Assessment and Optimization of a Microwave Imaging System for Brain Stroke Detection and Monitoring

Brain stroke is a leading cause of death and a significant contributor to adult disability. To reduce its damaging effects resulting from brain cell death, it is fundamental to start appropriate treatment as soon as possible, making an early and accurate diagnosis critical. Ideally, the diagnosis should be rapid and, if feasible, performed directly in the ambulance or at the trauma site. Microwave Imaging (MWI) can be an alternative to traditional diagnostic techniques like Computed Tomography and Magnetic Resonance Imaging, which both have drawbacks. MWI is safe due to the non-ionizing nature of the electromagnetic waves used, making it suitable for ongoing monitoring. Additionally, MWI Systems (MWISs) can be portable, enabling use in ambulances or directly at trauma sites, and relatively cheap. MWI for brain strokes relies on the dielectric contrast between materials at around 1 GHz, allowing it to differentiate between healthy and pathological tissues. This thesis aims to assess and optimize a MWIS designed for the detection and monitoring of brain stroke. It consists of an antenna array placed around a head, a Vector Network Analyzer (VNA) producing and receiving the signal, and a switching matrix directing the signal of the VNA to the antennas and vice versa. A laptop functions as the user interface, managing the system's operation and executing the imaging algorithms. The algorithm used resolves a non-linear and ill-posed inverse scattering problem, utilizing the measured scattering parameters as input. The MWIS in analysis employs the distorted Born approximation and the Truncated Singular Value Decompositions technique to solve it. The research activity begins with a brief review of the public reception of MWI in medical applications, as well as the refinement of system parameters and components. It then focuses on designing custom components for the head phantom used for the experiments, developing a user-friendly application for its operation, and evaluating its performance through experimental validation on a realistic head phantom. Initially, two types of switching matrices are compared: an electromechanical matrix and a solid-state one. The comparison aims to determine the optimal setup for MWIS's portability, considering both scanning speed and imaging accuracy. Next, the effects of the VNA settings, in particular the intermediate frequency filter and the number of averaged measurements, are analyzed to identify the best measurement speed and dynamic range trade-off. Then, the design and development of a custom-fit cup for the head phantom and ad-hoc containers for test objects that simulate strokes are described. These create a more controlled environment for mimicking diagnosis and monitoring tasks. Finally, the focus is on the development of the Wavision desktop interface, which offers an intuitive experience to both expert and non-expert users. It allows them to easily compare the impact of the parameters on the measurements, integrates various tasks, and ensures parameter control. In conclusion, this thesis represents a significant step toward the real-world application of the MWIS prototype for stroke detection and monitoring. Future research should prioritize trials with human volunteers to facilitate the inclusion of MWI into clinical practice.