Enhancing Precision in Window to the Brain Modeling: Methodology and Implementation of Hybrid Digital Twins

[EN] The Window to the Brain (WttB) is a novel cranial implant designed to enhance therapeutic procedures involving brain tissue. Previous computational models studying the effectiveness of the WttB exhibited some discrepancies with experimental results and inconsistencies in certain parameter value...

Descripción completa

Detalles Bibliográficos
Autores: Llamazares-López, Marcos, Trujillo Guillen, Macarena|||0000-0003-4145-2188, Cortés, J.-C.|||0000-0002-6528-2155, Villanueva Micó, Rafael Jacinto|||0000-0002-0131-0532
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:dnet:riunet______::3868e2c2cc5fd13581f791bcbe22efb2
Acceso en línea:https://riunet.upv.es/handle/10251/234917
Access Level:acceso embargado
Palabra clave:Hybrid digital twins
Particle swarm optimization
Grammatical evolution algorithm
Mathematical computer modeling
Photo-thermal laser effect
Window to the brain
Descripción
Sumario:[EN] The Window to the Brain (WttB) is a novel cranial implant designed to enhance therapeutic procedures involving brain tissue. Previous computational models studying the effectiveness of the WttB exhibited some discrepancies with experimental results and inconsistencies in certain parameter values. To overcome these drawbacks, the following steps are followed. We first perform a domain reduction where the model is solved via the finite element method. Then, model parameters are calibrated using asynchronous random Particle Swarm Optimization (arPSO) algorithm. A statistical identifiability analysis is performed to evaluate how accurately model parameters are estimated based on the quantity and quality of experimental data. Afterward, we implement Hybrid Digital Twins (HDT) using Grammatical Evolution and Lexicase Selection to improve the model fitting keeping the model complexity. The outcomes demonstrate a complete alignment between experimental and computational results, as well as reasonable values for all model parameters. The final optimized model achieved a mean absolute error of 0.1871, with a standard deviation of 0.0013 and a 95% confidence interval (CI) of [0.1866, 0.1876], indicating a very low residual error and high stability across simulations. Our computational approach enhances the results from previous studies, which can be more useful for improving clinical practice.