Validation of In-silico Pace Mapping for Guiding Ventricular Tachycardia Ablation in Clinical Practice

Fernando Campos1, Ursula Rohrer2, Karli Gillette3, Ali-Razak Rashid2, Iulia Nazarov2, Janneke Burger4, Pranav Bhagirath5, Luca Azzolin6, Aurel Neic7, Christopher Aldo Rinaldi2, Gernot Plank8, John Whitaker2, Martin Bishop2
1School of Biomedical Engineering and Imaging Sciences, King's College London, 2King's College London, 3Gottfried Schatz Research Center - Medical University of Graz, 4Amsterdam University Medical Center, 5Kings College London, 6Karlsruhe Institute of Technology, 7NumeriCor GmbH, 8Medical University of Graz


Abstract

Background: In-silico pace mapping uses patient-specific ventricular-torso computational models and ECG recordings to generate virtual maps that can assist in pre-procedural planning for ventricular tachycardia (VT) ablation. Despite its promise as a non-invasive tool, its clinical effectiveness remains to be validated. This study sought to evaluate the ability of in-silico pace mapping to localize focal activation sites during clinical pacing.

Methods: Personalized ventricular-torso models were created for 15 patients with ischemic and non-ischemic cardiomyopathy based on contrast-enhanced CT and late gadolinium-enhanced cardiac MRI. During clinical ablation procedures, 12-lead ECGs were recorded during ventricular pacing at multiple sites (mean 12 ± 9 per patient), and corresponding locations were extracted from the electroanatomical map-ping system to serve as ground truth. Simulated ECGs were then generated by pacing at 1,000 evenly distributed sites on the ventricular endocardium within each model. For each clinical pacing site, the similarity between recorded and simulated ECGs was quantified using correlation analysis, and the virtual site with the highest correlation was considered the model's predicted origin. Electrophysiological simulations were carried out with CARPentry (NumeriCor GmbH, Graz, Austria).

Results: In the representative case shown in the Figure, the model-predicted site showed a correlation coefficient of 0.90 and was located 12.2mm (yellow star) from the true pacing site (red star). The median distance for all clinical pacing sites for this patient was 18mm (IQR: 10.4-21.9mm, N = 5). Across all 12 patients, the median spatial error was 19.5mm (IQR: 17.2–26.2mm).

Conclusions: These findings demonstrate that in-silico pace mapping can estimate the site of ventricular activation with clinically relevant precision. With further validation, this approach could support VT ablation procedures by offering a rapid, non-invasive planning strategy to guide targeted therapy.