Machine-Learning based model order reduction of a biomechanical model of the human tongue

Background and Objectives: This paper presents the results of a Machine-Learning based Model Order Reduction (MOR) method applied to complex 3D Finite Element (FE) biomechanical model human tongue, in order create Digital Twin (DTM) that enables real-time simulations. The DTM is designed for future inclusion computer assisted protocol tongue surgery planning. Methods: proposed uses an “a posteriori” MOR allows, from limited number simulations with FE model, predict real time mechanical responses muscle activations. Results: evaluated associated separate single It shown be able account sub-millimetric spatial accuracy non-linear dynamical behavior observed these Conclusion: Further evaluations will include movements induced by multiple At this stage our offers promising perspectives use clinical context impact on mobility. As long term application, could used functional consequences terms speech production swallowing.