Abstract
Contact mapping is currently used to guide catheter ablation of scar-related ventricular tachycardia (VT) but usually provides incomplete assessment of 3D re-entry circuits and their arrhythmogenic substrates. This study investigates the feasibility of non-invasive electrocardiographic imaging (ECGi) in mapping scar substrates and re-entry circuits throughout the epicardium and endocardium.
Four patients undergoing endocardial and epicardial mapping and ablation of scar-related VT had computed tomography scans and a 120-lead electrocardiograms, which were used to compute patient-specific ventricular epicardial and endocardial unipolar electrograms (CEGMs). Native-rhythm CEGMs were used to identify sites of myocardial scar and signal fractionation. Computed electrograms of induced VT were used to localize re-entrant circuits and exit sites. Results were compared to in vivo contact mapping data and epicardium-based ECGi solutions. During native rhythm, an average of 493 ± 18 CEGMs were analysed on each patient. Identified regions of scar and fractionation comprised, respectively, 25 ± 4% and 2 ± 1% of the ventricular surface area. Using a linear mixed-effects model grouped at the level of an individual patient, CEGM voltage and duration were significantly associated with contact bipolar voltage. During induced VT, the inclusion of endocardial layer in ECGi made it possible to identify two epicardial vs. three endocardial VT exit sites among five reconstructed re-entry circuits.
Electrocardiographic imaging may be used to reveal sites of signal fractionation and to map short-lived VT circuits. Its capacity to map throughout epicardial and endocardial layers may improve the delineation of 3D re-entry circuits and their arrhythmogenic substrates.
This article include the following key novel contributions:
We present detailed comparisons of simultaneous epicardial-endocardial electrocardiographic imaging (ECGi) with epicardial ECGi in substrate mapping and activation mapping for scar-related ventricular tachycardia (VT). We demonstrate that the inclusion of endocardial layers could add important information regarding the scar substrate and arrhythmia circuit, which may be helpful for differentiating epicardial vs. endocardial VT exits and for revealing the 3D involvement of a macro-reentrant circuit.
We demonstrate the ability of ECGi to characterize low-voltage fractionated electrograms within myocardial scar. This was a first step towards investigating the potential of ECGi beyond the delineation of myocardial scar and into the identification of substrate within the scar that could be important for specific ablation targeting.
We show the qualitative and quantitative relationship between phase- and electrogram-based VT mapping, a topic that has remained little understood despite the increasing use of phase mapping in tracking cardiac arrhythmias.
Introduction
The majority of monomorphic ventricular tachycardia (VT) involves re-entry circuits formed by narrow channels of surviving tissue inside the scar, which can be effectively treated with catheter ablation.1 Contact catheter mapping can be used to identify the arrhythmogenic scar substrate. If the VT is haemodynamically stable, the exit and isthmus of the re-entry circuit can be identified from limited activation mapping in combination with entrainment pacing manoeuvres.1 Otherwise, substrate mapping during sinus or paced rhythm can identify regions of myocardial scar as well as potential critical sites to the VT, characterized by local abnormal electrograms such as fractionated, isolated, and late potentials that may be endocardial, epicardial, or intramural.2 However, there are challenges to point-by-point contact mapping, in part, due to its invasive nature. First, it does not allow non-invasive planning prior to the ablation procedure. Second, specific VT mapping is only feasible in a small proportion of patients due to multiple morphologies, non-sustained arrhythmias, and haemodynamic instability. Finally, simultaneous epicardial and endocardial mapping is not routinely available at all centres.
Electrocardiographic imaging (ECGi) has the potential to overcome these challenges by offering a non-invasive approach to reconstruct cardiac electrical signals from high-density surface electrocardiogram (ECG) in combination with heart-torso tomographic scans. It can be safely performed prior to a procedure for treatment planning. It can rapidly map arrhythmia episodes that are too short-lived for contact mapping. Furthermore, while ECGi is most-widely adopted and validated at the epicardial level,3,4 increasing literature has demonstrated its ability to simultaneously reconstruct epicardial and endocardial,5,6 or even transmural,7,8 electrical signals.
However, the potential of ECGi in scar-related VT remains largely unexplored. For mapping VT substrates, ECGi was used to electrically delineate myocardial scar in post-infarction hearts.3,7 It remains unknown whether ECGi can reveal low-amplitude signal characteristics, such as signal fractionation, that are suggestive of potential critical sites within and around the scar. For mapping VT circuits, investigations included two case studies of post-infarction canine hearts placed inside a torso tank9,10 and three human studies,3,4,11 all limited to epicardial VT mapping. Electrocardiographic imaging beyond epicardial mapping of scar-related VT was limited to one recent case study of transmural ECGi.8
This article investigates the use of simultaneous epicardial and endocardial ECGi (epi-endo ECGi) for substrate and activation mapping of scar-related VT. We specifically focus on the feasibility of extracting sites of signal fractionation during ECGi substrate mapping, the use of phase mapping for analysing re-entry circuits during ECGi VT mapping, and the comparison of epi-endo ECGi to epicardial-ECGi in the above tasks. Our hypotheses are: (i) during native rhythm, ECGi can reveal not only regions of myocardial scar but also local abnormal electrograms indicative of potential critical sites for VT; (ii) ECGi can map unstable VT re-entry circuits; and (iii) epi-endo ECGi can improve the delineation of 3D re-entry circuits and their arrhythmogenic substrates. Experiments were performed on a small case series of four patients who underwent catheter ablation of scar-related VT.3 Electrocardiographic imaging results were evaluated by in vivo contact mapping data and VT exits localized from successful ablation procedures.
Methods
Clinical procedures
Four consecutive consenting patients undergoing epicardial and/or endocardial catheter mapping and ablation of VT were enrolled. In a protocol approved by the institutional Research Ethics Board, patients underwent axial computed tomography (CT) imaging within 24 h before the procedure. Immediately before the ablation procedure, body surface electrodes were applied and a 120-lead ECG were recorded with 2000-Hz sampling rate during (i) sinus or implantable cardioverter-defibrillator (ICD) paced rhythm and (ii) induced clinical VT.
Catheter mapping and ablation were performed according to usual clinical protocols. Left ventricular endocardial mapping was performed by the retrograde aortic approach. The pericardial space was entered percutaneously using previously published techniques.12 A 3D electroanatomic mapping system was utilized to acquire substrate maps during sinus or ICD-paced rhythm. Ventricular tachycardia was induced by programmed stimulation from the right ventricular apex or outflow tract. When VT was inducible and tolerated, the site of VT exit was identified via a combination of activation and entrainment mapping. When VT was not inducible, non-sustained, or haemodynamically unstable, the potential site of VT exit was identified using 12/12 pace-match during pace-mapping. At the end of the procedure, non-inducibility of the clinical VT was confirmed using a standard programmed stimulation protocol with up to triple-extrastimuli.
Non-invasive electrocardiographic imaging
Figure 1A summarizes the workflow for ECGi. Using custom research software, noises in 120-lead ECGs were removed using wavelet filters: while filtering parameters were adjustable through the software interface, a level-8 decomposition was most-commonly used with a symlet-8 wavelet. To correct baseline drift, a polynomial of degree 6–10 was fit to and then subtracted from the ECG waveform. The process of de-nosing and baseline correction is illustrated in Figure 1B. Finally, faulty leads were discarded and representative QRST complexes were selected for ECGi.
(A) Overview of ECG imaging workflow. (B) Examples of ECG pre-processing using custom research software, which allow adjustment of filtering parameters through sliding bars, drop-down menus, and text boxes. CT, computed tomography; ECG, electrocardiographic.
Heart and torso models of each patient were created from axial CT scans using custom-written Matlab routines. Epicardial and endocardial contours of left and right ventricles were manually segmented to construct the biventricular surface mesh. To reduce the ill-posedness of the inverse problem, the resolution of the biventricular mesh was kept at a moderate level of 7.75 ± 2.20 mm with 198 ± 40 nodes on the epicardium and 258 ± 42 nodes on the endocardium. The torso was modelled as a homogeneous triangulated surface mesh, with 120 nodes representing surface electrodes at a resolution of 39.34 ± 10.12 mm.
The ventricular and torso mesh of each subject was input to the open-source SCIRun toolkit13 to construct the forward operator H using the boundary element method with the torso conductivity set as 0.2 Sm−1.7 This forward operator relates unipolar potential φv at ventricular epicardial and endocardial surfaces to body-surface potential φb in the form of φb = Hφv at each time instant. Given φb at all time instants from an ECG beat, φv was calculated independently at each time instant by solving the second-order Tikhonov regularization14: , where the first term minimizes the error of fitting surface ECG data and the second term enforces the smoothness of the solution. The regularization matrix L is the surface Laplacian operator and the regularization parameter λ was fixed at 0.5 for the results presented in this article.
In addition, epicardial-ECGi was performed on epicardial ventricular models independently obtained from each patient in previously published work.3 The rest of the ECGi pipeline was kept identical in the comparison study to isolate the effect of including endocardial surfaces in ECGi.
Substrate mapping and quantitative evaluation
From ECGi computed electrograms (CEGMs) during native rhythm, sites of scar were identified based on the presence of wide or large Q waves (>0.04 s in duration or deeper than one-fourth of the amplitude of the following R wave) and small R wave amplitude as described in Ref.10 Fractionated signals were identified as those with low R amplitude and QRS-complexes that comprise more than two positive deflections.
Electrocardiographic imaging identified scar was quantitatively evaluated against electroanatomic bipolar voltage map where bipolar voltage represented the difference between the maximum and minimum magnitude of a bipolar electrogram: myocardial scar core was identified with bipolar voltage <0.5 mV and scare border with bipolar voltage between 0.5 and 1.5 mV. These electroanatomic maps were post-operatively registered to CT-derived ventricular meshes using a rigid transformation based on visual inspection of anatomical landmarks, followed by a non-rigid registration method known as coherent point drift.7 On each heart, the size of identified scar and fractionated CEGMs were calculated. The overlap of ECGi-identified scar with electroanatomic low-voltage scar was evaluated using three metrics: DICE = 2TP/(2TP + FP + FN), sensitivity = TP/(TP + FN), and specificity = TN/(TN + FP), where TP, FP, TN, and FN denote the region of true-positives, false-positives, true-negatives, and false-positives in ECGi scar identification compared with electroanatomic maps. In addition, to study the association of CEGM and electroanatomic data, the following features were extracted from each CEGM: voltage was defined as the difference between the maximum and minimum magnitude of a CEGM signal, and duration was measured from the onset to the end of the CEGM deflections. Linear regression was used to model the association between electroanatomic bipolar voltage as the dependent variable, and CEGM voltage and duration as the independent variables. Instead of using a fixed-effect model on data pooled across the three subjects, we considered potential inter-subject differences and utilized a mixed-effect model to group the data by each individual subject.15 Therefore, the model has shared fixed effects across the subjects, and has random effects accounting for individual-level differences.
Ventricular tachycardia mapping
Phase mapping has been used to track the spatiotemporal progression of atrial and ventricular arrhythmias.16 However, the quantitative relation between phase- and EGM-determined activation sequences remains unclear, which will be examined in this study. To determine activation sequence from CEGM signals, we utilized a previously published method17 that considered the maximum negative gradient in temporal CEGM signals weighted by the spatial gradient. To calculate instantaneous phase signals from CEGM signals, because the CEGM during monomorphic VT showed a clear periodic pattern, the Hilbert Transform (HT) was directly applied to the CEGM after removal of the mean from the signal; instantaneous phase signal at each spatial location was then calculated from each CEGM signal and its time-shifted version resulting from the HT, as described in Ref.16 The time of activation is determined at the time of phase = π/2 as suggested by a recent study.18
The results were qualitatively evaluated due to the lack of quantitative reference data. The earliest activation site in the reconstructed circuit was visually compared to the general anatomical location of the VT exit identified from the ablation procedure. The morphology of the reconstructed circuit was analysed with respect to the location of the electroanatomic map scar.
Results
Substrate mapping
Electrocardiographic imaging was performed during sinus rhythm in patients 1 and 2, and during ICD-paced rhythm in patient 3. Patient 4 was removed from this analysis because excessive noise in ECG data precluded analysis of low amplitude CEGMs.
Comparison to in vivo voltage data and epicardial electrocardiographic imaging
Figure 2 compares side-by-side the scar obtained by epi-endo ECGi, epicardial-ECGi, and electroanatomic voltage maps registered to the CT-derived bi-ventricular mesh on each patient. In terms of scar identification, both ECGi methods were consistent with electroanatomic voltage data in the general location of the myocardial scar, although the quantitative overlap between ECGi and electroanatomic scars could be limited. This was especially true when the scar distribution was heterogeneous and extensive, such as in patient 3, where both ECGi methods appeared to overestimate the region of inferior scar whereas the epicardial-ECGi solution also under-estimated the heterogeneous lateral scar. On the epicardial level, it appeared that the two ECGi solutions were more similar to each other than to electroanatomic data. On the endocardial level, epi-endo ECGi was able to reveal the presence (patient 1 and 3) and absence (patient 2) of endocardial scar: in patient 1, the location of the endocardial scar was consistent to electroanatomic data; while endocardial electroanatomic data were not available in patients 2 and 3, the presence of endocardial scar was expected in patient 3 since a mid-septum endocardial VT exit site was found during the ablation.
Substrate mapping using ECGi. (A1/B1/C1) Comparison of electroanatomic bipolar voltage maps (row 1), epi-endo ECGi scar (row 2), and epicardial ECGi scar (row 3). Colour maps for electroanatomic map data encode the bipolar voltage with a cut-off of >1.5 mV for healthy tissue (blue) and <0.5 mV for scar core (red). Colour maps for ECGi data encode the label of ECGi-identified tissue where blue represents healthy tissue, red represents scar, and yellow represents sites of fractionation. Sites of signal fractionation were also highlighted using white dots and superimposed with electroanatomic and epi-endo ECGi scar maps. (A2/B2/C3) Examples of ECGi-computed electrograms. In each electrogram plot, the x-axis represents time in ms and the y-axis represents amplitude of the signal in mV. ECGi, electrocardiographic imaging; LV lateral, left ventricular lateral.
In terms of sites of signal fractionation, epi-endo ECGi identified a small number of fractionated signals inside the scar in each patient (denoted by white dots in Figure 2). In patients 1 and 2, two clusters of fractionated sits were found: one cluster was located at the heterogeneous region of the electroanatomic map scar near the location where a VT exit was identified during ablation (the approximate location of the exit was labelled by a white scar in all cases); the other cluster was located at the margin of the electroanatomic map scar. In patient 3, where the scar was the most heterogeneous, more fractionated signals were found at the heterogeneous region and margin of the electroanatomic map scar. One site of signal fractionation was also found on the endocardium of this patient, although it was localized on the basal-anterior region while the VT exit was found on the mid-septum of the left ventricle (LV). Examples of CEGMs reconstructed by ECGi are illustrated in Figure 2 for each patient.
Quantitative and statistical analysis
An average of 493 ± 18 CEGMs were analysed for each patient. Identified region of scar had an average area of 129 ± 20 cm2 and 25 ± 4% of the overall ventricular surface. Identified sites of signal fractionation had an average area of 13 ± 7 cm2 and 2 ± 1% of the ventricular surface. Table 1 summarizes the quantitative metrics of ECGi-detected scar in comparison to electroanatomic data, as well as the average CEGM voltage, duration, and voltage/duration ratio in healthy and fractionated sites for each patient. The final linear mixed-effects model showed that CEGM voltage (β = 0.16 mV/mV, P = 0.03) and duration (β = 0.19 mV/ms, P = 0.01) were significantly associated with electroanatomic bipolar voltage, where β is the coefficient associated with the respective independent variable in the regression model.
Results of ECGi substrate mapping
| Patient # | 1 | 2 | 3 | ||
|---|---|---|---|---|---|
| Surface considered | Epi (%) | Endo (%) | Epi (%) | Epi (%) | |
| Sizing error | Epi ECGi | −6 | NA | −15 | −20 |
| Sizing error | Epi-endo ECGi | 0 | 5 | −4 | 0 |
| DICE | 52 | 34 | 59 | 67 | |
| Sensitivity | 54 | 43 | 57 | 57 | |
| Specificity | 72 | 86 | 58 | 48 | |
| Surface considered | Epi + endo | Epi + endo | Epi + endo | ||
| CEGM voltage (mV) | Healthy | 3.61 ± 1.50 | 3.36 ± 1.30 | 3.57 ± 1.70 | |
| Fractionated | 0.85 ± 0.23 | 0.68 ± 0.18 | 3.57 ± 1.70 | ||
| CEGM duration (ms) | Healthy | 138 ± 30 | 120 ± 21 | 158 ± 26 | |
| Fractionated | 158 ± 26 | 147 ± 9 | 178 ± 30 | ||
| Voltage/duration ratio (mV/ms) | Healthy | 0.027 ± 0.014 | 0.029 ± 0.021 | 0.022 ± 0.010 | |
| Fractionated | 0.005 ± 0.002 | 0.005 ± 0.001 | 0.006 ± 0.002 | ||
| Patient # | 1 | 2 | 3 | ||
|---|---|---|---|---|---|
| Surface considered | Epi (%) | Endo (%) | Epi (%) | Epi (%) | |
| Sizing error | Epi ECGi | −6 | NA | −15 | −20 |
| Sizing error | Epi-endo ECGi | 0 | 5 | −4 | 0 |
| DICE | 52 | 34 | 59 | 67 | |
| Sensitivity | 54 | 43 | 57 | 57 | |
| Specificity | 72 | 86 | 58 | 48 | |
| Surface considered | Epi + endo | Epi + endo | Epi + endo | ||
| CEGM voltage (mV) | Healthy | 3.61 ± 1.50 | 3.36 ± 1.30 | 3.57 ± 1.70 | |
| Fractionated | 0.85 ± 0.23 | 0.68 ± 0.18 | 3.57 ± 1.70 | ||
| CEGM duration (ms) | Healthy | 138 ± 30 | 120 ± 21 | 158 ± 26 | |
| Fractionated | 158 ± 26 | 147 ± 9 | 178 ± 30 | ||
| Voltage/duration ratio (mV/ms) | Healthy | 0.027 ± 0.014 | 0.029 ± 0.021 | 0.022 ± 0.010 | |
| Fractionated | 0.005 ± 0.002 | 0.005 ± 0.001 | 0.006 ± 0.002 | ||
Top: quantitative metrics of ECGi-identified scar in comparison to electroanatomic map low-voltage scar. Bottom: CEGM voltage, duration, and voltage/duration ratio at healthy myocardium and sites of signal fractionation.
CEGM, computed electrogram; ECGi, electrocardiographic imaging.
Results of ECGi substrate mapping
| Patient # | 1 | 2 | 3 | ||
|---|---|---|---|---|---|
| Surface considered | Epi (%) | Endo (%) | Epi (%) | Epi (%) | |
| Sizing error | Epi ECGi | −6 | NA | −15 | −20 |
| Sizing error | Epi-endo ECGi | 0 | 5 | −4 | 0 |
| DICE | 52 | 34 | 59 | 67 | |
| Sensitivity | 54 | 43 | 57 | 57 | |
| Specificity | 72 | 86 | 58 | 48 | |
| Surface considered | Epi + endo | Epi + endo | Epi + endo | ||
| CEGM voltage (mV) | Healthy | 3.61 ± 1.50 | 3.36 ± 1.30 | 3.57 ± 1.70 | |
| Fractionated | 0.85 ± 0.23 | 0.68 ± 0.18 | 3.57 ± 1.70 | ||
| CEGM duration (ms) | Healthy | 138 ± 30 | 120 ± 21 | 158 ± 26 | |
| Fractionated | 158 ± 26 | 147 ± 9 | 178 ± 30 | ||
| Voltage/duration ratio (mV/ms) | Healthy | 0.027 ± 0.014 | 0.029 ± 0.021 | 0.022 ± 0.010 | |
| Fractionated | 0.005 ± 0.002 | 0.005 ± 0.001 | 0.006 ± 0.002 | ||
| Patient # | 1 | 2 | 3 | ||
|---|---|---|---|---|---|
| Surface considered | Epi (%) | Endo (%) | Epi (%) | Epi (%) | |
| Sizing error | Epi ECGi | −6 | NA | −15 | −20 |
| Sizing error | Epi-endo ECGi | 0 | 5 | −4 | 0 |
| DICE | 52 | 34 | 59 | 67 | |
| Sensitivity | 54 | 43 | 57 | 57 | |
| Specificity | 72 | 86 | 58 | 48 | |
| Surface considered | Epi + endo | Epi + endo | Epi + endo | ||
| CEGM voltage (mV) | Healthy | 3.61 ± 1.50 | 3.36 ± 1.30 | 3.57 ± 1.70 | |
| Fractionated | 0.85 ± 0.23 | 0.68 ± 0.18 | 3.57 ± 1.70 | ||
| CEGM duration (ms) | Healthy | 138 ± 30 | 120 ± 21 | 158 ± 26 | |
| Fractionated | 158 ± 26 | 147 ± 9 | 178 ± 30 | ||
| Voltage/duration ratio (mV/ms) | Healthy | 0.027 ± 0.014 | 0.029 ± 0.021 | 0.022 ± 0.010 | |
| Fractionated | 0.005 ± 0.002 | 0.005 ± 0.001 | 0.006 ± 0.002 | ||
Top: quantitative metrics of ECGi-identified scar in comparison to electroanatomic map low-voltage scar. Bottom: CEGM voltage, duration, and voltage/duration ratio at healthy myocardium and sites of signal fractionation.
CEGM, computed electrogram; ECGi, electrocardiographic imaging.
Ventricular tachycardia mapping
During induced VT, ECGi reconstructed five re-entry circuits among the four patients.
Activation sequence from phase vs. computed electrograms
Figure 3A and B compares an activation sequence determined from CEGM and phase signals. Figure 3C provides examples of local CEGM (blue) and phase (red) signals selected in their order of activation from this sequence: on each signal, the corresponding time of activation is marked by a dot with a vertical bar. As shown, while the CEGM-derived time of activation closely approximated the time of 0.5π in phase signals at some locations (examples 2 and 5), the discrepancy was larger at other locations (examples 1, 3, and 4). The general order of activation, however, was concordant between CEGM- and phase-derived activation sequences as noted by the black arrow in Figure 2. The resulting maps of activation thus showed a qualitatively similar conduction pattern in the form of a counter-clock-wise rotation on the lateral wall of the LV with similar locations of the earliest activation. There, however, existed a quantitative discrepancy. This observation was consistent among all five VT circuits being studied (see section S-I in Supplementary material online). Across the five VT circuits analysed, the difference between CEGM-derived activation time and that derived from phase = 0.5π was 21.06 ± 28.33 ms, and the value of phase at the time of CEGM-derived activation was 0.44π ± 0.16π.
Comparison of activation sequence extracted from phase signals and electrogram signals. (A) Activation time map extracted from phase signals at the time of phase = π/2. (B) Activation time map extracted from electrogram signals. Both colour maps in A and B encode the value of activation time. (C) Selected electrogram signals (blue) and phase signals (red) at locations as annotated in (A and B). Blue and red dots indicate the time of activation determined, respectively, from the time of steepest downslope on electrogram signals and from the time of instantaneous phase = π/2 on phase signals. In each plot, the x-axis represents time in ms. Two y-axes are used. The left y-axis (for phase signal) represents the value of phase. The right y-axis (for electrogram signal) represents the amplitude of the signal in mV. EGM, electrogram.
Comparison with epicardial electrocardiographic imaging
Figure 4 summarizes activation maps of all five circuits obtained from phase mapping using epi-endo ECGi (row 1) vs. epicardial-ECGi (row 2): the first view was selected to best represent the overall morphology of the circuit obtained by epi-endo ECGi, followed by standard anterior and inferior epicardial views to facilitate comparison with epicardial-ECGi.
Five VT activation maps obtained by epi-endo ECGi (row 1 in each panel), in comparison to those obtained by epicardial-ECGi (row 2 in each panel). Colour maps encode the value of activation time. VT exits determined from ablation procedures are noted in the left bottom panel of the figure. VT exits determined from ECGi are labelled by white stars. Each circuit is represented in three views: the first represents the best view to summarize the morphology of the epi-endo ECGi circuit, followed by standard anterior and inferior views to facilitate comparison with epicardial-ECGi. VT, ventricular tachycardia.
In VT1, both methods revealed an epicardial exit at the basal-lateral region of the LV, along with a counter-clock-wise rotation anchored at the epicardium of the left ventricular lateral wall. This estimated location of exit was consistent with that determined from the ablation procedure.
In VT2 and VT4, epi-endo ECGi revealed re-entrant circuits that exited, respectively, at high basal-inferior epicardium and the basal inferolateral region of the left ventricular endocardium. Both locations were consistent with those determined from the ablation procedure. Activation maps obtained by epicardial-ECGi were similar to that obtained by epi-endo ECGi on the epicardium. Without the endocardial view, however, an epicardial exit would be suggested in both cases, with the exit in VT5 corresponding to the site of epicardial breakthrough from the endocardial exit. In both cases, the absence of an endocardial view also suggested a focal rather than re-entrant mechanism (indicated by the white arrows).
In VT3 and VT5, epi-endo ECGi revealed re-entrant circuit exiting, respectively, at the mid-septum and basal anterior region of the left endocardium, consistent with the general anatomical regions determined from the ablation procedure. Epicardial activation patterns obtained by epicardial-ECGi again resembled those from epi-endo ECGi. Using epicardial views only, however, an epicardial anterior-basal exit would be suggested in both cases, reflecting the epicardial breakthrough from two different endocardial exits. While the epicardial activation pattern suggested a focal mechanism, interestingly, a ‘projection’ of the endocardial activity generated from epi-endo ECGi appeared to be obtained on the basal view of epicardial-ECGi—an artificial surface closing the atrioventricular plane.
Examples of detailed phase map sequence of VT1 and VT5 obtained by epi-endo ECGi are provided in Figure 5 (see complete videos of all reconstructed VT circuits in section S-IV in the Supplementary material online). Of note, in VT1, a slowing of conduction (crowding of isochrones) was observed along the boundary of the electroanatomic map scar at left ventricular lateral wall, and a zig-zag course of propagation and very-slow conduction was observed inside the scar. In VT5, the conduction was fast at the lateral and inferior epicardium where there was minimal evidence of scar on electroanatomic mapping, whereas the conduction was slow at basal-anterior epicardium where there was a large electroanatomically mapped scar. These examples show, qualitatively, the relationship between ECGi-reconstructed circuit morphology and the underlying scar substrates.
Examples of phase map sequence from ECGi of two re-entry circuits. The depolarization wave front is tracked by white arrows (at phase = π/2) to facilitate visualization. The timing of each snapshot with respect to the onset of surface QRS is labelled and tracked on the ECG trace of lead V2 (i.e. −20 ms = 20 ms before the QRS onset). Electroanatomic map scar is shown as a light-grey patch. (A) VT1 of patient 1: a counter-clock-wise macro-reentry centred at the inferior lateral region of the epicardium, with an epicardial exit at the basal inferolateral epicardium. (B) VT5 of patient 4: a macro re-entry, with an exit at the basal-anterior endocardium.
Discussion
This study demonstrated the feasibility of simultaneous epi-endo ECGi for mapping scar substrates and re-entry circuits in scar-related VT. For substrate mapping, ECGi reasonably agreed with electroanatomic mapping data but the overlap was quantitatively limited. This suggested that the true potential of ECGi, rather than in electrically delineating myocardial scar, may lie in the identification of local abnormal electrograms suggestive of critical sites for VT. While this small case series provided initial examples where ECGi-detected sites of signal fractionation was located near an VT exit or heterogeneous scar margin, further studies are required to establish this finding in a larger sample size as well as to determine whether ECGi could reveal additional low-amplitude characteristics such as isolated potentials and other local abnormal activities. For VT mapping, this study showed that the earliest sites of activation determined by epi-endo ECGi were qualitatively consistent with the anatomical region of the VT exits determined from the clinical procedure, including both epicardial and endocardial exits. Without an endocardial view, the earliest site of activation from epicardial-ECGi may reflect the epicardial breakthrough from the endocardial exit. In addition, while epi-endo ECGi revealed macro-reentry in all five cases, epicardial-ECGi in some cases revealed a focal rather than re-entrant mechanism, and in other cases indicated a re-entrant circuit in the artificial surface closing the atrioventricular plane that resembled a projection of the endocardial activity reconstructed by epi-endo ECGi.
Generally, phase signals reflected the relative order of activation in space. If local CEGM morphologies were similar across the space, the phase-generated time of activation would be similar to those generated by CEGM with a constant time shift. However, since the actual CEGMs differed substantially in morphology over space, there was no clear theoretical or experimental evidence supporting the existence of an exact phase value at which the time would correspond to CEGM-derived time of activation. Our results showed that the general order of spatial activation was qualitatively concordant between CEGM- and phase-derived activation, indicating that phase mapping may be suitable for supporting the analysis of the general morphology of an activation circuit. For quantitative analysis, the use of phase mapping warrants further investigations.
Human validation of ECGi in scar-related VT remains a significant challenge. In previous work, focal VT were studied in healthy rabbit19 and canine models20 without structural diseases; in the setting of scar-related VT, case studies of animal models were reported utilizing animal hearts placed in a torso tank.9,10 In the limited existing human studies,3,4 epicardial mapping of re-entry circuits was evaluated qualitatively to contact mapping data similar to this study. This study extended the work by Sapp et al.3 on the same patients to ECGi study of the endocardium. Similar to previous studies,3,4 this study was limited in the validation data and could be improved by activation/entrainment mapping data of stable VT or 3D DGE-MRI data of myocardial scar. This requires substantial effort in experimental design and data collection, although case studies have emerged8 to utilize joint DCE-MRI and catheter mapping to evaluate ECGi in scar-related VT.
Several factors may affect the analysis of ECGi and its use in clinical practice. First, the challenge of intra-operative registration between MRI/CT and electroanatomic maps has to be tackled in order to merge ECGi maps into the catheter mapping space. Second, as a feasibility study of epi-endo ECGi in scar-related VT, the classic Tikhonov regularization was used for the inverse reconstruction of the ventricular electrograms. This method can be affected by the choice of the regularization parameter as discussed in previous work (see section S-II in Supplementary material online).8 Future studies should consider alternative inverse methods that are more robust to regularization parameters. Finally, the presented ventricular model did not include the right ventricular outflow tract (RVOT) that is less frequently relevant in scar-related VT. The effect of excluding the RVOT in ECGi should be investigated in future studies (see section S-III in Supplementary material online).
Conclusion
This study demonstrated that (i) beyond the delineation of myocardial scar, ECGi may reveal critical sites of fractionated electrograms within the scar during native rhythm; (ii) ECGi could map short-lived episodes of VT and identify the exit sites; and (iii) the inclusion of endocardial layer in ECGi may improve the delineation of the 3D construct of a re-entry circuit and its arrhythmogenic substrate.
Acknowledgements
We would like to thank Jaume Coll-Font and Dana H. Brooks at the Northeastern University for participating in the discussion regarding the relationship between electrogram- and phase-derived activation sequence, and for sharing their software to extract time of activation from unipolar electrograms.
Funding
This work was supported by the US National Institutes of Health [R21HL125998 and R01HL116280], the US National Science Foundation [ACI-1350374], the Canadian Institutes of Health Research, and the Heart & Stroke Foundation of Nova Scotia. J.L.S is also a Canadian Arrhythmia Network Investigator.
Conflict of interest: S.N. is a scientific advisor to CardioSolv, St. Jude medical, Biosense Webster, and Simens and receive funding from Biosense Webster, Siemens, and Imricor. J.L.S. has served as a consultant to Biosense Webster, has received research funding from Biosense, St. Jude and Philips and has received speaker honoraria from Medtronic and St. Jude. This work has no relationship with industry.
