High-fidelity postmyocardial infarction ventricular tachycardia simulation for intraprocedure ablation guidance

Postinfarction ventricular tachycardia (VT) arises from reentrant circuits within partially conductive myocardial tissue. Despite antiarrhythmic medication, many patients require ablation therapy, yet recurrence rates remain as high as 50% at 5 years. Virtual-heart arrhythmia ablation targeting (VAAT) methods aim to improve outcomes by simulating VT activity preoperatively, but they are computationally intensive and often lack validation against clinical VT and electrocardiographic (ECG) data.

This study aimed to optimize a VAAT pipeline for speed and clinical relevance, enabling real-time use during ablation and validating its predictive accuracy against experimental VT and ECG data.

We developed a digital twin framework incorporating semiautomatic anatomic model generation, a simplified Mitchell-Schaeffer action potential model, a graphics processing unit-accelerated lattice-Boltzmann solver for the monodomain equation, fast ECG computation via the boundary element method, and semiautomated VT inducibility analysis. The framework was applied to 9 swine models of postmyocardial infarction VT across 3 stages: before ablation, after ablation, and 1-week follow-up.

The digital twin predicted VT inducibility in all clinically inducible cases and reproduced VT morphologies with polarity match in ≥11 of 12 ECG leads in 15 of 15 cases (12 by inducibility, 3 by pace mapping). It also identified 5 nonclinical VTs not inducible in vivo, suggesting its ability to uncover potential future VTs. A reduced parameter set enabled simulation of all unique VT morphologies while reducing execution time by 90%, supporting feasibility for intraoperative use.

The optimized VAAT pipeline achieved robust predictive performance and significant computational gains, potentially enabling real-time assessment of ablation completeness and enhanced procedural guidance.