Microwave Characterization of Brain Stroke Through a Mild Data-Driven Inversion Strategy

The role of microwaves in the early diagnosis of brain stroke is rapidly emerging. Several systems have been developed, and some of them are already under clinical evaluation. However, the quantitative imaging of the dielectric properties of the head is an extremely challenging task, and a continuous improvement in inversion algorithms is necessary to address the clinical needs. This paper proposes a mild datadriven strategy to solve the inverse scattering problem for microwave brain stroke characterization. The approach aims at joining the benefits of data-driven inversion and a Newtonbased solution procedure. Through a set of numerical simulations, promising dielectric reconstructions of cross sections of healthy and stroke-affected heads are obtained.