Reduction of the model noise in non-linear reconstruction via an efficient calculation of the incident field: application to a 434 MHz scanner

Microwave tomography has been drastically boosted by the development of efficient reconstruction algorithms based on an iterative solution of the corresponding non-linear inverse problem. The accuracy of the electric field radiated by the antennas of a microwave scanner, inside the target area, has been shown to play a significant role on the overall image quality. Taking into account the antenna environment is of prime importance, especially when operating at low frequency. For instance, the wall of a 60 cm diameter whole-body microwave scanner cannot be neglected at 434 MHz, even when using the immersion technique consisting of putting the target in water. Indeed, at such a frequency, the attenuation introduced by water is not sufficient to avoid multiple reflections on the scanner boundary walls. Consequently, the method of calculating the incident field constitutes a key factor in iteratively solving non-linear inverse problems. The selected technique must accommodate high accuracy while maintaining acceptable calculation complexity. In this paper, three distinct techniques are analysed. They are based on the use of i) free-space and ii) non free-space Green's function, and iii) a FDTD approach. All these techniques have been firstly investigated for their 2D version, being used in 2D reconstruction algorithms. However, the scattered field data are collected in a 3D scanner. For assessing the validity of the previous 2D techniques, their results have been compared to both experimentally and 3D-FDTD results.