Multimodality digital twins for cardiac arrhythmia created by anatomic and electrophysiological data fusion

Abstract Background Patient-specific digital twins (DTs) are a promising tool for understanding cardiac arrhythmias, testing therapies and predicting outcomes. Although they usually incorporate anatomical data, they often lack personalised cellular models, despite the fact that individual electrophysiology is crucial for the effective management of AF, among other arrhythmias. Purpose The aim of this work is to develop DTs for AF patients that include both anatomical and electrophysiological data. We hypothesize that merging data from anatomical and electrophysiological modalities will enable the development of more realistic and predictive digital twins. Methods We created twelve electrophysiologically-informed DTs (EP-DTs) of the left atrium (LA) by fusing the CT-based geometry with electrophysiological and local biomarkers (fig. 1). Local measures of cycle length (CL) and conduction velocity (CV) were obtained from intracavitary electrical recordings and converted into simulable parameters, such as electrical remodelling and diffusion, via a calibration study. Multiple simulations of the AF condition were initiated using pacing protocols (fig. 2a) and the ability of the EP-DTs to reproduce clinical biomarkers was evaluated. The EP-DTs were also compared with state-of-the-art DTs constructed using only anatomical information (anatomical DTs). The EP-DTs and Anatomic-DTs were then used to identify proarrhythmic regions and were matched against areas where ablation had acutely terminated AF. Results EP-DTs closely matched patient CL and CV values both globally and locally (e.g., CL: 178 ms vs. 172 ms in the LA appendage; 220 ms vs. 206 ms in the anterior wall; CV: 293 mm/s vs. 272 mm/s in the LA floor; 477 mm/s vs. 462 mm/s at the LA appendage base). EP-DTs also captured inter- and intra-patient variability more effectively than anatomic-DTs, significantly reducing mean errors in CL (0.14 ± 0.07 vs. 0.23 ± 0.09, p < 0.05) and CV reproduction (0.37 ± 0.12 vs. 0.54 ± 0.15, p < 0.05). Lastly, EP-DTs demonstrated greater specificity in identifying proarrhythmic regions (Fig. 2b–c), revealing a smaller area of AF-driving regions (9 ± 8% vs. 13 ± 7%) and exhibiting comparable sensitivity in identifying fibrillatory drivers whose ablation enabled the immediate termination of AF. Conclusion EP-DTs characterised using intracavitary CL and CV measurements more accurately reflect individual arrhythmia phenotypes. This has the corresponding benefit of creating more accurate digital twins, and therefore, could aid diagnosis and the selection of personalised therapies.WorkflowEP-Digital Twin