From Imaging-Guided to Digital Twin–Assisted Ventricular Tachycardia Ablation: Translational Integration of Anatomical and Functional Substrate Characterization

Ventricular tachycardia (VT) in the setting of structural heart disease remains a major clinical challenge. Despite progress in ablation technology, recurrence rates remain substantial, largely due to incomplete understanding and modification of the arrhythmogenic substrate. Traditional electroanatomical mapping (EAM) provides valuable functional information but suffers from limited spatial resolution, operator dependency, and procedural complexity. Conversely, advanced cardiac imaging offers exquisite structural detail yet lacks direct functional insight. Bridging these two domains is essential to improve the efficacy, safety, and reproducibility of VT ablation. This doctoral research explores a translational continuum—from anatomical imaging to computational modeling—aimed at enhancing both substrate characterization and ablation guidance. The work is structured in four main steps. First, the VOYAGE clinical trial evaluated imaging-guided ablation strategies integrating cardiac magnetic resonance (CMR) into the procedural workflow. Two imaging-based approaches were analyzed: a CMR-guided arm, in which ablation relied exclusively on imaging-derived scar channels, and a CMR-aided arm combining imaging with conventional EAM. The results demonstrated that both strategies are feasible, safe, and efficient, reducing procedural time and radiofrequency delivery while allowing satisfactory success rates. Second, cardiac computed tomography (CT) was investigated as an alternative imaging modality for substrate definition, focusing on the detection of lipomatous metaplasia (LM)—a marker of fibrofatty tissue replacement within chronic myocardial scar. In a cohort of patients undergoing first VT ablation, LM outperformed wall thinning and CT channels in predicting key electrophysiological features such as deceleration zones (DZs) and hidden slow conduction electrograms (HSC-EGMs). LM emerged as the only independent predictor of functional substrate localization, establishing CT as a practical and widely available tool for pre-procedural planning. Third, a digital twin model (CardioMat) was developed and validated to simulate patient-specific ventricular activation directly from LGE-CMR data. Using GPU-accelerated computation, CardioMat generated three-dimensional simulated activation maps. When compared with invasive ILAM recordings, simulated DZs showed high sensitivity and specificity in predicting functional substrate, confirming the ability of digital twins to reproduce patient-specific conduction patterns non-invasively. Finally, an in-silico ablation experiment was conducted using the same patient models to compare two therapeutic strategies: targeted DZ-only ablation versus extensive scar homogenization. Both approaches significantly reduced simulated VT inducibility. While scar homogenization achieved maximal non-inducibility, DZ-only ablation reached similar efficiency with one-third of the lesion volume, highlighting the potential of functional, imaging-informed strategies to optimize tissue preservation and procedural safety. Collectively, these studies demonstrate the feasibility and translational value of integrating multimodal imaging and computational modeling in VT ablation. The research establishes a conceptual and technical bridge from imaging-guided to model-guided electrophysiology, paving the way toward real-time digital twin–assisted procedures that could revolutionize pre-procedural planning, operator training, and personalized therapy in cardiac electrophysiology.