Prediction of EM Field in a Simple Homogenous Phantom at UHF MRI Using Physics-Informed Neural Networks (PINNs): Methodology in Data Generation

Motivation: Specific Absorption Rate (SAR) calculation is the most crucial safety analysis at ultra-high-field (UHF) 7T MRI. Current SAR computation methods rely on computationally intensive simulations, which are often impractically long for real-time clinical use. Goal(s): This study aims to develop a physics-informed neural network (PINN) capable of predicting electromagnetic (EM) field distribution at 7T MRI. Approach: A neural network is trained using data generated from EM simulations. One of Maxwell's equations is implemented as a physical constraint within the neural network to improve the accuracy of the field prediction. Results: Introducing physics into neural networks enhances EM field prediction. Impact: This study proposes a deep learning-based method for EM field prediction, which, by significantly reducing the computational time, can enable safer and more accessible 7T MRI.