Microwave Imaging for the Detection and Localization of Breast Tissue Malignancies Using Time Reversal Beamforming

Abstract

Breast cancer is the most common cancer suffered by Australian women. Early detection of cancer provides the best chance of survival to the victims. Microwave imaging has shown the potential to be a complimentary imaging modality to the existing breast cancer imaging techniques such as mammography, MRI and ultrasound. Microwave imaging can overcome the drawbacks of conventional imaging techniques such as patient discomfort and ionizing radiation hazard. The principle of microwave imaging for breast cancer detection is based on the dielectric property contrast between healthy breast tissues and malignant tissues. However, in dense breasts that have high amounts of dense fibro-glandular tissue content, the dielectric property contrast between tumor and surrounding healthy glandular tissues can be quite low. To overcome the problems arising from imaging in low contrast scenario, contrast enhancing agents and hybrid imaging modalities have been proposed in the literature. But, such complex modalities not only complicate the screening process but also add to patients discomfort and cost. Moreover, such techniques may still fail to detect multiple tumors unambiguously in highly dense breasts.

In this thesis, we investigate the use of computational time reversal imaging techniques for breast cancer detection and localization using anatomically realistic numerical breast phantoms. Both radar imaging and tomography imaging techniques have been applied for breast cancer detection. Microwave tomography cannot detect abrupt change in dielectric properties when contrast is low. On the other hand radar imaging can reveal the target location information even under low contrast scenario but suffers from clutter and noise in the medium. Time reversal microwave imaging can be considered to be a variant of radar imaging. Time reversal uses medium heterogeneity as an advantage and is highly suitable for imaging in heterogeneous medium. However, the performance of time reversal can also be affected by low dielectric property contrast between target and surrounding tissue clutter. To overcome the effects of clutter interference on target detection and localization, in this thesis, we propose novel beamforming techniques for time reversal microwave imaging. Firstly, we extend beamspace processing for time reversal imaging technique with an aim to reduce clutter effects and improve robustness of imaging. However, when we use ultra-wideband microwave pulses for imaging, a coherent approach is necessary to overcome problems due to random phase variations arising in each frequency bin. We propose two different novel coherent beamspace time reversal imaging techniques for breast cancer screening. The focusing matrix based coherent signal subspace processing is found to be more suitable for subspace and maximum likelihood based time reversal imaging techniques whereas the focusing matrix based on wavefield modelling method is found more suitable for time reversal minimum variance imaging. We propose to combine coherent focusing with beamspace processing (C-B) to obtain superior imaging localization performance. We have also derived Cramer Rao Lower Bound (CRLB) for beamspace time reversal imaging. We have proposed Coherent beamspace DORT (C-B-DORT), CB-TR-MUSIC, C-B-TR-RCB, C-B time reversal maximum likelihood (C-B-TR-ML) methods to detect small single and multiple tumors in highly dense breasts where conventional techniques are prone to fail. Our investigations have revealed that C-B-TR-ML imaging has superior performance compared to other techniques in detecting three small sized tumors embedded in a highly dense breast phantom.