Evaluation of prior information in microwave tomography experiments for brain stroke detection

This work examines the impact of prior information (or “initial guess”) for calibrating a microwave tomography system for brain stroke detection and differentiation, using a multi-layered, anatomically complex head phantom. The imaging algorithm applies the distorted Born iterative method (DBIM) combined with the two-step iterative shrinkage thresholding (TwIST) method. The initial guess for the algorithm is based on two models with different available information: one filled with the dielectric properties of average brain tissue, and one with a more accurate representation of the true head phantom. Our initial results demonstrate that the addition of thin head tissue layers (such as CSF) in the forward model is not critical for the successful reconstruction of the target’s dielectric properties. As expected, however, we achieve more accurate results with the multi-layer initial guess in challenging cases such as detecting an ischemic stroke-like target in the presence of a six-layers Zubal head phantom.