Comprehensive use of cardiac computed tomography to guide left ventricular lead placement in cardiac resynchronization therapy

Background Optimal lead positioning is an important determinant of cardiac resynchronization therapy (CRT) response. Objective The purpose of this study was to evaluate cardiac computed tomography (CT) selection of the optimal epicardial vein for left ventricular (LV) lead placement by targeting regions of late mechanical activation and avoiding myocardial scar. Methods Eighteen patients undergoing CRT upgrade with existing pacing systems underwent preimplant electrocardiogram-gated cardiac CT to assess wall thickness, hypoperfusion, late mechanical activation, and regions of myocardial scar by the derivation of the stretch quantifier for endocardial engraved zones (SQUEEZ) algorithm. Cardiac venous anatomy was mapped to individualized American Heart Association (AHA) bull’s-eye plots to identify the optimal venous target and compared with acute hemodynamic response (AHR) in each coronary venous target using an LV pressure wire. Results Fifteen data sets were evaluable. CT-SQUEEZ–derived targets produced a similar mean AHR compared with the best achievable AHR (20.4% ± 13.7% vs 24.9% ± 11.1%; P = .36). SQUEEZ-derived guidance produced a positive AHR in 92% of target segments, and pacing in a CT-SQUEEZ target vein produced a greater clinical response rate vs nontarget segments (90% vs 60%). Conclusion Preprocedural CT-SQUEEZ–derived target selection may be a valuable tool to predict the optimal venous site for LV lead placement in patients undergoing CRT upgrade.


Introduction
Patients with existing pacing systems, left ventricular (LV) systolic impairment, and a high proportion of right ventricular (RV) pacing benefit from cardiac resynchronization therapy (CRT). 1 CRT nonresponse occurs because of suboptimal LV lead positioning in myocardial scar with persistent dyssynchrony. 2,3 Cardiac magnetic resonance (CMR) can guide LV lead placement by avoiding scar and targeting late mechanical activation (LMA) 4 ; however, 28% of patients undergoing CRT have existing pacing systems unsuitable for CMR. 5 Cardiac computed tomography (CT) has the potential to guide LV lead placement in patients with existing pacing systems. 6 Rapid acquisition of 3-dimensional, isotropic, whole heart data sets with submillimeter spatial resolution can accurately delineate the coronary venous tree, 7 noninvasively assess regional and global LV function, 8 and detect regional hypoperfusion/myocardial scar. 9 Recently, CT has evaluated regional and global LV dyssynchrony and areas of LMA by calculating the stretch of the endocardial surface throughout the cardiac cycle (stretch quantifier for endocardial engraved zones [SQUEEZ]). 10 In patients with existing pacing systems undergoing CRT, we hypothesized that preprocedural cardiac CT-SQUEEZ by targeting areas of LMA and avoiding myocardial scar could guide LV lead placement through identification of the optimal venous target.

Methods
The study complied with the Declaration of Helsinki, and the protocol was approved by the local ethics committee. Informed consent was obtained from each patient. Between September 2014 and July 2016, we prospectively recruited 18 patients with a preexisting pacemaker/implantable cardioverter-defibrillator, persistent heart failure symptoms on optimal medical therapy, LV ejection fraction ,45%, and .50% RV pacing.

Preassessment
Patients underwent clinical assessment (New York Heart Association score and Minnesota Living with Heart Failure questionnaire), 6-minute walk test, cardiopulmonary exercise test, and 2-dimensional transthoracic echocardiography. Ischemic cardiomyopathy (ICM) was defined by prior myocardial infarction, coronary angiography demonstrating severe coronary disease and subsequent revascularization, and/or CMR evidence of myocardial fibrosis. Absence of these features inferred non-ICM.

Cardiac CT
Patients underwent cardiac CT using the Brilliance iCT 256-slice MDCT scanner (Philips Healthcare, Best, The Netherlands) before upgrade. Intravenous metoprolol was used to achieve a heart rate of ,65 beats/min in sinus rhythm and ,100 beats/min in atrial fibrillation. A total of 120 mL of intravenous contrast (Omnipaque, GE Healthcare, Princeton, NJ) was injected (5 mL/s) via a power injector into the antecubital vein. Descending aorta contrast-triggered (180 Hounsfield units [HU]), electrocardiogram (ECG)-gated scanning was performed with single breath-hold technique after a 10-to 12-second delay. Scanning parameters included a heart rate-dependent pitch of 0.2-0.45, a gantry rotation time of 270 ms, a tube voltage of 100 or 120 kVp depending on the patient's body mass index, and a tube current of 125-300 mA depending on the thoracic circumference. A second single-phase ECG-gated scan was acquired 12 minutes after the initial contrast bolus for myocardial scar imaging; the tube voltage was reduced by 20 kVp while the tube current was increased proportionally to account for an increase in image noise. 11 Initial retrospective ECG-gated scans were reconstructed in 10% phase increments throughout the cardiac cycle using iterative reconstruction, with 1-mm slice thickness, 0.5-mm slice increment, 250mm field of view, 512 ! 512 matrix, and an Xres smooth reconstruction kernel. Iterative reconstruction using the iDose4 algorithm (range 1-7) was used to reduce image noise and radiation dose. The cardiac CT scan was evaluated by an independent Society of Cardiovascular Computed Tomography Level III cardiac CT expert (R.R.). First-pass contrast-enhanced sequences were analyzed using standardized multiplanar reconstruction windows according to the AHA nomenclature for regional segmentation. End-diastolic myocardial wall thickness was evaluated reviewing each myocardial region in both short-and long-axis views. Areas of hypoperfusion were evaluated systematically with slice width thickness increased to 5 mm and display window and level settings adjusted to 100 and 200 HU, respectively. Abnormal perfusion was defined as myocardium exhibiting significantly reduced contrast distribution visually compared against attenuation of normal myocardium in each patient used as an internal reference. Where discrepancies existed, a 50 HU difference between normal and hypoperfused myocardium was used 12,13 along with consensus opinion from 2 individual experts in cardiac CT (AHA/ACC Level III experience, blinded to the clinical data).
We adopted a pragmatic approach for identifying regions of delayed enhancement as previously described. 11 This qualitative identification of scar where the myocardium is brighter reflects the current lack of standard criteria in the literature mainly because the Hounsfield unit attenuation varies significantly between patients with no accepted value set as a cutoff for fibrosis. Where scar detection was difficult, surrogate markers of a wall thickness of ,6 mm and regional hypoperfusion were used to infer scar.

CT-derived SQUEEZ
High-contrast disparity between LV blood pool and myocardium permits the identification and tracking of finely engraved endocardial surface features. The SQUEEZ method uses these features to track endocardial material points over the heart cycle to calculate regional cardiac function using the following formula: where A(v, 0) is the area of the small triangular patch (v) on the endocardial mesh at end diastole and A(v, t) is the area of the same patch at time t. The SQUEEZ metric is calculated for each of the triangular patches across the endocardium through cardiac phases. Thus, a high-resolution regional map of endocardial strain can be computed as SQUEEZ 2 1 at each point ( Figure 1). This metric has been correlated with circumferential strain (E cc ), the criterion standard for noninvasive regional strain, using tagged CMR sequences in a canine model of myocardial infarction. 14 SQUEEZ-derived regional function data were merged with individual patient anatomy. Septal segments were excluded as targets. SQUEEZ-derived strain curves with low amplitude strain (LAS) ,10% shortening were judged nonviable and excluded as targets (analogous to echocardiographic data with poor CRT response with LV lead placement in segments with LAS ,9.8%). 11,12 Regional time to peak strain (minimal SQUEEZ value) was calculated using the individual heart rate/cycle length for the CT acquisition ( Figure 2). Given the distribution of the coronary veins, the Figure 1 A: Computed tomography mid-ventricular short-axis images of the left ventricle at 2 time points in the R-R interval (0% 5 end diastole; 50% 5 end systole). Anterior and anteroseptal regions are akinetic and seen not to move throughout the cardiac cycle compared with the inferolateral segments that move inward by end systole. Inevitable beam hardening artifact from the existing pacing system noted in the right ventricle. B: Stretch quantifier for endocardial engraved zones values (y axis) vs cardiac cycle length (%) across 16 AHA segments demonstrate akinetic regions (red box) and late activating inferior/inferolateral walls (green box representing an ideal target for left ventricular lead placement). time to peak strain of myocardial segments overlying the same coronary vein was averaged to produce anterior, lateral, and inferior values for mechanical activation delay and compared with hemodynamic data.

CRT implantation
CRT upgrade was performed with a quadripolar LV lead in 17 patients (94%) and a bipolar lead in 1 because of unfavorable venous anatomy. Nine patients (50%) received CRT with defibrillator; 1 patient required an additional RV shock lead; and the remaining 8 already having preexisting implantable cardioverter-defibrillators. Identification of the optimal hemodynamic site for LV stimulation was performed by measuring acute hemodynamic response (AHR) using an 0.014-in high-fidelity Certus RADI PressureWire (St. Jude Medical, St. Paul, MN) in the LV cavity via a retrograde arterial approach as previously described. 13 Atrial pacing 10 beats/min above the intrinsic rate or RV pacing (DDDRV) for patients with no underlying rhythm was baseline. Atrioventricular delays were fixed at 100 ms and ventriculoventricular delay at 0 ms. AHR for each venous site compared biventricular pacing with baseline (% change, dP/dt max, mm Hg/s). AHR was compared with CT indices and the sensed QLV interval at each pacing site. 14 Patients underwent 6-month follow-up to identify clinical responders via the Packer score 15 and echocardiographic response defined as reduction in LV end-systolic volume .15%. The primary end point was cardiac CT-derived regional endocardial strain analysis (SQUEEZ) prediction of the pacing site achieving the optimal AHR.

Statistics
Continuous data were presented as mean 6 SD. Those with Gaussian distribution were compared using a paired t test; those with a non-Gaussian distribution were compared using the Wilcoxon matched-pair signed rank test. Categorical data were presented as absolute number of occurrences and associated frequency (%). Analysis of variance was used to compare .2 groups. The results were considered significant at a P value of ,.05.

Results
The patient demographic characteristics are summarized in Table 1. Patients had a high percentage of RV pacing with a mean QRS duration of 173 6 21 ms. Eight patients (44%) had ICM.

Quality of CT data sets
All CT scans were successfully completed (mean heart rate 64 6 7 beats/min; mean radiation dose-area product 1194 6 419 mGy$cm 2 ). Patients were supine for 15 6 1 minutes. Studies were independently assessed using the Society of Cardiovascular Computed Tomography quality score 16 (mean score 3.4 6 1.3 out of 5). The main reason for reduced scoring was beam hardening artifact from existing pacing wires. Wall thinning/hypoperfusion (inferring scar) was present in 7 of 8 patients with ICM; however, late contrast enhancement was limited to 1 patient.
We demonstrate the novel utility of preprocedural CT for LV lead guidance to regions of LMA devoid of LAS. Figures 1 and 2 show that cardiac CT sequences were able to generate functional data sets with sufficient temporal resolution to differentiate the region of LMA. In addition to wall thinning and hypoperfusion, the SQUEEZ algorithm inferred regions of scar on the basis of LAS in keeping with echocardiographic studies (defined LAS as radial strain ,10%), resulting in suboptimal response to CRT. 15 Our CT protocol included a late enhancement sequence to identify fibrosis; however, we identified only late enhancement in 1 patient. Our CT-SQUEEZ-derived algorithm predicted regions with an AHR within 2.5% of the maximum in 11 of 12 cases (92%). In addition, patients paced in CT targets had a more favorable response (Figure 4). In 1 case, the mismatch between the CT target and the vein with the best AHR was significant (22.2%). In 2 cases, CT-SQUEEZ targeted the posterior wall; however, the patients had no epicardial vein overlying this site.
The clinical utility of AHR .10% in predicting CRT response has been demonstrated, 16 and using this cutoff, a high proportion of CT targets had a positive AHR. There was a trend toward greater AHR in sites out of scar vs CT inferred scar (14.8% 6 12% vs 6.8% 6 8%; P 5 .17). The mean AHR was similar in CT targets compared with the QLV interval (Figure 4), reflecting a correlation between electrical and mechanical dyssynchrony. However, regions with electrical latency (QLV interval .100 ms) in scar resulted in suboptimal AHR, suggesting that the QLV interval alone may fail to identify the optimal stimulation site in the presence of scar.   We previously demonstrated that CMR-derived scar and dyssynchrony can guide LV lead implantation 18 with excellent CRT response if pacing in a CMR target in keeping with the present findings. Similarly, Laksman et al 19 used CMRderived scar and dyssynchrony-guided lead placement with echocardiographic super-response in 58% of patients. Currently there are limited clinical data describing the use of cardiac CT LV lead guidance. 20 Two randomized controlled studies TARGET 21 and STARTER 22 have shown benefit with LV lead guidance using speckle tracking echocardiography.
Scar avoidance using CMR-derived late gadolinium enhancement is associated with improved cardiovascular outcomes. 23 The ability of CMR to accurately define scar is superior to CT; however, almost one-third of patients undergoing CRT have existing pacemakers unsuitable for CMR. 5 Cardiac CT offers several potential advantages over CMR. Images of superior spatial resolution within a 3-dimensional isotropic data set are acquired within a matter of seconds; the new generation of CT scanners produce images in a single heartbeat. In addition, an exquisite differentiation of endocardium and blood pool enables accurate tracking of regional surfaces over the duration of the cardiac cycle. Delayed triggered contrast-enhanced sequences facilitate good opacification of the coronary venous tree, and volume-rendered reconstructions can delineate the course of potential venous targets before implantation. This may be particularly advantageous as coronary venous anatomy is highly variable and may impact successful CRT delivery. 24 Furthermore, identification of coronary sinus valves and highly angulated/tortuous vessels may help in patients with a previously failed implant. While targeting regions of Values are presented as mean 6 SD. 2D 5 2-dimensional; CPET 5 cardiopulmonary exercise test; CRT 5 cardiac resynchronization therapy; LV 5 left ventricular; LVEF 5 left ventricular ejection fraction; MLHF 5 Minnesota Living with Heart Failure questionnaire; NT-pro-BNP 5 N-terminal pro-brain natriuretic peptide; NYHA 5 New York Heart Association; TTE 5 transthoracic echocardiogram; VO 2 max 5 maximum oxygen consumption. Figure 6 Clinical response rates in CT-SQUEEZ target (n 5 10) vs nontarget (n 5 5) (P , .001) and ICM (n 5 8) vs NICM (n 5 10) (P 5 .07). CT 5 computed tomography; ICM 5 ischemic cardiomyopathy; NICM 5 nonischemic cardiomyopathy; SQUEEZ 5 stretch quantifier for endocardial engraved zones.
LMA is scientifically sound, LV lead placement is restricted by coronary vein anatomy that may not always overly the optimal region. In a cardiac CT study of 121 postmortem hearts, 29% had no coronary vein overlying the posterolateral region. 17 Lack of a venous target identified by preprocedural CT may allow alternative forms of LV stimulation (multisite/ endocardial) to be considered. 25,26 Study limitations Modern single energy source CT scanners are currently unable to accurately delineate extracellular myocardial fibrosis through late contrast enhancement. While iodinated contrast displays similar kinetic properties to gadolinium-DTPA (diethylenetriamine penta-acetic acid) and can demonstrate acute hypoperfusion from ischemic injury, reliable visualization of chronic fibrosis with late enhancement occurred in only 1 patient. We therefore used local wall thinning (,6 mm) and/or hypoperfusion to infer scar, and the low AHR in these segments is in keeping with scar despite the lack of visualization (Figures 4 and 5). The presence of existing pacing systems resulting in beam hardening artifact and degradation of signal in myocardial tissue local to the pacing leads may explain this. One study reporting a good correlation between late enhancement and scar in histological macroscopy in a chronic porcine mode 25 used higher-contrast doses than our study (145 6 35 mL vs 120 6 0 mL; P 5 .005), and the animals had no preexisting pacing systems. The development of dual energy source CT scanners holds promise in improving differentiation between subtle soft tissue characteristics and may be able to more reliably demonstrate myocardial fibrosis. 27 The temporal resolution of cardiac CT in this study (70-100 ms) is inferior to echocardiography (20 ms) and CMR (35-50 ms, obtained over multiple beats), and cardiac CT may be less sensitive to subtle regional motion changes. While the SQUEEZ algorithm detects motion abnormalities with high resolution, only lower-resolution estimates (16 standard AHA segments) were needed. Furthermore, the CT-SQUEEZ-derived metric has been shown to correlate well with circumferential strain (E cc ) in an animal model, suggesting that it is sensitive enough to demonstrate local regional motion differences and remain a useful tool to assess dyssynchrony. 14 The utility of AHR in predicting CRT response 16 is limited, and the results of a large multicenter randomized trial of AHR are awaited (RADI-CRT, Clinical Trial Registration No.: NCT01464502).

Conclusion
Preprocedural CT-SQUEEZ-derived target selection may be a valuable tool to predict the optimal venous site for LV lead placement by guiding the implanter toward late activating regions, away from areas of scar.