Accurate Conduction Velocity Maps and their Association with Scar Distribution on Magnetic Resonance Imaging in Patients with Post-Infarction Ventricular Tachycardias

Abstract

Background -

Characterizing myocardial conduction velocity (CV) in patients with ischemic cardiomyopathy (ICM) and ventricular tachycardia (VT) is important for understanding the patient-specific pro-arrhythmic substrate of VTs and therapeutic planning. The objective of this study is to accurately assess the relation between CV and myocardial fibrosis density on late gadolinium-enhanced cardiac magnetic resonance (LGE-CMR) imaging in patients with ICM.

Methods -

We enrolled 6 patients with ICM undergoing VT ablation and 5 with structurally normal left ventricles (controls) undergoing PVC or VT ablation. All patients underwent LGE-CMR and electro-anatomical mapping (EAM) in sinus rhythm (2,960 EAM points analyzed). We estimated CV from EAM local activation time using the triangulation method, that provides an accurate estimate of CV as it accounts for the direction of wavefront propagation. We evaluated for the association between LGE-CMR intensity and CV with multi-level linear mixed models.

Results -

Median CV in ICM patients and controls was 0.41 m/s and 0.65 m/s respectively. In ICM patients, CV in areas with no visible fibrosis was 0.81 m/s (95%CI: 0.59–1.12 m/s). For each 25% increase in normalized LGE intensity CV decreased by 1.34-fold (95%CI: 1.25–1.43). Dense scar areas have on average 1.97–2.66-fold slower CV compared to areas without dense scar. Ablation lesions that terminated VTs were localized in areas of slow conduction on CV maps.

Conclusions -

CV is inversely associated with LGE-CMR fibrosis density in patients with ICM. Non-invasive derivation of CV maps from LGE-CMR is feasible. Integration of non-invasive CV maps with EAM during substrate mapping has the potential to improve procedural planning and outcomes.

Graphical Abstract

Introduction

Knowledge of the distribution of myocardial conduction velocity (CV) in patients with ischemic cardiomyopathy (ICM) and ventricular tachycardias (VT) is critical for understanding the patient-specific pro-arrhythmic substrate and for planning therapeutic interventions. Abnormal slowed conduction through regions of surviving myocyte bundles is an important substrate causing VT,¹ and areas of slow conduction have been used as potential VT ablation targets in patients with ICM.^2–5 Areas of slow conduction can be identified with substrate mapping and pace mapping techniques.^3–9

However, the association between myocardial CV and fibrosis density on late gadolinium enhanced cardiac magnetic resonance imaging (LGE-CMR) has not been established yet. Establishing this relation is important for the following reasons: First, substrate mapping is increasingly utilized in clinical practice for identification of VT ablation targets in patients with ICM,^{5, 10} however it is often limited by poor spatial sampling density in the setting of complex intramural scar distribution and ambiguities in correlating electrical maps with anatomical structures.¹¹ Image integration of LGE-CMR with electroanatomical maps (EAM) improves substrate identification and characterization.^{6, 12, 13} Understanding the relation between LGE-CMR fibrosis and CV will augment the role of CMR image integration on substrate mapping. Second, CMR-based virtual heart modeling has emerged as a powerful platform for non-invasive VT risk assessment,^{14, 15} localization,¹⁶ and ablation planning¹⁷ in ICM patients. CV is an important parameter in virtual heart models as it determines arrhythmia localization and dynamics.¹⁸ In the absence of clinical data on the relation between LGE-CMR fibrosis and CV, the CV distribution that is ascribed to virtual heart models may not represent accurately that of the remodeled myocardium in ICM patients.^{1, 19–21}

Previous clinical studies have assessed the association between CV and LGE-CMR fibrosis density only indirectly. Conduction slowing in areas of fibrosis by LGE-CMR has been demonstrated using pacing maneuvers.⁶ CV over areas of hypoperfusion on contrast-enhanced computed tomography (a surrogate of fibrosis) has been estimated using the local activation time (LAT) between EAM point pairs.⁷ Calculating CV from EAM point pairs may not provide an accurate estimate of myocardial CV as the calculation does not account for the direction of wavefront propagation relative to the recorded EAM point pairs. A more accurate technique, the triangulation technique for CV estimation, uses the LAT of three points and rules of trigonometry to calculate local CV, accounting for the direction of wavefront propagation. However, it has been used for the purposes of assessing ventricular CV only in a porcine model of post-infarction VT.²²

In this study we aim to extend the triangulation technique to the clinical setting and provide an accurate, quantitative characterization of CV in patients with ICM using electroanatomical mapping. We aim to establish the association between CV and fibrosis density on LGE-CMR, and to evaluate the CV at the sites of successful VT ablation. The results of this study have important implications for patients with ICM and VT because they will: (a) improve our understanding on their electroanatomical substrate, (b) augment the adjunct role of CMR image integration in substrate mapping, and (c) enhance virtual heart modeling approaches and their translational potential.

Materials and Methods

Data Transparency Statement

The authors share the code used for the analysis presented herein and a sample dataset at GitHub and can be accessed at https://github.com/rheedaali/CV-LGE_LV. Further data that support the findings of this study are available from the corresponding author upon request.

Study Participants

This is a retrospective study. We identified participants of this study from the Johns Hopkins (JH) VT registry, that is an ongoing registry that includes all patients with PVCs/VT that had undergone a PVC/VT ablation procedure at JH. We included all patients that were enrolled in the JH VT registry and met the following criteria: (a) underwent catheter ablation for left-sided monomorphic VT or PVCs, (b) had a CMR within 1 year of the procedure, (c) had a LV activation map in sinus rhythm during the procedure and (d) had either ischemic cardiomyopathy (ICM) with scar present on CMR, or structurally normal left ventricles with no scar on CMR (controls). Control participants were patients who underwent VT ablation procedure for left-sided monomorphic VT, had no history of myocardial infarction and had no scar on CMR. We excluded patients that (a) had CMR of inadequate quality, (b) had an implantable cardioverter-defibrillator that caused any artifact on the heart, and (c) had fewer than 40 points collected during LV mapping in sinus rhythm. The study was approved by the local institutional review board (IRB) and all patients gave written informed consent to participate in the JH VT registry.

The details on the LGE-CMR protocol and the electrophysiological assessment performed to this study patients have been previously described.⁶ We summarize these details, along with the post-processing steps that we performed in the Online Supplemental Materials and Methods. A normalized LGE-CMR intensity > 50% was considered dense scar in patients with ICM.

CV Vector Calculation

We calculated the CV vector using the triangulation technique,²³ previously validated in animal studies.^{22, 23} This technique is superior to methods that calculate CV using EAM point pairs, as it accounts for the direction of wavefront propagation. A detailed description of this technique can be found in the “CV Vector Calculation” section of the Online Supplemental Materials and Methods. Briefly, we created triangular elements at the endocardial surface of the ventricle using points of the outer surface of the EAM cloud, and we calculated the local CV vector using principles of trigonometry (supplemental figure 1). The approach that we used to assign the CV calculated at each triangular element to the corresponding LGE-CMR intensity is described in the “Co-registration of CV Estimation and LGE CMR” section of the Online Supplemental Materials and Methods.

Statistical Analysis

A detailed description of our statistical analysis is provided in the “Statistical Analysis” section of the Online Supplemental Materials and Methods. Briefly, we performed linear regression analysis to evaluate for associations between normalized LGE intensity and CV, for each individual study participant. Since CV followed a right skewed distribution, we used the natural logarithm of CV (ln-CV) in the regression analysis. We performed hierarchical linear mixed-effects models to evaluate the association of CV with normalized LGE-CMR intensity in patients with ICM and controls. Hierarchical mixed models account for the fact that multiple CV measurements were nested within the same patient.

Results

Patient Characteristics

Eleven patients met the inclusion and exclusion criteria (ICM N=6, control N=5). We show their baseline characteristics in table 1, and a flow-chart summarizing patient selection in supplemental figure 2. All patients with ICM had VT and 3/5 patients with SNLV had VT (2/5 had PVCs). One patient with ICM had VT that was not related to ICM (ICM3, left posterior fascicular VT). Median age at the time of the procedure was 60.8 years with an interquartile range of 9.1 years and was similar between ICM and control patients. Both sexes were equally represented in this study. Patients had ablation procedure performed during 2011–2015 and CMR was performed at a median of 4 days before ablation (25^th-75^th percentile range: 0 to 13.5 days). A total of 2,960 points were analyzed. The median number of points collected from each patient during EAM was 101 (interquartile range of 297). The number of EAM points collected in patients with ICM was 305 (interquartile range: 468) and in control patients 62 (interquartile range: 36). The median edge length of the triangular elements that we used in this analysis was 10.6 mm (inter-quartile range 8.2 mm), which is in alignment to the spatial resolution of the CMR. The median CV in patients with ICM was 0.41 m/sec with a 5^th-95^th percentile range of 0.09 to 2.17 m/sec, while the median CV in control patients was 0.65 m/sec with a 5^th-95^th percentile range of 0.09 to 1.98 m/sec.

Table 1.

Baseline characteristics of patients enrolled in this study, overall and by ICM status

	Overall (N=11)	ICM (N=6)	Control (N=5)	P-value
Age (years)	60.84 (9.10)	57.85 (7.47)	62.11 (3.52)	0.465
Male	5 (45.5)	3 (50)	2 (40)	1.000
EAM Points (number)	101 (297)	305 (468)	62 (36)	0.201
Triangles (number)	81 (235)	194 (331)	68 (21)	0.068
EDVI (ml/m2)	91.56 (44.87)	101.3 (22.03)	71.27 (67.18)	0.297
ESVI (ml/m2)	44.04 (16.18)	44.51 (12.51)	32.51 (15.72)	0.053
SVI (ml/m2)	47.48 (19.9)	55.52 (16.96)	44.645 (5.67)	0.439
CI (ml/min/m2)	2.68 (0.34)	2.58 (0.16)	3.01 (0.39)	0.121
EF (%)	44.18 (19.43)	44.18 (3.17)	51.46 (29.96)	0.807
Beta-blockers	8 (72.73)	4 (66.67)	4 (80)	1.000
Amiodarone	3 (27.27)	3 (50)	0 (0)	0.182

Continuous variables are summarized as median (25th-75th Interquartile range); p-values calculated with the Wilcoxon rank-sum test. Categorical variables are summarized as count (percentage); p-values calculated with Fischer’s exact test. Abbreviations: ICM: ischemic cardiomyopathy; EAM: electroanatomic mapping; EDVI: end-diastolic volume index; ESVI: end-systolic volume index; SVI: stroke volume index; CI: cardiac index; EF: ejection fraction

CV slowing is proportional to LGE intensity on CMR in patients with ICM

Our results demonstrate an inverse association between CV and normalized mean LGE intensity in all patients with ICM (table 2) but not in controls (supplemental table 1). Standardized regression coefficients of ln-CV on normalized LGE intensity in patients with ICM ranged from −0.17 to −0.39 (p<0.01). Scatter plot analysis suggested a weak correlation between CV and normalized LGE in patients with ICM and no correlation in the control patients, as scar was not present (supplemental figure 3). The CV distribution of all patients with ICM, stratified by normalized LGE-CMR intensity quartiles is shown in figure 1. Median CV decreased from 0.5 m/sec in the lowest LGE-CMR quartile to 0.3 m/sec in the highest LGE-CMR quartile. Mixed model analysis demonstrated that CV at areas with no visible fibrosis on CMR (0% normalized LGE intensity) at a transmural depth set at 2.5mm is 0.81 m/s, with a 95% confidence interval of 0.59 to 1.12 m/sec. For each 25% increase of normalized intensity, CV decreased by 1.34-fold (95% confidence interval of 1.25 to 1.43). We obtained similar results when the transmural depth was set to 5, 7.5, and 10mm (table 3).

Table 2.

Correlation of conduction velocity with normalized intensity of LGE MRI in patients with ICM (N=6).

Depth:	2.5 mm			5.0 mm			7.5 mm			10.0 mm
Participant	T	β	p-value	T	β	p-value	T	β	p-value	T	β	p-value
ICM1	529	−0.17	<0.001	531	−0.20	<0.001	531	−0.19	<0.001	531	−0.19	<0.001
ICM2	60	−0.29	0.028	60	−0.30	0.022	61	−0.29	0.027	61	−0.28	0.035
ICM3	85	−0.29	0.007	85	−0.29	0.007	85	−0.27	0.012	85	−0.27	0.014
ICM4	395	−0.27	<0.001	397	−0.22	<0.001	399	−0.17	0.001	399	−0.14	0.005
ICM5	274	−0.24	0.001	283	−0.29	<0.001	286	−0.26	<0.001	290	−0.23	<0.001
ICM6	79	−0.39	0.002	79	−0.37	0.002	79	−0.14	0.234	79	−0.14	0.227

Linear regression of the normalized LGE MRI intensity on the natural logarithm of conduction velocity. T is the number of triangular elements used in analysis. Beta is standardized regression coefficient. P-values in bold denote statistical significance as defined by the pre-specified alpha criterion of p<0.05.

Figure 1.

Violin plots showing the CV distribution in control patients (all LGE-CMR intensities) and all patients with ICM stratified for normalized LGE-CMR intensity. In each violin plot the white bar represents the median, the black box -- the 25^th-75^th interquartile range, the vertical line -- the lower to upper adjacent value range, and the curved contours -- the probability density function estimated using an Epanechnikov kernel. In each violin plot the numbers on the y-axis represent (lower to upper) the minimal, median, and maximal CV value. The same letter on top of violin plots represents absence of statistically significant difference at the 0.05 criterion, derived by analysis of variance and Duncan’s correction for multiple comparisons. Similar results were demonstrated when averaging LGE-CMR intensities at 5.0, 7.5, and 10.0 mm (supplemental figure 5).

Table 3.

Overall relationship between CV and NI in patients with ICM

Depth	CV* (m/sec)	95% CI			p-value	Fold Change*	95% CI			p-value
2.5 mm	0.81	0.59	−	1.12	0.206	1.34	1.25	−	1.43	<0.001
5.0 mm	0.76	0.56	−	1.02	0.070	1.33	1.25	−	1.42	<0.001
7.5 mm	0.69	0.51	−	0.92	0.011	1.26	1.19	−	1.34	<0.001
10.0 mm	0.67	0.50	−	0.89	0.007	1.24	1.17	−	1.32	<0.001

Results from mixed model analysis in patients with ICM. CV* represents the conduction velocity in m/sec of the myocardium that has 0% normalized LGE-MRI intensity (intercept). CV Fold-change* represents the fold-decrease of conduction velocity, for every 25% increase in NI. For all depths, CV decreases proportionally to NI by 1.24- to 1.34-fold for every 25% increase in NI. Abbreviations: CV: conduction velocity; ICM: ischemic cardiomyopathy; NI: normalized intensity; 95% CI: 95% confidence interval of the prediction. P-values in bold denote statistical significance as defined by the pre-specified alpha criterion of p<0.05.

The magnitude of the inverse association between LGE intensity and CV is not affected by the transmural depth of fibrosis

In patients with ICM, the magnitude of the inverse association between CV and normalized mean LGE intensity was not affected by the distance from the endocardial surface used to average LGE intensities (table 2, figure 2). There was only 1 out of 6 patients (ICM6) for whom the inverse association between CV and normalized LGE-CMR intensity became statistically not significant at distances 7.5 and 10 mm from the endocardium.

Figure 2.

Spaghetti plots showing the inverse relationship between CV and normalized LGE-CMR intensity in patients with ICM. Each plot corresponds to one ICM patient. Different lines represent different distances away from the endocardial surface, over which LGE intensities were averaged.

CV is significantly slower over areas of dense scar in patients with ICM

In patients with ICM, areas of dense scar (LGE-CMR intensity > 50%) had on average 1.97 to 2.66-fold slower CV compared to areas without dense scar (supplemental table 2). In areas without dense scar, CV declined steeply with increasing LGE-CMR. For every 25% increase in normalized intensity there was a 1.42 to 1.68-fold decline in CV. In areas with dense scar, however, the CV decline was significantly less steep (1.31 to 1.71-fold less steep) compared to areas without dense scar.

CV of non-fibrotic myocardium is similar between patients with ICM and controls

When comparing patients with ICM and controls, both groups had similar CVs over the myocardium with no visible fibrosis on LGE-CMR (p-values for fixed-effect intercept > 0.2, figure3, supplemental table 3 and figure 1). ICM patients, however, had a 1.16 to 1.29-fold steeper decline in CV for increasing (normalized) LGE-CMR intensity. In fully-adjusted mixed-models analysis including interaction terms for depth there was no effect of depth on the intercept and slope of the ln-CV vs normalized LGE-CMR relation (effect of depth on ln-CV p=0.81, interaction of depth with model slope p=0.90). There was no modification of the effect that the ICM status had on the ln-CV vs normalized LGE-CMR slope by depth (three-way interaction of depth with effect of ICM status on model slope p=0.55). There was no significant change in model goodness-of-fit with inclusion of depth as a covariate (Log-likelihood changed from −8070.13 to −8067.37).

Figure 3.

Spaghetti plots showing the relationship between CV and normalized LGE CMR in all patients with ICM at a depth of 2.5 mm from the endocardial surface. Similar results were demonstrated when averaging LGE-CMR intensities at 5.0, 7.5, and 10.0 mm (supplemental figure 5).

Positive association of EAM voltage with CV and LGE-CMR in patients with ICM

In patients with ICM, there was a positive association between CV and bipolar and unipolar voltage amplitude, with standardized regression coefficients ranging from 0.21 to 0.32 for bipolar voltage, and from 0.22 to 0.29 for unipolar voltage (supplemental table 4). Median CV decreased from 0.52 m/sec to 0.23 m/sec in areas defined as non-scar (≥1.5 mV) and as dense scar (<0.5 mV) using bipolar voltage criteria. Using unipolar voltage criteria median CV decreased from 0.52 m/sec to 0.23 m/sec in areas defined as non-scar (≥8 mV) and dense scar (<5 mV) respectively (supplemental figure 4). The presence of low voltage amplitude is a surrogate for the presence of fibrotic tissue. In ICM, areas of low voltage amplitude exhibit moderate to good correlation with areas of fibrosis (by imaging or histology) in pre-clinical^24–26 and clinical^{6, 27, 28} studies. The association between voltage amplitude and CV is not affected by co-registration, since both voltage amplitude and CV are estimated from the same EAM. The standardized regression coefficients of bipolar and unipolar voltage with CV are of similar magnitude to those between CV and LGE-CMR. Bipolar and unipolar voltage amplitude were inversely associated with LGE-CMR intensity, with correlation coefficients ranging from −0.16 to −0.38 for bipolar voltage and −0.30 to −0.54 for unipolar voltage, respectively (supplemental table 5).

Ablation targets are localized within the areas of CV slowing in patients with ICM

While CV maps were not available during the ablation procedure, off-line analysis demonstrated that the ablation lesion set that was delivered to render the patients non-inducible for VT co-localized with slow CV on CV maps. We present the spatial correlation of ablation lesions that terminated VT with CV maps (derived off-line) and bipolar voltage maps (created intra-procedurally) in figure 4. The clinical strategy that we used to localize VT ablation targets for each patient is summarized in the Online Supplemental Results. Patient ICM3 had VT not related to scar and is not shown here. Areas of CV slowing (<0.5 m/sec) on the CV map co-localized with areas of low voltage on the bipolar voltage maps and areas of high normalized LGE intensity on 3D rendered CMR images. Ablation targets that rendered the patients non-inducible for VT were found to co-localize with areas of slow conduction on the CV maps in all 5 patients.

Figure 4.

Bipolar voltage maps (left column) juxtaposed to the CV maps (middle column) and the 3D rendered CMR image showing normalized LGE intensity (right column) in each patient with ICM. Red spheres represent the ablation lesions delivered during catheter ablation to render the patients non-inducible for VT. Areas of low voltage amplitude on bipolar voltage maps correspond to areas of CV slowing on CV maps and areas of high normalized LGE-CMR intensity on the 3D rendered images. The ablation lesions are located in the areas of CV slowing.

Discussion

Main findings

The aim of this study was to provide an accurate, quantitative characterization of CV in patients with ICM using EAM, and to establish the association between CV slowing and fibrosis density on LGE-CMR and whether it co-localizes with sites of successful VT ablation. Our primary finding is that CV slowing is proportional to fibrosis density on LGE-CMR in patients with ICM. In these patients, each 25% increase in LGE-CMR scar density is associated with a 1.34-fold decrease in CV. Areas of dense scar have on average 1.97 to 2.66-fold slower CV compared to areas without dense scar. There is a biphasic relation between CV slowing and increasing fibrotic density, with the greatest slowing occurring as normalized LGE-CMR intensity increases from 0 to 50%, while at areas of dense scar, CV continues to decrease but at a slower rate. This suggests that there might be a critical mass of surviving myocardial cells within the scar, below which CV is no longer affected. Additional findings from this study are: (1) CV slowing estimated from endocardial EAM reflects the presence of fibrosis, not only in the sub-endocardial layer, but also deeper in the myocardium; (2) the non-fibrotic myocardium of patients with ICM has similar CV compared to myocardium from structurally-normal left ventricles, which suggests that CV slowing in patients with ICM occurs with accumulation of fibrosis; and (3) lesions delivered during catheter ablation to eliminate infarct-related VTs co-localize with areas of CV slowing, suggesting that CV mapping could be used for identification of ablation targets.

Comparison with results of pre-clinical studies

The estimated CV of the non-fibrotic myocardium of ICM patients in our study (95%CI for all depths: 0.50–1.12 m/s) is considerably faster than that described in animal studies (0.4–0.65 m/s).^{19, 22} CV in conducting channels of surviving myocardium within the scar has been reported to be 0.28–0.62 m/s in swine post-infarction models.⁹ Last, the estimated CV of myocardial tissue with ≥ 75% LGE-CMR normalized intensity is slightly faster (95%CI: 0.28–0.41 m/s) than that obtained from human ex-vivo histologic preparations of surviving myocardial bundles within infarcted areas (0.2–0.3 m/s).^{22, 29} This highlights the importance of clinical assessment of fundamental EP properties such as CV, since values derived from animal models or human ex-vivo preparations might differ from those measured in patients.

Comparison with results of clinical studies

The CV slowing in areas of LGE-CMR fibrosis that we report herein is consistent with findings of previous clinical studies. Sasaki et al. were the first to describe the association between conduction slowing and scar transmurality on LGE-CMR in patients with ICM (n=23).⁶ In this study conduction slowing was defined as S-QRS > 40 msec but a quantitative assessment of local CV values was not performed.⁶ Ustunkaya et al. evaluated the association of local CV with myocardial hypoperfusion (as a surrogate of scar) on cardiac computed tomography in 14 patients with ICM.⁷ CV was estimated as the average linear distance between 5 pairs of EAM points divided by the difference in activation times of each pair. Hypoperfusion was assessed by hypoattenuation during contrast-enhanced cardiac computed tomography. Mean CV was 0.53±0.40 m/s and CV was log-linearly and inversely associated with hypoperfusion, ranging from 0.18 m/s for the areas with least perfusion to 1.28 m/s in areas of best perfusion;⁷ these results are comparable to ours. Last, CV in conducting channels of surviving myocardium within the scar has been reported to be 0.55m/s in clinical studies using Ripple mapping.⁵ These values are within the CV range we report in this study. However, the spatial resolution of LGE-CMR and EAM precludes us from specifically assessing CV in conducting channels of surviving myocardium. Considering the size of the triangular elements that we used for CV calculation, the CV that we report over areas of fibrosis represents an average CV.

Significance for substrate mapping

This study demonstrates that LGE-CMR can be used to non-invasively generate CV maps using regression models that summarize the association between LGE-CMR fibrosis density and invasively measured CV. Image-based, non-invasive derivation of CV maps has direct clinical implications since critical sites of post-infarction VTs are confined to areas of CV slowing, which are then targeted for ablation.^2–5 Non-invasive CV maps can be used prior to VT ablation for procedural planning. LGE-CMR-based CV maps can be integrated with EAM during substrate mapping improving detection of VT ablation targets. In a recent study, image integration with EAM motivated additional mapping and epicardial access in 57% and 33% of patients with VT, respectfully (n=112).³⁰ In this study, although CV maps were not available during the VT ablation procedure, the ablation lesions delivered during conventional VT mapping co-localized in areas of slow CV on CV mapping. However, considering the small sample size, LGE-CMR resolution, and study design, this study does not demonstrate or prove spatial specificity to guide VT ablation procedures, but is generated hypothesis for future studies. Non-invasive CV maps have the potential to compensate for the limited sampling density of EAM. Incorporation of non-invasively acquired CV maps in VT ablation procedures can potentially reduce procedural and fluoroscopy time devoted to substrate characterization, overall improving the safety of the procedure. With improvement of CMR technology allowing for increased spatial resolution, image-based CV mapping may be used to distinguish the critical conduction slowing within channels critical for VT, further improving image-based substrate mapping.

Significance for virtual heart modeling

The results of this study can be used to inform ICM virtual heart model development by providing an accurate patient-derived assessment of CV and its relation to LGE-CMR fibrosis density in patients with ICM. Our results can be directly used to assign realistic CV values to LGE-CMR-derived virtual heart models. Our study characterizes the continuum in the relationship between LGE-CMR and CV; CV can now be modeled as a continuous function of LGE-CMR intensity, rather than as discrete categories of “normal”, “peri-infarct zone”, and “dense scar”.^{14, 17} Incorporating realistic CV assessment in virtual heart models has the potential to improve characterization of VT dynamics in patients with ICM, but this needs to be tested in future studies. In this study we also describe the within-subject and between-subject CV variability for different levels of LGE-CMR intensity. These variabilities can be used in sensitivity analyses to assess the robustness of virtual heart models in VT localization and ablation planning. Such a sensitivity analysis is a critical step in virtual heart model development as it will allow for quantification of model precision and robustness.

Limitations

Our study has several limitations. First, it is a single-center retrospective observational study with a relatively small sample size. Second, LGE-CMR and EAM have limited spatial resolution for detailed characterization of fine structures within scar, such as VT isthmuses. For this we cannot assess for differences in myocardial CV within or outside the areas of critical VT isthmuses. However, the spatial resolution of the EAM map that we used here is in alignment with what has been used in contemporary studies^{7, 22, 31} assessing the relation between CV and different electrogram properties with features derived from cardiac CT or CMR. Third, even with careful segmentation, small amounts of endoluminal contrast or periventricular adipose tissue might have been included within myocardial contours. This unintentional error would manifest as augmentation or attenuation of myocardial LGE-CMR intensity and would thus weaken the strength of correlation between LGE-CMR and CV. Fourth, results may also be limited by potential positional errors when registering EAM points to corresponding sectors on LGE-CMR based on the registration information obtained by the EAM software. However, the strength of correlation between CV and Bipolar/Unipolar voltage (as a surrogate of fibrosis) was similar to that between CV and LGE-CMR, suggesting a minimal effect of any segmentation and registration error in CV and LGE-CMR association. Fifth, EGM parameters on EAM can be affected by the contact or the orientation of the mapping catheter³², which may account for some unexplained variance in the models. However, by manually reviewing the quality of all electrograms and by using the single-conforming, compact boundary triangulation method we minimized the likelihood of including points with poor contact in our analysis. Sixth, the triangulation algorithm that we used to estimate CV assumes a locally planar wave-front. In general, considering the relatively small size of triangles that we used, this is a valid assumption. However, in regions of scar, there may be heterogenous, curvilinear, “zig-zag” conduction, violating the assumption of a planar wave-front and resulting in an average (rather than exact) CV calculation. Last, mapping from the endocardial surface does not account for intramural activation, and CV may be underestimated if the activation wavefront dives underneath sub-endocardial scar and re-emerges to the surface.

Conclusions

In this study we provide an accurate quantitative assessment of CV in patients with ICM. We demonstrate that CV is inversely associated with fibrosis density and that areas of conduction velocity slowing co-localize with sites of successful ablation of VTs. We demonstrate that CV maps can be non-invasively derived from LGE-CMR imaging. These results contribute to our understanding of the pro-arrhythmic substrate in patients with ischemic cardiomyopathy and strengthen the adjunct role of CMR image integration in substrate mapping during VT ablation procedures. Our results can be directly incorporated in virtual heart models providing a realistic, patient-derived CV assessment across the continuum of LGE-CMR intensity. Utilization of noninvasively acquired CV maps has the potential to improve procedure design and outcomes, but this needs to be tested in future studies.

What is known:

• Conduction slowing is an important substrate for ventricular tachycardias (VT) ischemic cardiomyopathy (ICM) patients.

• Substrate mapping is increasingly utilized for identification of VT ablation targets and image integration of late gadolinium-enhanced cardiac magnetic resonance imaging (LGE-CMR) with electroanatomical maps improves substrate identification and characterization.

• CMR-based virtual heart modeling has emerged as a powerful platform for non-invasive VT risk assessment localization, and ablation planning in ICM patients.

What this study adds:

• Conduction slowing is proportional to fibrosis density on LGE-CMR in patients with ICM.

• LGE-CMR can be used to non-invasively generate conduction velocity maps, augmenting the adjunct role of CMR image integration in substrate mapping.

• The results presented can be directly used to assign realistic conduction velocity values to LGE-CMR-derived virtual heart models. This has the potential to enhance virtual heart modeling approaches.

Non-standard Abbreviations and Acronym:

CV

conduction velocity

EAM

electroanatomic mapping

ICM

ischemic cardiomyopathy

LAT

local activation time

LGE-CMR

late gadolinium enhanced cardiac magnetic resonance imaging

NI

normalized intensity

VT

ventricular tachycardias