Improved Neighborhood-Based Algorithm to Facilitate the Reduction of Skin Reflections in Radar-Based Microwave Imaging
Michael R. Smith
Elise C. Fear
Low power, near-field (NF) radar imaging techniques have been proposed for breast cancer detection and long-term monitoring. It is important to optimize the data processing paths required for NF image reconstruction given the inherent resolution limitations of microwave compared to MRI or X-ray imaging. A key limitation in obtaining internal tumour and breast feature information is the reflection from the skin surface physically close to the antenna. Typically, algorithms to remove this dominant reflection involve subtracting an estimate of the time domain signal for the skin reflection from one antenna location using information from other locations. A key challenge in these approaches is determining the portion of the signal, the skin dominant window (SDW), to use to determinethe weights applied to nearby antenna signals when calculating the skin reflection estimate. Equipment limitations and breast characteristics impact the amount of data that can be captured, leading to the well-known Gibbs' ringing distortionsin the time domain signals. We suggestthat the Gibbs' ringing from the magnitude larger skin reflection has caused the length of the SDW to be over-estimated in previous determinations. Since this distorted signal now overlaps the time signals from the tumour and breast responses, removing the skin reflection estimatemay result in attenuation of tumour responses. In this contribution, two alternative strategies for designing the SDW are proposed. One minimized the first skin peak in the SDW, i.e., the furthest from the breast feature signals, and the other minimized the main, i.e., largest, skin peak within the SDW. Both new approaches were shown to effectively suppress the skin signal on simulated and patient data while allowing recovery of the missing portions of the desired internal breast feature signals leading to an increase in the overall intensity of the images and preserving the tumour response. However, we provided reasons why we considered that basing the suppression on the largest skin signal peak would provide a more consistent improvement in the breast feature signals.
