Abstract
This work introduces advanced Physics Informed Neural Networks (PINNs) algorithms to model temperature distributions within the 2D breast model. It also proposes an enhanced prediction strategy employing PINNs for the computation of partial differential equations (PDE) featuring intricate initial and boundary conditions. Under the circumstances of fast progress of physics informed neural networks, benchmark analysis of PINN and FEM was run. Through the simulation of temperature distribution in breast tissues, the identification of potential abnormal areas and the indication of tumors have been achieved. The accuracy of the PINN methods, presented in this study, was validated by Finite Element Analysis (FEA), demonstrating good agreement of the two approaches within the range of free parameters for estimating temperature distributions on the skin surface of an actual breast model. This comparison validates the PINN approach and affirms its precision.