Convergence and stability assessment of Newton-Kantorovich reconstruction algorithms for microwave tomography

For newly developed iterative Newton-Kantorovitch reconstruction techniques, the quality of the final image depends on both experimental and model noise. Experimental noise is inherent to any experimental acquisition scheme, while model noise refers to the accuracy of the numerical model, used in the reconstruction process, to reproduce the experimental setup. This paper provides a systematic assessment of the major sources of experimental and model noise on the quality of the final image. This assessment is conducted from experimental data obtained with a microwave circular scanner operating at 2.33 GHz. Targets to be imaged include realistic biological structures, such as a human forearm, as well as calibrated samples for the sake of accuracy evaluation. The results provide a quantitative estimation of the effect of experimental factors, such as temperature of the immersion medium, frequency, signal-to-noise ratio, and various numerical parameters.