Physical knowledge improves prediction of EM Fields

2025-03-12

AI Summary

This study addresses the challenge of high-fidelity spatial prediction of electromagnetic (EM) fields inside human-body radiofrequency (RF) coils at 7 Tesla MRI. We propose U-Net Phys, a physics-informed deep learning model that for the first time incorporates magnetic flux continuity (∇·B = 0) as a finite-difference–based regularization term directly into the U-Net loss function, enabling end-to-end integration of physical constraints with data-driven learning. The model takes as input RF coil excitation parameters (phase, amplitude, position) and tissue-specific EM properties (permittivity, conductivity, density). A 3D U-Net architecture is trained on CST-simulated data, incorporating multi-physics parameterization and finite-difference regularization. Experimental results demonstrate that U-Net Phys reduces mean relative error in internal EM field prediction by 32% compared to the standard U-Net, while substantially improving generalizability and physical consistency—particularly adherence to Maxwell’s equations.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Low Level & Physics-based Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We propose a 3D U-Net model to predict the spatial distribution of electromagnetic fields inside a radio-frequency (RF) coil with a subject present, using the phase, amplitude, and position of the coils, along with the density, permittivity, and conductivity of the surrounding medium as inputs. To improve accuracy, we introduce a physics-augmented variant, U-Net Phys, which incorporates Gauss's law of magnetism into the loss function using finite differences. We train our models on electromagnetic field simulations from CST Studio Suite for an eight-channel dipole array RF coil at 7T MRI. Experimental results show that U-Net Phys significantly outperforms the standard U-Net, particularly in predicting fields within the subject, demonstrating the advantage of integrating physical constraints into deep learning-based field prediction.

Problem

• Predicts electromagnetic fields in RF coils with subjects
• Integrates physical laws to enhance deep learning accuracy
• Improves field prediction within subjects using physics-augmented U-Net

Innovation

• 3D U-Net predicts EM fields using coil data
• U-Net Phys integrates Gauss's law for accuracy
• Trained on CST Studio Suite simulations for 7T MRI