Predicting in-silico ventricular tachycardia induction on fuzzy cardiac digital twins

Ventricular tachycardia (VT) is a leading cause of sudden cardiac death (SCD). To date, clinical VT risk assessment is either inaccurate, invasive or expensive. Patient-specific computer modeling of heart electrophysiology (EP) is a promising non-invasive risk assessment approach. It has three major limitations: computational cost, inaccurate arrhythmogenic substrate geometry and unknown patient-specific tissue EP properties. We present a pacing outcome predictor (POP) that quickly identifies the conditions needed to induce VT in three-dimensional personalized cardiac computer models. The POP is used to test VT induction across multiple model versions of a patient, covering scar geometry and EP property variations, with an added layer of model reduction to further accelerate calculations. The results are combined into a personalized VT risk metric based on in-silico VT inducibility across the patient's models. In a retrospective 10-patient repaired Tetralogy of Fallot cohort, the metric showed complete separation (10/10) between patients with and without clinically inducible VT, suggesting potential for non-invasive VT risk assessment in this population.