A personalized real-time virtual model of whole heart electrophysiology

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): This research has also received funding from the European Union’s Horizon 2020 research and innovation programme under the ERA-NET co-fund action No. 680969 (ERA-CVD SICVALVES) funded by the Austrian Science Fund (FWF), Grant I 4652-B to CMA. This project has also received funding MedalCare 18HLT07 from the EMPIR programme co-financed by the Participating States and from the European Union's Horizon 2020 research and innovation programme. Background Biophysics-based Cardiac Digital Twin (CDT) models strive to encode known physics and physiology in mathematical equations and tune these to represent individual patients. Owing to their mechanistic nature, when calibrated with high-fidelity CDTs show high potential to aide in clinical diagnostics, treatment planning, and prognostics. Challenges in building CDTs limit uptake and widespread in the clinics. Generation of CDT models and their personalisation based on measurable clinical data (i.e. the 12-lead electrocardiogram (ECG)) must be performed fast and automated, to be applicable at a clinical scale, and CDTs must replicate a patient’s electrophysiology (EP) with high fidelity to offer predictive capabilities for clinical use. Purpose We aimed to create the first personalized real-time in silico model of whole heart EP capable of both replicating the 12-lead ECG of a single subject under healthy sinus rhythm and subsequently capturing the complex mechanisms of various cardiac diseases. Method We constructed an automated pipeline for the generation of a ventricular-torso model personalised according to the 12-lead ECG during normal sinus rhythm (NSR), fitted with a physiologically-detailed cardiac conduction system facilitating retrograde activation, and included atrial and atrio-ventricular dynamics. Predictive capabilities of the mechanistic model were probed by interrupting conduction in the left and right bundle branches (LBBB, RBBB), by creating accessory paths, and by pacing at the right-ventricular (RV) apex. EP simulations were carried out with real-time performance. Goodness of fit was assessed for sinus rhythm in the 12 lead ECG. Computed pathological ECGs were evaluated with standard clinical diagnostic criteria. Results The computed 12-lead ECG of the personalised whole heart EP model under NSR showed close correspondence with the measured 12-lead ECG of the male subject. The 12 lead ECGs under the various pathologies manifested most morphological features in agreement with diagnostic criteria. Conclusions We report on a personalized real-time in silico model of whole heart EP for a single male subject that closely replicated the 12 lead ECG. Simulated pathological showed most known diagnostic features, but some features were less pronounced as not all factors of the underlying pathology were considered. The importance of dilation or slower cell-to-cell conduction in clinical LBBB aetiology, for example, was not considered. While the CDT shows promise for patient care in a single subject, broader validation with clinical and experimental data is needed to demonstrate agreement between models and physical reality. The pipeline must also be applied to additional subjects of both sexes.