Microwave breast tumor detection exploiting wideband Jacobians

Gradient-based iterative algorithms are preferred in microwave tomography due to their fast convergence despite the fact that the solution is only local. The drawback is that response Jacobians are not available with current commercial electromagnetic solvers and reconstruction algorithms have to be equipped with in-house codes specifically designed to produce response derivatives in addition to the responses themselves. Even with in-house codes, the computation of the response Jacobians in a wide frequency band is a challenge. Here, we present an efficient approach to the computation of wideband response Jacobians with time-domain solvers. Through this approach, the Jacobian distributions (maps) of imaged regions can be computed with relatively small memory requirements and negligible computational overhead. This self-adjoint computation is reduced to a simple post-process of the field solution which can be applied with any commercial FDTD-based solver. We show that the detection of scatterers such as breast tumors can be achieved with a single system analysis by computing the wideband Jacobian maps of high-fidelity patient-specific tissue models.