Impact of Increasing Antenna Model Complexity on Microwave Tomography Using DBIM

In microwave tomography (MWT), reconstruction accuracy is challenged by modeling error, namely the mismatch between the numerical representation and the actual experiment. Accurate antenna modeling is perceived as an important step toward reducing this error, but the actual benefit of increasing antenna model complexity has not been analyzed in the literature. This work fills this gap by conducting a rigorous numerical analysis of the issue using two popular algorithms for its study: the finite-difference time-domain (FDTD) method for antenna and forward-problem modeling, and the distorted Born iterative method (DBIM) for implementing the iterative inversion algorithm. We consider various FDTD tools of increasing complexity to improve the agreement between the FDTD forward solver and an accurate numerical model implemented in commercial software. After validating these models for different antennas, we perform reconstructions for a stroke-detection scenario. Our results show that in a practical setting, sophisticated antenna modeling in the forward solver does not necessarily improve reconstruction accuracy for monopole-type antennas widely used in MWT. Our model-error analysis confirms that calibration is always necessary in practice and that its impact supersedes efforts to model the antenna more faithfully.