Deep Learning For Tumor Margin Identification In Electromagnetic Imaging

This paper focuses on adapting a new architecture, DeepTMI, to iterative optimizations utilized for nonlinear electromagnetic inverse scattering problems. The proposed method is a learned technique which divides the electromagnetic images into two regions, i.e. tumor and normal, with high accuracy and approximately characterizes the multi-scattering physical mechanism. The performance of the proposed method is validated by an experimental demonstration, showing that the method is a promising tool for efficiently tackling nonlinear inverse scattering problems.