Brain stroke is the second cause of death and the third of disabilities,for this it requires timely diagnosis and treatment to reduce the risk of death or further damages.The most adopted clinical techniques are Computed Tomography and Magnetic Resonance Imaging,but,although accurate and with high resolution,they are expensive,not portable,not comfortable and not suitable for a continuous monitoring,essential after a stroke onset.Moreover,the Computed Tomography with ionizing radiation as X-Rays is even harmful and not appropriate for repeated examinations.This leads to the necessity to develop a diagnostic device based on Microwave Imaging(MWI)technology,which is low cost,not harmful(it uses non-ionizing radiation),not invasive and usable for bedside monitoring and frequent examinations.MWI exploits the fact that the stroke has different dielectric properties(relative permittivity and conductivity)respect to brain healthy tissues at MW frequencies and,evaluating this dielectric contrast,it is possible to obtain an image as the location of brain stroke.A MWI prototype for brain stroke monitoring,realized at the LACE is composed by 24 printed monopole antennas,placed conformal to a head phantom,filled with a liquid mixture whose dielectric properties are those of an averaged brain at MW frequencies.Each antenna,immersed in a brick of graphite-silicon(coupling medium),works as a transmitter to illuminate the phantom and as a receiver to collect the field scattered by the head.A 24x24 switching matrix and a Vector Network Analyzer measure the 24x24 scattering matrix useful for the image reconstruction.The image is obtained through a differential approach:the difference between two scattering matrices related to two different time instant(one with a target-stroke inside the phantom and one without it)is in input to a reconstruction algorithm based on the Born approximation and the truncated singular value decomposition,so that the inverse scattering problem is solved and the contrast reconstructed.This thesis focuses on the experimental assessment of this system through a measurement campaign,with different types of targets to simulate the presence of a stroke.First measurements have been made with a small plastic ball with a precise geometry,easier to locate inside the phantom:although inaccuracies(especially in matching the right position),the system shows the ability to image the object.Next acquisitions have been made with a target whose dielectric properties are the same of the blood so it could simulate a hemorrhagic stroke;an immediate case is a balloon filled with a liquid,which mimics the blood,inserted in the phantom,while a more refined target is a cylindrical case,put in different positions of the head.The images obtained with the cylindrical-case are less noisy than the ones with the balloon and the target is more realistic and easier to manage;also in these cases,the device is able to see the object,even if with difficulty in catching the right position.The limits reside in the spatial resolution and in a sensitivity to perturbations, as temperature or measurement noise.Calibration procedures are investigated,placing two of the 24 antennas(reference channel) external to the phantom,to understand if their signal,not dependent on the target,can help to correct the measurements.Next possible steps:smaller target,simulation of ischemic stroke, simulation of monitoring(rotating the target).Final challenge:3D wearable MWI helmet..