Tomographic Microwave Imaging With Incorporated Prior Spatial Information

We have implemented a soft prior regularization technique for microwave tomographic imaging that exploits spatial prior information from alternative imaging modalities. We have previously demonstrated in both simulation and phantom experiments that this approach is capable of improved property recovery compared with our conventional Tikhonov regularized method. An important concern with this type of integration is that the spatial information could be implemented in such a way as to overly influence or even bias the final results. For this reason, we have performed a rigorous quantitative analysis of this approach using both simulation and phantom experiments to investigate the sensitivity to inaccurate or even false spatial information. The majority of cases tested here involved simple targets to easily assess problems when the prior shapes are incorrect with respect to size and location or even when there is an extra target that does not exist in the actual imaging situation. In addition, we have also performed a simulated experiment utilizing anthropomorphic breast structural information to explore the capabilities in more challenging situations. The results are encouraging and demonstrate that the soft prior regularization can be a powerful tool, especially where the goal is specificity instead of sensitivity.