Deep Learning Enhanced Contrast Source Inversion for Microwave Breast Cancer Imaging Modality

Abstract

This study presents a deep-learning (DL) based contrast source inversion (CSI) algorithm for quantitative microwave breast cancer imaging. Inverse scattering analysis for quantitative dielectric profile reconstruction is promising for a higher recognition rate for cancer detection, especially for malignant breast tumors. We focus on CSI as a low complexity approach, and implement a deep convolutional autoencorder (CAE) scheme using radar raw-data, which enhances the convergence speed and reconstruction accuracy. Numerical tests using MRI-derived realistic phantoms demonstrate that the proposed method significantly enhances the reconstruction performance of the CSI.