Description
Traction force microscopy (TFM) is an important tool to measure the force transmitted between the cell and the external microenvironment. However, calculating the stress from the displacement of markers is a challenging task because it is an ill-posed inverse problem. Most of the TFM experiments to date thus are performed on a two-dimensional flat geometry, which is usually solved by incorporating the theory of linear elasticity with regularization. Nonetheless, neural network-based machine learning has been shown to be a promising alternative for solving such inverse problems. Here, we propose a workflow to perform three-dimensional TFM by a machine learning based approach, which combines physics-informed neural network and the finite element method to solve the equations of elasticity. Specifically, the implementation of the training dataset and the boundary conditions associated with the three-dimensional TFM setup are clarified in this proposal.
