Downloaded from orbit.dtu.

dk on: May 02, 2016

Prototype of Microwave Imaging System for Breast-Cancer Screening

Rubk, Tonny; Zhurbenko, Vitaliy

Published in:
Proceedings of ANTEM/URSI 2009
DOI:
10.1109/ANTEMURSI.2009.4805106
Publication date:
2009
Document Version
Publisher's PDF, also known as Version of record
Link to publication

Citation (APA):
Rubk, T., & Zhurbenko, V. (2009). Prototype of Microwave Imaging System for Breast-Cancer Screening. In
Proceedings of ANTEM/URSI 2009. (pp. 1-4). IEEE. 10.1109/ANTEMURSI.2009.4805106

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
You may not further distribute the material or use it for any profit-making activity or commercial gain
You may freely distribute the URL identifying the publication in the public portal ?
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.

2009 13th International Symposium on Antenna Technology and Applied Electromagnetics and the Canadian Radio Sciences Meeting

Prototype of Microwave Imaging System for

Breast-Cancer Screening
Tonny Rubk and Vitaliy Zhurbenko
Department of Electrical Engineering
Technical University of Denmark
DK-2800 Kgs. Lyngby, Denmark
Email: [tru/vz]@elektro.dtu.dk

AbstractMicrowave imaging for breast-cancer detection has

received the attention of a large number of research groups in
the last decade. In this paper, the imaging system currently being
developed at the Technical university of Denmark is presented.
This includes a description of the antenna system, the microwave
hardware, and the imaging algorithm.

I. I NTRODUCTION
An increasing number of research groups are pursuing
microwave imaging for biomedical applications, especially
breast-cancer detection [1][5]. The use of microwave imaging
for breast-cancer detection is based on the contrast in electromagnetic parameters between the healthy and the cancerous
tissue which has been reported by several authors [6][8].
Both nonlinear inverse scattering (also known as microwave
tomography) [3], [4], [9] and UWB radar techniques [1],
[2], [5] have been proposed for microwave imaging of the
breast. When using UWB-based techniques, the resulting
images show the point-of-origin of the reflections caused by
the tumors while the inverse scattering algorithms reconstruct
the spatial distribution of the constitutive electromagnetic
parameters, i.e., permittivity and conductivity, of the breast.
At the Technical University of Denmark, a microwave imaging system for breast-cancer screening based on a 3D nonlinear
inverse scattering algorithm is currently being developed, and
the prototype of this system is described in this paper. The
principle of the imaging system is illustrated in Fig. 1. During
the examination, the patient is to lie prone atop an examination
table with her breast suspended through an aperture in the
table. Beneath this aperture, a measurement tank filled with a
liquid (glycerin-water mixture) with constitutive parameters
close to those of the breast to maximize the coupling of
microwave energy to the interior of the breast is positioned.
The breast is then irradiated by a single antenna at a time
using a single-frequency signal and the response measured on
the remaining antennas of the system. By using each antenna
in the system as both a transmitter and a receiver, a large
number of measurements may be obtained and used as input
in an inversion algorithm in which the forward model is based
on Maxwells equations.
This paper is organized as follows: In Section II, the antenna
system used in the system is described and in Section III, the
microwave hardware which has been developed for performing

Fig. 1. Imaging principle used in the microwave imaging system. One antenna
at a time transmits a sinusoidal signal and the phase and amplitude of the
resulting signals are measured by the remaining antennas.

the measurements is presented. Finally, in Section IV, the

imaging algorithm is described.
II. A NTENNA S YSTEM
A schematic of the antenna system used in the imaging
system is shown in Fig. 2. It consists of 32 horizontally
oriented monopole antennas positioned in a cylindrical setup
with a radius of 8 cm. Although only a single frequency is
used in the reconstruction algorithm, the system has been
designed for operating at any given frequency in the range
from 500 MHz to 3 GHz, thereby allowing for choosing the
optimum frequency for the reconstruction.
When designing the antenna setup, three major issues have
to be addressed: The first is how to ensure that the imaging
system is sensitive to changes in the entire imaging domain,
i.e., to ensure that a change in the constitutive parameters of
the breast will be detected whether the change is close to the
chest wall of the patient or close to the nipple. The second
issue concerns the modelling of the field from the antennas: In
order to reduce the artifact level in the images, the field used in
the imaging algorithm should be identical to the field radiated
by the antennas. Any differences will result in an increased
artifact level in the reconstructed images. The third and final
thing to take into account is the mechanical complexity and

978-1-4244-2980-6/09/$25.00 2009 IEEE

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 4, 2009 at 03:40 from IEEE Xplore. Restrictions apply.

Fig. 2. Schematic of the antenna setup. The antennas are indicated by the
black and white lines (white lines indicating the stripped part of the outer
conductor). The lid of the measurement tank is indicated by the grid at z =
0 cm and the imaging domain is indicated by the light-gray hemisphere.

Fig. 3. One of the 32 monopole antennas used in the imaging system.The

antenna consists of a coaxial cable from which the outermost 3.5 cm of the
outer conductor has been stripped.

size of the antennas. Since a large number of antennas must be

positioned in a relatively small volume, the antennas must be
small and to prevent the antennas from becoming prohibitively
expensive to manufacture, as well as to ensure that the 32
antennas are identical, the antennas should be kept simple.
To fulfill these requirements, the design using a cylindrical
setup with 32 horizontally oriented monopole antennas has
been adapted. A photo of one of the antennas is shown in
Fig. 3. This type of antenna was first reported used for microwave imaging in [10] and is easy to manufacture, small in
size, and computationally inexpensive to include in numerical
models. When inserted in the lossy coupling liquid, the return
loss is better than 6 dB throughout the applied frequency range.
The horizontal orientation of the antennas has been chosen
to ensure full coverage of the imaging domain. Although a
setup using vertically oriented antennas has also been considered due to the increased coupling of the microwave energy
between the antennas in such a setup, it has been found that
vertically oriented monopoles are less sensitive to changes in
the upper part of the imaging domain, close to the chest wall.
A photo of the antenna system mounted in the measurement
tank is shown in Fig. 4. The antennas are positioned in a
cylindrical measurement tank which is filled with the glycerinwater based coupling liquid mimicking the constitutive parameters of the breast. This liquid is lossy with an effective
conductivity of approximately 0.05 - 0.2 S/m in the frequency
range used by the system, with increasing conductivity with

Fig. 4. Photo of the antenna setup used in the imaging system. The monopoles
are seen beneath the metallic lid. In the clinical-test system, the measurement
tank will be cut off immediately above the metallic lid to allow for the patient
to lie prone with her breast suspended through the aperture in the lid.

increasing frequency. Although this is a relatively low loss it

is still sufficient to ensure that any signals from the outside
of the measurement tank and reflected signals from the wall
of the tank are attenuated and hence will not influence the
measurements.
At the top of the measurement tank, towards the chest wall
of the patient, a metallic lid is used to obtain a known and
well-defined boundary. The shielding effect of the metallic
lid implies that the measured signals are only affected by that
part of the chest wall which is immediately above the aperture.
Hence, the imaging domain needs only to be extended slightly
into the half space above the lid, as indicated in Fig. 2.
III. M ICROWAVE H ARDWARE
Each antenna has been equipped with its own highsensitivity transceiver module to allow for the large dynamic
range required by the system for imaging in the lossy coupling
liquid. These transceiver modules are seen as aluminum boxes
attached to the antennas in the photo of the antenna setup in
Fig. 4.
A schematic of a transceiver modules is shown in Fig. 5.
Each module contains a low-noise amplifier (LNA) and a radio
frequency (RF) amplifier which amplifies the signal received
by the antenna before it reaches the mixer. The local oscillator
(LO) generator is operating at a frequency which is offset
1 KHz from the RF frequency and feeds the 32 transceiver
modules through a 32-way power divider. After the down
conversion, the 1 KHz intermediate frequency (IF) signal is

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 4, 2009 at 03:40 from IEEE Xplore. Restrictions apply.

SPDT switch

Mixer
To ADC

To ant.
LNA

RF

IF

IF

SPST switches

To RF gen.

To LO gen.

Fig. 5. Schematic of transceiver module. Each transceiver module contains an

LNA, an RF amplifier and two IF amplifiers to achieve the necessary dynamic
range. The mixer is fed through a power-divider circuitry while the RF
generator is connected to the transceiver modules through a switching network
which, in combination with the switches in the module itself, minimizes the
leakage from the transmitting to the receiving channel.

amplified by two IF amplifiers and fed to an 18 bit analogto-digital converter (ADC) with a built-in 10 kHz analog lowpass filter.
The digitized signal is processed using a third-order Chebyshev band-pass filter and the amplitude and phase of the
resulting sinusoidal signal stored on a computer for later use
in the reconstruction algorithm.
The multiplexing between the transmitting and receiving
mode of the individual antennas are achieved using a singlepole double-throw (SPDT) switch. However, this switch in
itself does not provide sufficient isolation between the signal
from the RF generator and the signal from the antenna when
this is operating in receive mode. To remedy this, two additional single-pole single-throw (SPST) switches are included in
the transceiver modules. Additionally, the RF generator feeds
the transceiver modules through a switching network and not
a power-divider network as the LO signal.
In addition to the isolation obtained by inserting the additional switches, special attention has been payed to avoid
signal leakage within the transceiver modules. This includes
filtering of the power supply lines and of the digital lines
used for controlling the switches, the use of shielding fences
between the receive and transmit blocks in the modules and
the use of a minimized PCB enclosure.
The signal which leaks from the SPDT switch to the LNA
when a given antenna is acting as transmitter is sampled
simultaneously with the signals from the receiving antennas by
the ADC. This allows for using the leakage signal as reference
when determining the phase of the signals measured by the
receiving antennas.
To illustrate the performance of the transceiver modules
1000 measurements of the amplitude and phase of the signal
received with an antenna on the opposite side of the imaging
domain from the transmitting antenna has been analyzed.
During the measurements the system was operating at 2.0 GHz
and the measurement tank contained only the coupling liquid,
i.e., there were no scattering objects in the imaging system.
The 1000 measurements were performed during a time period
of approximately 2 hours and 40 minutes since pauses of approximately 10 seconds were inserted between the individual
measurements. The ADC was operating with a sampling rate

of 100 kHz and used 8000 samples pr. measurement.

The average amplitude of the 1 KHz IF signal measured
by the ADC was 0.446 V corresponding to a power level at
the terminal of the antenna of approximately -117 dBm. This
is 130 dB below the 13 dBm signal which is available from
the RF generator and clearly illustrates the need for the very
large dynamic range of the microwave hardware. The standard
deviation of the 1000 measurements is 5.83 mV, corresponding
to 1.3% of the mean amplitude. The mean of the measured
phase was 159.57 with a standard deviation of 0.713 . In
the preliminary tests, this level of accuracy has been found to
be adequate although it may be increased by increasing the
number of samples used by the ADC.
The total measurement time (32 transmitters with 31 receivers each) during which the breast of the patient must be
suspended through the aperture is approximately 2 minutes pr.
breast, including the time needed to fill the measurement tank
with the coupling liquid.
IV. I MAGING A LGORITHM
To represent the distribution of the constitutive parameters
in the imaging domain, the domain is divided into cubic
cells. In each cell, the constitutive parameters are assumed
constant and the size of the cells is chosen on the basis of
the expected size of the objects to be imaged as well as the
wave length in the coupling liquid. Typically, a cube side
length of between one tenth and one twentieth of a wave
length is used, corresponding to a side length in the order of
2 to 5 mm, depending on the chosen frequency. To represent
the permittivity and conductivity of the individual cells, the
squared complex wave number is applied. Using the time
notation eit , the squared wave number is given by
k 2 (r) = 2 0 (r) + i0 (r)

(1)

wherein i is the imaginary unit, is the angular frequency, and

0 is the free-space permeability. The conductivity is given by
and the permittivity by with r indicating the position.
A nonlinear inverse-scattering algorithm is applied in the
system to reconstruct the three-dimensional distribution of
the constitutive electromagnetic parameters in the imaging
domain. In this algorithm, the distribution is reconstructed by
solving the nonlinear minimization problem


2 2

k 2 = argmin S meas S calc(k )
2
o
n
2
= argmin kS res k2
(2)
using an iterative Newton algorithm. Herein, the column
vector S meas holds the measured data and S calc holds the
corresponding data calculated assuming the distribution of
squared wave numbers given by k 2 . The column vector S res
is the residual vector. The minimization problem is formulated
using the log-magnitude and unwrapped phase formulation
introduced in [11] which has been shown to improve the
performance of the algorithm by emphasizing large relative

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 4, 2009 at 03:40 from IEEE Xplore. Restrictions apply.

changes and allowing for reconstructions to span multiple

Riemann sheets.
In each of the N iterations of the Newton algorithm, the
following three steps are performed (with n indicating the
iteration number):
1) Calculate the signals resulting from the distribution of
squared wave numbers given by the column vector k 2n
and the corresponding Jacobian matrix Jn . To this end, a
method-of-moments code utilizing the adaptive integral
method and higher-order basis functions [12] is used.
2) Solve the linear problem

2 

2
2
res
k n = argmin J n k n S n
2

subj. to regularization.

(3)

k 2n .

to determine the update vector

This problem is
ill-posed in the classical sense and regularization must
be applied to achieve a suitable solution. To this end
an algorithm using the conjugated gradient least squares
algorithm [13, Sec. 6.3] in combination with a trustregion [14], the L-curve criterion [13, Sec. 7.5], and a
Euclidean-distance penalty term [15] has been applied.
3) Update the distribution of complex squared wave numbers using
(4)
k 2n+1 = k 2n + n k 2n
with the scalar n being the Newton step [16]. The
Newton step is most often unity, although in some
circumstances it is less than one to ensure that the phase
of the calculated signal is unwrapped properly [4].
The most time-consuming task in the reconstruction is solving
the forward problem. This may take as much as 30 minutes to complete, although 15 minutes is more typical for
a standard discretization of the imaging domain. After the
forward solution S calc , and thus S res , has been found, the
calculation of the Jacobian and the update of the distribution is
found in less than half a minute. The reconstruction algorithm
typically converges within the first 12 iterations, yielding a
total reconstruction time of 3 - 6 hours depending on the time
used solving the forward problem.
Preliminary studies have shown that the system is capable of
detecting spherical scatterers with a diameter less than 5 mm
and a number of reconstructions of phantom measurements
will be presented at the conference.
V. C ONCLUSION
A prototype system for microwave imaging of the breast
was presented. The system consists of three major parts; an
antenna system, a set of microwave transceiver modules and
an ADC for storing the measured data, and, finally, a reconstruction algorithm. The antenna system has been designed
to achieve a near-uniform coverage of the imaging domain
and minimum influence from the surroundings while the
transceiver modules has been designed for quick examination
with a dynamic range of more than 130 dB. The nonlinear inversion algorithm applies the log-magnitude unwrapped phase

formulation to improve its performance as well as a methodof-moments algorithm which has been optimized for use in
the reconstruction.
ACKNOWLEDGMENT
This work is supported by the Villum Kann Rasmussen
Foundation, grant no. VKR020891.
R EFERENCES
[1] W. C. Khor, et al., An experimental and theoretical investigation into
capabilities of a UWB microwave imaging radar system to detect breast
cancer, EUROCON, 2007. The International Conference on Computer
as a Tool, pp. 771776, 2007.
[2] M. Klemm, et al., Experimental and clinical results of breast cancer
detection using UWB microwave radar, Proceedings of the IEEE Antennas and Propagation Society International Symposium, 2008, 2008.
[3] J. Johnson, T. Takenaka, and T. Tanaka, Two-dimensional time-domain
inverse scattering for quantitative analysis of breast composition, IEEE
Transactions on Biomedical Engineering, vol. 55, no. 8, pp. 19411945,
2008.
[4] T. Rubk, et al., Nonlinear microwave imaging for breast-cancer
screening using Gauss-Newtons method and the CGLS inversion algorithm, IEEE Transactions on Antennas and Propagation, vol. 55,
no. 8, pp. 23202331, 2007.
[5] T. Williams, J. Sill, and E. Fear, Breast surface estimation for radarbased breast imaging systems, IEEE Transactions on Biomedical Engineering, vol. 55, no. 6, pp. 16781686, 2008.
[6] W. T. Joines, et al., The measured electrical properties of normal and
malignant human tissues from 50 to 900 MHz, Medical Physics, vol. 21,
no. 4, pp. 54750, 1994.
[7] S. P. Poplack, et al., Electromagnetic breast imaging: Results of
a pilot study in women with abnormal mammograms, Radiology Radiological Society of North America, vol. 243, no. 2, pp. 350359,
2007.
[8] M. Lazebnik, et al., A large-scale study of the ultrawideband microwave
dielectric properties of normal, benign and malignant breast tissues
obtained from cancer surgeries, Physics in Medicine and Biology,
vol. 52, no. 20, pp. 60936115, 2007.
[9] P. Hashemzadeh, et al., Experimental results of an optimization approach to two-dimensional time domain electromagnetic inverse problem
with application to microwave breast tomography, Proceedings of the
3rd IASTED International Conference on Biomedical Engineering, 2005,
pp. 192196, 2005.
[10] P. Meaney, et al., An active microwave imaging system for reconstruction of 2-D electrical property distributions, IEEE Transactions
on Biomedical Engineering, vol. 42, no. 10, pp. 10171026, 1995.
[11] , Microwave image reconstruction utilizing log-magnitude and
unwrapped phase to improve high-contrast object recovery, IEEE
Transactions on Medical Imaging, vol. 20, no. 2, pp. 104116, 2001.
[12] O. S. Kim and P. Meincke, Adaptive integral method for higher order
method of moments, IEEE Transactions on Antennas and Propagation,
vol. 56, no. 81, pp. 22982305, 2008.
[13] P. C. Hansen, Rank-Deficient and Discrete Ill-Posed Problems: Numerical Aspects of Linear Inversion, ser. Monographs on Mathematical
Modeling and Computation. SIAM, 1998.
[14] D. C. Sorensen, Newtons method with a model trust region modification, SIAM Journal on Numerical Analysis, vol. 19, no. 2, pp. 409426,
1982.
[15] D. Li, P. Meaney, and K. Paulsen, Conformal imaging with a noncontacting microwave antenna array, Proceedings of the IEEE MTT-S
International Microwave Symposium, 2001, vol. 1, pp. 563566, 2001.
[16] C. T. Kelley, Iterative Methods for Linear and Nonlinear Equations,
ser. Frontiers in Applied Mathematics. Philadelphia, PA: SIAM, 1995,
vol. 16.

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 4, 2009 at 03:40 from IEEE Xplore. Restrictions apply.