MRI-guided Ventricular Tachycardia Ablation in Patients With Structural Heart Disease

BACKGROUND AND RATIONALE

Sudden cardiac death (SCD) remains a major global public health burden, and ventricular tachycardia (VT) and ventricular fibrillation account for the majority of cases. In patients with structural heart disease (SHD), VT typically arises from re-entrant circuits located within or adjacent to myocardial scar. Ischemic cardiomyopathy (ICM) is usually characterized by dense infarct-related scar, whereas non-ischemic cardiomyopathy (NICM) is characterized by more diffuse fibrosis. Radiofrequency catheter ablation has become a cornerstone therapy for reducing VT recurrence and implantable cardioverter-defibrillator (ICD) interventions in both ICM and NICM patients. However, important challenges persist: long-term success rates remain modest (40-60%), the procedure carries a risk of complications, and conventional three-dimensional electroanatomical mapping (EAM) has inherent limitations. Current EAM strategies rely on voltage-based surrogates of scar, which lack the spatial resolution to delineate complex three-dimensional scar architecture and may fail to identify critical arrhythmogenic substrates, particularly during hemodynamically unstable VT.

Cardiac magnetic resonance imaging (MRI) with late gadolinium enhancement (LGE) has emerged as the gold standard for three-dimensional scar characterization. By exploiting differences in gadolinium washout kinetics between normal and scarred tissue, LGE-MRI distinguishes viable myocardium from scar with high accuracy and spatial resolution and has proven value for arrhythmic risk stratification. Pre-procedural LGE-MRI can precisely localize the scar core and border zone (BZ), the key substrates of re-entry; yet in current workflows these data are rarely used for real-time procedural guidance. Emerging evidence suggests that integrating MRI with EAM-guided substrate ablation may overcome the limitations of voltage mapping by improving ablation lesion precision and reducing unnecessary tissue injury. Whether an LGE-MRI-assisted strategy further improves ablation success in SHD-related VT, however, remains unproven. This randomized controlled trial evaluates whether a VT ablation strategy based on pre-procedural LGE-MRI scar analysis and intraprocedural real-time image registration can further reduce VT recurrence and improve long-term outcomes in patients with SHD-related, scar-dependent VT.

STUDY OBJECTIVES AND HYPOTHESIS

The study aims to demonstrate the efficacy and safety of LGE-MRI-integrated EAM-guided catheter ablation compared with conventional EAM-guided catheter ablation in patients with SHD and documented sustained VT. The investigators hypothesize that MRI-guided ablation will reduce the 12-month composite of VT recurrence, cardiovascular hospitalization, or death, without increasing procedure-related complications.

STUDY DESIGN AND RANDOMIZATION

This is a prospective, multi-center, single-blind, randomized controlled trial. All patients with documented VT in whom the underlying cause is suspected cardiomyopathy (i.e., scar-related VT) undergo cardiac MRI screening. Patients with positive LGE findings are enrolled and randomly assigned in a 1:1 ratio using stratified centralized block randomization via an automated telephone system: the experimental group receives LGE-MRI-image-integrated EAM-guided ablation, and the control group receives conventional EAM-guided ablation without image integration, with operators/mappers blinded to the LGE-MRI findings. Patients with LGE-negative MRI scans are enrolled in a separate ablation registry cohort for prospective observational analysis (to be published separately); comparison of this registry with the randomized patients serves as an additional prespecified outcome measure. A total of 526 patients will be enrolled (263 per group).

CARDIOVASCULAR MAGNETIC RESONANCE (CMR) PROTOCOL

All gadolinium-enhanced CMR examinations are performed on a 1.5 T scanner (Magnetom Avanto, Siemens Healthineers, Erlangen, Germany) with electrocardiographic gating and respiratory motion compensation. Before imaging, all patients undergo standardized breath-hold training to minimize motion artifacts. The standardized protocol sequentially acquires: (1) transverse and sagittal black-blood T2-weighted images for anatomical localization; (2) balanced steady-state free precession (bSSFP) cine sequences in long-axis (two-, three-, and four-chamber) orientations and contiguous short-axis slices (8 mm thickness, 2 mm gap) covering the entire left ventricle (LV) for functional assessment; and (3) phase-sensitive inversion recovery (PSIR) Turbo FLASH sequences in long-/short-axis orientations matched to the cine images for LGE acquisition. Post-processing with commercial software (CVI42 v5.14.1, Circle Cardiovascular Imaging) quantifies LV structural and functional parameters and LGE extent, defined as myocardial regions with signal intensity exceeding 6 standard deviations above normal tissue - a validated threshold that correlates with histopathological fibrosis and ensures reproducibility. LGE extent is expressed as the percentage of enhanced myocardium relative to LV mass, and enhanced myocardium at the ventricular insertion points is systematically included in the quantification. Semi-automated contouring with manual correction mitigates partial-volume effects and artifacts, maintaining interobserver variability below 5%.

ADAS-3D IMAGE RECONSTRUCTION

The ADAS-3D platform (ADAS-3D Medical, Barcelona, Spain) processes CMR datasets through an integrated workflow combining machine-learning algorithms with operator-supervised refinement to generate patient-specific three-dimensional ventricular models with tissue characterization. After DICOM import, the software performs automated segmentation of LV endocardial and epicardial borders using deep-learning-based edge detection; operators then validate and manually correct the segmentation via an interactive interface to ensure anatomical fidelity. For patients with cardiac implantable electronic devices (CIEDs), in whom CMR image quality may be compromised, manual segmentation is used. The validated contours drive automatic construction of a three-dimensional mesh model of the entire LV, subdivided into nine transmural layers (10-90% wall thickness) to quantify fibrosis distribution across myocardial depth. Tissue characterization is achieved by coregistering LGE pixel signal intensity (PSI) maps onto the three-dimensional geometry using a standardized color schema. Operators interactively adjust BZ/scar thresholds based on histogram analysis of regional signal distribution (using 40% ± 5% and 60% ± 5% of maximum intensity as thresholds for the border zone and scar core, respectively), after which the software computes quantitative substrate parameters including total scar mass (g), dense scar volume (cm3), border-zone percentage (%), and the number and morphology of border-zone channels. On the PSI maps, heterogeneous tissue channels (HTCs) are defined as continuous corridors of border zone surrounded by scar core, or by scar core plus an anatomical barrier, connecting two areas of healthy tissue. HTCs are identified automatically by the ADAS-VT software and classified as sub-endocardial (10-50% transmural depth), sub-epicardial (60-90%), or transmural (10-90%). The reconstructed three-dimensional electrophysiological model is exported in STL format for intraprocedural integration with the electroanatomical mapping system, enabling dynamic visualization of the arrhythmogenic substrate during ablation.

ABLATION PROCEDURES

Standard placement of coronary sinus and right ventricular reference electrodes is performed in all patients. Three-dimensional EAM is guided by the EnSite X system using an HD-GRID catheter or a contact-force ablation catheter. Voltage mapping is performed in all patients during sinus rhythm, and ablation is performed with a saline-irrigated radiofrequency ablation catheter. Programmed or incremental ventricular stimulation is used to induce the clinical VT, continuing until the ventricular refractory period is reached or VT is induced.

Control group (conventional EAM-guided ablation): For patients with reproducibly inducible, hemodynamically stable VT, activation mapping and entrainment mapping are performed to identify the critical isthmuses of the re-entrant circuit, and ablation targets these isthmuses. For patients with hemodynamically unstable or non-inducible VT, a substrate-mapping strategy during sinus rhythm is adopted. The arrhythmogenic substrates of SHD-related VT in this study comprise low-voltage areas, fractionated potentials, and late potentials; substrate-based ablation aims to eliminate all such abnormal potentials. In patients with inducible clinical VT, abnormal potentials are additionally ablated to reinforce the lesion set.

Experimental group (LGE-MRI-integrated EAM-guided ablation): In addition to all of the above EAM-guided strategies, CMR information is used to focus mapping on the areas of interest (scar and HTCs on the PSI maps) to facilitate identification of ablation targets. Radiofrequency energy is delivered at the entrances of the HTCs identified on the PSI maps, regardless of whether the targeted sites display pathological electrograms. This integrated strategy leverages multimodal imaging to improve the precision and effectiveness of ablation.

Procedural endpoints: Post-ablation ventricular stimulation that fails to induce any VT is defined as successful ablation; induction of non-clinical VT is considered partial success; and persistent inducibility of clinical VT is defined as ablation failure. For patients whose VT could not be induced before ablation, the procedural endpoint is elimination of the abnormal potentials identified by EAM or CMR. ICD implantation is recommended for patients who meet guideline criteria.

POST-ABLATION MANAGEMENT AND FOLLOW-UP

All patients are monitored in hospital for at least 24 hours after the procedure. Outpatient follow-up evaluations are conducted at 3, 6, and 12 months. At each follow-up, patients wear a wearable smart ECG monitoring device for at least 7 consecutive days to detect VT recurrence; in patients with an ICD, device interrogation data are used for recurrence monitoring.

SAMPLE SIZE AND STATISTICAL ANALYSIS

The sample size is based on the 12-month composite primary endpoint. Based on published data, the estimated event rate is 0.35 in the MRI-guided group and 0.50 in the control group. With a two-sided alpha of 0.05, 90% power, and 1:1 allocation, 446 patients are required; allowing for a 15% dropout rate, the final sample size is 526 patients (263 per group).

Baseline demographic and clinical characteristics will be reported. Categorical variables will be summarized as frequencies/percentages and continuous variables as means with standard deviations or medians with interquartile ranges, as appropriate. Baseline comparisons will use chi-square and t-tests. Survival analysis will use the Kaplan-Meier method with the log-rank test. Multivariable and machine-learning models (including XGBoost, random forest, linear and logistic models, and Cox proportional-hazards models) will be used to determine independent associations between outcomes and influencing factors. Model performance will be evaluated using positive and negative predictive values, the C-statistic, calibration plots, decision-curve analysis, and the net reclassification improvement (NRI), with the DeLong test used to compare model performance. All statistical tests are two-sided with p < 0.05 considered significant. Analyses will be performed with R (version 4.1.3) and SAS 9.4.

ADVERSE EVENT REPORTING

All adverse medical events occurring during the study and follow-up are recorded and reported. Serious adverse events - including death, permanent or severe disability or loss of function, and events requiring hospitalization or prolongation of hospitalization - are graded according to NCI-CTCAE version 5.0 and reported to the clinical events committee and the ethics committee in accordance with local regulations. The clinical events committee adjudicates the primary endpoint events and safety endpoint events.