Simulation of Temperature Distribution During HIFU Therapy Using Physics Based Deep Learning Method

Abstract

Deep learning techniques has been employed recently to solve Partial Differential Equations (PDEs). A current approach known as Physics-Informed Neural Network (PINN), has evolved as a remarkable method to implement deep learning with the corresponding physics laws in the form of given linear or nonlinear PDEs. PDEs were commonly solved by using classical numerical methods like Finite Element Method or Finite Difference Method (FDM). However, it requires huge computational resources due to data set requirements, multiple dimensions or discretization. The solution of solving PDEs using PINN utilizes a mesh-free domain while still maintains high accuracy compared to conventional numerical methods. Comparing to FDM, PINN runs in less execution time with the same features and constraints. In addition, using PINN to estimate the solutions of PDEs can significantly reduce the tremendous discretized elements needed. In this paper, a PINN architecture is proposed, which employs the Bioheat Transfer Equation (BHTE) into a neural network to predict the temperature rise in a heterogeneous tissue. The thermal model simulates the heat conduction generated from the wave propagating from High Intensity Focused Ultrasound (HIFU) transducer.