Accounting for uncertainty in computational models of ventricular tachycardia improves ablation guidance.

Personalized 'digital twin' technology has the potential to revolutionize target guidance for catheter ablation therapy of ventricular tachycardia (VT); however, concerns regarding the practical implementation of these computationally intensive approaches, along with the robustness of simulation predictions given uncertainty in the interpretation and analysis of clinical imaging data used to reconstruct image-based models, are hindering clinical translation. Here we present a new technical framework that provides near-real-time guidance on VT ablation targets, inherently incorporating uncertainty in reconstructed scar anatomy. We demonstrate close agreement between simulated ECG 'fingerprint' signatures of VT circuits and clinical recordings in ischemic ablation patients, highlighting the need to account for variations in reconstructed scar anatomy to identify the best-matching simulated circuit. Finally, we introduce a robust method for integrating anatomical information across all viable simulated circuits from different model 'instances' into a simulated ablation target heatmap for practical clinical guidance.