icrowave imaging painful, and there are health concerns able exams.

Microwaves deposit power

for medical applica- related to exposure to ionizing radiation. in tissues, and this power deposition

m tions has been of

interest for many
years. Microwave
images are maps of
The concerns related to mammography
have generated interest in alternative
approaches to breast imaging.
Medical imaging methods other than
has been extensively studied for appli-
cations such as cellular telephones and
hyperthermia. Safety standards (e.g.
IEEE Standard C95.1 –1999) have been
the electrical property distributions in X-ray have been applied to breast imag- defined using maximum permissible
the body. The electrical properties of ing. For example, ultrasound is used exposure to quantities such as electric
various tissues may be related to their clinically to determine whether a lesion and magnetic fields, induced currents in
physiological state. For example, the detected on a mammogram is a liquid the body, and the specific absorption
properties of tissues change with tem- cyst or a solid tumor. Magnetic reso- rate (SAR). SAR quantifies the amount
perature. One application of microwave nance imaging (MRI) is useful for of power absorbed in a volume of tis-
imaging that has been proposed is mon- examining women with implants. sue, and is formally defined as:
itoring hyperthermia, which is the However, it is expensive and is current-

d dW
application of heat to tissue. In this ly in the testing stage as a screening S AR = =
case, the changing electrical properties tool. Alternative imaging methods for dt ρd V
indicate the successful deposition of where dW is incremental energy,
dV is volume element and ρ is
Microwaves for
heat in the tissue of interest.
Other changes in electrical proper- the density of the volume. SAR
ties may be caused by disease. There is has units of W/kg. Standard

breast cancer detection?

some evidence of changes in the prop- C95.1-1999 indicates that
erties of cancerous tissues when com- acceptable devices operating
pared to normal tissues. Cancer detec- between 100 kHz and 6 GHz
tion with microwave imaging is based have maximum SAR over the
on this contrast in electrical properties. body of less than 1.6 W/kg when
Recently, microwave imaging for averaged over 1g of tissue. To
breast cancer detection has interested evaluate compliance of an EM
many researchers. Elise C. Fear, system to the safety standard,
power deposition and expected
The gold standard Paul M. Meaney heating can be estimated with
and runners up computer models. For example,
The gold standard method of breast
imaging is mammography, an x-ray
& Maria A. Stuchly Gandhi and colleagues per-
formed studies of cellular tele-
image of a compressed breast. A mam- phones radiating in the vicini-
mogram is a map of the densities of the ty of human head models.
breast, and has proven to be quite sensi- taking a look at a possibly safer way to check Results show that radiated
tive to the presence of lesions in the power of 121 mW is required
breast. However, not all lesions are can- to obtain 1.6 W/kg SAR at
cerous. It may not be apparent which detecting breast cancer have been pro- 1900 MHz. We expect microwave
abnormalities are cancerous without posed. They are based on different breast cancer detection systems to
additional imaging or biopsy. Also, not physical properties, such as tissue elas- operate with power levels one or two
all lesions may be detected (up to 15% ticity, temperature, and optical or elec- orders of magnitude below those of a
may be missed), and it is difficult to trical characteristics. Electromagnetic cellular telephone. Evaluating SAR
image the 25% of women who have fields in generalæand microwaves in for specific systems with computer
dense breasts. Lastly, many women find particularæare utilized in techniques models ensures safety and compliance
mammography uncomfortable or based on electrical property contrasts. with standards. Thus, microwave
108
The success of a particular medical imaging is not expected to pose a
imaging approach for reliable detec- health risk to the patient.
Relative tion of small malignant tumors From a technical point of view,
106 Permittivity hinges on the presence of a signifi- microwave breast tumor detection has
cant and consistent contrast between the potential to detect small tumors. It
104 malignant and other breast tissues. It is also expected to be less expensive
muscle appears that this physical basis may than methods such as MRI and nuclear
2 fat be present for methods relying on medicine. This is because microwave
10
electromagnetic fields, including equipment costs a fraction of the
microwave imaging. equipment needed for MRI and
100 Conductivity nuclear medicine installations (hun-
Why microwaves? dreds of thousands vs. millions). The
10-2 From the patient’s viewpoint, imaging process is also anticipated to
100 102 104 106 108 1010 microwave breast imaging is attrac- be very rapid, sensitive (detect most
Frequency (Hz) tive because both ionizing radiation tumors in the breast) and specific
Fig. 1 Dielectric properties of low (fat) and and breast compressions are avoided. (detect only cancerous tumors). The
high (muscle) water content tissues The result is safer and more comfort- key to sensitivity, specificity and the
over a broad frequency range.

12 0278-6648/03/$17.00
Authorized licensed use limited to: Alliance University. ©11,2025
Downloaded on February 2003 IEEE IEEE POTENTIALS
at 05:20:33 UTC from IEEE Xplore. Restrictions apply.
 ability to detect small tumors is the alpha dispersion (kHz region). This conductivity estimated from images are
anticipated electrical property contrast change is caused by factors such as higher than indicated by previous ex
between malignancies and normal tis- ionic diffusion in the layer of charged vivo measurement studies. This may be
sues. Specifying whether a suspicious particles surrounding the cell. The beta due to a variety of factors including the
area is malignant or benign may be relaxation is responsible for the next heterogeneous nature of breast tissue,
possible with a contrast in the electrical decrease in permittivity, and results blood volume content and others, sug-
properties of these tissues. from charging of the interfaces between gesting that the situation is more com-
cell membranes, cell interiors and extra- plex than previous studies portray.
Electrical properties cellular suspensions. Conductivity also In order to assess the properties of
of tissues increases due to increased conduction specific normal, malignant and benign
At microwave frequencies, biologi- through the cell membrane. The gamma tissues, a large-scale measurement pro-
cal tissue interactions with the fields are relaxation occurs near 20 to 25 GHz, gram (at the University of Calgary and
defined by their complex permittivity, ε, and is primarily due to the dipolar relax- the University of Wisconsin at
which consists of the dielectric constant ation of water in the tissues. Madison) has been established. The
(the real part), ε´, and the loss factor Figure 1 indicates that water content measurement program aims to create a
(the imaginary part), ε´´. The dielectric is a major factor in determining permit- comprehensive database of the electrical
constant determines the ability of the tivity. Low water content tissues, such properties of specific breast tissues over
material to store the electric field ener- as bone, fat, lung and the outer layer of a broad frequency range. While this
gy, while the loss factor indicates how skin, have lower permittivity values added property variation might compli-
much of that energy is converted into than high water content tissues such as cate tumor detection and breast charac-
heat and dissipated. muscle, blood, brain, and internal terization somewhat, the predominance
Foster and Schwan discussed the organs. Measurements of the permittivi- of fatty tissue in the normal breast will
electrical properties of a variety of tis- ty of various tissues and tumors have most likely still act to facilitate a signifi-
indicated that typically tumors have a cant contrast mechanism between nor-
permittivity 10-20% greater than that of mal and malignant tissue. In addition,
the healthy host tissue. this wide property range may also pro-
Breast Several studies in the literature sug- vide a basis for assessing and possibly
Transmitter gest that a large electrical contrast exists even diagnosing benign and pre-cancer-
Receivers between normal breast tissues and ous conditions.
a)
malignant tumors. In these studies, nor-
mal breast tissues displayed properties Microwave imaging
similar to fatty tissues, while tumor Microwave imaging is defined as
Breast properties were similar to muscle."seeing" the internal structure of an
Tumor object by means of electromagnetic
However, the breast consists of both
Transmitter fields at microwave frequencies (300
fatty and glandular tissue. Previously
b) Receivers MHz – 30 GHz). The basic problem is
reported studies do not provide precise
illustrated in Fig. 2. A trans-
Fig. 2 a) The basic micro- mitter is used to illuminate
wave imaging prob- the breast with microwaves,
lem involves illumi- which travel through the
nating the breast breast and may be detected at
with microwaves receivers located on the oppo-
and detecting ener- site side of the breast.
gy traveling through Alternatively, reflections may
or reflected from the be recorded at the transmit-
breast. b) With a ting antenna.
tumor present, With a tumor present,
the reflected and waves traveling through the
transmitted waves breast encounter a change in
change compared electrical properties, causing
to the tumor-free the incident wave to scatter.
case. This scattering changes the
amounts of energy detected at
sues extensively in a 1989 Fig. 3 The microwave breast imaging system at the receivers and the transmit-
review. Various tissue types Dartmouth College. ter, as shown in Fig. 2. Images
have different permittivity values, as details on the particular types of normal are formed using information contained
shown in Fig. 1 for muscle and fat. This tissues examined. It is important to in the detected energies.
plot shows the distinct changes in per- assess the properties of each type of tis- This implies at least two approaches
mittivity values with frequency which sue and the contributions to electrical to creating microwave images. The first
are referred to as dispersions. At low properties. approach discussed in this paper is
frequencies, tissues have high permittiv- For example, a recent microwave transmission-reflection, and the second
ity due to the insulating cell membrane. imaging study by Meaney et al. sug- is reflection (radar). Other approaches,
The permittivity decreases with the gests that the in vivo permittivity and such as microwave radiometry and ther-

FEBRUARY/MARCH 2003
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply. 13
 antenna is changed since the Second World War.
and the process is Typically, a system sends a micro-
repeated. The wave pulse from a transmitter and
shape of the object receives it in the same location with the
and spatial distrib- same antenna, as illustrated in Fig. 2.
ution of the permit- There are, however, various other radar
tivity are obtained systems that may use modulated har-
from the transmit- monic microwave signals rather than a
ted (incident) and pulse, or use a different location of the
scattered (received) receiver from that of the transmitter.
Fig. 4 Reconstructed images at 900 MHz using measurement fields. Radar imaging in remote sensing cre-
data from a 10.7 cm diameter molasses cylinder with For penetrable, ates images similar to laser pictures
a 1.9 cm diameter saline inclusion as a function of i. e. non-metallic acquired either from a satellite or from
background medium. The top 3 images show the per- objects, that are an aircraft. Because of deeper penetra-
mittivity and the bottom 3 images show the conductiv- heterogeneous, the tion of microwaves (compared to light)
ity. The images on the left have saline (εr = 77.1, σ = transmitted waves into materials, images of soil moisture,
1.72 S/m) as background medium, the center images undergo numerous ice thickness, and conditions of vegeta-
have a 50:50 mixture of glycerine and water (εr= 55.9, reflections from tion and crops are routinely obtained. In
σ = 1.64 S/m), and the right images have a 60:40 glyc- b o u n d a r i e s case of these applications, the distance
erine:water mixture (εr= 47.9, σ = 1.35 S/m). between the object between the targets forming the images
parts that have dif- and the antennas is very large. Ground-
ferent complex penetrating radar for mine detection, on
permittivities. This the other hand, accomplishes detection
is because the with an antenna located typically just
boundaries are not above the ground. The same basic prin-
planar and often ciples used in ground penetrating radar
the internal objects are used also in medical imaging.
are smaller than the
wavelength. Thus, Why breast cancer?
the incident wave Both transmission and reflection
is scattered in all methods of imaging are more attractive
directions rather for breast cancer detection than for
than reflected in detection of other cancers. One reason is
one specific direc- the contrast between healthy and dis-
tion and likely eased breast tissues is likely much
undergoes multiple greater than the contrast in these permit-
reflections. tivity values for other tissues.
This poses an The small size and physical accessi-
inverse scattering bility of the breast compared to internal
problem that is dif- organs is also an advantage. For exam-
ficult to solve due ple, accessing the breast with the
to the nonlinear microwaves does not involve penetrat-
relationship ing layers of muscle, which results in
between the mea- scattering and attenuation of the inci-
sured scattered dent field before it reaches the object of
Fig. 5 Reconstructed permittivity and conductivity images at
fields and pattern interest (e.g. an internal organ).
600 MHz for the three imaging planes closest to the
(image) of the per- With microwave imaging in general,
chestwall of the left breast of a 53 year old woman with
mittivity. Also, the issues such as resolution, contrast,
scattered breast density for two background media: a)
evanescent waves object size, solution speed, robustness,
0.9% saline, and b) 60:40 mixture of glycerine and saline.
(waves close to the numerical stability, compatibility with
scatterer that are the object geometry, and amount of a
mo-acoustic computed tomography, strongly attenuated) are not measured, priori information, need to be success-
have also been proposed. and high spatial frequency information fully addressed. In medical applications,
is lost. These problems exist in any some of the issues are exacerbated by
Transmission-reflection application of microwave imaging. several factors. High resolution is
Transmission-reflection imaging, or required to detect small tumors, while
tomographic imaging, is similar to com- Reflections high frequencies are strongly attenuated
puted tomography. As shown in Fig. 2, The second approach to microwave in most tissues. Similarly, the object
a coherent microwave signal is transmit- breast cancer detection uses reflections size compared to the target (tumor) is
ted by one antenna. Numerous antennas to form images. Many ideas incorporat- typically rather large. Total time
surrounding the object being imaged ed in this approach are based on radar. required for data acquisition cannot be
receive waves transmitted through the Radar, "radio detection and ranging," excessive in order to minimize patient
breast. The position of the transmitting systems have been intensively studied motion. But, there are no limitations on

14 IEEE POTENTIALS
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply.
the post-processing time required to also enhanced feature detection while
Time
obtain the images. reducing unwanted artifacts. Finally, an
Compatibility of the system for illu- adjoint matrix building procedure has
mination and scattered field acquisition Antennas been implemented. It has reduced the
with clinically acceptable arrangements image reconstruction time to less than
also poses some challenges. In practice, Tumor one minute for a simple image with no
the object (human body) is most likely sacrifice in quality.
to be in the near field of the transmitting In terms of practical system imple-
and receiving antennas. Current Fig. 6 The principle of synthetic focus- mentation, the patient interface has
research in microwave breast imaging is ing. The signals recorded at undergone substantial evolution. The
directed at resolving these issues. each antenna are shown. A current interface consists of tank filled
focal point is identified (white with a liquid in which the breast is
Microwave tomography square or star) and the time
Several research groups have been taken to travel from each 3 Lossless immersion medium
investigating microwave tomography antenna to the focal point and 4
for breast cancer detection. Presently, back is calculated. The time-
5 Breast tissue
the only approach with an actual clinical delayed signals (shown with
system is being developed by a group at green square or red star) are 6

x (cm)
+ Tumor
Dartmouth College (Fig. 3). The 32- summed together. With the 7
channel system can acquire coherent sig- focal point at a tumor, the 8
nal data utilizing a standard heterodyne reflections from the tumor add 9
approach over a range of 500 MHz to 3 together, giving a much more Skin
10
GHz. All channels are capable of operat- significant result.
11
ing in both transmit and receive mode to Antenna
maximize the data collected. Channel to boundary element (BE) method for the 12
2 4 6 8 10
channel isolation is greater than 120 dB. surrounding lossy medium allows the y (cm)
Also, signals are routinely measured imposition of exact boundary conditions
down to –130 dBm with minimal impact for the fields radiating to infinity in a Fig. 7 Simple cylindrical model used for
on overall exam time. lossy medium. This hybrid approach initial investigations. The breast
The current patient interface is a tank also keeps computation costs low. model diameter is 6.8 cm, skin
filled with a mixture of glycerine and The transmitting antennas can be thickness is 2 mm and tumor
water to provide adequate electrical easily and accurately modeled as point diameter is 6 mm. The antenna is
property match to the breast. The sources. This allows one to compensate located 1 cm from the breast.
woman lies prone on the examination for field perturbations due to the pres- Electrical properties are as follows:
table with the breast pendant through an ence of the full array of antennas. In breast tissue with εr=9 , σ=0.4 S/m,
aperture into the tank. The array of 16 addition to improvements in the model skin with εr=36 , σ =4 S/m, tumor
monopole antennas protrudes into the match, techniques have been introduced with εr=50 , σ =4 S/m, immersion
tank from the lower plate. They can be within the imaging algorithm to both liquid with εr=9 , σ =0 S/m.
positioned at pre-determined heights by improve quality and reduce computation
a computer-controlled linear actuator to time. Analogous to X-ray computed Chest Wall
acquire data at multiple planes of the tomography (CT), the approach mini- x
breast. Signals are sequentially transmit- mizes the log-magnitude and phase z Antennas
ted by each antenna and measured at all errors directly. This is inherently a more
15 receivers for 12 frequencies over
the band and for seven vertical array nipple Fig. 8
positions. The total acquisition time is The hemispherical
14 minutes for both breasts combined. 80 breast model with diam
The image recovery process is an skin
70 eter of 14 cm. The breast
iterative model-based approach. It model and antenna are
z (mm)

60
works to minimize the error between immersed in a low loss
50
the measured fields and a forward liquid. a) The orientations
numerical solution of the electric 40
30
of the chest wall and
fields. The forward solution computes antenna locations are
20 40
fields using a model of the system glandular tissue 50 indicated. b) The interior
80
with estimated electrical properties, 160 140 120 100 100 of the model contains
which are updated at each iteration. 80 60 40 120
y (mm) objects representing
Central to this approach is that the x (mm) glandular tissue (blue,
numerical algorithm accurately green, white, pink and
reflects the physics of the illumination linear treatment of the data and has been immersed. yellow spheres and cylin
tank. Several important innovations shown to be more sensitive to internal Liquids of ders). The 6 mm diameter
have been implemented to accomplish breast structures. higher and tumor is the small red
this required accuracy. Utilizing the A strategy to conform the imaging lower con- phere. The remainder of
finite element (FE) method with the zone exactly to the breast contour has trasts with the space is filled with
fatty breast tissue.

FEBRUARY/MARCH 2003
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply. 15
 a)
MHz for a 10.7 cm diameter Figure 5 is a set of reconstructed
Ground plane
cylinder of molasses (ε r = images (three planes closest to the chest
16.0, σ = 0.36 S/m) with a 1.9 wall) at 600 MHz for the left breast of a
Tumor Tumor cm diameter, 0.9% saline (εr = 53 year old woman. They show scat-
PVC pipe
PVC pipe
77.1, σ = 1.72 S/m) inclusion. tered radiographic density and no known
y Antenna
Wall The background medium is abnormalities for both the saline and
thickness 10cm 92cm b) different for each set (upper XGlyc = 60% cases. The images for the
=0.5 cm
rpipe= 5 cm and lower) of images. The saline exhibit uneven breast contours
Antenna images on the left utilize a with exaggerated increased permittivity
z x 3cm 0.9% saline background and near their centers. 3D effects predomi-
Ground plane have diverged. The back- nantly impact the recovered conductivi-
grounds for the images in the ty, which is significantly reduced in size
Fig. 9 Test set-up with monopole antenna, PVC center and right are mixtures of compared to the permittivity counterpart
pipe and tumor model a) top-down view glycerine and water with XGlyc especially at positions 2 and 3.
and b) side view. = 50% and 60% where XGlyc is The comparable set of images for the
the fraction of glycerine in the glycerine:water mixture produces more
the breast tissue, namely saline and a glycerine:water mixture. The images uniform breast contours, with the con-
mixture of glycerine and water, have with the glycerine-water background ductivity perimeter much closer in size
been examined. Recent results utilizing delineate the phantom perimeter well to that of the breast permittivity distrib-
lower contrast coupling media have along with accurately detecting the ution. The much reduced permittivity
demonstrated both significant reduction inclusion. This tomographic approach is increase at the breast center for these
in 3D artifacts (due to signals propagat- not inherently limited by the half-wave- images may reflect the actual heteroge-
ing out of the imaging plane) and length Rayleigh criterion, but rather by neous distribution of adipose and glan-
enhanced feature extraction within a signal to noise. Resolution down to 1/12 dular tissues. These two tissues are
large scattering object such as the of a wavelength has been demonstrated composed of substantially different pro-
breast. and there generally is an improvement portions of water. Current clinical stud-
Figure 4 shows a set of relative per- in image accuracy with increased oper- ies, including assessing correlations
mittivity and conductivity images at 900 ating frequency. between electrical properties and physi-
ological factors such as age and breast
x 10-3
3
density along with studying property
Normalized voltage

Antenna response subtracted

Initial Signal: 2 Average signal subtracted fluctuations due to menstrual cycle vari-
• antenna excitation ations, are under way.
1
• reflections from skin,

Radar systems
breast tissue, tumor 0
1
0.8
-1
Normalized voltage

0.6
0.4
In 1998, Hagness and colleagues
0.2 -2
0 500 1000 1500 2000 2500 3000 3500 4000 introduced a pulsed radar system for
0
-0.2
Calibration: subtract Time step breast tumor detection. This was the
returns recorded After subtracting the antenna
-0.4
-0.6 without breast present
first medical application of technology
response, the skin reflection is
-0.8
the dominant component of that previously has been used predomi-
-1
0 500 1000 1500 2000 2500 3000 3500 4000 the signal. nantly in military radar applications.
Time step 1
0.8
This method avoids nonlinear inverse
Skin subtraction:
Normalized voltage

Integration subtract average of

0.6 scattering techniques that may be com-
0.4
signals from each signal 0.2 plex and computer intensive. However,
0
1
0.9 -0.2
it does not provide profiles of the per-
mittivity, but only identifies regions of
Normalized voltage

0.8 -0.4 Uncompensated

Compensated
0.7 -0.6
0.6 Integration: transform -0.8 increased scattering due to a small
0.5 -1
0.4
center of signal from 0 500 1000 1500 2000 2500 3000 3500 region of different permittivity.
zero to maximum
0.3
0.2
Time step To acquire the data, the breast is illu-
0.1 Comparison of signals before minated with an ultra-wideband pulse,
0 and after compensation shows
0 500 1000 1500 2000 2500 3000 3500 4000
suppression of clutter. and the same antenna collects the back-
Time step
Compensation: account scattered waves. This process is repeat-
for 1/r attenuation as wave ed for multiple antenna positions. A
propagates from source x 10-6
4
Tumor
time-shift-and-add algorithm is applied
Normalized voltage

No Tumor
3
to the set of pulses recorded at the anten-
2
na output in all positions. The time delay
1
for the round trip to a given point in the
0
Focus: 1) calculate time delay from each antenna test domain is computed for each anten-
-1
to focal point; 2) indentify corresponding part of na position. Next, the signals with the
processed signal at each antenna, and add -2

together; 3) scan focal point through

0 500 1000 1500 2000 2500 3000 3500 appropriate time delays are summed up.
Time step The focal point is scanned to a new loca-
breast to create image
Processed signals with and tion and the focusing process is repeated.
without tumor present in
breast model. As shown in Fig. 6, this procedure
ensures that signals from a given pixel
Fig. 10 Image reconstruction algorithm for radar-based approaches

16 IEEE POTENTIALS
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply.
 1
4
based on the same principle has also quency, so the upper frequency of the
0.9 been developed. Extensive numeri- band is limited by the need to provide
5 0.8 cal modeling and experimental feasi- adequate penetration into the tissue.
0.7 bility assessments indicate that this These two requirements result in an
6
0.6 system is also very promising. Since ultra-wideband signal. The ultra-wide-
x (cm)

7 0.5 the theoretical results in terms of band signal implies use of a specialized
0.4 detection of small tumors appear to antenna. Antennas are generally nar-
8 be comparable for the planar and our rowband devices, and one way of
0.3
9 system, the more detailed descrip- increasing the bandwidth is with resis-
0.2
tion that follows is limited to our tive loading. But, this increase in band-
10 0.1 system. width comes with a loss of efficiency.
0 Thus, we have investigated resistively
Tissue sensing adaptive
2 3 4 5 6 7 8
y (cm) loaded monopole antennas with Wu-
radar (TSAR) system King profiles for this application.
Fig. 11 Image of simple cylindrical The TSAR system was developed at The TSAR system has been studied
breast model created with 5 the Universities of Calgary and with computer models and simple exper-
rows of 5 antennas. the plane Victoria. The woman is in a prone posi- iments. The initial computer models are
shown is a cut through the cylin- tion with the breast pendulant and cylindrical in shape, and consist of breast
der perpendicular to the axis. immersed in a liquid medium. This tissue surrounded by a thin layer of skin
or voxel in the test domain add coher- arrangement is similar to that in the (Fig. 7). The breast tissue contains ran-
ently. Signals from scatterers in other clinical system developed at Dartmouth dom variation in electrical properties of
locations add incoherently, which tends College. The antenna is physically up to +/-10% over 4 mm cubes. A tumor
to reduce clutter. Thus, a small scatterer scanned to synthesize an array that of diameter 6 mm is embedded in the
can be picked up, if it has sufficiently encircles the breast. All antenna loca-
large contrast compared to a random het- tions are at a distance from the breast (1 0.03
erogeneity within the test domain. The cm in a liquid similar to breast tissue). 0.04

z (m)
same principle is used in several synthet- This allows a frame of reference (i.e. 0.05
ic aperture ground penetrating radar sys- known antenna locations) to be estab- 0.06
tems. One of the significant advantages lished, which is useful for image recon- 0.07
of this approach is that high resolution struction. However, this arrangement 0.04 0.06 0.08 0.1 0.12 0.14 0.16
can be obtained provided that a suffi- poses a major challenge due to strong ( )
Fig. 12 Image of realistic breast model
ciently wideband pulse can be used. reflections from the skin. These reflec- created with 90 antennas (9 rows
In the system introduced by Hagness tions cannot be gated out and have to be of 10 antennas). An additional
and colleagues, the woman to be accounted for in the algorithms used for step is added to the image for-
scanned lies on her back and antennas processing the return signals. mation algorithm, namely sub-
are placed on the flattened breast sur- The breast is illuminated with an traction of the average response
face. A comprehensive numerical evalu- ultra-wideband pulse in order to satisfy of the 5 antennas closest to the
ation of performance of this system has two conflicting requirements. For reso- antenna of interest.
been performed with the finite different lution of 1 cm in tissue, a bandwidth of
time domain (FDTD) method. Initial several GHz is required. However, tis- breast model at a depth of at least 3 cm.
work has suggested the feasibility of sue loss increases with increasing fre- The model is placed in a liquid simi-
detecting tumors as small as 2 mm at -0.15 1 lar to breast tissue, and scanned with a
depths of up to 5 cm. 0.9 resistively loaded dipole designed at 4
Resistively loaded bowtie antennas -0.1 0.8 GHz and excited with a differentiated
were developed for added sensitivity to 0.7 Gaussian pulse with significant frequen-
-0.05
small tumors. Observation of the cross- 0.6 cy content between 1 and 10 GHz. The
x (m)

polarization was proposed for detection 0 0.5 antenna is scanned to nine vertical loca-
of tumors in the presence of strongly 0.4 tions and five positions around the
scattering planar objects such as the 0.05 breast at each vertical location (45 loca-
0.3
chest wall. The use of ultra-wideband 0.1 0.2 tions in total).
signals provides ultra-wideband reflec- 0.1 A more complex hemispherical
tions, or microwave signatures. It was 0.15 model has also been investigated (Fig.
suggested that if various tumor shapes -0.15 -0.1 -0.05 0 0.05 0.1 0.15 8). This model contains a chest wall,
are associated with specific microwave y (m) nipple and variations representing glan-
signatures, then this characteristic may Fig. 13 Image of 1.2 cm diameter dular tissue. It is also significantly larger
be used to aid in diagnosis or detection. wood dowel located inside of than the simple cylinder (14 cm diameter
Further investigations for more realistic 10 cm diameter PVC pipe. In compared to 6.8 cm). The same antenna
MRI-based 2D models were carried out. this case, reflections were is scanned to nine rows of 10 antenna
The successful detection of small recorded at 16 antenna loca- locations, resulting in a conical array.
tumors in this more complex environ- tions around the pipe. Data From an experimental perspective, a
ment was demonstrated. Initial experi- were recorded in the frequen- simple model consisting of a PVC pipe
mental feasibility testing is ongoing. cy domain, then transformed to (representing skin) with a tumor model
An alternative cylindrical system the time domain using the (e.g. wood dowel) has been scanned
Fourier transform.

FEBRUARY/MARCH 2003
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply. 17
with a resistively loaded monopole National Research Council, 2001. Theory Tech., vol. 49, Sept. 2001, pp.
antenna. The setup is shown in Fig. 9. • E.C. Fear, S.C. Hagness, P.M. 1607-1613.
Reflections recorded at the illuminat- Meaney, M. Okoniewski and M.A.
ing antennas are used to form images Stuchly, “Near-field imaging for breast About the authors
with synthetic focusing methods intro- tumor detection,” IEEE Microwave Elise C. Fear received the Ph.D.
duced in Fig. 6. The general image for- Magazine, vol. 3, March 2002, pp. 48-56. degree in electrical engineering from
mation algorithm is summarized in Fig. • E.C. Fear, X. Li, S.C. Hagness and University of Victoria in 2001. She was
10. Results are included for the three M.A. Stuchly, “Confocal microwave an NSERC (Natural Sciences and
models described previously, and are imaging for breast tumor detection: Engineering Research Council of
shown in Figs. 11 to 13. In all cases, the localization of tumors in three dimen- Canada) Postdoctoral Fellow in electrical
key challenge is reducing the dominant sions,” IEEE Transactions on engineering at the University of Calgary,
reflections from the skin to allow for Biomedical Engineering, vol. 49, Aug. Calgary, AB, Canada from 2001-2002,
detection of the tumor. The algorithms 2002, pp. 812-822. and joined the same department as an
described in Fig. 10 effectively reduce • P.M. Meaney, M.W. Fanning, D. Li, assistant professor in July 2002. Her
this reflection, and the tumor response S. P. Poplack, and K.D Paulsen, “A clini- research interests involve the interaction
is evident in Figs. 11 to 13. Significant cal prototype for active microwave imag- of electromagnetic fields with living sys-
clutter remains in Fig. 12, suggesting ing of the breast,” IEEE Transactions on tems, including the interaction of low-
that the development of more effective Microwave Theory and Techniques, vol. frequency fields with biological cells.
clutter reduction algorithms is an impor- 48, 2000, pp. 1841-1853. Paul M. Meaney received the
tant avenue for future research. • C. Gabriel, S. Gabriel, and E. A.B.degrees in computer science and
Corthout. “The dielectric properties of electrical engineering from Brown
Conclusions biological tissues: I. Literature survey,” University, Providence, RI, in 1982,
Microwave systems for cancer detec- Physics in Medicine and Biology, vol. the M.S. degree in electrical engineer-
tion are coming of age. Enough promise 41, 1996, pp. 2231-2249. ing from the University of
is there to be optimistic that in the next • S. Gabriel, R. W. Lau, and C. Massachusetts, Amherst, in 1985, and
decade a viable system will become Gabriel, “The dielectric properties of the Ph.D. degree in biomedical engi-
available. It is not expected that such a biological tissues: II. Measurements on neering from Dartmouth College,
system or systems will replace x-ray the frequency range 10 Hz to 20 GHz,” Hanover, NH, in 1995. He was a
mammography as a screening tool. But Phys. Med. Biol., vol. 41, 1996, pp. Postdoctoral Fellow at Dartmouth
these systems used together can 2251-2269. College from 1995–1996 and a National
improve detection and limit false posi- • S. Gabriel, R. W. Lau, and C. Science Foundation (NSF)-NATO
tive findings. Perhaps for young women Gabriel. “The dielectric properties of Postdoctoral Fellow at the Royal
for whom mammography is not recom- biological tissues: III. Parametric mod- Marsden Hospital, Sutton, U.K., from
mended, a microwave system will els for the dielectric spectrum of tis- 1996–1997. He has been a Research
become a useful and painless low risk sues,” Phys. Med. Biol., vol. 41, 1996, Assistant Professor at Dartmouth
diagnostic aid. At present, there is still a pp. 2271-2293. College since 1997. His interests
great deal of research and engineering • K. R. Foster and H. P. Schwan, include developing microwave imaging
design to be completed before this opti- “Dielectric properties of tissues and bio- for biomedical applications—especially
mistic outlook becomes a reality. logical materials: a critical review,” breast imaging and hyperthermia moni-
Critical Reviews in Biomedical toring, along with elastography.
Acknowledgements Engineering, vol. 17, 1989, pp. 25-104. Maria A. Stuchly received the M.Sc.
This paper is adapted from a paper • S. S. Chaudhary, R. K. Mishra, A. in EE from the Warsaw Technical
appearing in the March 2002 edition of Swarup, and J. M. Thomas, “Dielectric University, Warsaw, Poland in 1962, and
the IEEE Microwave Magazine. We properties of normal and malignant the Ph.D. degree in EE from the Polish
would like to thank our co-authors on human breast tissues at radiowave and Academy of Sciences, Warsaw, Poland in
that paper, Dr. Susan Hagness, microwave frequencies,” Indian J. 1970. In 1970, she joined the University
University of Wisconsin and Dr. Michal Biochem. and Biophys., vol. 21, 1984, of Manitoba. In 1975, she joined the
Okoniewski, University of Calgary, for pp. 76-79. Bureau of Radiation and Medical Devices
their contributions. • A. J. Surowiec, S. S. Stuchly, J. R. in Health and Welfare, Ottawa, ON,
Barr, and A. Swarup, “Dielectric prop- Canada as a Research Scientist. In 1992,
Read more about it erties of breast carcinoma and the sur- she joined the University of Victoria,
• Samuel K. Moore, “Better Breast rounding tissues,” IEEE Trans. Biomed. Victoria, BC, Canada, as a Visiting
Cancer Detection,” IEEE Spectrum, Eng., vol. 35, 1988, pp. 257-263. Professor with the Dept. of Electrical
May 2001, p. 50-54. • W. T. Joines, Y. Z. Dhenxing, and R. and Comp. Engineering. Since January
• Mammography and Beyond: L. Jirtle. “The measured electrical proper- 1994, she has been a Professor and
Developing Technologies for the Early ties of normal and malignant human tis- Industrial Research Chairholder funded
Detection of Breast Cancer, sues from 50 to 900 MHz,” Medical by the Natural Sciences and Engineering
Committee on Technologies for the Physics, vol. 21, 1994, pp. 547-550. Research Council of Canada, BC Hydro
Early Detection of Breast Cancer, • O.P Gandhi, Qing-Xiang Li and and Trans Alta Utilities. Her research
Sharyl J. Nass, I. Craig Henderson, and Gang Kang, “Temperature rise for the interests are in numerical modeling of
Joyce C. Lashof, eds., National Cancer human head for cellular telephones and interactions of electromagnetic fields
Policy Board, Institute of Medicine, for peak SARs prescribed in safety with the human body and wireless com-
and Commission on Life Studies, guidelines,” IEEE Trans. Microw. munication antennas.

18 IEEE POTENTIALS
Authorized licensed use limited to: Alliance University. Downloaded on February 11,2025 at 05:20:33 UTC from IEEE Xplore. Restrictions apply.