Author + information
- Received November 9, 2012
- Revision received May 3, 2013
- Accepted May 6, 2013
- Published online November 1, 2013.
- Lian-Yu Lin, MD, PhD∗,
- Mao-Yuan M. Su, PhD†,
- Jien-Jiun Chen, MD‡,
- Ling-Ping Lai, MD, PhD∗,
- Juey-Jen Hwang, MD, PhD∗,
- Chuen-Den Tseng, MD, PhD∗,
- Yih-Sharng Chen, MD, PhD§,
- Hsi-Yu Yu, MD, PhD§,
- Wen-Yih I. Tseng, MD, PhD†,‖∗∗ ( and )
- Jiunn-Lee Lin, MD, PhD∗∗ ()
- ∗Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan
- †Department of Medical Imaging, National Taiwan University Hospital, Taipei, Taiwan
- ‡Cardiovascular Center, National Taiwan University Hospital Yun-Lin Branch, Douliou, Taiwan
- §Department of Surgery, National Taiwan University Hospital, Taipei, Taiwan
- ‖Center for Optoelectronic Biomedicine, National Taiwan University College of Medicine, Taipei, Taiwan
- ↵∗Reprint requests and correspondence:
Dr. Jiunn-Lee Lin, Department of Internal Medicine, National Taiwan University Hospital, No. 7, Chun-Shan S. Road, Taipei, Taiwan.
- ↵∗∗Dr. Wen-Yih I. Tseng, Center for Optoelectronic Biomedicine, National Taiwan University College of Medicine, No. 1, Jen-Ai Road, Section 1, Taipei, Taiwan.
Objectives This study evaluated whether the conductive channel (CC) identified by late gadolinium enhanced–cardiac magnetic resonance (LGE-CMR) is associated with ventricular tachycardia (VT) in patients with systolic heart failure (HF).
Background One recent study demonstrated that the CC formed by heterogeneous tissue within the core scar could be detected by LGE-CMR and that the CC is responsible for clinical VT. We hypothesized that the CC could help identify HF patients at risk for VT.
Methods A total of 63 patients from a CMR database with left ventricular ejection fraction (LVEF) below 50% and with hyperenhancement on LGE-CMR were included. The cine and LGE images were analyzed to derive the LV function and scar characteristics, and to identify the CC. The outcomes, including VT, ventricular fibrillation (VF), and total mortality, were obtained by reviewing medical records.
Results After a median 1,379 (interquartile range: 271 to 1,896) days of follow-up, 8 patients had VT/VF attacks and 14 patients died. Among the CMR-measured parameters, only the probability of identifying the CC by LGE-CMR was higher in patients with VT/VF than those without VT/VF (75.0% vs. 16.4%, p < 0.001). The probability of identifying the CC was also higher in the total mortality group than the survival group (50.0% vs. 16.3%, p = 0.004). The other LGE-CMR variables were not significantly different between the 2 groups. A univariate Cox regression model showed that CC identification was positively associated with VT/VF attacks (hazard ratio [HR]: 27.032, 95% confidence interval [CI]: 3.291 to 222.054, p = 0.002) and excess total mortality (HR: 4.766, 95% CI: 1.643 to 13.824, p = 0.004). The LVEF was inversely associated with VT/VF attacks (HR: 0.119, 95% CI: 0.015 to 0.977, p = 0.048) and excess total mortality (HR: 0.491, 95% CI: 0.261 to 0.925, p = 0.028) during follow-up.
Conclusions We demonstrated that CC identification using LGE-CMR can help identify HF patients at risk for VT/VF.
Patients with congestive heart failure (HF) carry an additional risk of morbidity and mortality (1). It is of clinical importance to identify these high-risk patients because optimal medical and device therapy could substantially lower this risk (2,3). Recently, late gadolinium enhanced–cardiac magnetic resonance (LGE-CMR) has become a powerful tool for predicting cardiac events (4–6). In contrast to the precise left ventricular (LV) systolic function measurement offered by cine cardiac magnetic resonance (CMR), LGE-CMR has excellent sensitivity and specificity for detecting small myocardial scars and quantifying the core scar and border zone (5,6).
More recently, researchers have found that LGE-CMR could help identify the abnormal substrate for ventricular tachycardia, which could therefore facilitate the ablation procedure (7,8). Through image integration, LGE-CMR has been shown to detect ventricular tachycardia (VT)-associated scars that could not be identified by invasive electroanatomical voltage mapping (EAVM) (7). In 1 study, Perez-David et al. (8) showed that with computer-assisted visual identification, LGE-CMR could help detect the conductive channel (CC) within the scars. Through concomitant EAVM and electrophysiology analysis, the researchers further demonstrated that a significant amount of the CCs were critical isthmuses for VT (8).
In the present study, our goal was to investigate whether the CC identified by LGE-CMR could be used to predict risk for patients with systolic HF.
This is a retrospective cohort study of an LGE-CMR database from 2004 to 2011. Patients with known renal dysfunction or estimated glomerular filtration rate <30 ml/min/1.73 m2 were excluded from the CMR examination according to the American College of Radiology manual of contrast media. We surveyed the LGE-CMR database of the National Taiwan University Hospital and selected patients with left ventricular ejection fraction (LVEF) below 50% and with hyperenhancement on LGE-CMR. A total of 130 patients with LVEF <50% had a LGE-CMR examination. All patients had clinical symptoms and signs of congestive HF. We excluded 10 subjects with previous acute myocardial infarction (MI) whose LGE-CMR images were acquired within 3 months post-infarction. To investigate the prognostic value of LGE-CMR image parameters, we excluded 21 patients whose LV function deteriorated or recovered dramatically during follow-up (LVEF with over 15% of variation); the LVEF measured by various image modalities (echocardiography or radionuclide angiography) should have remained stationary in subsequent visits even after interventions such as percutaneous coronary intervention, coronary artery bypass graft, valve replacement or repair surgery, and cardiac-resynchronization therapy. Of the remaining 99 subjects, only 63 subjects with available LGE-CMR raw data were included for analyses. The basic demographics; cardiovascular risk factors, including hypertension, diabetes mellitus, dyslipidemia, and chronic kidney disease (CKD), defined as estimated glomerular filtration rate <60 ml/min, etiologies of impaired HF; medications of the included subjects; and other cardiovascular disease, such as ischemic stroke or peripheral arterial occlusive disease, were obtained by reviewing medical records. The outcomes, including malignant ventricular tachycardia and mortality etiologies, were also obtained by reviewing medical records or the National Health Institute database in Taiwan.
Magnetic resonance imaging
The study was performed on a 1.5-T magnetic resonance imaging system (Sonata, Siemens, Erlangen, Germany) with a 4-channel phased array torso coil. Cine CMR was first performed using a segmented balanced steady-state free precession gradient echo pulse sequence with a retrospective electrocardiographic R-wave gating. The scanning parameters were repetition time, 3.9 ms; echo time, 1.6 ms; flip angle, 70°; bandwidth, 977 Hz/pixel; slice thickness, 7 mm; gap, 3 mm; matrix size, 256 × 159; and pixel size, 1.3 mm × 1.3 mm. Multiple short-axis slices were prescribed from the mitral orifice to the LV apex. A total of approximately 10 to 12 short-axis slices were obtained, depending on cardiac size. For each slice level, patients were required to hold their breath for approximately 10 s. The true temporal resolution was 49 ms, and 30 cardiac phases were reconstructed for each slice level.
The LGE-CMR image study was performed by slow infusion of a gadolinium-based contrast agent (Magnevist, Berlex Laboratories, Wayne, New Jersey) through the left antecubital vein; the dose was 0.2 mmol/kg body weight. Ten minutes after contrast medium injection, LGE images were acquired in the same short-axis planes as those used for cine images by using an electrocardiographic R-wave–triggered, phase-sensitive inversion recovery (PSIR) prepared segmented fast gradient echo pulse sequence (9). The scanning parameters were inversion time (TI)/repetition time/echo time, 250 ms/9.6 ms/4.2 ms; flip angle, 25°; bandwidth, 130 Hz/pixel; slice thickness, 7 mm; gap, 3 mm; matrix size, 256 × 192; and pixel size, 1.37 mm × 1.37 mm. We acquired LGE-CMR images using PSIR sequence with a fixed TI of 250 ms. The PSIR sequence takes into account the polarity of the z-magnetization, and so the contrast between the normal myocardium and the scar is kept relatively stable despite the mismatch of the TI with the null point of the normal myocardium (10). This feature allows us to complete the acquisition of the LGE-CMR images quickly, without spending time on adjusting the TI.
LV Function and Mass Analysis From Cine CMR
Detailed procedures regarding LV function and mass analysis have been described elsewhere (11). Briefly, the endocardial and epicardial contours of the LV in the short-axis view were determined at each slice level by a user-interface graphics program, which used a gray-level auto-contouring algorithm provided by MATLAB version 7.9 (MathWorks, Natick, Massachusetts). The area enclosed by each contour was computed. The LV volumes for each time point were then determined by Simpson's rule to obtain the volume–time curve of the LV. This is equivalent to the sum of the areas of the corresponding slice levels multiplied by the slice thickness plus gap. The end-diastolic volume and end-systolic volume were assessed from the maximal and minimal values of the volume–time curve. These values were used to compute the LVEF. LV mass was computed as the difference between LV epicardial volume at end-diastole and end-diastolic volume, multiplied by the density of the myocardium, which is 1.05 g/ml.
Quantification of the Core Scar and Gray Zone
The infarct and remote zones were determined on LGE-CMR images for each patient. On the LGE images, the endocardial and epicardial contours were manually traced to define the LV wall. The infarct zone was defined as the sector of the LV wall that contained hyperenhancement of any transmurality. The remote zone was defined as the remaining portion of the LV wall that did not contain hyperenhancement. In the infarct zone, a pixel was counted as the core scar if its signal intensity (SI) was higher than 3 SD above the average SI of the remote myocardium, and it was counted as the gray zone if the SI was between 2 and 3 SD above the average SI of the remote zone (12). The volumes of the core scar and of the gray zone were quantified based on the Simpson's rule: volume = (summation of the interested areas of all short-axis slices) × (slice thickness + gap). The volumes were normalized by an individual's total LV mass volume and are expressed in terms of percentages.
Identification of the Conductive Channels
The CC in the ventricular wall was identified and defined according to a previous study (8). Two investigators (L.Y.L. and M.Y.S.), who were blinded to the outcomes, analyzed the data. A custom-developed MATLAB program was used to create a computer-assisted image that was derived by subtracting the difference of the areas between regions with an SI over 2 SD and regions over 3 SD for each short-axis slice. (8). As illustrated in Figure 1, by visually reviewing each subtraction image, a CC is considered if a corridor of the heterogeneous tissue (blue zone) penetrates the scar (yellow zone) with its 1 end (Fig. 1A) or 2 ends (Fig. 1B) connecting to the normal myocardium (red zone). If the gray zone surrounds the scar without deep penetration into the scar (Fig. 1C), it is not considered a CC. Figures 2A and 2B are LGE-CMR images that were obtained from 2 patients with a CC in the infarct zones. Figure 2C shows images of a patient without a CC. The presence of a CC was only accepted if both investigators independently agreed on the outcome. In this study, we used the phase image of the PSIR as opposed to the magnitude image to quantify the myocardial scar and CC. The benefits of the reconstructed image include reduction of sensitivity in B1 field inhomogeneity, increased tolerance of TI, background noise reduction, and nearly complete magnetization recovery (9).
The normality of the variables was tested by the Shapiro-Wilk test. Continuous variables with normal distribution are expressed as means ± SD, whereas those that were not normally distributed are reported as medians and interquartile ranges (25% to 75%). Categorical variables were expressed as percentages. The demographics, etiology for HF, medication usage, and image variables were compared between patients with different outcomes by using the Student t test for normally distributed continuous variables and by using the Mann-Whitney U test for those that were not normally distributed. Categorical variables between patients with different outcomes were compared by the Fisher exact test. The predictors for outcomes, which included VT, ventricular fibrillation (VF), and total mortality, were analyzed by a univariate Cox regression model. The hazard ratio (HR) was calculated for each LGE-CMR variable. The presence of CC was entered as a categorical variable (yes or no). The LVEF, the LV mass, the core scar, and the gray zone were also entered as categorical variables by using their median values as cut points (≥ or < median values). For Kaplan-Meier survival analysis, patients were divided into 2 groups that were stratified according to the existence of the CC. Between-group differences in survival were tested by log-rank statistics. A p value <0.05 was considered statistically significant.
A total of 63 patients, age 27.3 to 83.3 years, were included. Of this total, 14 patients were women and 49 were men. The median number of follow-up days was 1,379 (interquartile range, 271 to 1,896) days. Three cases were lost during follow-up. During the follow-up period, 8 patients had VT/VF attacks, and 14 patients died. Among the 14 mortalities, 6 patients died of ventricular tachycardia, 4 died of congestive HF, 2 died of severe sepsis, and 2 died of respiratory failure. The basic demographics, risk factors, etiology for HF, comorbidities, medication usage, CMR image parameters, and the incidence of new stable angina or acute coronary syndrome (ACS) during follow-up between patients with and without VT/VF are summarized in Table 1. As shown in Table 1, patients with VT/VF were more likely to have CKD with borderline significance (62.5% vs. 25.9%, p = 0.050). Among the LGE-CMR image parameters, only the probability of identifying the CC by LGE-CMR was higher in patients with VT/VF (75.0% vs. 16.4%, p < 0.001). Other statistically insignificant variables between the 2 groups included age, sex, incidence of hypertension, diabetes mellitus, dyslipidemia, old MI, other cardiovascular diseases, etiologies for HF, the percentage of antiplatelet, beta-blocker, and angiotensin-converting enzyme inhibitor/angiotensin receptor blocker usage, LVEF, LV mass, the percentages of the core scar and gray zone, and the incidence of new stable angina or ACS during follow-up.
The basic demographics, risk factors, HF etiology, comorbidities, medication usage, CMR variables, and the incidence of new stable angina or ACS during follow-up between mortality and survival groups are summarized in Table 2. Among the variables, the incidence of CKD (57.1 vs. 22.9, p = 0.022), old MI (64.3% vs. 28.6%, p = 0.026), and the probability of the identification of the CC (50.0% vs. 16.3%, p = 0.004) were significantly higher in the mortality group. Other variables were not statistically significant between the mortality and survival groups.
The HR of each CMR image variable in predicting different outcomes is shown in Table 3. The median values for LVEF, LV mass, core scar, and gray zone are 37.5%, 148.5 g, 33.41%, and 12.9%, respectively. Among the CMR image parameters, the identification of the CC was positively associated with VT/VF attacks during follow-up (HR: 27.032, 95% confidence interval [CI]: 3.291 to 222.054, p = 0.002). The LVEF (HR: 0.119, 95% CI: 0.015 to 0.977, p = 0.048) was inversely associated with VT/VF attacks. Other parameters, which included LV mass and percentages of the core scar and gray zone, were not significantly associated with VT/VF attacks. As for the prediction of total mortality, the identification of the CC (HR: 4.766, 95% CI: 1.643 to 13.824, p = 0.004) and the LVEF (HR: 0.491, 95% CI: 0.261 to 0.925, p = 0.028) were also predictors for excess mortality.
Figure 3 shows the Kaplan-Meier survival curves for VT/VF attacks (Fig. 3A) and total mortality (Fig. 3B). Log-rank statistics showed that patients with a CC had a significantly poorer prognosis than those without a CC (p < 0.001 and p = 0.007 for VT/VF and total mortality, respectively).
Our results showed that among the LGE-CMR image parameters, only the presence of the CC and lower LVEF were significantly associated with VT/VF attacks and total mortality.
LGE-CMR has emerged as an important clinical tool in predicting cardiac events in patients with HF (5,6,12–18). Many studies have demonstrated that the extension of the scar measured by LGE-CMR is superior to traditional risk predictors such as LVEF and LV volumes in predicting future cardiac events (13–17). In the present study, we showed that the percentages of the core scar and gray zone are not associated with VT/VF attacks and total mortality. It is likely that the “quality” of the scar, such as tissue heterogeneity or presence of the CC, rather than the quantity plays a larger role in the generation of VT/VF.
Many previous reports have shown the importance of the scar heterogeneity, that is, the peri-scar gray zone in predicting cardiac events. These reports suggest that the residual peri-infarct myocardium might provide the substrate for arrhythmias and/or ischemia (5,6,12,14,18). In contrast to these previous studies, our data showed that the percentage of the gray zone is not a predictor for either VT/VF and total mortality. One possible explanation is that the criteria of SI to identify the core scar and the gray zone are different among studies. Another possibility is that the patients in the current analysis seldom (3.2% had new stable angina, and 9.5% had ACS) had any new ischemic episodes during follow-up. It is well known that revascularization of the ischemic regions could substantially reduce the incidence of ventricular arrhythmia in patients with ischemic cardiomyopathy (19,20). It is possible that a peri-infarct gray zone without ischemia is electrophysiologically more stable and is less likely to generate VT/VF. Finally, the critical site of VT/VF generation has been shown to be located predominately within the region of the core scar (8,21). This explains our main finding that the presence of the CC within the core scar is better than quantifying scar heterogeneity in predicting VT/VF attacks.
Human studies have shown that the slow conduction area or critical isthmus of the scar-related VT re-entrant circuit is composed of bundles of viable, but injured, myocytes embedded in the scar tissue (22,23). The heterogeneous tissue, although mainly located in the peri-scar region, is most likely more harmful when located within the core scar area, resulting in the formation of the so-called CC. The CC can now be identified invasively by endocardial voltage mapping and has become a standard procedure for scar-related VT ablation (24,25). In a recent cross-sectional design study, Perez-David et al. (8) showed that the CC could be identified noninvasively by computer-assisted visual identification of the LGE-CMR image. By concomitant EAVM and electrophysiology study, the authors also showed that each CC identified by noninvasive method corresponded to a similar channel detected by EAVM, and a majority of the channels were related to VT critical isthmuses (8). In the present study, we further demonstrated that CC could also help identify patients at risk of VT/VF. Furthermore, in the present study, 6 of 14 patients died of VT/VF; therefore, the CC also became a predictor for total mortality.
First, the study population is small, and the study cohort is retrospective. Second, because we did not perform EAVM for our patients, there were no invasive data to validate the CC identified by LGE-CMR. Third, in this study, all of the LGE images were acquired with an inversion recovery prepared segmented 2-dimensional (2D) turbo fast low angle shot sequence, not the 3-dimensional sequence. The LGE images acquired with the 2D sequence only allows us to identify corridors within each 2D image. We cannot perform 3-dimensional reconstruction of the LGE images and examine the vertical corridors as Perez-David et al. (8) did. Also, each LGE-CMR slice image was 3 mm apart, which may not be close enough to detect any potential CC located between the slices. However, the slices were uniformly distributed, and the coverage reached up to 70% of the whole LV myocardium. The impact of underestimating CC or scar should be insignificant. Despite this limitation, our results show that the conductive channels identified on 2D images still constitutes a good predictor of ventricular tachycardia in patients with HF. Finally, the identification of the CC is sensitive to the definition of the SI thresholds of core scar and gray zone. The main reason for using a 3 SD threshold to identify the infarct zone is because the area produced by this threshold matches closely with what is shown on the LGE images by visual inspection. The optimal definition should be validated by invasive 3-dimensional EAVM.
We demonstrated that a CC identified in an LGE-CMR image can help identify patients at risk of VT/VF. Large-scale studies are needed to confirm whether this finding could guide the physicians to choose proper therapy (for example, an implantable cardioverter-defibrillator) for their patients with HF.
This work was supported in part by the National Science Council, Taiwan (grant numbers: NSC-99-2314-B-002-116-MY3 and NSC-102-2314-B-002-087-MY2). The authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. L-Y Lin and Su contributed equally to this work.
- Abbreviations and Acronyms
- conductive channel
- cardiac magnetic resonance
- electroanatomical voltage mapping
- heart failure
- late gadolinium enhanced–cardiac magnetic resonance
- left ventricle/ventricular
- left ventricular ejection fraction
- myocardial infarction
- phase-sensitive inversion recovery
- signal intensity
- ventricular fibrillation
- ventricular tachycardia
- Received November 9, 2012.
- Revision received May 3, 2013.
- Accepted May 6, 2013.
- American College of Cardiology Foundation
- Schocken D.D.,
- Arrieta M.I.,
- Leaverton P.E.,
- Ross E.A.
- Cheong B.Y.,
- Muthupillai R.,
- Wilson J.M.,
- et al.
- Heidary S.,
- Patel H.,
- Chung J.,
- et al.
- Roes S.D.,
- Borleffs C.J.,
- van der Geest R.J.,
- et al.
- Wijnmaalen A.P.,
- van der Geest R.J.,
- van Huls van Taxis C.F.,
- et al.
- Perez-David E.,
- Arenal A.,
- Rubio-Guivernau J.L.,
- et al.
- Yan A.T.,
- Shayne A.J.,
- Brown K.A.,
- et al.
- Assomull R.G.,
- Prasad S.K.,
- Lyne J.,
- et al.
- Wu K.C.,
- Weiss R.G.,
- Thiemann D.R.,
- et al.
- Kwon D.H.,
- Halley C.M.,
- Carrigan T.P.,
- et al.
- Schmidt A.,
- Azevedo C.F.,
- Cheng A.,
- et al.
- Kelly P.,
- Ruskin J.N.,
- Vlahakes G.J.,
- Buckley M.J. Jr..,
- Freeman C.S.,
- Garan H.
- Brugada J.,
- Aguinaga L.,
- Mont L.,
- et al.
- de Bakker J.M.,
- van Capelle F.J.,
- Janse M.J.,
- et al.
- de Bakker J.M.,
- Coronel R.,
- Tasseron S.,
- et al.
- Arenal A.,
- del Castillo S.,
- Gonzalez-Torrecilla E.,
- et al.