Tumor Margin Identidication Based on Contrast Values in Millimeter-wave Imaging

This work proposes a novel method for tumor margin identification in millimeter-wave imaging in which a priori information about the statistical distribution of the dielectric properties of tissues is employed. In prior work, we presented a new millimeter-wave imaging modality that visualizes tumor profiles with high contrasts through the depth of skin with low cost. In this paper, we add the capability of estimating the margin of the tumor based on the contrast in millimeter-wave images. We propose a novel level set method in which the image function takes only two constant values. A learning-based method is employed to estimate the level set and corresponding reflectivity profile. Synthetic data are generated for objects within a 3-D imaging cavity and used for model training and testing. Furthermore, the method is tested on a numerical phantom. The reconstructed image indicates that the method can produce accurate estimates of object location, shape, and size while recovering the contrast with an error less than 5%.