Clinical validation of simulated VT morphologies with clinical ECGs demonstrates the potential of in-silico approaches for ablation guidance

Abstract Background Personalized computational 'digital twin' approaches integrate detailed structural data from patient imaging with functional physiological models, enabling the simulation of realistic cardiac electrophysiological dynamics. These models can simulate and identify all potential ventricular tachycardia (VT) circuits, including those that may be missed during the ablation procedure, as well as helping pinpoint optimal ablation targets, ensuring effective treatment. Despite their promise for guiding ablation, in-silico algorithms lack rigorous validation against clinical data, particularly in simulated ECG morphologies, limiting clinician and patient confidence. Purpose To clinically validate our novel in-silico VT ablation guidance algorithm by directly comparing with clinical arrythmias via ECGs, permitting robust and rapid assessment of the validity of predictions. Methods Commercially available software (ADAS3D) was used to generate scar maps from 3D wideband LGE-CMR images (Figure A) and mapped over to a personalized CT based biventricular computational model (Figure B). A reaction-Eikonal based tool for Virtual Induction and Treatment of Arrythmias (VITA) was implemented to identify all possible reentrant pathways that can sustain potential clinical VTs within the ventricular model (Figure C). Corresponding ECGs were simulated, utilizing whole-torso CT data to identify exact locations of ECG electrodes in the patient. VT ECGs that were simulated from VITA were compared with clinical VTs (Figure E), recorded following VT induction in the lab (n=3 different morphologies) and the mean correlation of the highest 10 leads noted. To account for uncertainty in the clinical data and ensure the closest fit of model to clinical data, different pixel intensity thresholds in the LGE data were swept-over, and the simulation pipeline repeated, to identify the effect of variations on correlation with clinical ECGs (Figure F). A summary of the pipeline is shown in Figure 1. Results The pipeline successfully identified all scar-related VT circuits, and was able to provide rapid clinical validation through comparison with clinical ECGs. The number of unique VT circuits identified ranged between 11-17. The highest mean correlation varied between 0.749 – 0.907, and an increase in mean correlation values was seen when the ADAS scar core threshold was increased from 0.6 to 0.7. However, an increase in the scar core threshold beyond 0.7 generally did not result in higher mean correlation values. An increase in the border-zone threshold from 0.3 to 0.5 did not contribute to a considerable change in mean correlation values. Conclusion The pipeline developed allows the rapid validation of all possible VT circuits identified using VITA with clinical ECGs. Close agreement was seen with the peak mean correlation value of 0.907 between clinical and simulated ECGs, demonstrating the potential of VITA for use within a clinical workflow to aid ablation guidance.Figure 1 Table 1