470 IEEE MICROWAVE AND GUIDED WAVE LETTERS, VOL. 9, NO. 11, NOVEMBER 1999 Microwave System for Breast Tumor Detection E. C. Fear, Student Member, IEEE, and M. A. Stuchly, Fellow, IEEE Abstract— A preliminary numerical analysis of a system for breast tumor detection, amenable to practical implementation, is described. The general idea was previously introduced and relies on ultrawide-band radar and confocal imaging. The system consists of an array of small antennas placed away from the breast. Simulations of each antenna in the array are performed with the finite-difference time-domain (FDTD) method. A tumor detection algorithm is applied to the data, incorporating skin return subtraction to enhance tumor returns. Index Terms—Microwave imaging, tumor detection, ultrawide- band radar. I. INTRODUCTION B REAST cancer affects many women, and early detection is an important part of management of this disease. Mammography, which exposes women to ionizing radiation, is typically used in breast imaging. Microwaves have the poten- tial to provide effective tumor detection due to the difference in dielectric properties of normal breast tissue and breast tumors at microwave frequencies. Previously proposed systems for microwave imaging reconstruct object profiles from measured scatter of a narrowband illumination signal by the object. Early systems employed diffraction tomography in reconstruction which did not provide adequate images. More recent systems use iterative nonlinear inverse-scattering approaches, provid- ing improved images but requiring computationally intensive image reconstruction [1], [2]. A new concept introduced by Hagness et al. uses a pulsed microwave confocal system to detect tumors [3], [4]. The ideas are closely related to optical confocal systems, however microwaves offer better penetration depth in tissue. Time-shifting recorded signals creates synthetic focal points. Returns from scatterers at the focal point add coherently, while returns from scatterers off focus add incoherently and are suppressed. Both two- and three-dimensional (2-D and 3-D) simulations were performed with the finite-difference time-domain (FDTD) method. In both cases antennas (monopoles in 2-D and bowties in 3-D) were placed on a flattened breast model. Promising results were obtained. The work presented here follows the general approach in [3] and [4], and presents a preliminary numerical evaluation of a detection system more amenable to practical implementation. Thus, we believe that this new approach offers a feasible alternative to traditional microwave imaging for breast tumour detection. Manuscript received July 23, 1999; revised September 7, 1999. This work was supported by NSERC. The authors are with the Department of Electrical and Computer Engineer- ing, University of Victoria, Victoria, BC V8W 3P6, Canada. Publisher Item Identifier S 1051-8207(99)09819-0. Fig. 1. Proposed system configuration. Configuration 1 (left): 9 antennas are placed concentric with the 10-cm-diameter breast model. A 5-mm-diameter tumour is located 1.25 cm under the skin. Configuration 2 (right): 11 antennas are positioned between 2 and 3 cm from the breast model. The tumor is 4 mm diameter and located 2 cm from the skin. In both cases, the electrical properties are as in [3], and the antennas are spaced circumferentially 1 cm apart. II. PROPOSED SYSTEM AND MODELING Our system is suitable for a routine scan of the breast, and its representation in one plane is shown in Fig. 1. The patient lies in a prone position with the breasts immersed in a liquid with electrical properties similar to those of fat. An array of antennas is placed in the liquid and positioned in an arc at a distance from the breast. For data acquisition, one antenna transmits an ultrawide-band pulse and the scattered returns are recorded at the same antenna. This is repeated sequentially for each antenna in the array. The antennas are spaced to reduce their coupling, however the array may be rotated to a new position for acquisition of additional data. Furthermore, translating the array vertically allows for scans of different cross sections through the breast. This antenna configuration differs from that described in [3] and [4] in two ways. First, the array is placed sufficiently far from the skin so that returns from the breast do not arrive during pulse transmission. This allows for recording of skin re- flections and their use in image processing. The antennas used here are resistively loaded dipoles of length 1.35 cm, while 8-cm-length bowties are presented in [4] and [5]. The smaller antennas are practical for implementation, as placement of an array of antennas near the breast is possible and results in shorter acquisition time due to electronic scanning of elements. In this preliminary study, the breast is modeled as a finite cylinder with the electrical properties of fat surrounded by an 1051–8207/99$10.00  1999 IEEE