Gêmeos digitais do coração para disfunções de condução ventricular

Abstract

The objective of this work is to validate a procedure for generating cardiac digital twins from clinical data and to propose an optimization algorithm based on Differential Evolution (DE) that reduces the computational cost in comparison with the traditional Bayesian approach. The procedure integrates geometric reconstruction from magnetic resonance imaging, fiber orientation, incorporation of a Purkinje network, and simulation of ventricular activation through the Eikonal model, allowing the calibration of physiological parameters, such as conduction velocities and early activation sites, with approximation of electrocardiogram (ECG) signals. Parameter inference is performed, by default, using Bayesian methods of the SMC-ABC type, and the quality of the fit between simulated and clinical ECGs is evaluated using metrics such as the Pearson correlation coefficient (PCC) and the root mean square error (RMSE). The study uses data from four patients from the University Hospital of the Federal University of Juiz de Fora, including two with left bundle branch block, one with masked block, and one without bundle branch block. The validation involves controlled stimulation protocols and sinus rhythm, analyzing the ability of the model to reproduce clinically observed patterns. For the proposed method, inference times were evaluated for different DE population sizes. The results show that Differential Evolution recovers activation patterns and conduction velocity values compatible with those obtained with SMC-ABC, while maintaining similar PCC and RMSE levels and simultaneously reducing inference time by up to 2.5 times. It is concluded that the procedure is robust for representing alterations in ventricular conduction and that DE constitutes a promising alternative for accelerating the personalization of cardiac digital twins in clinical applications.