Adaptive Combining via Correlation Exploration for Ultrawideband Breast Cancer Imaging

Abstract

A novel Adaptive Combining via Correlation Exploration (ACE) algorithm of ultrawideband (UWB) imaging for breast cancer detection is proposed. ACE explores and exploits the correlation between backscattered signals and local coherence reference signals generated within each group of neighboring antennas. High-correlation signals are adaptively selected, summed, and weighted by the product of their corresponding coefficients, forming the intensity of each pixel within an imaging area. The efficacy of proposed algorithm is validated on 3-D anatomically and dielectrically accurate finite-difference time-domain (FDTD) breast models. A variety of scenarios, for both homogenous and heterogeneous, sparse and moderately dense breast models, coupled with both ideal and practical artifact removal methods, are considered. The superior performance of ACE in identification of malignant tumors is demonstrated in comparison to delay-and-sum, delay-multiply-and-sum, and filtered delay-and-sum algorithms. The ACE algorithm is shown to be not only immune to early-stage artifact, but also capable of distinguishing responses from cancerous and fibro-glandular tissues, in which the other three techniques suffer significantly or completely fail. This shows the high promise of ACE for breast cancer screening.