Target Selection in Multistatic Microwave Breast Imaging Setup using Dielectric Lens Daniela M. Godinho * , Jo˜ ao M. Fel´ıcio †‡ , Carlos A. Fernandes ‡ , Raquel C. Conceic¸˜ ao * * Instituto de Biof´ısica e Engenharia Biom´ edica, Faculdade de Ciˆ encias da Universidade de Lisboa, 1749-016 Lisbon, Portugal, dgodinho94@gmail.com † Centro de Investigac¸˜ ao Naval (CINAV), Escola Naval, Almada, Portugal ‡ Instituto de Telecomunicac¸˜ oes, Instituto Superior T´ ecnico, Universidade de Lisboa, Lisbon, Portugal Abstract—Microwave Imaging (MWI) has been studied to aid early breast cancer detection. Current prototypes in more advanced stages of development include both monostatic or multi- static setups. However, multistatic configurations usually include a high number of antennas which consequently require complex and computationally-intensive signal processing algorithms to ensure a good target detection. We previously presented a novel approach using a dielectric lens which reduces the signal processing burden of multistatic setups, while ensuring good spatial resolution. In this paper, we evaluate this novel setup using an anatomically realistic breast phantom and its capability of selecting targets inside the breast. We show a successful detection of the targets using an artefact removal algorithm based on singular value decomposition when the Bessel beam is centered at the target location. Index Terms—bessel beam, breast cancer detection, dielectric lens, microwave imaging, multistatic setup, singular value decom- position. I. I NTRODUCTION Every year, more than 2 million new breast cancer cases are reported worldwide and this type of cancer is now the most common cancer among men and women [1]. Microwave Imaging (MWI) has shown potential to aid breast cancer detection and diagnosis [2], [3]. It has several advantages compared to currently used imaging modalities: it uses non- ionising radiation, it does not require breast compression and it is relatively low-cost [4]. Several radar MWI systems have been proposed in the last decades, either using monostatic (i.e. signals transmitted and received by the same antenna), or multistatic setups (i.e. signals transmitted by one antenna and received by the remaining ones). While monostatic systems may imply long acquisition times, the acquisition in multistatic systems may be faster. The latter type of systems may be susceptible to coupling effects between antennas which may hinder the target detection and therefore require more complex and computationally-intensive algorithms to allow a successful detection. In a previous study [5], we presented a feasibility study of a novel approach using a dielectric Bessel lens which illumi- nates the breast with a pencil beam, decreasing the coupling effects between the antennas. Promising results were obtained with a spherical breast phantom and an ideal calibration (i.e. subtracting simulated signals with and without targets). In this paper, we further evaluate this setup using a homogeneous anatomically realistic breast phantom and consider a realistic calibration where the target position is not known. We apply an automated artefact removal algorithm and compare the obtained 3D imaging results with an equivalent setup using just the lens primary feed as the active antenna, without the dielectric lens. II. SIMULATED SETUPS The proposed setup is designed assuming the patient is lying in the prone position, with the breast extending through an opening on an examination table. It is a dry imaging system, which means both the breast and the antennas are placed in air. The dielectric lens and the respective primary feed are placed below the breast, and a second single antenna scans around the breast in a circular configuration in a total of 12 positions. We use a planar slot-based single-layer printed antenna formed by two crossed exponential slots (in short, XETS), impedance-matched from 2 to 6 GHz. The XETS type antenna is used both for the stand-alone receiving antenna scanning around the breast, and for the primary feed of the dielectric lens. The assembly of XETS and the Bessel lens works as the transmitting antenna. Both the setup and the antenna are described in more detail in [5]. All these elements were designed and simulated using the Computer Simulation Technology (CST) Studio software [6]. A realistic breast phantom [7] is considered for the hereby presented tests. The phantom is homogeneous with relative permittivity  r =8 and dissipation factor tan(δ)=0.1, which approximately corresponds to the upper limit of fat permittivity [8]. Two spherical Perfect Electric Conductor (PEC) targets with a 5 mm radius are placed inside the realistic breast. They are placed asymmetrically, at coordinates (x, y) = (30, 0) and (x, y)=(-30, 10) mm and at the same z plane (50 mm from the nipple surface. These targets will be called T1 and T2 from now on, respectively. The results with the same setup with and without the lens (Fig. 1) are compared. The distance between the XETS behind the lens and the surface of the sphere is 255 mm, while between the stand-alone XETS (when it is used without the lens) and the surface is 115 mm. This ensures the same incident field level on the breast in both configurations. In both cases, the XETS below the breast is placed in three positions x =[-30; 0; 30] mm, in order to evaluate if this setup allows selecting only one target