Brain injuries represent a critical situation, where both detection and monitoring should be quick and accurate at the same time. Microwave techniques are thus gaining attention in the diagnostic process of these diseases. However, the detection of inhomogeneities and variations inside the human brain by using electromagnetic fields at microwave frequencies is a very challenging inverse problem. An innovative hybrid microwave imaging method is introduced in this contribution, which combines the benefits of a fast qualitative processing technique with an accurate tomographic reconstruction of the dielectric properties of the human head. This method has been successfully applied to obtain microwave images from both synthetic data and laboratory measurements. Numerical simulations involve three-dimensional realistic models of stroke-affected heads, whereas simplified cylindrical phantoms have been exploited for the experimental validation of the approach. In both conditions, the proposed technique yields promising results, which may be considered a preliminary step towards the realization of a clinical imaging prototype.