Investigation of noise effect on image quality in microwave head imaging systems

The effect of noise on the ability to detect strokes in the human head using microwave-based systems is investigated using both simulations and measurements. The simulations, which are implemented using the full-wave electromagnetic solver (CST), utilise a realistic numerical head model generated from magnetic resonance imaging slices. On the other hand, the experiments are performed using a wideband microwave system with a frequency range of 1-4 GHz. The experiments use a head phantom consisting of materials that accurately emulate (within 3% of reported properties of human brain tissues) the main tissues found in the human brain. It is shown, in both of the simulations and measurements, that a minimum signal-to-noise ratio of around 10 dB is required to accurately detect the presence of a stroke. Below this level, the microwave-based system either does not detect or falsely indicates the location of the stroke. Based on the presented results, the required microwave power to achieve acceptable detection is well within the recommended safe levels.