Two-step reconstruction process for microwave tomography without a priori information

Paper in proceeding, 2016

While our log transform reconstruction algorithm provides significant advantages over competing algorithms because of its ability to converge to a unique, global solution without utilizing a priori information, the process still requires a modest amount of regularization which naturally tends to smooth the image and sometimes blur important features. We have integrated this overall technique with a novel 2-step reconstruction to produce significantly improved images with respect to feature definition without inducing unwanted artifacts. The critical challenge in this scenario is determining the regularization parameter for the Euclidean distance penalty term in the second step. In this case, because the algorithm is quite fast, the regularization term is computed using a simple linear search. Initial phantom experiment results indicate that targets as small as 5mm diameter are detectable and that the algorithm behaves as an efficient estimator according to standard parameter estimation metrics for a wide range of targets. Results also suggest that there are equally significant improvements in reconstructions from actual patient exams. © 2016 European Association of Antennas and Propagation.