Comparison of Reconstruction Algorithms for Brain Stroke Microwave Imaging

Abstract

The aim of this paper is to describe and compare the performances of three image reconstruction algorithms that can be used for brain stroke microwave imaging. The algorithms belong to the class of non-linear iterative algorithms and are capable of providing a quantitative map of the imaged scenario. The first algorithm is the Contrast Source Inversion (CSI) method, which uses the Finite Element Method (FEM) to discretize the domain of interest. The second one is the Subspace-Based Optimization Method (SOM) that has some properties in common with the CSI method, and it also uses FEM to discretize the domain. The last one is the Distorted Born Iterative Method with the inverse solver Two-step Iterative Shrinkage/Thresholding (DBIM-TwIST), which exploits the forward Finite Difference Time Domain (FDTD) solver. The reconstruction examples are created with 3-D synthetic data modelling realistic brain tissues with the presence of a blood region, representing the stroke area in the brain, whereas the inversion step is carried out using a 2-D model.