Microwave Imaging Evaluation of Prior Structural Information on the Inversion-Kernel Building Apply to a Brain Stroke Monitoring Scenario

This work investigates the impact of prior structural information on the inversion kernel used to follow up on a brain stroke condition. To that end, we perform a numerical study that mimics an intracranial hemorrhage and aims to retrieve the morphological evolution of the stroke-affected area between different time instants via direct inversion based on the Born Approximation and the truncated singular value decomposition. Then, we consider different operators, imaging kernels, adding tissue shape information, and evaluating the imaging retrieval performance via the structural similarity index, the dice similarity coefficient, the normalized Hausdorff distance, and a sizebased metric, similarity metrics. The results confirm that more apriori information improves overall performance; however, more importantly, they show that even with approximated kernels, less information, and a more realistic clinical scenario, the imaging might perform well enough as a medical indication.