Speaker
Description
Recent advances in cardiac electrophysiological modelling, coupled with modern computational technology, enable fast in silico simulation of ventricular tachycardia (VT), showing strong potential for applications in arrhythmia risk stratification and therapy planning, such as catheter ablation. However, accurate patient-specific model personalisation remains a major challenge, limiting simulation fidelity and hindering clinical translation. In this work, we develop a multimodal personalisation framework that combines cardiac structural information derived from computed tomography (CT) with electrical characteristics extracted from 12-lead electrocardiograms (ECGs).
We studied six patients undergoing electrophysiological studies. Sinus rhythm 12-lead ECG signals were extracted from EP recordings and used as the objective for parameter optimisation. Pre-ablation CT scans were used to construct patient-specific biventricular simulation domains and to estimate ECG lead positions. Building on prior work in which in silico VT induction was achieved using EP models personalised from CT-derived myocardial wall thickness (WT), we implemented a two-stage optimisation framework consisting of early activation onset estimation followed by optimisation of EP model parameters. Following early-onset estimation, the simulated 12-lead ECG signals achieved an average QRS peak accuracy of 91.20%. With optimised parameters, the average QRS duration error was reduced to 3.94 ms, compared with 37.42 ms using baseline parameters. The complete in silico VT induction pipeline was subsequently applied to patients with adequate electroanatomical VT mapping data (n = 4). Using the personalised parameters, clinically observed VT patterns were successfully reproduced, with VT cycle lengths closer to the recorded values than those obtained with baseline parameterisation. More importantly, in-silico VT induction succeeded in one patient for whom it had previously failed when using baseline parameters.