Physics-Informed Regularization for Microwave Imaging in Biomedical Applications

Microwave tomography is a promising imaging modality in which the dielectric properties of an unknown object are reconstructed quantitatively. Microwave tomography requires solving the non-linear and ill-posed inverse scattering problem. A priori information is typically required to regularize the problem and generate useful images. In this work, electromagnetic power balance is introduced as a physics-informed regularizer. Electromagnetic power balance is incorporated with the conventional data mismatch cost function to produce a new multiplicative regularizer. The technique is validated with low loss dielectric cylinders and a simplified forearm model in simulation.