Confocal microwave imaging and artifact removal algorithms for the early detection of breast cancer

Abstract

Microwave imaging is an emerging imaging modality for the early detection of breast cancer. Early-time artifact removal and imaging algorithm are the two most important signal processing components of any Confocal Microwave Imaging (CMI) system. The artifact removal algorithm reduces the large undesired early-time reflections from the breast skin that could potentially mask the tumour response. The imaging algorithm generates images of the breast such that the tumour is a strong scatterer and clutter due to healthy breast tissues is suppressed. In this thesis, artifact removal and imaging algorithms have been investigated.

Several artifact removal algorithms for CMI along with an algorithm adapted from the Ground Penetrating Radar (GPR) have been evaluated in terms of their ability to reduce the artifact, while preserving the tumour response. The results from a comparative study have led to the development of a novel hybrid artifact removal algorithm that combines the best features of two existing artifact removal algorithms. The Hybrid Artifact Removal (HAR) algorithm has been shown to effectively reduce the early-time artifact while preserving the tumour response in 3D numerical breast phantoms. The HAR algorithm is then extended to a multistatic data acquisition approach. The proposed Multistatic Artifact Removal (MAR) algorithm has been shown to reduce the early-time artifact in selective multistatic signals, which improves the overall imaging quality compared to monostatic-only imaging.

Since different CMI prototypes use different scan configurations, the HAR algorithm, along with the Neighbourhood-based Skin Subtraction (NSS) algorithm, have been applied to the most common scan configurations used in CMI prototypes. Both algorithms have been shown to successfully reduce artifacts and produce similar quality images across all scan configurations examined. The NSS algorithm demonstrated better artifact suppression than the HAR. However, the HAR algorithm demonstrated better tumour response preservation particularly in experimental breast phantoms that were scanned with the patient-specific scan configuration of the second-generation TSAR prototype. Finally, a variety of imaging algorithms have been compared across patient data obtained from a small-scale patient study at the University of Calgary. The Delay Multiply and Sum (DMAS) imaging algorithm has been shown to provide the best quality images, with correct detection and localisation of breast lesions in most cases.