ARTIFACT REMOVAL ALGORITHMS FOR MICROWAVE IMAGING OF THE BREAST

Abstract

One of the most important components of any Confocal Microwave Imaging (CMI) system for breast cancer detection is the early-stage artifact removal algorithm. The early-stage artifact is composed of the incident pulse combined with the re∞ection from the skin-breast interface and residual antenna reverberation, and must be removed from the received signal at each antenna before further processing can take place. If the early-stage artifacts are not removed, they could potentially mask energy re∞ected from shallow tumors located close to the surface of the skin, and also hinder the identiflcation of tumors located deeper within the breast. Many existing artifact removal algorithms are based on variants of the assumption that the artifact in a particular channel can be estimated and efiectively removed by creating a reference waveform. This reference waveform is typically based on the average of the artifact in all channels. The artifact in a particular channel is then removed by subtracting this reference waveform from the recorded signal. More sophisticated algorithms estimate the artifact in each channel as a flltered combination of all the artifacts, and have been shown to be more robust to normal variations in skin thickness. However, increased underlying dielectric heterogeneity, as highlighted by Lazebnik etal., could result in greater variation in the early-stage artifact, making the artifact removal process much more di-cult. In this paper, several existing artifact removal are examined in this context of increased dielectric heterogeneity, and based on these results, suggestions for future work are presented. More than 40,000 women die annually in the United States from breast cancer, making it the leading cause of death in American women. One of the most promising alternate breast imaging modalities is microwave imaging. The physical basis for microwave imaging is the dielectric contrast between the constituent tissues of the breast and cancerous tissue at microwave frequencies. The Confocal Microwave Imaging (CMI) approach involves illuminating the breast with a UWB pulse, recording the backscattered signals and then using these signals to identify and locate signiflcant dielectric scatterers within the breast. Regions of high energy within the resultant image may suggest the presence of tumours. However, recent studies have found the breast to be dielectrically heterogeneous. Signiflcantly, Lazebnik etal. (1) found a very signiflcant dielectric contrast between normal adipose and flbroglandular tissue within the breast. Comparison studies have examined the performance of several UWB beamforming algorithms in the dielectrically heterogeneous breast. However, no previous study has compared the performance of early-stage artifact removal algo- rithms. The early-stage artifact is composed of the input signal, the re∞ection from the skin-fat interface and any antenna reverberation. This artifact is typically several orders of magnitude greater than than the re∞ections from any tumours present within the breast. If the artifact is not removed efiectively, it could easily mask tumours present within the breast. In this paper sev- eral existing artifact removal algorithms are described and compared, before suggestions for future development are presented.