Tailoring Process for the Regional Personalization of Atrial Fibrillation with a Novel Cardiac Model

Clara Herrero Martín1, Carlos Fambuena Santos2, Maria de la Salud Guillem Sánchez1, Andreu M. Climent1, Ismael Hernández-Romero3
1Universitat Politècnica de València, 2UPV, 3ITACA Institute, Universitat Politècnica de València


Background. Personalization of mathematical models has the goal to help in the identification of optimal antiarrhythmic therapies for each patient. Nevertheless, their need of high computational resources and long running times move them far away from real-time clinical practice. Methods. In this study, a novel cellular automata model is presented. This model captures the essential dynamics seen in real myocardium tissue while requiring low computational resources and short simulation times. Moreover, we present a tailoring process based on regional regression curves which allow the regional personalization of some of the automata parameters thus it can mimic a desired electrophysiological behavior. These regression curves linked the wished electrophysiological properties (the regional conduction velocities and the duration of action potentials) with their corresponding adjusted automata parameters. Results. Both the automata and the tailoring procedure have been compared against an already validated detailed-mathematical during diverse cardiac rhythms. A high performance was obtained for regular rhythms simulations (as sinus rhythm and Atrial Flutter (AFL) cases). During Atrial Fibrillation (AF) same general dynamics (similar regional location of singularity points) than the reference model were obtained. A systematic comparison between the detailed-model and the CA indicate that mean absolute relative error of CV were 5.56%, 11.73% and 37.96% for the sinus rhythm, AFL and AF models respectively. Regarding the mean absolute relative error of APD80, errors were 1.02% for sinus rhythm, 5.17% for AFL and 9.63% for A.F. Conclusion. This novel automata model and its personalization frame-work may be applied to produce fast and personalized simulations from clinical data.