Abstract
In this work we study the limits for the applicability condition and a method to improve the structural reconstruction algorithm for breast lesions detection and characterization reported in [1]. The strategy shown in this reference permits to detect and characterize the tumors of 1 cm diameter situated deeply inside the dense breast if the contrast between dielectric permittivity of tumor and surrounding tissue is at least 1.5. However, in some cases [2], experimentally, it has been reported that the contrast can be lower. Therefore, we consider cases in which low contrast leads to more difficult reconstructions. Our study of limitations of algorithm shows that by mean of additional regularization by using prior average values of dielectric parameters in human breast, even in case of contrast 1.2, the detection of 1 cm tumors is robust. Our contribution is to show that statistical criteria during the tumor search in cases when the average dielectric properties are not available, leads to localization of tumor and its separation from reconstruction ‘ghosts’. © 2011 American Institute of Physics.
