricles compared with failing ventricles.

View larger version (20K): [in this window] [in a new window]   Fig. 8.   A: raw data from a representative animal listing sequentially (read in columns from left to right) the no. of noncontiguous wavefronts present at 10-ms intervals for a 2-s period during early-onset VF. B: FFT analysis on the raw data to provide an estimate of the turnover rate of fibrillation wavefronts, i.e., the relative rate at which wavefronts emerge and dissipate.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Methodology. Gornick et al. (14) demonstrated that the system used in this study provides anatomically accurate endocardial maps of the LV that allow identification and location of endocardial pacing sites to within 4.0 ± 3.2 mm of their pathologically confirmed source. Correlation of timing and morphology of computed electrograms compared with contact electrograms was 0.97, and timing difference between computed and contact electrograms for best fit was 0.64 ± 2.48 ms. This methodology has also been validated for endocardial mapping of the right atrium during sinus rhythm, atrial flutter, and atrial fibrillation (21), where the correlation coefficients between contact and virtual electrograms were 0.80 ± 0.12, 0.85 ± 0.17, and 0.81 ± 0.18, respectively, without any correction for timing differences between systems. Our correlation coefficient between virtual and directly recorded endocardial electrograms during VF was slightly lower (0.72). This may be due in part to the larger size of the ventricle compared with the atrium, for the ability of the system to accurately reproduce electrograms is expected to diminish as the distance between the source and the recording electrodes increases. This is supported by the increase in correlation coefficient we noted when analysis was limited to sites within 34 mm of the center of the EnSite catheter (0.78). It is also possible that degeneration of the magnitude of electrograms with progressive VF could contribute to variation between recorded and virtual electrograms, and these correlations were calculated without any corrections for possible timing differences between the directly recorded and calculated (virtual) electrograms.

Effects of CHF on frequency components of VF waveforms. FFT of electrograms recorded during early-onset VF from two representative surface leads showed a peak frequency in the power spectrum similar to that of four virtual endocardial sites. The peak frequency at these six sites ranged from 6.2 to 7.4 Hz. These peaks occur in the same range as those reported by others for dogs (13, 29) and swine (5) and did not differ between normal animals and those with CHF.

Wavefront activity. Mapping of myocardial activation has been examined by recordings from right and left ventricular epicardial (2, 19), endocardial (49), and intramural (35) electrode arrays and more recently by use of optical recordings from epicardial surfaces stained with voltage-sensitive dye (17, 52). Such studies have excluded VF as being random and rather have described varying degrees of organization with constantly changing wavefronts that evolve as VF persists, leading to postulation that VF is "deterministic chaos" (11). The occasional disparity we noted, where the surface electrocardiogram appears disorganized whereas the activation patterns recorded directly from the heart appear well organized, is consistent with the previous findings of Ideker et al. (18). We have not analyzed our data to determine whether CHF hearts differ from normal hearts in specific patterns of wavefront activity such as spirals (9, 33) or rotors (16, 17), the degree to which the complexity of activation patterns differ (39), or the extent to which reentrant versus nonreentrant pathways appear (20, 25). However, our method of mapping over the full endocardial surface allowed us to quantitate ventricular electrical activity in VF in a manner previously not possible. As a result, we were able to determine that the number of active wavefronts declines over 5 min of VF and that fewer wavefronts are active at any given time during VF in CHF than in normal hearts. A comparison of our activation maps to isochronal maps obtained from portions of intact ventricles or isolated tissue blocks is difficult at best. However, if the sizes of the distinct wavefronts of other studies are extrapolated to a full surface of ventricle in proportion to the area of tissue studied, a quantification of wavefronts produces numbers consistent with our findings. For example, Witkowski et al. (52) recorded from ~30% of the surface of the heart, and their isochronal maps show one or two distinct wavefronts present in the example frames. This would suggest that up to six wavefronts may be present at one time over the entire epicardial surface. Of course we recognize that the right ventricle may be importantly involved in the onset and/or maintenance of VF, and we did not include the right ventricle in our analysis.

Turnover rate of wavefronts. We have described the dominant frequency in the FFT performed on the number of wavefronts present as they emerge and dissipate with time during VF as the "turnover rate." This turnover rate is similar to the peak frequency of the power spectrum performed on the endocardial electrograms and surface leads. It is not surprising that the rate at which wavefronts emerge and dissipate on the endocardial surface is linked to the fundamental frequency at which individual areas of the endocardium activate. Like the peak and centroid frequencies of the virtual electrograms, the turnover rate was no different in CHF than in normal hearts, despite the fact that fewer wavefronts were present in CHF.

Conclusions. The effectiveness of internal defibrillators in preventing sudden death in high-risk patients with CHF emphasizes the need to better understand the mechanisms involved in initiation and maintenance of VF. The finding in this study that wavefront activity is decreased in CHF in a way similar to that seen late in progression of VF in normal hearts is provocative. With fewer active wavefronts at the onset of VF, CHF hearts appear to begin VF with the same disadvantage for therapy as hearts that have already persisted in VF for ~5 min. The exact metabolic/electrophysiological changes causing the decrease in wavefront activity and resistance to defibrillation remain to be elucidated.