Learning Nonlinearity of Microwave Imaging Through Deep Learning

The inherent presence of nonlinearity in microwave imaging makes the inversion problem challenging. To address this issue, we propose to use a machine learning method to learn the nonlinearity. In particular, a deep-learning based method is employed to correctly estimate the total electric field from an approximation that is generated by a Born-approximated solution. The learned total field is then used to estimate permittivity and conductivity. The applicability of our method is demonstrated for a 2D breast microwave imaging problem. The results show that the proposed method can predict the total electric field with high fidelity to the true total electric field and then can improve contrast recovery.