Title:Physics informed neural network for forward and inverse modeling of low grade brain tumors

Abstract:A low grade tumor is a slow growing tumor with a lower likelihood of spreading compared to high grade tumors. Mathematical modeling using partial differential equations (PDEs) plays a crucial role in describing tumor behavior, growth and progression. This study employs the Burgess and extended Fisher Kolmogorov equations to model low-grade brain tumors. We utilize Physics Informed Neural Networks (PINNs) based algorithm to develop an automated numerical solver for these models and explore their application in solving forward and inverse problems in brain tumor modeling. The study aims to demonstrate that the PINN based algorithms serve as advanced methodologies for modeling brain tumor dynamics by integrating deep learning with physics-informed principles. Additionally, we establish generalized error bounds in terms of training and quadrature errors. The convergence and stability of the neural network are derived for both models. Numerical tests confirm the accuracy and efficiency of the algorithms in both linear and nonlinear cases. Additionally, a statistical analysis of the numerical results is presented.