Abbreviations
CAD | coronary artery disease |
CT | computed tomography |
EAVM | electroanatomic voltage mapping |
EP | electrophysiology |
FDG | fluorodeoxyglucose |
LGE | late gadolinium enhancement |
MRI | magnetic resonance imaging |
PET | positron emission tomography |
TTE | transthoracic echocardiography |
VT | ventricular tachycardia |
SCD | sudden cardiac death |
SPECT | single photon emission computed tomography |
Introduction
With the widespread adaptation of substrate-based ablation for ventricular tachycardia (VT), the electrophysiology (EP) laboratory has become increasingly integrative. In addition to fluoroscopic guidance, standard pacing maneuvers, and direct anatomic visualization with intracardiac ultrasonography, electroanatomic voltage mapping (EAVM) tools have enabled electrophysiologists to create three-dimensional renderings of myocardial scar in real time. These scar maps, performed in sinus or paced rhythms, can guide ablation, and therefore are highly valuable in the setting of VTs associated with hemodynamic instability, multiple morphologies, and inconsistent inducibility. Nonetheless, despite various ablative approaches to substrate modification, the overall success rates of these strategies for VT have been modest. This has led to interest in complementary methods of delineating scar.
Noninvasive methods of myocardial scar assessment, magnetic resonance imaging (MRI) in particular, have proven to be highly accurate. However, whether these promising modalities will demonstrate sufficient additive value to be incorporated into an already complex EP laboratory environment is uncertain. This review is designed to explain our current state of knowledge of myocardial scar assessment and its implications in the treatment of patients with ventricular arrhythmias. Various noninvasive modalities for myocardial scar assessment will be described, as will correlative information about how such scar relates to voltage mapping descriptions of substrate. A detailed discussion of integrating scar imaging into VT ablation will follow. Finally, the promising role of noninvasive imaging of scar to optimize risk stratification for sudden cardiac death (SCD) will be addressed.
Multimodality Assessment of Myocardial Scar
From a pathology standpoint, “myocardial scar” refers to the fibrotic replacement of normal tissue that occurs after injury and associated necrosis. The presence of scar is not always apparent by imaging assessments of morphology. In the case of extensive infarct, a region of thinned myocardium or aneurysm may be observed by transthoracic echocardiography (TTE), computed tomography (CT), or MRI. In some cases, fatty deposition in the subendocardium can also occur. This lipomatous metaplasia may be observed by cardiac CT or MRI. Although it is typically observed in patients with a chronic infarct, it has also been described in idiopathic dilated cardiomyopathy [1]. Even wall motion assessment by TTE is not a consistent method to detect scar. In regions of dysfunction, it may be difficult to discern infarct from ischemia because of stunned or hibernating myocardium.
In the setting of coronary artery disease (CAD) the presence or absence of extensive scar can be determined by a variety of noninvasive tests that assess viability. Viability studies assess the potential of dysfunctional regions of the heart to produce functional recovery after revascularization. With dobutamine stress echocardiography, viable myocardium demonstrates contractile reserve in regions of dysfunction. Infarct, on the other hand, remains akinetic with progressively increasing dobutamine doses. The sensitivity of this technique to detect fully viable segments that will recovery function after revascularization is approximately 80%; however, the ability to detect partially viable segments with 25–50% scar is lower [2].
Another modality for scar detection is rest-redistribution single photon emission CT (SPECT) with thallium. Thallium-201, an analog of potassium, is actively taken up by myocytes with preserved membrane function. Although at rest perfusion in ischemic myocytes may be poor, they will demonstrate tracer uptake on delayed imaging so long as there are viable. Viable regions defined by this modality have a high (>85%) chance of functional improvement with revascularization, whereas nonviable regions with persistent defects fail to demonstrate functional improvement in more than 75% of cases [3].
Positron emission tomography (PET) with [18F]fluorodeoxyglucose (FDG), a glucose analog, has a higher resolution than SPECT. Viable myocardium, with its relative affinity for glucose, demonstrates increased FDG uptake in regions of poor perfusion (i.e., a metabolic-perfusion mismatch). Nonviable or infarcted myocardium demonstrates poor FDG uptake, with matched perfusion and metabolic defects on PET. Compared with thallium scintigraphy, PET performs similarly in terms of predicting left ventricular functional improvement, although it likely has higher yield when there is significant left ventricular dysfunction [4]. In addition, PET has been shown to detect fibrosis in nonischemic cardiomyopathy [5]. Although all three forms of viability assessment provide a macroscopic view of large territories that might not benefit from revascularization, none of them directly visualize scar the way that MRI does.
Direct Visualization of Scar with MRI
Cardiac MRI is the most robust noninvasive modality for scar detection. The MRI technique of late gadolinium enhancement (LGE) makes use of gadolinium’s inability to cross intact cell membranes in normal tissue and its tendency to pool in regions of myocardial cell damage and fibrosis where the extracellular space has expanded. Because gadolinium alters the T1 relaxation properties of the surrounding tissue in a magnetic environment, regions of fibrosis demonstrate brightening or hyperenhancement, also known as LGE or delayed gadolinium enhancement, several minutes after contrast medium administration. In comparison, normal myocardium, because of its failure to retain contrast medium, appears blackened or nulled. In animal infarct models, regions of LGE closely match scar on pathology [6]. In patients with myocardial infarction, the difference in image intensity between abnormal and normal myocardium has been reported as more than six standard deviations with a high in-plane resolution in the range of 1.5 mm [7].
MRI also allows a full-thickness assessment of fibrosis within ventricular myocardium. MRI provides information about the presence and size of LGE, its location, and its pattern of distribution (i.e., CAD or non-CAD). In the presence of CAD, LGE is described as either subendocardial or transmural because of the wavefront physiology of ischemia. In nonischemic heart diseases, LGE classically spares the subendocardium and tends to be epicardial, mid-wall, or global, but may also be absent. For example, in idiopathic dilated cardiomyopathy, 59% of patients demonstrate no enhancement and 28% demonstrate mid-wall striae, whereas 13% demonstrate patterns typical of CAD [8]. Approximately half of patients with nonischemic cardiomyopathies who present for VT ablation demonstrate scar on MRI [9]. If present, these regions are frequently home to critical substrate.
In some scenarios, LGE MRI can aid in identifying the cause of cardiomyopathy, which may have electrophysiological implications. The left ventricular hypertrophy in hypertrophic cardiomyopathy, for example, can appear concentric or symmetric in up to 42% of patients [10, 11], sometimes making this diagnosis difficult by TTE. The demonstration of a characteristic pattern of hyperenhancement at right ventricular septal insertion points or in regions of hypertrophy would support the diagnosis of hypertrophic cardiomyopathy. Classic findings have been described in other disease states as well. There is patchy intramural or epicardial involvement in cardiac sarcoidosis, myocarditis, Anderson-Fabry disease, and Chagas disease. More global or diffuse LGE may be seen in systemic conditions or infiltrative diseases such as amyloidosis, systemic sclerosis, and cardiac sarcoidosis. Figure 1 demonstrates examples of the various types of scar seen by MRI.
Scar Assessment with EAVM
Although MRI is the optimal noninvasive assessment method for scar, the invasive assessment of scar with EAVM serves a unique purpose in EP. Mapping is performed in real time in the EP laboratory with the goal of identifying a critical region of arrhythmic substrate. To create an electroanatomic voltage map, a catheter makes point contact with a given region of tissue, and with use of reference points, a three-dimensional rendering of the touched surface typically with voltage information is created. If a region of critical tissue is identified, the same catheter can often be used to perform confirmatory EP maneuvers and direct ablation.
In EAVM, myocardium may be characterized as healthy/normal tissue, dense scar, or a border zone. These distinctions typically rely on the operator’s choice of voltage cutoffs and interpretation of the electrogram signals; therefore, they are not absolute. Some commonly used voltage cutoffs are derived from studies of VT entrainment, such as the study by Hsia et al. [12]. They reported that all entrance and isthmus sites corresponded to regions of abnormal bipolar voltage (<1.5 mV), with most of the sites being located in regions of electrically defined dense scar (≤0.5 mV) rather than in its border zones (0.5–1.5 mV).
Although EAVM is good at defining general regions of interest, it has its drawbacks. Achieving a sufficient point density of electrogram data can be tedious and time-consuming. It usually involves radiation exposure and occasionally can provoke arrhythmias. Most importantly, voltage mapping is subject to error. Normal tissue may yield a low-amplitude electrogram if contact is inadequate or where there is overlying fat, such as on the epicardium. Conversely, abnormal tissue may appear to produce a high-amplitude signal if its far-field qualities are not recognized, or the catheter slips between a normal and an abnormal structure. This can result, for example, if a catheter inadvertently touches a normal papillary muscle instead of an underlying region of inferior infarction. In addition, electrograms mapped from the endocardium do not represent the full myocardial thickness. Epicardial mapping is one approach to this problem; however, it has known risks and may still fail to identify mid-myocardial substrate.
Integrating MRI in the EP Laboratory
The great potential of integrating noninvasive scar data into EAVM would be to limit precise point mapping to only critical regions of substrate. Various methods to merge scar data into EAVM systems have been described [13–21]. It is possible to merge MRI and voltage datasets with vendor-integrated segmentation tools during the procedure. In one common approach, the endocardial volume on MRI can be registered as a shell. Then with the same dataset, the endocardial border of scar for each MRI short-axis image can be traced manually or in a semiautomated fashion with a signal intensity of two or three standard deviations above the threshold to define scar. A cutting tool can be used to extract the planimetered LGE regions to create a full-thickness slab of scar. This can be fused with an endocardial shell to create an endocardial map of scar. Then this LGE map can be fused with the electroanatomic voltage map through a landmark registration process [15]. Landmarks might include the left ventricular apex, left main coronary artery, aortic valve, or mitral annulus [18, 19]. Although LGE maps commonly display the presence or absence of endocardial scar, advanced maps based on signal intensity and transmurality, as shown in Figure 2, have also been described.
In some cases, the most challenging step in this integration process may be obtaining the preprocedure MRI image. In patients with conventional implantable cardiac defibrillators, MRI is generally contraindicated. There is a concern for damaging the device, inadvertent harm to the patient, and lead-related artifacts. Relying on the growing body of experience of MRI being safely performed in patients with implanted devices, some groups have performed preablation MRI in patients with defibrillators. In addition, it is likely that the emerging generation of MRI-conditional defibrillator systems will potentially resolve the issue of MRI safety. In the meantime, integration of CT late contrast enhancement or PET-CT datasets may provide alternatives [20, 21].
Correlating MRI and Voltage Data
Integration studies of MRI and voltage data have yielded insights into how voltage mapping findings relate to LGE abnormalities. Sites with low voltage correlate with the presence of LGE [9, 13, 14]. In one study of postinfarction VT ablation, Desjardins et al. [14] reported that sites with a bipolar voltage of 1.0 mV or less demonstrate 89% sensitivity and 84% specificity in detection of scar as defined by LGE. In addition, a threshold of 1.3 mV seemed to yield the best correlation between infarct size assessment by EAVM and LGE. Furthermore, all identifiable isthmus sites were located in areas of LGE. More than 70% of those isthmus sites were located in the core zone of the infarct (where the MRI signal intensity is three or more standard deviations greater than the reference), whereas the remaining sites localized to the peripheral gray zone (where LGE signal intensity was less). Transmurality was associated with lower voltages in general but had a wide range of overlap with subendocardial infarcts.
Despite the relative correlation, considerable mismatch has been reported. Codreanu et al. [13] reported that with a 1.5-mV bipolar cutoff, there was a 20% mismatch in infarct size measurements by EAVM and LGE. Specific regions of mismatch were influenced by the choice of transseptal or transaortic ablative approaches. Another study reported good correlation overall but with a tendency to underestimate infarct size by EAVM in inferior infarcts [16].
These findings suggest that EAVM by amplitude alone only roughly approximates scar. Abnormal electrogram characteristics besides low voltage, including prolonged duration, fragmentation or fractionation, sharp or “spiky” potentials, and isolated or late potentials, further correspond with the presence of LGE [13, 14, 22]. Desjardins et al. [14] reported that 89% of patients with fragmented electrograms and 95% of patients with isolated, late potentials had associated LGE.
In addition, regions of epicardial and intramural scar pose special problems from a voltage mapping perspective. These regions of interest may not be readily identified endocardially by standard voltage thresholds. One study reported that when epicardial regions of LGE were mapped from the endocardium, higher bipolar voltages (1.52±1.41 mV) were seen compared with the bipolar voltages for endocardial regions of LGE (0.94±1.07 mV) [14]. Although epicardial mapping, when feasible, is beneficial in these situations, intramural or mid-myocardial scars represent blind spots that can go undetected by both epicardial and endocardial approaches. In a population of patients with nonischemic cardiomyopathies of various causes, Bogun et al. [9] found that endocardial voltage mapping failed to detect the presence of mid-myocardial scar that was seen by LGE. Furthermore, the presence of isolated mid-myocardial LGE was associated with a much higher rate of procedural failure.
MRI Insights into Infarct Architecture
In addition to scar location, scar heterogeneity is thought to be conducive to electrical dispersion and areas of slow conduction that are a substrate for arrhythmia. Within the core zone of an infarct, MRI identifies heterogeneous areas that do not demonstrate maximal contrast enhancement. These gray zones correspond histologically to surviving muscle bundles within the scar tissue of an infarct, and it has been posited that they are more likely to contain critical conducting channels than regions of homogeneous scar [22–24]. In a study of heart failure patients by Lin et al. [25], the presence of these conductive channels, corridors of heterogeneous scar with relatively higher signal intensity within dense scar, was associated with a higher rate of ventricular fibrillation/VT attacks and mortality. In one integration study where LGE signal intensity maps were used to identify “channels” of heterogeneous tissue amid dense scar, these foci of heterogeneity were demonstrated to be critically related to clinical VT on the basis of their electrophysiological characteristics in 83% of patients [26].
Similarly, in a study of VT patients, critical isthmus sites, electrically defined by excellent pace map profiles, evidence of concealed entrainment, or termination during ablation, tended to cluster in regions with near transmurality (more than 75% wall thickness) or at the transition between core and border zone regions, suggestive of heterogeneity. Sites fulfilling both criteria contained all isthmus sites defined by concealed entrainment, 77% of VT termination sites, and 56% identified by pace mapping [27].
MRI and Procedural Planning
Although real-time integration of MRI scar maps in the EP laboratory is the ideal application of this modality, MRI has demonstrated additional utility in the areas of procedural planning and the assessment of unsuccessful ablations. In a study of ablative strategies for ischemic VT, patients were divided on the basis of whether transmural or subendocardial LGE was seen on MRI. Acosta et al. [23] compared endocardial-only ablation in patients with subendocardial scar, endocardial-only ablation in patients with transmural scar, and an endocardial-epicardial approach in patients with transmural scar. Patients with transmural scar who did not have epicardial ablation had a significantly higher rate of VT recurrence (41%) compared with the transmural scar group in which the endocardial-epicardial approach was used (12%). The ablative efficacy in the transmural scar group in which the endocardial-epicardial approach was used was very similar to the efficacy in the subendocardial group that underwent endocardial ablation alone. The presence of epicardial or intramural scar, rather than transmurality, was the criterion used in another study of patients undergoing repeated VT ablation to plan an ablation strategy either with or without epicardial access. Njeim et al. [28] were able to maintain the strategy in 95% of patients. They concluded that MRI-guided decision making was preferable to an empiric epicardial approach in repeated ablations.
MRI can also be used to confirm effective and properly localized ablative lesion formation. Ilg et al. [29] assessed radiofrequency lesion formation in 35 patients before ablation and at more than 1 year after ablation of VT or premature ventricular contractions, in nine of whom previous ablations had failed. With 83% reported ablation efficacy in the overall group, new scar corresponding to ablative lesions was seen in 71% of the total group. In this group, unsuccessful ablations were more common when papillary muscles were targeted, with ablation failure in one third of papillary VT cases. In these cases, MRI demonstrated LGE in the adjacent endocardium rather than on the papillary itself. This suggests poor tissue contact during attempts at papillary muscle ablation. In an additional third of unsuccessful ablations, where an initial endocardial approach failed, an epicardial site was the successful ablation target in a repeated procedure. The absence of scar, as was noted more often in right ventricular outflow tract sites, did not necessarily predict an unsuccessful ablation. Not surprisingly, larger regions of LGE indicated more complex ablative substrate.
MRI in Risk Stratification
Beyond VT ablation, MRI has an even greater potential in the risk stratification for SCD. The presence of LGE, its extent, and its heterogeneity have all demonstrated correlations with cardiac outcomes [30–33]. In a study of patients with a left ventricular ejection fraction greater than 35% and nonsustained VT, Dawson et al. [31] reported that the presence of scar was associated with more than a threefold increased risk of the combined end point of SCD, sustained VT, and defibrillator firing. A meta-analysis in 2747 nonischemic subjects reported that the risk of major adverse cardiac events was threefold higher when scar was present, and in 1367 of these patients, the risk of arrhythmic events was fivefold higher [32]. Additional studies suggest that the extent of scar is predictive of adverse events [29]. In combined cohorts of ischemic and nonischemic patients, it seems a critical volume of scar may portend risk, with values in the range of 5% or greater of left ventricular mass suggested as possible cutoffs [33].
Scar heterogeneity, in addition to its role in reentrant VT, likely portends greater risk of malignant arrhythmias. More extensive tissue heterogeneity has been found to correlate with increased inducibility by programmed EP study [33]. In addition, in a study of patients with ischemic cardiomyopathy who underwent MRI and subsequent defibrillator implantation, the extent of the infarct gray zone was a stronger predictor of spontaneous ventricular arrhythmias than were total or core infarct sizes [34]. It is important to note, though, that in ischemic heart disease, the data are mixed as to whether the extent of total scar, core infarct, or peri-infarct is most predictive of adverse cardiovascular events.
The incremental risk associated with LGE in the setting of left ventricular systolic dysfunction is a critical issue in risk stratification. Klem et al. [35] attempted to answer this question by analyzing their cohort of patients with a range of left ventricular ejection fractions. Their study showed that patients with a left ventricular ejection fraction greater than 30% and significant scar (>5%) had higher risk of death or implantable cardiac defibrillator discharge than those with minimal scar (hazard ratio 6.3, 95% confidence interval 1.4–28), but that this risk was similar to that of the group with a left ventricular ejection fraction of 30% or less. Among patients with a left ventricular ejection fraction of 30% or less, those with scar demonstrated greater risk than those without scar (hazard ratio 3.9; 95% confidence interval 1.2–13.1). Those with a left ventricular ejection fraction of 30% or less and minimal scarring had risk similar to that of patients with a left ventricular ejection fraction greater than 30%. These strategies need to be evaluated prospectively in larger cohorts but certainly suggest an important role for MRI in future SCD risk assessments.
Conclusion and Take-Home Messages
There is an emerging role for integrating noninvasive scar assessment, particularly MRI, in the treatment of patients with ventricular arrhythmia.
LGE with MRI allows the electrophysiologist to begin a procedure with an understanding of the topography and extent of myocardial scar.
With the use of MRI integration techniques, the precise nature of EAVM can be focused on critical regions of interest, which can save time and effort.
An epicardial or intramural location of scar can meaningfully change the procedural approach.
Postprocedure MRI may help the next procedure determine whether ablative lesions were appropriately targeted.
The location, extent, and heterogeneity of LGE seem to add to our current rudimentary methods for SCD assessment.