Imaging Electrocardiographic Dispersion of Depolarization and Repolarization During Ischemia
Simultaneous Body Surface and Epicardial Mapping
Background— Myocardial ischemia creates abnormal electrophysiological substrates that result in life-threatening ventricular arrhythmias. Identifying patients at risk of such abnormalities by use of body surface electrical measures is controversial. We investigated the sensitivity of torso measures, recorded simultaneously with epicardial electrograms, to changes in dispersion of depolarization and repolarization during localized ventricular ischemia.
Methods and Results— Ventricular epicardial electrograms were recorded from 5 anesthetized pigs with a 127-electrode sock. A controllable suture snare was used to ligate the left anterior descending coronary artery (LAD). The chest was reclosed, and a vest with 256 ECG electrodes was fitted to the torso. Simultaneous arrays of epicardial electrograms and torso ECGs were recorded during LAD occlusion and reperfusion. Activation-recovery intervals (ARIs), QTu and RTu dispersion (where u indicates upstroke), and QRST integrals were calculated, and these data were fitted to anatomically customized computational models of the swine ventricular epicardium and torso. LAD occlusion caused the epicardial ARI dispersion to steadily increase, whereas the location of shortest ARI shifted from the posterobasal ventricular tissue (control) to the anteroapical myocardium, distal to the suture snare. These changes were associated with a steady increase in the torso RTu dispersion as the shortest RTu interval moved from the right shoulder (control) to the sternum. QTu and RTu dispersion determined from the 12-lead ECG did not consistently reflect the myocardial changes.
Conclusions— Although changes in myocardial repolarization dispersion resulting from localized ischemia are not reliably reflected in temporal indices derived from the 12-lead ECG, they can be readily identified with high-resolution torso ECG mapping.
Received November 21, 2002; revision received January 31, 2003; accepted February 4, 2003.
Myocardial ischemia and infarction can create abnormal electrophysiological substrates that trigger life-threatening ventricular arrhythmias. Identifying patients at risk of such abnormalities is clearly highly desirable in managing their condition. Noninvasive body surface electrical recordings have been used to define indices thought to reflect abnormal electrophysiological substrate.1,2 However, the reliability of body surface measures (for example, recovery time dispersion, QT dispersion, and QRST integral mapping) to accurately detect myocardial substrate is viewed as being useful,3–6 marginal,2,7,8 or having no predictive power.9–11 Indeed, recent work has shown that indices derived from simulated body surface signals were inadequate in reflecting the underlying measured epicardial activation-recovery intervals (ARIs).12 This has further called into question the clinical usefulness of body surface measures to detect dispersion of depolarization and repolarization,13,14 although it should be noted that of the above-mentioned studies, just of four1–3,12 incorporated dense-array body surface mapping, whereas the other reports were based solely on the standard 12-lead ECG.
Failure of the 12-lead ECG to provide a reliable index of the dispersion of myocardial repolarization9–11 may be a result of poor spatial resolution, measurement error,15 or the choice of leads from which indices are determined.16 The purpose of this study was to test whether changes in the dispersion of ventricular repolarization were reflected simultaneously in indices determined from high-resolution body surface ECG mapping. Importantly, we tested whether body surface measures could accurately detect heterogeneity of myocardial repolarization during localized ventricular ischemia that may not be reliably identified with the 12-lead ECG.
Anesthesia, intensive care, and surgical methods have been fully described previously17 but are briefly summarized here. The investigation conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health and was performed in accordance with the British Home Office license requirements.
Anesthesia, Monitoring, and Surgery
Five domestic pigs (the Pig Improvement Company, Abingdon, Oxfordshire, UK) weighing 28 to 36 kg were anesthetized with 2% to 3% halothane (Fluothane, ICI) for induction, tracheotomized, and artificially ventilated (Oxford Mark II ventilator, Penlon). For maintenance, bolus infusions of α-chloralose (Sigma, 100 mg/kg IV) were repeated approximately every 2 hours as required. Femoral arteries and veins were cannulated, and the arterial blood pressure (ABP) and lead II ECG were monitored. Core temperature and arterial blood gases and pH were maintained at physiological values. ABP was measured with a saline-filled pressure transducer (SensoNor 840) connected to a real-time data acquisition system (MP 100, Biopac Systems Inc) with Acqknowledge 3.0 software for the Macintosh (Macintosh Quadra 950), and heart rate was computed with this software.
Animals were thoracotomized and pericardectomized, and the left anterior descending coronary artery (LAD) was ligated (but not occluded at this stage) equatorially with a suture ligature (Ethicon, 3-0 mersilk). An elasticized sock containing 127 unipolar stainless steel electrodes (interelectrode spacing ≈7 mm; Biomedical Instruments Designers) was slipped over the ventricles. The chest was then reclosed (the epicardial electrode wires and suture ligature exited near the diaphragm) and filled with saline to eliminate air pockets. A custom-made vest containing 256 electrodes (interelectrode spacing ≈15 mm) was then fitted to the animal.
Anatomically Accurate Computational Models of the Torso and Ventricles
Formulation and customization of anatomicocomputational models of the heart and torso have been fully described previously.18,19 In brief, smoothly continuous mathematical descriptions of the torso and ventricular surfaces were fitted20 to geometric data digitized from computed tomography and 3D echocardiographic images, respectively. Anatomic torso landmarks were digitized with a mechanical digitizing arm (FARO Technologies) and used with a nonlinear host mesh fitting procedure to customize the computed tomography–derived (generic) torso model to the animal specific geometry. The FARO arm was used to record the position and orientation of the echocardiographic probe so that the ventricular model could be registered with the torso geometry.
Approximately 85% of the torso electrode locations were digitized with the FARO arm. The remaining inaccessible electrodes (on the back of the animal) were interpolated from their nearest neighbors with the known vest electrode topography. Vest electrode locations were projected onto the anatomic torso model. Epicardial electrode locations were estimated from the known electrode sock topography and projected onto the anatomic ventricular model by aligning landmarks on the sock with the posterior descending artery and the LAD. Electrode projections were used together with the anatomicocomputational models to interpret the spatial variation in torso ECGs and epicardial electrograms.
Simultaneous Torso and Ventricular Epicardial Recordings
Electrodes on the epicardial sock and torso vest were connected to a 448-channel UnEmap cardiac mapping system (Auckland Uniservices Ltd). Simultaneous arrays of torso and epicardial unipolar signals were sampled with 12-bit resolution at 2 kHz, subject to a band-pass filter of 0.05 to 900 Hz (fifth order, −100 dB/decade, linear phase, low-pass filter). All signals were referred to the Wilson’s central terminal (electrical average of the signals recorded from both front limbs and the left rear limb), whereas the right rear limb was driven by the negative of the Wilson’s central terminal signal (“right leg drive”) to increase the signal-to-noise ratio.
Signal Analysis and Mapping
Epicardial activation mapping methods have been fully described previously17 and are outlined schematically in Supplementary Figure S1. According to Haws and Lux,21 ventricular epicardial activation times were computed for each epicardial electrogram with the steepest negative intrinsic deflection during the QRS complex (minimum dV/dt; see Figure S1A). Activation times were arbitrarily referred to the earliest activation time across all electrograms (the datum) and used together with the estimated electrode projections to fit a scalar field to the anatomic ventricular model. Dispersion of epicardial activation was computed from the range of activation times and indicated with a red-to-blue rainbow spectrum corresponding to earliest-to-latest activation, respectively (Figure S1B). Black isoactivation time contour bands were superimposed on the ventricular map to indicate the activation sequence. Dispersion of ventricular epicardial repolarization was analyzed in a similar manner, except that the recovery time was defined as the steepest positive intrinsic deflection (maximum dV/dt) during the T wave of the epicardial electrograms21 (Figure S1A and S1B). Recovery times for each epicardial electrode were referred to the same fixed datum (namely, the earliest activation time) to which the activation times were referred. Ventricular epicardial ARI maps were generated by subtracting the recovery time from activation time for each epicardial electrogram,21 fitting a scalar field to these data, and then superimposing a red-to-blue rainbow spectrum to represent shortest-to-longest ARIs, respectively (Figure S1B).
In an analogous manner, body surface QTu and RTu (where u indicates upstroke) interval maps were fitted to data derived from the individual torso ECGs. Specifically, the torso QTu interval was defined as the time difference between the steepest positive intrinsic deflection (maximum dV/dt) during the T wave and the onset of the QRS complex. To contrast with the QTu analysis, we also quantified torso RTu intervals, which were defined to have the same T-wave end points, but the RTu interval onset times were defined as steepest negative intrinsic deflections during the QRS complexes (minimum dV/dt; similar to the epicardial analysis). In addition, a 12-lead ECG was derived from the full torso lead set and used to quantify conventional QTu and RTu dispersion. To this end, 9 electrodes from the body surface vest array were selected to best approximate the geometric configuration of the standard clinical 12-lead ECG. We did not correct QTu or RTu intervals for heart rate (eg, with Bazett’s formula) because the interval measures were used in calculations of QTu or RTu dispersion, for which R-R interval correction may be misleading.22
To quantify changes in signal magnitudes during ischemia, the individual epicardial electrograms and torso ECGs were integrated over their entire QRST interval (Figure S1C). Epicardial and torso QRST integral maps were fitted to electrode data (Figure S1D) to illustrate the spatial changes during ischemia. Recordings from electrodes with poor contact, a low signal-to-noise ratio, and ballooning ST segments were excluded from the analysis. Among the 5 animals, 10±3% of epicardial recordings and 12±3% of the torso recordings were rejected from the analysis.
Control electrograms and ECGs were recorded before LAD occlusion. The LAD was then occluded for 2 to 4 minutes, during which time ventricular epicardial electrograms and body surface ECGs were sampled simultaneously at 20-second intervals. In 4 animals, recordings continued at 20-second intervals for 1 to 11 minutes after LAD reperfusion. The remaining animal went into terminal ventricular fibrillation during LAD occlusion.
All data are expressed as mean±SEM. Single-factor ANOVA (Microsoft Excel 2000) was used to determine differences in indices derived from the sets of epicardial, torso, and 12-lead recordings. Post hoc comparisons by use of Student’s t test were performed to elucidate differences, and statistical significance was accepted for a value of P<0.05.
For all studies (n=5), LAD occlusion caused the intrinsic heart rate to increase by 8±2 bpm (from the baseline of 145±16 bpm), whereas the ABP remained relatively constant (control: 113±7/75±5 mm Hg versus occlusion: 105±6/70±4 mm Hg).
The most dramatic changes were seen after 4 minutes of LAD occlusion, although similar trends were observed during 2 and 3 minutes of LAD occlusion in different animals. Because of the variability in snare location (resulting in different-size regions of ischemic myocardium) and the range of occlusion times, it was not appropriate to compare results among animals quantitatively. Therefore, one detailed case study (animal X) is presented in this report.
For the case study X, ST-segment elevation was observed on leads V1 and V2 of the 12-lead ECG (see Supplementary Figure S2) after 1 minute (ΔSTV1=0.21 mV; ΔSTV2=0.14 mV) and 4 minutes (ΔSTV1=0.48 mV; ΔSTV2=0.32 mV) of LAD occlusion. By this latter stage, the T waves of leads V3, V4, and V5 had inverted. All other standard leads showed little change during occlusion, and the observed changes had reversed after 6 minutes of reperfusion. A wide complex ventricular arrhythmia was first observed after 163 seconds of LAD occlusion. Immediately after this, ectopic beats occurred every 7 to 15 cycles, and this decreased to 1 ectopic beat every 3 to 7 cycles after 240 seconds of LAD occlusion. No ectopic beats were observed after coronary reperfusion.
Ventricular Epicardial Activation Sequence
Figure 1A illustrates the activation sequence changes compared with control caused by 4 minutes of LAD occlusion. Epicardial activation propagation was slowed distal to the ligature, and this was associated with a marked increase in the dispersion of activation times from 17 ms (control) to 36 ms, as the region of latest epicardial activation (blue in Figure 1A) moved from the posterobasal ventricular muscle to the center of the ischemic zone. Dispersion of epicardial activation time increased steadily during the ischemic period (Figure 1B), and this was rapidly reversed after coronary reperfusion. The activation sequence returned to the control state after 6 minutes of reperfusion.
Epicardial Recovery Sequence
Under control conditions, the sequence of ventricular epicardial recovery closely followed the epicardial activation sequence (compare Figures 1A and 2⇓A), with earliest recovery occurring 161 ms after the earliest epicardial activation (the datum). However, the dispersion of epicardial recovery (defined by its range) was 3 times greater (51 ms) than the activation dispersion (17 ms) in the control state. The earliest site of epicardial recovery shifted from the mid–left ventricular myocardium (control) to the center of the ischemic zone during LAD occlusion (Figure 2A), and this was associated with a marked increase in the dispersion of ventricular epicardial recovery from 51 ms (control) to 88 ms after 4 minutes of occlusion. This increase was primarily a result of the decrease in the earliest epicardial recovery time (calculated over all epicardial electrograms). In contrast, the latest epicardial recovery time remained relatively unchanged during LAD occlusion, although it increased temporarily after reperfusion. The epicardial recovery sequence and dispersion measures had returned to control after 6 minutes of coronary reperfusion.
Epicardial Activation-Recovery Intervals and Torso QTu and RTu Dispersion
Figure 3 illustrates the evolution of epicardial ARIs and torso QTu and RTu dispersion (computed across 113 epicardial and 225 torso recordings, respectively) during ischemia and reperfusion, together with a similar analysis for the 12-lead ECG. An increase in the epicardial ARI dispersion was associated with an increase in the torso QTu interval dispersion during LAD occlusion (Figure 3A). Compared with the QTu analysis, the temporal evolution of the torso RTu interval dispersion provided a better estimate of the epicardial ARI changes during the protocol (Figure 3B). Similar trends were seen in the other animals, as illustrated in Figure 4. For each animal, the time at which the epicardial ARI dispersion was maximal was identified (eg, 160 seconds for case study X in Figure 3B), and data from these time points were compared statistically with control values among all animals (n=5). The mean epicardial ARI dispersion increased significantly, from 58±5 ms (control) to a maximum of 147±14 ms during LAD occlusion (P<0.05). The corresponding mean torso RTu dispersion also increased significantly, from 56±6 to 113±16 ms (P<0.05). In contrast, the mean 12-lead ECG RTu dispersion did not increase significantly (control: 46±3 ms to 79±18 ms; P=0.093) because of the large variance in the 12-lead ECG RTu during LAD occlusion, although there was a trend in the data. Compared with the 12-lead ECG, the torso array was more sensitive to the underlying myocardial changes, and this is most likely because of the increased spatial sampling of the ECG activity.
The spatial variation in epicardial ARIs (Figure 5A) was similar to the epicardial recovery sequence during the entire protocol. In case study X, LAD occlusion caused the minimum epicardial ARI to decrease steadily from 152 ms (control) to 93 ms after 4 minutes of occlusion as the location of this minimum shifted from the mid–left ventricular free wall (control) to the middle of the ischemic zone. These changes were reversed after 6 minutes of coronary reperfusion. The maximum epicardial ARI remained relatively unchanged during LAD occlusion (data not shown). Torso RTu interval mapping (Figure 5B) exhibited similar changes, with the minimum steadily shortening from 152 ms near the right shoulder (control) to 112 ms near the sternum after 4 minutes of LAD occlusion, whereas the maximum torso RTu interval remained relatively unchanged (data not shown).
QRST Integral Mapping
During LAD occlusion, epicardial QRST integral values (Figure 6) increased markedly across the ischemic zone because of elevations of the ST segments and T-wave peaks observed in the adjacent epicardial electrode recordings. During LAD occlusion, the positive region of QRST integral observed on the anterior ventricular surface (distal to the suture snare) increased steadily in both area and magnitude. The epicardial QRST integral map had returned to the control state after 6 minutes of reperfusion.
Although only small changes (compared with control) to the torso QRST integral map were observed after 20 seconds of LAD occlusion (Figure 6), elevated QRST integral values were clearly evident on the torso chest after 40 seconds of occlusion. This region of positivity increased in magnitude and area during the occlusion, consistent with the observed epicardial QRST integral changes across the ischemic zone.
The reliability of the standard 12-lead ECG to accurately assess the heterogeneity of myocardial repolarization has received considerable attention and is now regarded by several leading research groups as being inadequate.9–14 A variety of studies have yielded conflicting results regarding the sensitivity of body surface dispersion measures to changes in the dispersion of myocardial repolarization.2,4,6 The conclusions from these reports were based on in vitro or nonconcurrent recordings of body surface and epicardial activity, and just 2 of these studies2,12 incorporated dense arrays of body surface signals (which were simulated in the latter12). Using arrays of simultaneously sampled torso and epicardial electrograms in vivo, we have confirmed that changes in myocardial ARI dispersion caused by regional ventricular ischemia may not be readily identifiable by use of temporal indices derived from the 12-lead ECG. However, our observations provide the first direct evidence that dispersion indices derived from high-density torso mapping can reliably detect changes in the underlying myocardial dispersion of depolarization and repolarization. Furthermore, regional variations in epicardial QRST integrals during localized ventricular ischemia were well preserved on the body surface.
Recent work12 using a computational model of the torso and epicardium to compute body surface ECGs from epicardial electrograms that had been recorded previously from an open-chest dog (a “forward problem”) demonstrated the inadequacy of body surface measures to determine underlying myocardial activity. Localized alterations in myocardial repolarization were induced with regional epicardial warming that resulted in a shortening of the ventricular epicardial ARIs and increased ARI dispersion. Our results and those of others23 who have studied myocardial ischemia report similar trends in ARIs and their dispersion. Of particular interest in the computational study12 was the counterintuitive prediction that epicardial warming led to a shortening in the minimum QT interval and a decrease in the QT interval dispersion calculated from both the 64-lead body surface and standard 12-lead ECG sets (see Figure 4 of Burnes et al12) We did not observe this trend in torso measures during localized ventricular ischemia. Instead, the increase in epicardial ARI dispersion was well reflected in the QTu and RTu interval measures determined from simultaneous high-density torso mapping, whereas temporal indices determined from the 12-lead ECG were not sensitive to the underlying myocardial changes. This latter finding is consistent with other studies.9–11
Body surface QRST integral mapping is thought to identify abnormalities in the sequence of myocardial depolarization or repolarization caused by various pathological states, including post–myocardial infarction,24 idiopathic ventricular fibrillation,25 and exercise-induced ischemia.26 However, the conclusions of these studies have not been directly supported with simultaneous myocardial recordings. Our results indicate that the QRST integral distribution on the epicardium is well preserved in that observed on the body surface during the evolution of localized ventricular ischemia. Although it was not readily apparent after 20 seconds of LAD occlusion, torso QRST integral changes subsequently reflected the epicardial activity, with a distinct area of positivity on the torso chest adjacent to that observed on the anterior ventricular epicardium. Furthermore, the torso QRST integral maps remained dipolar throughout the protocol. These findings are consistent with reported body surface changes during myocardial ischemia.26 The 12-lead ECG clearly detected ST-segment and T-wave changes on several leads during LAD occlusion, consistent with the changes in the torso QRST integral maps.
Our results would support the notion that spatial distributions of epicardial potentials are reflected at the body surface, allowing ventricular repolarization dispersion to be well represented on the torso. Importantly, body surface temporal and potential measures reliably detected the evolution of electrical changes brought about by regional ventricular ischemia. In this context, indices of ventricular repolarization heterogeneity and ischemia (ARIs and QRST integrals, respectively) were well preserved in the corresponding body surface measures (QTu and RTu interval dispersion and QRST integrals). This raises the possibility that high-density body surface ECG mapping may elucidate electrophysiological substrate underlying abnormal dispersion of ventricular repolarization. Techniques that robustly and accurately reconstruct myocardial electrical activity with good spatial resolution from body surface potential recordings (an “ECG inverse”) could provide a noninvasive imaging modality to screen patients at risk of repolarization abnormalities.
This research was funded by The Wellcome Trust and the British Heart Foundation. Dr Nash was also supported by an Edward Penley Abraham Cephalosporin Research Fellowship at Lady Margaret Hall, Oxford. We appreciate the technical assistance of Christopher Hirst and Vivienne Harris and thank Dr Simon Golding for editorial assistance.
The online figures and movies are available in an online-only Data Supplement at http://www.circulationaha.org.
Spargias KS, Lindsay SJ, Kawar GI, et al. QT dispersion as a predictor of long-term mortality in patients with acute myocardial infarction and clinical evidence of heart failure. Eur Heart J. 1999; 20: 1158–1165.
Brendorp B, Elming H, Jun L, et al. QT dispersion has no prognostic information for patients with advanced congestive heart failure and reduced left ventricular systolic function. Circulation. 2001; 103: 831–835.
Burnes JE, Ghanem RN, Waldo AL, et al. Imaging dispersion of myocardial repolarization, I: comparison of body-surface and epicardial measures. Circulation. 2001; 104: 1299–1305.
Kors JA, van Herpen G. Measurement error as a source of QT dispersion: a computerised analysis. Heart. 1998; 80: 453–458.
Nash MP, Thornton JM, Sears CE, et al. Ventricular activation during sympathetic imbalance and its computational reconstruction. J Appl Physiol. 2001; 90: 287–298.
Haws CW, Lux RL. Correlation between in vivo transmembrane action potential durations and activation-recovery intervals from electrograms: effects of interventions that alter repolarization time. Circulation. 1990; 81: 281–288.
Ejima J, Martin D, Engle C, et al. Ability of activation recovery intervals to assess action potential duration during acute no-flow ischemia in the in situ porcine heart. Experimental Cardiology Group, University of North Carolina at Chapel Hill. J Cardiovasc Electrophysiol. 1998; 9: 832–844.
Dambrink JH, SippensGroenewegen A, van-Gilst WH, et al. Association of left ventricular remodeling and nonuniform electrical recovery expressed by nondipolar QRST integral map patterns in survivors of a first anterior myocardial infarction. Captopril and Thrombolysis Study Investigators. Circulation. 1995; 92: 300–310.
Ahmmed GU, Miyakoda H, Shigemasa C. Isointegral analysis of body surface electrocardiographic mapping for assessing exercise-induced changes in repolarization properties in patients with coronary artery disease. Yonaga Acta Medica. 1997; 40: 107–116.