Evaluation of Automata-Based Simulations for Atrial Fibrillation in 2D/3D Geometries Reproducing Disease Progression

• Main Authors: Romitti, Giada S., Liberos, Alejandro, Termenon-Rivas, Maria, de Arcaya, Javier Barrios-Alvarez, Romero, Pau, Serra, Dolors, Garcia-Fernandez, Ignacio, Lozano, Miguel, Sebastian, Rafael, Rodrigo, Miguel
• Format: Conference Proceeding
• Language: English
• Subjects: Biological system modeling; Computational modeling; Geometry; Mathematical models; Rhythm; Solid modeling; Three-dimensional displays
• ISSN: 2325-887X
• DOI: 10.22489/CinC.2023.243

Atrial biophysical simulations demand high computational resources due to the large number of equations to be solved and the small space and time discretization, which make them impractical for clinical use. Cellular Automaton (CA), by taking a finite number of states, reduces computational time and can provide interesting insights for subsequent digital twin models. This study explores a CA that successfully simulates atrial electrophysiology in patho-physiological conditions, when compared with biophysical simulations. CA was trained at cellular and tissue level from biophysical simulations in tissue slabs under controlled pacing conditions. Then, trained CA was used to simulate complex AF patterns in a 2D sheet of atrial tissue and a complete 3D atrial geometry that were compared against biophysical simulations. In the former, self-sustained arrhythmia under different degrees of electrophysiological remodeling gave a mean difference error of 6.25\pm 2.87ms in cycle length. In the 3D geometry, the value of local activation times differed by 9.6\pm 7.6ms for sinus rhythm and provided comparable patterns for chronic AF. Results exhibit near-real-time computational speed enhancements. These findings indicate CA models' efficiency in replicating patient-specific atrial electrophysiology in clinical practice.