Continuous neural networks and finite element application for the tissue deformation reconstruction dynamic

This paper presents the nonparametric modeling based in differential neural networks (DNN) of soft tissue deformation dynamic under a single external pressure force. The construction of the DNN-adaptive model is based on the finite element method (FEM), the proposal is to make that every element be approximate by a DNN. The DNN input is taken by the nodes information collected from real experimental data captured from a variable-velocity electro-mechanical platform applying a single-point force to a tissue sample, in this way, an assembled DNN is used to join the element DNNs to obtain the complete system modeling. To verify the qualitative behavior of the suggested methodology, here the estimated trajectories are compared with the Motion Capture spatial position vector of the surface of the sample tissue. The adaptive laws for weights ensure the closeness of DNN trajectories to the tissue dynamics.