h
de
nd
as b
nna
sses
nd
bili
icat
image in which the distinct dielectric properties of malig-
nant tissue are potentially visible.
This process depends on:
bandwidth of approximately 77%.
As described in subsequent sections, initial antenna
DTD tech-
t time both
al measure-
equivalent
rements
n modelled
d the patch,
slot in the
inductance
n to restrict
y designed
r the layer† achieving high resolution
† overcoming the high attenuation in human tissue, to
permit the detection of relatively deep-seated tumours
† preventing reflections from skin, bones and other anato-
mical features (clutter) obscuring the signals from tumours.
design and optimisation were carried out using F
niques. The paper discusses in detail for the firs
the FDTD modelling and the subsequent practic
ments conducted in contact with a biological
medium.
2 FDTD modelling and practical measu
Figs. 1 and 2 show the stacked-patch configuratio
using FDTD. A microstrip line was used to fee
employing electromagnetic coupling through a
antenna ground plane, thereby avoiding the
associated with a probe feed, which is well know
bandwidth.
Stacked-patch antennas are conventionall
using low permittivity materials, especially fo
# The Institution of Engineering and Technology 2007
doi:10.1049/iet-map:20050189
Paper first received 10th August 2005 and in revised form 12th January 2006
R. Nilavalan, I.J. Craddock and J. Leendertz are with the Department of
Electrical & Electronic Engineering, University of Bristol, UK
A. Preece is with the Department of Medical Physics, University of Bristol, UK
R. Benjamin is with 13 Bellhouse Walk, Bristol BS11 0UE, UK
E-mail: ian.craddock@bristol.ac.ukWideband microstrip patc
for breast cancer tumour
R. Nilavalan, I.J. Craddock, A. Preece, J. Leendertz a
Abstract: A patch antenna is presented which h
4–9.5 GHz into human breast tissue. The ante
simulation and practical measurements to po
that remain largely consistent over the ba
Consideration is also given to the antenna’s a
also found to be suitable for the proposed appl
1 Introduction
Breast cancer is the most common cancer in women. X-ray
mammography is currently the most widely-used detection
technique [1]. However the X-ray contrast between a
tumour and the surrounding tissue is of the order of a few
percent and as a result it suffers from relatively high missed-
detection and false-detection rates. X-rays are also ionising
and, hence, not generally suited to frequent screening. They
also require uncomfortable compression of the breast.
Microwave detection of breast tumours is a nonionising
and indeed potentially low-cost alternative. The high
contrast between the dielectric properties of a malignant
tumour and the normal breast should manifest itself in
terms of lower numbers of missed-detections and false-
positives. This potential has led to the exploration of detec-
tion techniques based on microwave-radar by a number of
groups around the world [2, 3].
Research at Bristol employs a post-reception
synthetically-focussed detection method originally devel-
oped for landmine detection [4, 5]. All elements of an
antenna array transmit a broadband signal in turn, the
elements sharing a field of view with the current transmit
element, then record the received signal. By predicting the
path delay between the transmit and receive antennas via
any desired point in the breast, it is then possible to
extract and time-align all the signals from that point.
Repeated for all points in the breast, this yields a 3DIET Microw. Antennas Propag., 2007, 1, (2), pp. 277–281antenna design
tection
R. Benjamin
een designed to radiate frequencies in the range
is shown by means of previously unpublished
s a wide input bandwidth, radiation patterns
of interest and a good front-to-back ratio.
ty to radiate a pulse, and in this respect it is
ion.
Achieving high resolutions and good anticlutter perform-
ance requires wide bandwidth operation, and an operating
bandwidth of 4–9.5 GHz is the objective herein. Over this
bandwidth the antenna design employed must exhibit
good performance, both in terms of input match and radi-
ation pattern (low sidelobes, a beam that is wide enough
to encompass a reasonable portion of the breast, good
front-to-back ratio and consistent patterns). Furthermore, a
compact, low-profile antenna design is additionally desir-
able in order to reduce the complexities of the physical
array structure and to achieve a degree of conformality
with the body. The wideband bowtie antenna employed
at Bristol in the past for landmine-detection research is
therefore not ideal [6]. Various different types of antennas
are being considered by research groups involved in
tissue-sensing applications using pulsed radar techniques.
Typical examples of such antennas include the resistively-
loaded bowtie [7], slotline bowtie [8], ridged pyramidal-
horn [9], resistively loaded dipole [10] and microstrip
Archimedean spiral [11].
This paper presents a low-profile stacked-patch antenna
design that can operate over the necessary wide bandwidth
for this application. While stacked-patch antennas are well
known to have good operating bandwidths, the bandwidths
achieved are usually of the order of 20% [12]. The stacked-
patch antenna presented here has been designed to radiate
directly into a medium [13] which has similar dielectric
properties to breast tissues, and furthermore achieves a277

Table 1: Antenna dimensions
Parameter Dimension, mmthat separates the upper and lower patches. However, appre-
ciating that this antenna was intended to radiate into a
medium of 1r ¼ 9.8 (the approximate dielectric properties
of healthy breast tissue), rather than air, the decision was
taken to use higher-dielectric materials.
The antenna, therefore, consists of two stacked patches
printed on a dielectric substrate of 1r ¼ 2.2 and separated
from the ground plane by a second substrate of 1r ¼ 10.2.
The initial estimates of patch sizes (Fig. 2) were each
chosen based on achieving a lowest-order resonance at
either end of the desired operating frequency band. These
dimensions were then optimised for the correct frequency
of operation using an FDTD model. With this established,
the slot length, width and length of the microstrip stub
section were then manually optimised, guided by an
FDTD model, in order to achieve a good match and good
near-field patterns. The final values for the antenna dimen-
sions are shown in Table 1.
The antenna input responses achieved through FDTD
simulation are shown in Fig. 3 The initial model did not
incorporate losses in the medium. However the second
FDTD curve shows the relatively minor perturbation
Microstrip Patch 2
Metallic back
plane
y
2y
1
y
3
er = 10.2
z2
x2x1 x3
z3
z1
Patch 1
Slot
Breast tissue
er = 10.2
er = 2.2
y
4
Fig. 2 Antenna configuration
Microstrip
Metallic Back Plane
Patch 1
Patch 2 Slot
Dielectric
Blocks
Fig. 1 Stacked-patch antenna
278resulting from including 2 dB/cm attenuation in the breast
medium. Practical measurements were carried out with
the antenna radiating into a dielectric phantom [13] and
are also shown in Fig. 3.
These results show that the antenna has a 210 dB
antenna-feed match of 4–9 GHz, with the exception of a
small (1–2 dB) midband mismatch centred at 6.5 GHz,
which has not been found to be important in imaging exper-
iments. Agreement between FDTD and experimental results
in Fig. 3 is very pleasing, especially given the inevitable
manufacturing tolerances involved in defining the small
patches and 0.66-mm feed, and then undertaking the
blind-alignment and adhesion of the substrate layers.
An input match is only one requirement for the antenna;
radiation patterns are also important over the full band of
operation. The practical pattern measurement is, however,
rather difficult, since the antenna must be measured in the
medium, rather than in air (as in an anechoic chamber).
Due to the practical difficulties associated with undertaking
practical measurements and given the good agreement
obtained with FDTD for both the shape and levels of the
input response, it was felt reasonable to rely instead upon
FDTD simulations to estimate the pattern characteristics,
again assuming radiation into a dielectric medium with
1r ¼ 9.8.
The far-field radiation patterns of the stacked-patch
antenna were found by postprocessing the FDTD data at
specific frequencies. The calculated copolar radiation
patterns for the principal planes are shown in Figs. 4
and 5. Cross-polar levels (not shown) were 40 dB lower.
Figs. 4 and 5 show beamwidths of approximately +408
in the f ¼ 08 plane and +308 in the f ¼ 908 plane at the
midpoint frequency of 6.5 GHz. The patterns are relatively
x1 0.66
x2 6.0
x3 9.0
y1 0.64
y2 1.9
y3 0.8
y4 1.27
z1 6.5
z2 6.0
z3 3.0
-50
-40
-30
-20
-10
0
10
2 3 4 5 6 7 8 9 10
Frequency, GHz
R
e
tu
rn
lo
ss
,
dB
Measured FDTD FDTD - Lossy
Fig. 3 Antenna input response
IET Microw. Antennas Propag., Vol. 1, No. 2, April 2007