Microwave Breast Cancer Imaging: Simulation, Experimental Data, Reconstruction and Classification

This work concerns the microwave imaging (MWI) for breast cancer. The full process to develop an experimental phantom is detailed. The models used in the simulation stage are presented in an increasing complexity. Starting from a simplified homogeneous breast where only the tumor is placed in a background medium, moving to an intermediate complexity model where a rugged fibroglandular structure other than tumor has been placed and reaching a realistic breast model derived from the nuclear magnetic resonance phantoms. The reconstruction is performed in 2D using the linear TR-MUSIC algorithm tested in the monostatic and multistatic approaches. The description of the developed phantom and the instruments involved are detailed along with the already planned improvements. The simulated and experimental results are compared. Finally a classification stage based on the leading technique known as “deep learning”, an improved branch of the machine learning, is adopted using mammographic images.