Real-Time X-MRI-Guided Left Ventricular Lead Implantation for Targeted Delivery of Cardiac Resynchronization Therapy

OBJECTIVES This study sought to test the feasibility of a purpose-built, integrated software platform to process, analyze, and overlay cardiac magnetic resonance (CMR) data in real time within a cardiac catheter laboratory and magnetic resonance imaging scanner in the same facility with the ability to transfer patients from one to the other (X-MRI) environment to guide left ventricular (LV) lead implantation. BACKGROUND Suboptimal LV lead position is a major determinant of poor cardiac resynchronization therapy (CRT) response, and the optimal site is highly patient speci ﬁ c. Pacing myocardial scar is associated with poorer outcomes; conversely, targeting latest mechanical activation (LMA) may improve them. METHODS Fourteen patients (age 74 (cid:1) 5.1 years; New York Heart Association functional class: 2.7 (cid:1) 0.4; 86% ischemic with ejection fraction 27 (cid:1) 7.6%; QRSd: 157 (cid:1) 19 ms) underwent CMR followed by immediate CRT implantation using derived scar and dyssynchrony data, overlaid onto ﬂ uoroscopy in an X-MRI suite. Rapid LV segmentation enabled detailed scar quanti ﬁ cation, identi ﬁ cation of LMA segments, and selection of myocardial targets. At coronary venography, the CMR-derived 3-dimensional shell was

C ardiac resynchronization therapy (CRT) is a highly efficacious treatment for symptomatic patients with heart failure, severe left ventricular (LV) dysfunction, and broad QRS duration (1,2). Despite 2 decades of delivering biventricular pacing for selected patients with heart failure, 30% to 50% fail to undergo LV remodeling (3), with most implanters using empirical placement of the LV lead on the posterolateral wall. However, the optimal site for LV stimulation is highly patient specific (4) and suboptimal lead position is a major determinant of poor response (5). LV lead positioning in or near areas of myocardial fibrosis is associated with poorer outcomes (6)(7)(8). Furthermore, targeting LV myocardial segments with latest mechanical activation using speckle tracking echocardiography has demonstrated utility in improving CRT outcomes in single-center randomized studies (9,10). Echocardiography, however, is a highly user-dependent imaging modality and the reproducibility of dyssynchrony metrics is limited (11,12). Cardiac magnetic resonance (CMR) imaging has recently emerged as a highly valuable modality providing unparalleled image quality and information on etiology for patients with heart failure (6); furthermore, it can delineate the location and burden of myocardial scar through late gadolinium enhancement sequences (13) and provide valuable data on mechanical dyssynchrony (14) and LV contraction patterns (15), both of which can inform appropriate placement of the LV lead. We have previously demonstrated the ability to overlay CMRderived anatomy, scar, and dyssynchrony data onto fluoroscopy for guiding the placement of the LV lead (16,17). This resulted in improved acute response and chronic echocardiographic response above the rates in a standard nonguided approach. However, the large quantity of data, associated lengthy computational processing time, and highly manual process, as well as software limitations at that time meant CMR scans were performed at least 2 weeks before the implant.
The ability to process and display such information in real time would represent a significant advance in the ability to use CMR guidance. In this paper, we report the first demonstration of a real-time, purpose-built user interface in a hybrid cardiac catheter laboratory and magnetic resonance imaging scanner in the same facility with the ability to transfer patients from one to the other (X-MRI) facility to enable the processing, analysis and fusion of CMR-derived data to guide the implanting physician in optimizing the deployment of an LV lead for CRT delivery. Using this technique, the implanting physician can use contemporaneous gold standard myocardial imaging to avoid regions of scar while targeting late activating segments, thereby permitting imaging-guided LV lead implantation in a single procedure. The platform was installed on a dedicated prototype workstation connected to the biplane x-ray system. It was designed for fast, automated data processing to enable information to be extracted from MRI and visualized in the time it takes to transfer the patient from the MRI to angiography suite to minimize any disruption. The platform includes automatic intra-and inter-MRI protocol slice registration, automatic LV segmentation, semiautomatic scar segmentation and metrics for mechanical dyssynchrony, scar distribution, burden, and transmurality. After each automated task, the clinician is given the opportunity to verify and manually adjust the results.

METHODS
PATIENT FLOW AND CRT IMPLANT. Respiratory and cardiac-gated CMR images were acquired. Two-, 3-, and 4-chamber and multiple slice short-axis balanced steady-state free-precession images were acquired Real-Time CMR-Guided CRT Implantation -2 0 Healthcare GmbH) and prepared for CRT implantation, which began immediately after the scan.
Simultaneously, the CMR data were uploaded, processed, and analyzed on the prototype platform using the following steps ( Figure 1): 1. LV epicardial and endocardial automated segmentation with manual adjustment where necessary.
2. Registration of cine and LGE sequences.
3. Delineation of myocardial fibrosis.   Real-Time CMR-Guided CRT Implantation -2 0 1 7 : --degrees of freedom, translation, and rotation ensured the accuracy of the registration in orthogonal planes.
Following this process, the CMR-derived 3D model is registered to the x-ray coordinate system; subsequent x-ray acquisitions during the case are displayed with instantaneous overlay of the correctly oriented 3D model ( Figure 3).
Steps 1 through 4 were completed while the patient was being prepared for the implantation.     given the association between posterolateral wall scar and poor outcomes (7). CMR in comparison has the ability to identify and quantify both scar and dyssynchrony and may represent the optimal imaging modality for CRT guidance (6). A meta-analysis of 511 patients using STARTER, TARGET, and another prospective cohort study (20), confirmed the efficacy of an image-guided approach compared with conventional implant with a higher odds ratio of response to CRT (odds ratio: 2.1; 95% confidence interval:  Values are mean AE SD, n (%), or %. *Scar burden calculated using prototype platform.
ACE ¼ angiotensin-converting enzyme; BNP ¼ B-type natriuretic peptide; 3D ¼ 3-dimensional; MLHFQ ¼ Minnesota Living with Heart Failure Questionnaire; MRA ¼ magnetic resonance angiography; NYHA ¼ New York Heart Association; peak VO2 ¼ peak oxygen uptake on cardiopulmonary exercise testing.  Furthermore, in cases in which LV lead placement in or adjacent to scar was inevitable because of the distribution of coronary venous anatomy, the CMR scar mask was able to guide the positioning the poles (from the multipolar LV lead) away from islands of scar to achieve more favorable testing parameters.
LIMITATIONS AND CHALLENGES. This is a small proof of principle study; however, a larger, randomized controlled study would be necessary to demonstrate whether an image-guided approach is superior to the standard of care, which we plan to perform.
Given that 2 previous studies using echocardiogramderived markers (9,10) and 2 recent studies using multimodality imaging (27,28)  We anticipate future iterations of the platform to include scar analysis from LGE sequences with higher spatial resolution (higher number of slices or 3D scar) and calculation of dyssynchrony through a range of different metrics including cine sequence-derived strain, as described previously (27). In this study, of importance was that the spatial and temporal resolution were the same as for routine scanning; however, more elaborate and refined sequences may be able to be incorporated in the near future. Last, we intentionally studied a predominantly ischemic population, given that they have the most to gain from a technique that may be able to improve CRT response. Response rates in nonischemic patients are indeed significantly higher; we would therefore advocate studying a larger number of these patients to evaluate the platform from the perspective of image guidance using dyssynchrony parameters alone.
Coronary venous anatomy has been described using 3D whole heart sequences (electrocardiogramtriggered, respiratory-navigated, steady-state freeprecession inversion recovery) applied to the whole heart over a short period (e.g., 60 to 80 ms), obtaining spatial resolution of 1.5 Â 1.5 Â 2 mm (30). However, this process adds at least 30 min of scanning time and, furthermore, the coronary veins require segmentation to extract onto the 3D mesh model, which can be time consuming and limit the applicability of data utilization in real time. Furthermore, heart failure patients may have worsened breathing when lying supine; therefore, we wanted to keep scanning time to a minimum, before the implant, which also requires them to lie flat.  (32). Furthermore, cardiac CT may provide a feasible alternative to CMR; measures of strain have shown to correlate between the 2 imaging modalities in an animal model (33) and scar can be demonstrated (34). We envisage using a similar guidance platform with CT data in a similar way in the future.
Given the expense of CMR in some centers, this technique may not be cost-effective; however, the data could be processed offsite and fed back to the core site. The current technique of single-sitting CMR and implant requires an X-MRI laboratory, which is not available in all centers. It is envisaged, however, that in the future, with improved image registration techniques not requiring fiducial markers, the CMR platform could be used in an offline manner with scanning and implantation in separate locations and times.
Finally, it is notable that only 71% of patients were able to have the LV lead delivered to the target segment; this is similar to our previous study of CMR guidance (15). This usually occurs from the lack of a coronary vein subtending the target segments. This may be viewed not as a limitation but as an advantage of the system. One may argue that in those cases in which the target segments are not subtended by viable epicardial veins, the conventional epicardial approach may be switched to an endocardial LV approach at the outset. Endocardial pacing has been shown to be of use in nonresponders to conventional CRT (35); however, the optimal site of stimulation varies greatly between patients (4,36,37). The value of such a system may be to both identify patients in whom such an approach is required and to then use the system to perform targeted endocardial LV lead implantation. Recently, we have shown that such a targeted approach for endocardial LV stimulation using CMR guidance to avoid scar results in improved hemodynamic response (38).