Our objective was to establish a novel model for the study of ventricular fibrillation (VF) in humans. We adopted the established techniques of optical mapping to human ventricles for the first time to determine whether human VF is the result of wave breaks and singularity point formation and is maintained by high-frequency rotors and fibrillatory conduction. We describe the technique of acquiring optical signals in human hearts during VF, their characteristics, and the feasibility of possible analyses that could be performed to elucidate mechanisms of human VF. We used explanted hearts from five cardiomyopathic patients who underwent transplantation. The hearts were Langendorff perfused with Tyrode solution (95% O2-5% CO2), and the potentiometric dye di-4-ANEPPS was injected as a bolus into the coronary circulation. Fluorescence was excited at 531 ± 20 nm with a 150-W halogen light source; the emission signal was long-pass filtered at 610 nm and recorded with a mapping camera. Fractional change of fluorescence varied between 2% and 12%. Average signal-to-noise ratio was 40 dB. The mean velocity of VF wave fronts was 0.25 ± 0.04 m/s. Submillimetric spatial resolution (0.65–0.85 mm), activation mapping, and transformation of the data to phase-based analysis revealed reentrant, colliding, and fractionating wave fronts in human VF. On many occasions the VF wave fronts were as large as the entire vertical length (8 cm) of the mapping field, suggesting that there are a limited number of wave fronts on the human heart during VF. Phase transformation of the optical signals allowed the first demonstration ever of phase singularity point, wave breaks, and rotor formation in human VF. This method provides opportunities for potential analyses toward elucidation of the mechanisms of VF and defibrillation in humans.
- action potential duration
- conduction block
understanding the mechanisms of human ventricular fibrillation (VF) has been limited because of the constraints of using conventional mapping techniques and detection of local activation, as detailed in a recent editorial (6). The maximal spatial resolution that can be attained is limited by the number of electrodes that can be positioned at any given time to make contact with the myocardium. Conventional contact multielectrode mapping has been successful in mapping activation sequences and elucidating mechanisms of ventricular tachycardia in humans (3). In VF, where wave fronts can fractionate and collide, complex electrograms are registered at the recording unipolar electrodes that may be due to far field activity, graded responses, and subthreshold electrotonic propagation in addition to local activation. Assigning local activations of complex electrograms based on phase, short Fourier transform, or even first time derivative could be problematic (6). Although bipolar electrical mapping may provide information on local electrical activity, it has the drawback of directional limitation in VF, where wave fronts that are perpendicular to the recording axis of the bipole may not register an electrogram.
Optical mapping has allowed scientists to reliably measure local activation and repolarization (12). Methods have been developed to perform optical mapping in Langendorff-perfused animal hearts (13), and techniques have been devised to ameliorate the problems of motion artifact and dye toxicity (4). The practical application of this method has been the transformation of the data into phase-based trajectories for mapping electrical activity and elucidating mechanisms of VF (2, 5, 14). This has allowed the understanding of the mechanism of rotor activity and its role in VF in animal hearts under normal (14, 15) and ischemic (17) conditions. Thus optical imaging is a powerful tool to study the mechanism of VF. If optical mapping can be applied to human hearts, it may provide us with an additional tool to elucidate the mechanisms of human VF. Hence, for the first time, we have used optical imaging of ventricular electrical wave propagation in intact human hearts in a Langendorff setup. Here we provide details on the methodology of acquiring optical signals, their quality, and the feasibility of possible analyses that could be performed for elucidation of the mechanisms of VF. In addition, we provide the first demonstration in the human heart that VF is associated with wave breaks and singularity point formation and is maintained by high-frequency rotors and fibrillatory conduction.
MATERIALS AND METHODS
This protocol was approved by the University Health Network ethics committee, and informed consent was obtained from each patient. The hearts studied were explanted from five cardiomyopathic patients (2 females, 3 males) who underwent transplantation. The mean age was 45 ± 10 yr. All patients had ejection fraction <20%. Both female patients were taking amiodarone at the time the hearts were explanted. Immediately after the heart was explanted from the recipient, it was placed in cold Tyrode solution, transported to an adjacent room <5 min away, and flushed thoroughly to remove blood particles. The explanted human hearts have minimal aortic tissue, making retrograde perfusion of coronaries not possible. Thus we selectively cannulated the right and left main coronary arteries in all hearts, anchored them with 2-0 silk sutures, and perfused the hearts. To keep adequate coronary perfusion the flow rate was maintained at 0.9–1.1 ml·g−1·min−1. The hearts varied in weight between 400 and 625 g; thus the flow rate varied between 0.44 and 0.7 l/min. The height of the perfusion reservoir was then adjusted to maintain a perfusion pressure of 60–65 mmHg. The temperature was maintained at 37°C and continuously monitored. In two of the five hearts, to exclude the epicardial fat layer, we mapped the endocardium while selectively perfusing the right and left coronary arteries. For endocardial mapping, ventricles were opened by an incision parallel to the posterior interventricular septum and the free walls were immobilized. Once the preparation was stabilized for 10 min, the mapping protocol was completed within 30 min.
The hearts were Langendorff perfused with Tyrode solution (95% O2-5% CO2), and potentiometric dye (0.2 mg di-4-ANEPPS dissolved in 5 ml of DMSO and further diluted into 20 ml of Tyrode solution) was injected as a single bolus into the coronary circulation over a period of 5 min. During this time flow rate was turned down in the Langendorff pump. The imaged region was physically immobilized to minimize motion. Immobilization was performed only just before the images were acquired, by holding the preparation between two clear Plexiglas sheets as described previously (14, 15). Fluorescence was excited at 531 ± 20 nm (FF01-531/40-25, Semrock, Rochester, NY) with a 150-W halogen light source (MHF-G150LR, Moritek Japan). The emission signal was long-pass filtered at 610 nm and recorded with a charge-coupled device and CMOS camera [MiCAM02 (n = 2) or Ultima (n = 3), BrainVision]. The spatial resolution, depending on the mapping window, was 0.65–0.85 mm, and the temporal resolution was as fast as 1 kHz. The heart was paced at a cycle length of 600 ms at twice diastolic threshold from the posterior right ventricular (RV) epicardium with hook electrodes. The fractional change of fluorescence (or dF/F) is linearly proportional to the voltage across the membrane of excitable cells and was measured by BV-analyzer software (BrainVision). We measured an average signal-to-noise ratio (SNR) by taking the value in decibels of the dominant frequency (DF) of action potential (AP) signals and subtracting the noise floor of the power spectrum. VF was induced by briefly touching the heart with the two poles of a 9-V battery. Three to five episodes of VF were recorded per heart for 2 s each, 5 s after VF had stabilized into an apparently chaotic rhythm with no contraction of the myocardium. The changes in dye signal were recorded and analyzed with BV-analyzer software (BrainVision). Phase singularities and rotors were detected as detailed previously (14).
Maximum and minimum ventricular conduction velocities (CVmax and CVmin) were quantified by pacing the RV free wall. In each case, the magnitude of local conduction vectors was determined on the basis of the activation times of surrounding pixels, as described elsewhere (9a). Briefly, optical movies of paced activity were signal averaged in the absence of pharmacological motion reduction. The time of the optical AP upstroke was marked at the maximum time derivative of the fluorescence change, and a diastolic baseline was established immediately prior to the upstroke. Activation times (defined as the time at which each signal reached 50% of its upstroke amplitude) were determined for each pixel. Local conduction vectors were calculated for each pixel on the basis of the activation times of surrounding pixels. Only those vectors that proceeded in an orderly centrifugal fashion from the pacing electrode were analyzed. Similarly, simultaneously activating pixels close to the stimulating electrodes were ignored. Vectors were averaged in 30° bins of orientation, and CVmax and CVmin were was recorded from each activation sequence (9a).
Examples of APs on the endocardial aspect of RV free wall during ventricular pacing are illustrated in Fig. 1A. dF/F varied between 2% and 12% for the hearts studied. The average SNR was 40 dB. The mean AP duration, while the RV epicardium was paced at 600 ms, varied between 205 ± 20 and 264 ± 50 ms depending on the surface mapped for the five hearts studied. The conduction velocity of the paced wave front varied between 0.41 (CVmin) and 0.87 (CVmax) m/s. Examples of the VF recordings are illustrated in Fig. 1B with corresponding APs for three of the pixels.
Figure 2A shows an image of the left ventricular (LV) interventricular septum with an overlay of an area that was mapped electrically. We performed electrical mapping during VF, using a 112 multielectrode contact plaque array with 2-mm spatial resolution. The same septal region was optically mapped soon thereafter. Representative electrograms in VF obtained from four extracellular electrodes are presented in Fig. 2B. These electrograms are broad and of low amplitude and low frequency. In this particular example, it is difficult to identify local activation accurately and consistently from contact electrical mapping for the purposes of constructing isochrones. The complex activations seen by extracellular multielectrode mapping in VF, due to reentrant, colliding wave fronts and electrotonic interactions, are impossible to discern from true local activation. The clear advantage of optical mapping over extracellular electrode mapping in determining local activation is demonstrated in Fig. 2, C and D. The optical signals acquired during VF show distinct upstrokes indicating local activation. Supplemental Video 1 created from these activation sequences shows wave fronts that invade the septum, collide, fractionate, and reenter. (The online version of this article contains supplemental data.) The direction of wave fronts in the region mapped was centripetal. The mean velocity of VF wave fronts was 0.25 ± 0.04 m/s.
The identification of local activation lends itself to construction of isochronal maps. The directional information of the wave fronts is discerned from the isochrones seen in Fig. 2, E and F. Figure 2E reveals an example of a VF activation front that is large and demonstrates a figure eight progression depicted in 10-ms isochrones. Sites 1–5 indicate the sites from which the five optical signals seen in Fig. 2C were acquired over a 140-ms sequence from a1 to a2. On many occasions the VF wave fronts were large, sweeping out the entire vertical length (8 cm) of the interventricular septum. Similarly, Fig. 2F is another 10-ms isochrone map acquired 3 s later with a smaller figure eight activation sequence in a different location. Sites 1–4 indicate the locations from which the optical signals seen in Fig. 2D were acquired over a 140-ms sequence indicated in the time window b1–b2. While a figure eight reentry is seen in one area, multiple dissociated activations are seen simultaneously in other areas, consistent with wave breaks.
The assessment of rotors in human VF has been limited by the spatial resolution achievable in human tissue (6). Also, the optical AP recordings are better suited for phase-based analysis (1, 5). Supplemental Video 2 presents phase-transformed data that reveal transiently erupting rotors, the prominent one lasting four rotations in the top right corner. Figure 3A reveals a series of snapshots of this rotor showing the cycle length to be 155 ms. The DF map of the corresponding VF episode is shown in Fig. 3B. This reveals that during VF the region mapped does not demonstrate similar activation rate but rather is made up of domains with spatially distributed activation rates ranging from 5.4 to 6.8 Hz. As in previous animal experiments, the regions at which rotors were observed as seen in Supplemental Video 2 correspond to the domains with the highest activation rate of 6.8 Hz in the DF map.
We were also able to map rotor activity on the epicardium. Supplemental Video 3 is an example of epicardial mapping on another heart in which we identified the base of the posterior papillary muscle on the endocardium. Using this identification, we located the corresponding region on the epicardium and mapped the region centered at this location. This video illustrates the posterior LV and RV, with the posterior descending artery in the middle of the mapping field. It demonstrates a transient rotor lasting five rotations, before it is destabilized by a wave front that invades from the RV. In this case the rotor localized to the area over the base of the posterior papillary muscle on the epicardium, at the junction of the posterior LV free wall and the posterior interventricular septum.
We also studied the potential effects of the dye on conduction velocity by comparing values obtained from conventional electrical mapping techniques performed before and after optical mapping measurements. In one experiment, high-resolution multielectrode mapping showed that conduction velocity did not change appreciably on infusion of the dye. Before infusion CVmax was 0.81 m/s; after infusion it remained at 0.87 m/s.
This is the first report of optical mapping of ventricular activation in fibrillating human hearts. The most important new results of the study are as follows. 1) We have demonstrated that the technique is feasible and provides opportunities for potential analyses that will shed light on the mechanisms of VF in humans. 2) High-resolution optical mapping of the endocardial surface allowed for the first time an accurate demonstration that the mean velocity of VF wave fronts was 0.25 ± 0.04 m/s, which is significantly slower than both CVmin (0.41 m/s) and CVmax (0.87 m/s) of paced wave fronts. 3) Submillimetric spatial resolution, activation mapping, and transformation of the data to phase-based analysis revealed reentrant, colliding, and fractionating wave fronts in human VF. 4) Phase transformation of the optical signals allowed for assessment of phase singularity points and rotor formation. 5) In some experiments sustained high-frequency rotors were demonstrated to maintain the overall fibrillatory activity within the optical mapping field. 6) On many occasions the VF wave fronts were as large as the entire vertical length (8 cm) of the mapping field, suggesting that it is likely that there are a limited number of wave fronts on the human heart during VF.
With regard to signal quality and dynamic range of optical signals, recent publications do not detail dF/F, except for some seminal papers (8, 9, 13). While using di-4-ANEPPS, Loew and colleagues found that dF/F ranges between 7% and 10%, depending on the preparation (13). During optical mapping of perfused guinea pig hearts, they found (9) a dF/F of 5%, placing perfused heart preparations in the bottom end of that sensitivity range. Knisley and Hill (8) measured AP between 0.8% and 6.4% dF/F on rabbit Langendorff-perfused hearts. Such numbers reported in the literature are comparable to the range of dF/F values of 2–12% in our experiments. A relatively large range of dF/F could be seen depending on the preparation (insufficient staining, shadow effect caused by the papillary muscle) or the heart condition (infarcted areas). However, similar to Loew and colleagues, we did not observe a significant change of dF/F with dye concentration as long as it was maintained within a workable range (13). Consequently, the fluorescence level seen by the camera can be set close to saturation in order to maximize signals. However, because of this relatively small dynamic range, the use of a high-resolution recording device is imperative. For example a 10-bit high-speed digital camera can only record 210, or 1,024, levels of light. A signal size of 5% could then be represented, at best, with only 51 levels or 6 bits, which is clearly not sufficient to accurately depict APs. Both cameras we used had 14-bit resolution, giving 16,384 × 5% = 819 levels or 10 bits of useful information. However, Salama and Choi (13) observed AP amplitudes between 8% and 15%. The latter is the highest value that we found for heart preparations reported in the literature (13).
We measured an average SNR of 40 dB by taking the value in decibels of the DF of AP signals and subtracting the noise floor of the power spectrum. A higher-resolution camera contributes to reducing quantification noise, but other noise sources must be minimized. Light source fluctuations and motion artifacts can appear as a change in fluorescence (13). The latter presents a formidable challenge while mapping beating human hearts. On electrical recordings, especially with bipolar channels, SNR of 60 dB can be achieved under the best of conditions. Thus the SNR of 40 dB we obtained for human optical mapping is acceptable, but not superior to optimal bipolar electrical recordings. However, it is important to factor in the caveat that the SNR was measured without any filtering whatsoever. Filtering techniques time and/or space filtering will help decrease the noise level and increase the SNR.
In previous studies limitations in spatial resolution precluded us from coming to definite conclusions based on the size of wave fronts (10). Our study shows the presence of large wave fronts that reenter, collide, and fractionate similar to in vivo VF in humans (10). In this study with submillimetric spatial resolution, the finding of large sweeping wave fronts suggests that there are probably at any given instance a few wave fronts on the human heart during VF as opposed to multiple wave fronts. It is clear that optical mapping provides significant clarity in studying local activation in VF. The video images from the unequivocal local activation detection illustrate the value of this technique in the study of the source of the wave fronts. In Supplemental Video 1 the wave fronts that exited the mapping field were less than the wave fronts that entered the mapping field. This suggests that the interventricular septum is not the source of the VF wave fronts for this segment of VF. Analysis of this sort over multiple regions with the aid of a panoramic mapping system (7) will allow us to ascertain the source of wave fronts in VF.
The phase-transformed data allow for assessment of phase singularity points and rotors. Analysis of rotors and activation rate as shown in Fig. 3 will allow us to study their potential role in maintenance of human VF. In Fig. 3 the region with the prominent rotor had the activation rate that was the fastest compared with other regions mapped. Together the DF map and the phase analysis provide support to the mechanistic notion of high-frequency source and domains during VF (14), which needs further exploration in humans. The phase-based analysis also allowed us to demonstrate the feasibility of studying reentrant activity and directly relating it to anatomic structures. Indeed, it was previously demonstrated that the base of a papillary muscle may anchor and sustain rotors in rabbits (16). The specific anatomic localization of the rotor by this method will allow us to relate it in the future to anatomic structures in humans that may anchor and influence rotor behavior. Sampling myocytes and tissues in the regions where rotors stabilize or wave breaks occur will enable us to study the ion channels responsible for these events and design drugs to prevent progression of VF in humans.
Although this was the first study to demonstrate the feasibility of studying VF in an intact human heart, Efimov et al. (4a) demonstrated the feasibility of using optical mapping to study wave propagation in isolated, superfused human atria a decade ago. In that study, the optical dye did not change the electrical properties of the atrial tissue; consistent with that study, our study suggests that conduction velocity is not altered significantly.
Limitation of the model.
These VF episodes were mapped in unloaded, denervated hearts. However, the human Langendorff model has been used to study activation sequences under a variety of physiological conditions. The fact that the tissue has to be immobilized either physically, or by using uncoupling drugs that have been shown to have a significant effect on tissue and cell electrophysiology, is a major limitation at this point (11). There are significant challenges to mapping a human heart optically. Removing or minimizing motion artifact is an important challenge in sinus rhythm or during pacing but less of a challenge in VF, where the mechanical artifact is minimal. Activation detection fortunately is not affected by this because it precedes mechanical activity temporally. The biggest challenge in mapping human hearts is the variable presence of the epicardial fat layer, which obstructs visual access to the myocardium. In this study, the presence of such a layer precluded mapping two of the five hearts epicardially.
K. Nanthakumar is a recipient of the Clinician-Scientist Award from the Canadian Institutes of Health Research. This study was supported by Canadian Institutes of Health Research Grant NA 777687 to K. Nanthakumar and by National Heart, Lung, and Blood Institute Grants P01-HL-39707, R01-HL-70074, and R01-HL-60843 to J. Jalife.
We thank Brady Okura from Sci Media, USA, Ltd., for technical assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 by the American Physiological Society