Microwave-based detection of breast cancer using the generalized likelihood ratio test

In this paper the generalized likelihood ratio test (GLRT) is applied to the problem of detecting breast cancer via ultrawideband microwave radar. We propose a model for the covariance matrix of the clutter due to breast tissue heterogeneity and show that the GLRT with clutter whitening easily detects and localizes tumors under diverse circumstances. We illustrate the performance of the GLRT and clutter whitening transformation with simulated data from a 2-D realistic numerical breast model. A statistical analysis of the GLRT under matched and mismatched conditions characterizes the power of the test and we provide examples of detectors with mismatch in tumor size and tumor location.