Model-based inverse scattering for microwave breast imaging: An analysis of tissue feature sensitivity versus model error sensitivity

Quantitative microwave imaging based on inverse scattering shows promise for breast cancer risk assessment, detection, and treatment monitoring due to its low cost, non-ionizing nature, and fully three-dimensional imaging potential. Solving for the unknown tissue dielectric properties distribution is commonly achieved using iterative model-based inverse scattering algorithms. These algorithms function by comparing measured scattered fields to scattered fields calculated from a forward simulation. The underlying assumption is that the best match between simulated and measured signals will occur when the simulated model of the object, i.e. the breast, has the same dielectric properties as the true object, and that differences between the measured and simulated signals are dominated by differences between the actual and estimated properties of the object. For any array, uncertainty in the material properties of the “known” environment outside the object represents an additional source of signal discrepancy, and thus it is of interest to evaluate the desired sensitivity to breast tissue features relative to undesired sensitivity to uncertainty in the properties of the environment.