Regional Differences in Ventricular Fibrillation in the Open-Chest Porcine Left Ventricle
It has been hypothesized that during ventricular fibrillation (VF), the fastest activating region, the dominant domain, contains a stable reentrant circuit called a mother rotor. This hypothesis postulates that the mother rotor spawns wavefronts that propagate to maintain VF elsewhere and implies that the ratio of wavefronts propagating off a region to those propagating onto it (propoff/propon) should be >1 for the dominant domain but <1 elsewhere. To test this prediction in the left ventricular (LV) epicardium of a large animal, most of the LV free wall was mapped with 1008 electrodes in 7 pigs. VF activation rate was faster in the posterior than in the anterior LV (10.0±1.3Hz versus 9.3±1.3Hz; P<0.001). The anterior LV had a higher fraction of wavefronts that blocked than did the posterior LV and had a propoff/propon ratio <1 (P<0.001). The mean conduction velocity vectors of the VF wavefronts pointed in the direction from the posterior to the anterior LV. Although these findings favor a dominant domain in the posterior LV, the facts that the anterior LV had a higher incidence of reentry than did the posterior LV and that the posterior LV did not have propoff/propon significantly different from 1 do not. Thus, quantitative regional differences are present over the porcine LV epicardium during VF. Although these differences are not totally consistent with the presence of a dominant domain within the LV free wall, the mean conduction velocity vector is consistent with one in the septum.
Various hypotheses have been presented to explain the complex activation sequences during ventricular fibrillation (VF). One hypothesis is drawn from a computer model of atrial fibrillation where “wandering wavelets” encounter tissue in which refractoriness is spatially dispersed heterogeneously, leading to reentry that changes pathways from cycle to cycle.1 Another hypothesis states that instead of the spatial dispersion of refractoriness, the temporal restitution proportion of refractoriness causes block that sustains VF.2,3⇓ A third hypothesis is that a single stable dominant rotor, the “mother rotor,” spawns wavefronts that spread away from it and block to form complex activation sequences that maintain fibrillation in the rest of the myocardium.4–9⇓⇓⇓⇓⇓
Recent evidence for the mother rotor hypothesis comes from studies of optical mapping that imaged isolated rabbit hearts,6,7⇓ slabs of sheep right ventricular (RV) and left ventricular (LV),8 and guinea pig hearts.9 This last study reported that the mother rotor was localized to the anterior LV epicardium and was responsible for maintaining VF throughout both ventricles. However, others investigating the same location in the same species did not find evidence of a mother rotor.10 In the intact swine heart, epicardial reentry also is uncommon in the anterior LV.11 Optical mapping of slabs of sheep hearts suggests a mother rotor might be intramural.8 Conversely, optical mapping of the cut transmural surface of slabs of swine myocardium did not find evidence of a dominant rotor intramurally.12 However, the cut surface could have altered the boundary conditions of the tissue, which may have prevented the formation of an intramural rotor whose filament terminated on the cut surface.
The mother rotor hypothesis implies that the ratio of wavefronts that propagate off the region containing the mother rotor to those that propagate on to it (propoff/propon) should be >1, but that this ratio should be <1 elsewhere. It also implies that if the mother rotor is epicardial, the epicardial region containing the mother rotor should have a shorter refractory period, a higher VF activation rate, a greater incidence of reentry during VF, and a smaller amount of block during VF than the remainder of the ventricles. Global epicardial recordings from swine have demonstrated that the basal LV has a higher activation rate than the remainder of the LV and RV free walls.13 Epicardial mapping demonstrated that VF wavefronts tend to propagate from the anterior LV to the anterior RV.14 These findings suggest that the LV could be the site for a dominant domain in the swine, if one exists. Therefore, we investigated the LV free wall in an open-chest model in swine for signs of a dominant domain.
Due to the large number of electrodes needed to cover the epicardial surface of a large animal heart with sufficiently close spacing to map the VF activation sequence, exploring the mother rotor hypothesis with global electrical mapping has not previously been possible.15 By using 1008 electrodes on soft plaques that were sutured to the beating heart, we could map VF activation sequences over 55 cm2 of the LV epicardium of porcine hearts.
We determined whether differences exist between the anterior LV and the posterior LV with regard to refractoriness and during VF with regard to activation rate, propoff/propon, incidence of reentry, incidence of block, directionality of wavefronts, and other quantitative characteristics of VF activation.
Materials and Methods
Animal Preparation and Experimental Setup
The study was approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee. Seven farm pigs (University of Alabama at Birmingham Animal Resources Program, Birmingham, Ala) were anesthetized using 25 mg/kg of intravenous thiopental sodium, and anesthesia was maintained using Isoflurane in 100% oxygen. Normal metabolic state was maintained by correcting monitored levels of Po2, pH, base excess, calcium, potassium, carbon dioxide, and bicarbonate, as needed.
The chest was opened via a left lateral thoracotomy at the 5th intercostal space. A plaque containing 1008 electrodes (42×24 array) was sutured to the beating heart, covering a region from the posterior descending artery (PDA) to the left anterior descending coronary artery (LAD) (Figure 1). The plaque covered 55 cm2, which measurement indicated was about 84% of the left ventricular free wall. The electrodes in the plaque consisted of chlorodized silver spheres, 0.75 mm in diameter, with 2-mm interelectrode spacing. The ground reference for the unipolar recordings was a silver chloride electrode attached to the chest wall.
Mapping System and Data Acquisition
Two mapping systems were synchronized by a common clock for time alignment of the two mapping data streams, each recording simultaneously from 504 of the 1008 electrodes. The unipolar electrograms were band-pass filtered with a high-pass filter of 0.05 Hz and a low-pass filter of 500 Hz. The data were sampled at 2 kHz and recorded digitally with 14-bit resolution.16
VF was induced using a 9V battery applied to the base of the LV. VF was allowed to last 30 seconds before defibrillation. Twenty-five seconds of data were acquired starting from the induction of VF. Five episodes of VF were induced per animal. A minimum of 7 minutes was allowed between VF episodes. Each 25 second episode was analyzed in 5-second epochs.
Quantification of VF
We used analysis algorithms previously developed in our laboratory for quantifying VF activation patterns.17,18⇓ By grouping together active samples that were adjacent in space and time, individual wavefronts were isolated. The overall activation rate was calculated as the mean number of activations per epicardial recording site per second.
Collisions and fractionations of wavefronts were identified. A wavefront remaining after two or more wavefronts collided was classified as a new wavefront. When localized conduction block caused a wavefront to fractionate into two or more segments, each segment was classified as a new wavefront. Events were identified in which generalized conduction block caused a wavefront to totally disappear within the mapped region. Events in which a wavefront first appeared in the plaque at an electrode that was not at the edge of the plaque and was not as a result of collision or fractionation were defined as breakthrough. If a wavefront first appeared at one of the edge electrodes of the plaque, it was deemed to have propagated onto the plaque (propon) from outside the mapped tissue. If a wavefront disappeared at an edge electrode of the plaque it was assumed to have propagated off the plaque (propoff).
We used 0.5-second data segments beginning at each 5-second epoch to compute the following variables,17,18⇓ which were tabulated separately for the entire LV free wall mapped, as well as for the anterior and posterior halves of the LV free wall mapped: (1) total number of wavefronts; (2) mean activation rate; (3) mean epicardial area swept out by each wavefront; (4) multiplicity—the number of distinct activation paths in the VF pattern; (5) repeatability—the average number of wavefronts that propagated in each of the distinct pathways; (6) the fraction of wavefronts that collided; (7) the fraction of wavefronts that fractionated; (8) the fraction of wavefronts that blocked; and (9) the fraction of wavefronts that broke through.
We also calculated the directionality of VF from the propagation velocity of each wavefront. The wavefront velocity was estimated by computing the location of the centroid of the wavefront at each time sample of the data stream. The velocity of each wavefront centroid was separated into x and y vector components. The mean weighted velocity vectors of the wavefronts in the x direction (posterior to anterior) and y direction (apex to base) were calculated for each segment as described elsewhere.19
We used the entire 5-second data epoch for the analysis of reentry. This automated analysis identified and quantified wavefront components that completed at least one cycle of reentry.11 A component consists of all wavefronts that interact through fractionation or collisions. Epicardial reentry incidence was defined as the fraction of components that were reentrant. A network optimization algorithm was applied to follow the path of the tip of the reentrant circuits to quantify the mean number of reentrant cycles completed, the mean perimeter and area circumscribed by the tip of the circuit, and the mean duration of the reentrant cycles.11,17,18⇓⇓ These variables were tabulated for the anterior and posterior LV.
In addition to the mean number of activation fronts passing each epicardial site per second, we used a second method to assess activation rate. The fast Fourier transform (FFT) was computed using Welch’s method on 2-second epochs with 1.024-second overlapping Hanning windows and 0.5-Hz resolution.20 The peak of the computed power spectrum was used as an estimate of the mean activation rate for the tissue beneath each electrode in the anterior LV and the posterior LV.
The ratio of wavefronts propagating off the plaque to the wavefronts propagating onto the plaque (propoff/propon) was calculated for the entire region mapped, as well as for the anterior LV and posterior LV. We also calculated this ratio for each of the 4 edges of the entire plaque.
Activation Recovery Interval
To estimate the refractory period throughout the mapped area, the activation recovery interval (ARI) was calculated at each site.21 ARI has been shown to be correlated with the effective refractory period.
Values are shown as mean±95% confidence intervals. Statistical comparisons were performed using the software package Statview, SAS Institute. Student’s t test, single group, paired, unpaired, and standard MANOVA and, if significant, univariate ANOVA were used as appropriate. Eight nonreentrant VF variables were analyzed with multivariate and univariate analyses of variance. The variables were included in a multivariate analysis of variance with factors of animal, location (anterior versus posterior), VF episode, and epoch. A complete model with up to 3-way interactions was constructed, to determine which factors contributed to the overall effect in all of the variables. Finally, each of the variables was analyzed separately with a univariate analysis of variance using the same factors as in the final multivariate model.
An expanded Materials and Methods section can be found in the online data supplement available at http://www.circresaha.org.
The animals weighed 47.6±2.3 kg (mean±SD). The hearts weighed 231±22 (mean±SD) grams. In total, 175 epochs from 35 episodes of VF in the 7 animals were analyzed. Examples of electrograms and a single frame of an activation map animation are shown in Figure 2. From the multivariate model, the significant within-animal effects found were location and epoch. No interactions were significant between location (anterior and posterior) and epoch, indicating that the effect of location on the variables was independent of the effect of VF duration on these variables. Therefore, because our primary intent was to investigate the effect of location on VF, results for all epochs are presented together for each location.
The quantitative variables of VF for the anterior and posterior halves of the mapped region are shown in Figure 3. The number of wavefronts was higher for the posterior than the anterior LV (27 versus 34; Figure 3A). The VF activation rate was significantly higher for the posterior than the anterior LV, both by counting the number of activations at each electrode (9.3 versus 10 Hz; Figure 3B) and by FFT analysis (9.0 versus 9.3 Hz). The area swept out by the wavefronts for the anterior was greater than the posterior LV (475 versus 425 mm2; Figure 3C). The wavefronts were less complex for the anterior LV as evidenced by the lower multiplicity index (4.6 versus 5.7; Figure 3D) and the lower incidence of fractionation (0.15 versus 0.17; Figure 3G). The incidence of block was higher for the anterior compared with the posterior LV (0.26 versus 0.21; Figure 3H). There was no significant difference between the two regions with regard to the repeatability index (5.4 versus 5.5; Figure 3E) and the incidence of wavefront collision (0.16 versus 0.17; Figure 3F) or of breakthrough (0.24 versus 0.23; Figure 3I).
Even though the number of wavefronts and the activation rate were lower in the anterior LV, the incidence of reentry was significantly higher there than in the posterior LV (0.11 versus 0.08; Figure 4A). The area within (55 versus 68 mm2; Figure 4B) and the perimeter around (32 versus 36 mm; Figure 4C) the path of the central tip of the reentrant wavefronts were significantly higher for the posterior LV than for the anterior LV. Although there were no significant differences in the number of reentrant cycles (1.4 versus 1.5), the duration of a reentrant cycle was longer in the posterior than the anterior LV (69 versus 74 ms; Figure 4D). Within animal effects of VF episode and epoch were not considered for this part of the analysis, because reentrant cycles were not present in some VF epochs.
Directionality of VF
The decomposition of the weighted velocity vectors into x and y components over the entire mapped region yielded an x-velocity (parallel to the AV groove from PDA to LAD) of 0.07 m/s, which was significantly different from 0 (Figure 5A). This relationship held true for 6 of the 7 animals (Figure 5B). However the mean weighted y-velocity vector (perpendicular to the AV grove) was not significantly different from 0. There were no significant differences between the anterior and posterior LV with regard to the magnitude and angle of the velocity vectors. Figure 5C is a series of snapshots of activation maps illustrating a typical example of wavefronts propagating primarily parallel to the AV groove from PDA to LAD.
The propoff/propon ratio for the entire region mapped was not significantly different from 1 (Figure 6A). The propoff/propon ratio for the posterior LV was 1.04, which was significantly higher than for the anterior LV (0.92; Figure 6B). Although the ratio for the anterior LV was significantly different from 1, the ratio for the posterior was not (Figure 6B). Figures 6C and 6D illustrate the distribution of this ratio for the anterior LV and posterior LV, respectively. We correlated the ratio of propoff/propon to possible signatures of a dominant domain, ie, reentry on the epicardium and breakthrough to the epicardium. Propoff/propon was not related to either the incidence of reentry (r2=0.001) or the incidence of epicardial breakthrough (r2=0.075).
We also calculated the ratio of the wavefronts that propagated off to those that propagated on for each of the 4 edges of the plaque individually. The posterior edge of the plaque, adjacent to the PDA, had a ratio <1, and the anterior edge of the plaque adjacent to the LAD had a ratio >1. The ratios for the basal edge adjacent to the AV groove and the apical edge adjacent to the LV apex were not statistically different from 1 (Figure 7). These observations are consistent with the directionality of the velocity vector pointing from the PDA to the LAD and both findings indicate a significant trend for VF wavefronts to enter the mapped region from the posterior border of the mapped region and to propagate off the anterior border.
Activation Recovery Interval
The mean ARI was significantly longer for the anterior than for the posterior LV (275 versus 268 ms). This finding was present in all animals except one. There was a significant correlation between the activation rate by the FFT method and the mean ARI (r2=0.57).
This study demonstrates significant differences in quantitative characteristics of epicardial VF activation between the anterior and posterior LV free walls in open-chest swine. The posterior LV is able to sustain higher VF rates, and activation sequences are more complex than in the anterior LV. The mean velocity vector in the LV is not random, instead it points from the region of higher activation rate and greater complexity to the slower region. The incidence of block is higher in the anterior LV, the region of slower activation. The faster region, the posterior LV, does not generate more wavefronts than it receives. Although more wavefronts propagate onto the anterior LV than propagate off it, this region has the higher incidence of reentry. Moreover, the incidence of reentry and breakthrough do not correlate with the ratio of propoff/propon.
Although some of these findings are consistent with a dominant domain in the LV free wall, specifically the posterior wall, others are not (Table). The findings are consistent with at least two possible mechanisms of VF maintenance: (1) a dominant domain containing a mother rotor could be located in the interventricular septum or the posterior RV giving rise to activation fronts that propagate into the posterior LV and then the anterior LV; and (2) short-lived reentry circuits and wandering wavelets could be present throughout all of the ventricular tissue and responsible for VF maintenance, with the activation rate proportional to the refractory period in each region and with the tendency for activation fronts in the faster activating regions to spread into slower activating regions.
Varying Activation Rates Within the Ventricle
The use of FFT analysis to estimate activation rates is common in the recent literature,7–10⇓⇓⇓ in spite of little validation of this technique for this purpose.22 A recent publication points to the dilemma of sacrificing resolution of the FFT to obtain a single dominant peak and the resultant ambiguity in using the FFT for describing activation rates.12 Therefore, we also used the method of identifying activations based on a dV/dt threshold to estimate the VF activation rate. Both methods demonstrate a higher global activation rate in the posterior than the anterior LV. Others have demonstrated an almost 2-fold difference within the ventricle in the activation rates of rabbit hearts,7 slabs of sheep RV and LV,6 and guinea pig hearts.9 This study and a study by Newton et al13 demonstrate only a 7% difference between the fastest and the slowest activating regions, suggesting that the large regional differences observed in smaller hearts7–9⇓⇓ may not be present in larger hearts.
Varying Organization of VF Within the Ventricle
The posterior LV sustains a greater number of wavefronts than the anterior LV. This finding is related to the findings that the area swept out by the wavefronts is smaller and the refractory period is shorter in the posterior than the anterior LV so that the VF activation rate is faster. VF in the posterior LV is more complex than in the anterior LV by some measures, including a higher multiplicity index and fractionation, but not by others, including the number of wavefronts that follow each activation pathway (repeatability), block incidence, and collision incidence. Rogers et al14 reported that VF is slower and is more organized in the RV than the LV, containing fewer, larger wavefronts that follow fewer distinct pathways and are less likely to fractionate or collide with other wavefronts. Our data are consistent with the above study in that the region with the slower activation rate has less complex VF activation sequences exhibiting fewer, larger wavefronts that are less likely to fragment. The LV region with the more complex VF and the higher activation rate also has a lower incidence of reentry, which argues against the hypothesis that stable epicardial reentry in this region sustains VF.
Our results differ from those of Samie et al9 who reported that in the isolated, perfused guinea pig heart, the fastest activating region was in the anterior LV not the posterior LV and that a dominant domain was present in the anterior LV giving rise to VF wavefronts that propagated to the remainder of the ventricular myocardium including the posterior LV. The differences in the results could have been caused by the difference in species, the difference in size of the hearts, or the effects of removing and perfusing the isolated guinea pig hearts compared with the intact pig hearts that became ischemic during VF because of the lack of perfusion.
The significant differences in propoff/propon, activation rate, and quantitative characteristics of VF between the anterior LV and the posterior LV suggest the presence of domains on the LV epicardium. Chen et al7 used an approach similar to the propoff/propon ratio to suggest the presence of a dominant domain in rabbit hearts. They analyzed wavefronts entering and leaving the optical imaging field during episodes of VF and when a periodic source of wavelets (a rotor or breakthrough) was present, the number of wavefronts leaving was greater than the number entering the field. Chen et al used a model in which the isolated heart was perfused during VF, whereas our model consisted of the in vivo heart of a larger mammal that was not perfused during VF so that ischemia developed. In our pig model, the anterior LV had more wavefronts propagate onto it than propagate off it. However, the posterior LV did not have a propoff/propon ratio significantly >1, suggesting that it merely conducted and sustained the number of wavefronts that propagated onto it without housing a mother rotor. This assumption is consistent with our finding that the direction of wavefront propagation is primarily from PDA to LAD. Also the incidence of reentry and breakthrough did not correlate with the propoff/propon ratio, which does not support the presence of a dominant domain in the area mapped, approximately 80% of the LV epicardial free wall.
Propoff/propon was <1 for the posterior edge of the plaque and >1 for the anterior edge, suggesting that the net flow of the wavefronts entered from the posterior edge adjacent to the posterior septum and exited from the anterior edge. The mean conduction velocity vector of the wavefronts pointed from the PDA toward the LAD, which also suggests this conclusion. Although our observations do not support the dominant domain hypothesis in the extensive LV epicardial region mapped, they do not disprove the presence of a dominant domain in areas not mapped such as the posterior interventricular septum, posterior RV, or intramurally. However, the epicardial activation rate on the posterior RV has been demonstrated to be slower11 than the LV, and a recent plunge needle study has demonstrated that the epicardial activation rate is faster than either the midmyocardial or endocardial activation rate,23 suggesting that if a mother rotor exists it is probably in the interventricular septum. This possibility is supported by the findings of Ikeda et al24 who studied endocardial activation patterns during VF in the isolated, perfused canine RV free wall, interventricular septum, and LV free wall. Greater numbers of wavelets were seen in the septum than in the RV or LV free walls at baseline, and VF in the septum displayed a shorter cycle length than in the RV or LV free walls. A smaller critical mass was required to maintain VF in the septum than in the RV or LV free walls. All these findings suggest that the septum is important for VF maintenance.24
In conclusion, even though the fastest activating portion of the epicardial surface during VF is the basal half of the LV free wall, consistent with a dominant domain in this region,13 high-density electrical mapping of the region suggests that, if a dominant domain exists, it is not in the LV base but may be in the septum.
If a dominant domain is indeed in the septum, it remains to be determined why, in the large majority of cases, activation fronts propagate to the posterior LV free wall but not the anterior. The reason may be related to the fact that the ARI estimate of the refractory period in the anterior LV was significantly longer than in the posterior LV so that conduction block of activation fronts exiting the system may have been more likely in the anterior than in the posterior direction.
Differences in Refractoriness Within the Ventricle
ARIs have been shown to correlate with refractory periods21 and action potential durations.25 We found that the global ARI measures for the anterior and posterior LV moderately correlated with the mean activation rates for the corresponding regions. Although the ARI measurements were made during sinus rhythm, the hypotension resulting from VF may have activated cardiac sympathetic efferents that preferentially innervate the posterior LV,26 resulting in further shortening in refractoriness and elevation of activation rate, especially in the posterior LV during VF. Regional differences in restitution kinetics are another reason that could potentially explain the differences observed in this study. Laurita et al27 demonstrated that the restitution slope was shallower in the RV than in the LV using optical mapping in the intact guinea pig heart.
The region mapped in our study did not include the LV apex. However, it is unlikely that the apex housed a dominant domain, as neither the propoff/propon ratio of the apical edge of the plaque nor the directionality of the conduction velocity vector of VF was consistent with that possibility. The lack of mapping data from the septum and transmural region of the LV free wall is also a limitation, as the analysis performed could not detect the presence of a mother rotor in those regions.
This work was supported by NIH grants HL66256 and HL28429. K.N. is supported by the clinician scientist program of the Canadian Institute of Health Research. We would like to thank Dr Cheryl Killingsworth for expert advice on the experimental setup. We also acknowledge Kate Sreenan for assistance with manuscript preparation.
Original received June 7, 2002; revision received September 12, 2002; accepted September 16, 2002.
- ↵Moe GK. On the multiple wavelet hypothesis of atrial fibrillation. Arch Int Pharmacodyn Ther. 1962; 140: 183–188.
- ↵Garfinkel A, Kim YH, Voroshilovsky O, Qu Z, Kil JR, Lee MH, Karagueuzian HS, Weiss JN, Chen PS. Preventing ventricular fibrillation by flattening cardiac restitution. Proc Natl Acad Sci U S A. 2000; 97: 6061–6066.
- ↵Lewis T. The Mechanism and Graphic Registration of the Heart Beat, 3rd ed. London, UK: Shaw and Sons Ltd; 1925.
- ↵Gurvich NL. Fibrillation, and Defibrillation of the Heart. Moscow, USSR: Moscow Medical Publishing; 1957.
- ↵Chen J, Mandapati R, Berenfeld O, Skanes AC, Jalife J. High-frequency periodic sources underlie ventricular fibrillation in the isolated rabbit heart. Circ Res. 2000; 86: 86–93.
- ↵Zaitsev AV, Berenfeld O, Mironov SF, Jalife J, Pertsov AM. Distribution of excitation frequencies on the epicardial and endocardial surfaces of fibrillating ventricular wall of the sheep heart. Circ Res. 2000; 86: 408–417.
- ↵Samie FH, Berenfeld O, Anumonwo J, Mironov SF, Udassi S, Beaumont J, Taffet S, Pertsov AM, Jalife J. Rectification of the background potassium current: a determinant of rotor dynamics in ventricular fibrillation. Circ Res. 2001; 89: 1216–1223.
- ↵Choi BR, Liu T, Salama G. The distribution of refractory periods influences the dynamics of ventricular fibrillation. Circ Res. 2001; 16: 88:e49–e58.
- ↵Rogers JM, Huang J, Smith WM, Ideker RE. Incidence, evolution, and spatial distribution of functional re-entry during ventricular fibrillation in pigs. Circ Res. 1999; 84: 945–954.
- ↵Valderrabano M, Yang J, Omichi C, Kil J, Lamp ST, Qu Z, Lin S-F, Karaguezian H, Garfinkel A, Chen P-S, Weiss JN. Frequency analysis of ventricular fibrillation in swine ventricles. Circ Res. 2002; 90: 213–222.
- ↵Newton JC, Evans FG, Chattipakorn N, Rogers JM, Gray RA, Ideker RE. The estimated global epicardial distribution of activation rate and conduction block during porcine ventricular fibrillation. J Cardiovasc Electrophysiol. In press.
- ↵Wolf PD, Rollins DL, Simpson EV, Smith WM, Ideker RE. A 528 channel system for the acquisition and display of defibrillation and electrocardiographic potentials. In: Murray A, Arzbaecher R, eds. ProcComputers in Cardiology. Los Alamitos, Calif: IEEE Computer Society Press; 1993: 125–128.
- ↵Millar CK, Kralios FA, Lux RL. Correlation between refractory periods and activation-recovery intervals from electrograms: effects of rate and adrenergic interventions. Circulation. 1985; 72: 1372–1379.
- ↵KenKnight BH, Bayly PV, Chattipakorn N, Windecker S, Usui M, Rogers JM, Johnson CR, Ideker RE, Smith WM. Efficient frequency domain characterization of myocardial activation dynamics during ventricular fibrillation. In: Proceedings of 17th Annual International Conference of IEEE/EMBS, September 1995. Piscataway, NJ: IEEE Press; 1995: 349–350.
- ↵Newton JC. Underlying Mechanisms and Regional Control of Ventricular Fibrillation. Birmingham, Ala: University of Alabama at Birmingham; 2002. Dissertation.
- ↵Haws CW, Lux RL. Correlation between in vivo transmembrane action potential durations and activation-recovery intervals from electrograms. Circulation. 1990; 81: 281–288.
- ↵Opthof T, Misier AR, Coronel R, Vermeulen JT, Verberne HJ, Frank RG, Moulijn AC, van Capelle FJ, Janse MJ. Dispersion of refractoriness in canine ventricular myocardium. Effects of sympathetic stimulation. Circ Res. 1991; 68: 1204–1215.
- ↵Laurita KR, Girouard SD, Rosenbaum D. Modulation of ventricular repolarization by a premature stimulus: Role of epicardial dispersion of repolarization kinetics demonstrated by optical mapping of the intact guinea pig heart. Circ Res. 79: 493–503.