Assessment and Development of Microwave Imaging for Breast Cancer Detection

At the Technical University of Denmark (DTU), a 3D tomographic microwave imaging system is currently being developed with the aim of using nonlinear microwave imaging for breast-cancer detection. The imaging algorithm used in the system is based on an iterative Newton-type scheme. In this algorithm, the distribution of the constitutive parameters is updated in each iteration based on a comparison between the measured signals and the signals computed by a full-wave electromagnetic solver for the assumed distribution of parameters.
In this work, a study on development and improvement of the imaging algorithm used in the microwave tomographic imaging system is presented.
Non-linear microwave tomographic imaging of the breast is a challenging computational problem. The breast is heterogeneous and contains several high-contrast and lossy regions, resulting in large differences in the measured signal levels. This implies that special care must be taken when the imaging problem is formulated. Under such conditions, microwave imaging systems will most often be considerably more sensitive to changes in the electromagnetic properties in certain regions of the breast. The result is that the parameters might not be reconstructed correctly in the less sensitive regions.
In this study, a combination of two different methods for obtaining a more uniform sensitivity throughout the breast are investigated. The first is a reformulation of the minimization problem. Here, three different formulations are investigated, namely the real-imaginary formulation, the log-phase formulation, and the relative formulation. The second method for obtaining uniform sensitivity throughout the imaging domain is a scaling of the parameters to be reconstructed. This scaling, based on the norms of the columns of the Jacobian, is also introduced as a measure of the sensitivity of the algorithm. The reformulation of the problem and the scaling of the parameters are shown to improve performance of the microwave imaging system – both when applied for reconstruction of images from 2D simulated data and when used on measured data from a 3D imaging system.
Additionally, the use of multiple frequencies in the nonlinear microwave algorithm is considered. Using multiple frequencies allows for obtaining the improved resolution available at the higher frequencies while retaining the regularizing effects of the lower frequencies. However, a number of different challenges arise when using data from multiple frequencies for imaging of biological targets. The performance of a nonlinear microwave tomography algorithm is tested using simulated data from anatomically realistic breast phantoms. These tests include several different anatomically correct breast models from the University of Wisconsin-Madison (UWCEM) repository with and without tumors inserted.