Você está na página 1de 15

RADIO SCIENCE, VOL. 43, RS6S06, doi:10.

1029/2007RS003807, 2008
Click
Here
for
Full
Article
Evaluation of a hemi-spherical wideband
antenna array for breast cancer imaging
M. Klemm,1 I. J. Craddock,1 A. Preece,1 J. Leendertz,1 and R. Benjamin2
Received 10 December 2007; revised 11 June 2008; accepted 1 October 2008; published 19 December 2008.
[1] Using similar techniques to ground penetrating radars, microwave detection of breast
tumors is a potential nonionizing and noninvasive alternative to traditional body-imaging
techniques. In order to develop an imaging system, the team at Bristol have been
working on a number of antenna array prototypes, based around a stacked-patch element,
starting with simple pairs of elements and progressing to fully populated planar arrays.
As the system commences human subject trials, a curved breast phantom has been
developed along with an approximately hemi-spherical conformal array. This contribution
will present details of the conformal array design and initial results from this unique
experimental imaging system as applied to an anatomically shaped breast phantom.
Citation: Klemm, M., I. J. Craddock, A. Preece, J. Leendertz, and R. Benjamin (2008), Evaluation of a hemi-spherical
wideband antenna array for breast cancer imaging, Radio Sci., 43, RS6S06, doi:10.1029/2007RS003807.

1. Introduction [4] To date most of the work on the breast cancer


detection have been based on the computer simulations
[2] Breast cancer is the most common cancer in [Kosmas et al., 2004; Abas et al., 2007]. There have been
women. X-ray mammography is currently the most only a few experimental breast-imaging radar systems
effective detection technique; however, it suffers from reported in the open literature [Craddock et al., 2005; Sill
relatively high missed- and false-detection rates, involves and Fear, 2005]. In this paper we present for the first
uncomfortable compression of the breast and also entails time an assessment of techniques to de-embed the tumor
exposure to ionizing radiation. Microwave detection of response from real experimental data.
breast tumors is a potential nonionizing alternative being
investigated by a number of groups [Hagness et al.,
1998; Fear et al., 2003; Fear, 2005; Bialkowski and Wee, 2. Development of an Experimental System
2007]. In these microwave-based systems, in a similar 2.1. Antenna Design
fashion to ground penetrating radars, microwaves are
transmitted from an antenna or antenna array, and the [5] A prerequisite for all microwave imaging systems
received signals, which contain reflections from tumors, is a suitable antenna array. Initial work concentrated on
are recorded and analyzed. developing a simple but low-profile and wide-band
[3] This contribution presents details of the conformal antenna that would cover the 4 – 10 GHz frequency
radar-based breast cancer detection system. Unlike the range. An aperture stacked-patch antenna was designed
other published work on the subject, the developed for this purpose. The antenna used herein is a modified
experimental radar system operates in a multistatic mode, version of the antenna presented in [Nilavalan et al.,
originally proposed for breast cancer and land mine 2007], where it was employed in a planar array for breast
detection by [Benjamin, 1996]. Compared to the mono- imaging.
static approaches, a multistatic approach with a fully [6] For the conformal array, the antenna was rede-
populated antenna array enables far more data to be signed. The final antenna design in presented in Figure 1.
gathered. The antenna cross-section, dielectric materials and size
of individual patches were kept the same as in [Nilavalan
et al., 2007]. Only the ground plane size (and hence a
1
feeding line substrate) was substantially reduced to 28 
Centre for Communications Research, Department of Electrical
and Electronic Engineering, University of Bristol, Bristol, UK.
17 mm2. Dimensions of two dielectric substrates where
2
Bristol, UK. patches are printed are 17  17 mm2. Additionally, as we
learned from the experience of using the planar array
Copyright 2008 by the American Geophysical Union. [Craddock et al., 2005] it is better to shield the antenna
0048-6604/08/2007RS003807$11.00 from the surrounding environment, therefore we added a
RS6S06 1 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 1. Cavity backed aperture stacked-patch antenna for breast cancer detection.

cavity at the back of the antenna. The cavity has a planar bandwidth of the transmit transfer function is about 7
inner dimensions of 18  11 mm2 and is 12 mm long. To GHz (from 3.5 to 10.5 GHz).
absorb the back radiation of the antenna and avoid any [8] Additionally, we performed a transmission measure-
resonances the cavity was lined with a broad-band ment between two antennas (face to face, 10 cm separa-
absorbing material (Eccosorb FGM-40 from Emmerson tion) immersed in a lossy matching liquid (described in
& Cumming). the following paragraph). As an input pulse we chose the
[7] In Figure 2 we present the measured antenna input waveform presented in Figure 4, which covers a fre-
match (S11), which shows that the antenna is matched quency range between 4 and 9 GHz on a 3 dB level. As
(S11 < 5 dB) between 4 and 10 GHz. The good described by Hines and Stinehelfer [1974], this type of
transient performance in this frequency range is visible pulse is suitable for time domain analysis of microwave
when looking at the simulated (FDTD) transmit transfer systems when performing measurements in the frequency-
function of the antenna shown in Figure 3. The 10 dB domain. The resulting pulse transmitted between our

Figure 2. Measured S11 (input match) characteristic of the antenna.


2 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 3. Simulated (FDTD) transmit transfer function (at boresite) of the antenna.

antennas and its spectrum is shown in Figure 5. The 2.2. Conformal Symmetrical Antenna Array Design
transmitted pulse is clearly longer than the input pulse,
[9] Given the effort in designing and constructing a
due to the antenna’s response but also due to a lossy and
conformal hemi-spherical array, the intention from the
dispersive immersion medium. Comparing the simulated
outset was to design not only an array for laboratory use
transfer function of the single antenna (see Figure 4) with
on a realistic, curved phantom, but also one that would
the measured two-antenna transfer function (see Figure 5),
serve as an initial clinical prototype. Approximately 20
we can clearly see the effect of the lossy medium. The
female volunteers came forward from the University and
10 dB bandwidth of the measured transfer function was
the fit between their breasts and various plastic spherical
reduced to 3 GHz (3.5 – 6.5 GHz), for the 10 cm distance
sections was assessed with them lying in a prone position
between antennas. Obviously, as the distance between
- the prone (face-down) position being felt to offer the
antennas with change, also the transfer function will have
best chance of the breast forming a gently and uniformly
different 10 dB bandwidth and also its maximum value
curved shape. Following this assessment, the dimensions
will be at slightly different frequency. From these trans-
of the array were input into a 3-D CAD model, along
mission experiments it is apparent that dispersive losses
with the antenna elements and all supporting metalwork.
of the normal breast tissue will have a biggest impact on
[10] The resulting antenna array is formed around
limiting the achievable temporal resolution.
lower part of a 78 mm-radius sphere, in four rows of
four antennas. The side view of the array and breast

Figure 4. Synthetic pulse used as an antenna excitation: (left) time domain waveform and (right)
spectrum.
3 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 5. Two antenna transfer function (along boreside direction). Antennas were immersed in
the lossy matching liquid, 10 cm distance.

model is shown in Figure 6. The arrangement of antenna [13] Next, a curved skin phantom was developed. The
elements (top view) seen in Figure 7 gives enough skin layer is 2 mm thick, it is a part of a 58 mm-radius
clearance for the cables and connectors, which pass hemi-sphere. When the skin phantom is fitted into the
between the elements of adjacent rows. The partly array, as shown in Figure 9, it lies 20 mm above the
constructed array is shown in Figure 8. antenna elements. This standoff between antennas and
[11] The clear advantage of the designed curved array breast provides a reasonable coverage of a breast by an
is its conformity to the breast’s shape, providing a good antenna radiation pattern. The electrical parameters of
breast coverage by antennas radiation patterns. The real the skin layer were chosen again according to the
aperture array, together with the switching network, give previously published data: the material is dispersive
a fast data acquisition. In a single scan there is no and at 6 GHz it has a relative dielectric constant of 30
mechanical scanning involved, saving a lot of measure- and attenuation of 16 dB/cm.
ment time (it is important for measurement with real [14] After fitting the skin phantom into antenna array, a
patients). Another benefit of this radar system is the fact plastic tank is connected to the array as presented in
that it operates in a multistatic mode, what gives a far Figure 10. The tank is then filled with the normal breast-
greater spatial diversity compared with the monostatic fat equivalent liquid (the same as matching liquid).
operation. The main disadvantage of the presented real [15] A tumor phantom material with a relative dielec-
aperture radar system is the antenna coupling, as well as tric constant close to 50 and conductivity 7 S/m (at 6 GHz)
additional reflection from mechanical parts of the array. was also developed. This gives a contrast between
We will describe later in the paper how to deal with these dielectric properties of breast fat and tumor phantom
undesired signals.
2.3. Three-Dimensional Physical Breast Phantom
[12] For experimental testing we developed appropri-
ate materials and a 3-D breast phantom. As shown in
Figure 6, during the measurements the antennas are
immersed in a matching liquid, to reduce reflections
from the skin and for a more compact antenna design.
We decided that this matching liquid would be the same
as the material simulating properties of a normal breast-
fat, mainly for practical reasons (only one liquid required
in manufacturing). The developed matching and normal
breast tissue equivalent liquid [Leendertz et al., 2003]
has a relative dielectric constant of about 9.5 and
attenuation of 1.2 dB/cm at 6 GHz. This material is also
dispersive (see Craddock et al. [2005] for its frequency- Figure 6. Hemi-spherical antenna array and breast
dependent characteristics). phantom configuration: side view.
4 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 7. Positions of antenna elements in the array, looking from the top.

Figure 8. Partly constructed real antenna array.


5 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

[20] The characteristic equation of the improved DAS


algorithm is expressed as:
Z t
Fe ð x; y; zÞ ¼ QF ð x; y; zÞ 
0
!2
X
M
 wi ð x; y; zÞ  yi ðt  Ti ð x; y; zÞÞ dt
i¼1

ð1Þ

where M = N(N-1)/2 (N is the number of antennas in the


array), wi is the location dependent weight calculated
during preprocessing (steps 2 and 3 of the preproces-
sing), yi is the measured radar signal and Ti is the time
delay. The time delay Ti for a given transmitting and
Figure 9. Skin phantom fitted into the antenna array. receiving antenna is calculated based on the antenna’s
position, position of the focal point r = (x, y, z) as well as
an estimate of average wave propagation speed, which in
materials of around 1:5. Recently published data in our processing is simply assumed to be constant across
Lazebnik et al. [2007], based on a large clinical study, the band (though in our experiments it will not be). As
however suggest that the contrast between healthy and we recently presented in Klemm et al. [2008], by using
malignant breast tissues might be lower, at least in some this additional weighting factor QF, quality of images is
women. significantly improved due to the clutter reduction
capability of the new algorithm.
3. Focusing Algorithms
4. Array Evaluation Results and Imaging
[16] To obtain the 3-D image of the scattered energy,
we employ postreception synthetic focusing. We employ Results
a modified version of a classical delay-and-sum (DAS) 4.1. Tumor Response De-embedding Techniques
beamforming [Benjamin et al., 2001], briefly described
below. [21] Before applying the preprocessing step and the
focusing algorithm, the tumor response must be extracted
3.1. Preprocessing: Equalization from measured data. Measured data contain the tumor
response, as well as additional undesired signals (antenna
[17] Before applying the focusing algorithm we have coupling, reflections from the skin, reflections from
to perform a preprocessing step. This process aims at
equalization of scattered tumor responses for different
antenna pairs. Ideal preprocessing would result in all
received pulses being of the same shape, amplitude and
perfectly time-aligned. In our preprocessing the follow-
ing steps are performed: (1) extraction of the tumor
response from measured data, (2) equalization of tissue
losses, and (3) equalization of radial spread of the
spherical wavefront.
[18] In the work reported herein there will be frequency-
dependence both of the tissue losses and of the radiation
patterns of the antennas, however for simplicity we do
not attempt to correct for these frequency-dependencies
in our processing.
3.2. Modified Delay-and-Sum Algorithm
[19] The modified DAS algorithm uses an additional
weighting factor QF (quality factor), compared to the
standard DAS. QF can be interpreted as a quality factor
of the coherent focusing algorithm.
Figure 10. Antenna array together with the tank.
6 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 11. Comparison of two radar signals (raw measured data) used in background subtraction
technique. Both signals are for the same multistatic path (transmission between antennas 1 and 2, as
shown in Figure 8), with and without the tumor inside breast phantom.

mechanical parts of the array). To subtract all the [28] 4. Finally, while the array is in position AP2, the
unwanted signals, we employ two techniques: either tumor is removed from the breast phantom and the
(1) background subtraction, or (2) rotation subtraction. second background measurement BAP2 is obtained.
[22] In the background subtraction method the tumor [29] These four data sets provide a basis for obtaining
response is extracted using two measurements, with and three focused images of the phantom: two images using
without the tumor present in the breast. This method is the impractical but ideal background subtraction method
effective and useful in evaluating the array imaging and one using the more practical rotation subtraction
properties, as well as in array calibration, but cannot method. The tumor responses TR for respective images
possibly be used with real patients. are obtained using the following operations on the
[23] The second method, rotation subtraction, also recorded time domain signals (not on the images): (1)
employs two measurements, but with the tumor present Image 1: TRI1 = TAP1  BAP1; (2) Image 2: TRI2 = TAP2
in the breast in both cases. The first measurement is  TAP1; (3) Image 3: TRI3 =TAP2  BAP2.
performed with the array in a given position, then the
array is rotated (in a horizontal plane, around its central 4.2. Comparison of Tumor Response Extraction
vertical axis) and a second measurement is recorded. Methods: Evaluation of Experimental Radar Signals
Because the rotation subtraction method does not require [30] In this paragraph we compare the tumor response
a background measurement, it can potentially be used in extraction techniques described above, by investigating
realistic scenarios with breast cancer patients. individual multistatic radar signals. Additionally, we
[24] To obtain the relevant measured data for back- show how we minimize effects of antenna mutual
ground and rotation subtraction methods, the following couplings.
measurement steps were performed: [31] We compare effectiveness of two extraction tech-
[25] 1. First, in a given array position (AP1), a back- niques by looking at respective raw measured multistatic
ground measurement BAP1 is taken. There is no tumor radar signals. As an example, we chose a pulse trans-
present in the breast phantom. mitted between antennas 1 and 2 (see Figure 8 for
[26] Then, as the array stays in the same position AP1, antenna positions), where we can also observe the effect
we introduce a tumor into the breast phantom at a certain of mutual coupling between antennas. In Figure 11 we
location and a measurement TAP1 is recorded. can see two radar signals, without and with the tumor
[27] 3. Next, keeping the breast phantom with tumor inside the breast phantom, used in background subtrac-
fixed we rotate the array to a second position AP2. After tion method. As described in the previous paragraph, by
rotation we perform a second measurement of the breast subtracting these two signals we can easily extract tumor
with tumor (TAP2). response from measured signals. This method however is

7 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 12. Comparison of two radar signals (raw measured data) used in rotation subtraction
technique. Both signals are for the same multistatic path (transmission between antennas 1 and 2, as
shown in Figure 8), but at two array positions: AP1 and AP2.

not clinically useful, because the background measure- reflected from the skin and then the tumor response.
ment (without tumor inside a breast) will not be available As we can see, the tumor response is very small and
with real cancer patient. In the figure we indicated a hidden within the late-time portion of the received signal.
different parts of the radar signal. As this is a signal The tumor response can only be extracted by subtracting
transmitted between two adjacent antennas in the array, a the background signal (shown later in a paper in
first signal received in a direct signal coupled from the Figure 13). Other parts of the signal (antenna coupling
transmitting antenna. Next, we can identify a signal

Figure 13. Comparison of extracted tumor response when using background subtraction and
rotation subtraction methods.
8 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 14. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P1: x = 10, y = 0, z = 20: (a) 3-D focused image, (b) 2-D image through the
horizontal plane z = 15, (c) channel data at the location of the detected tumor (x = 9, y = 3, z =
15), (d) as Figure 14c but normalized. Images were obtained using background subtraction for
array position AP1. Two-dimensional contour plot shows signal energy on a linear scale,
normalized to maximum in the 3-D volume, values below 0.1 rendered as white.

and skin reflection) are almost identical for cases with line) provides a good tumor response, with very small
and without the tumor. distortions. The antenna coupling and skin reflection
[32] Next, in Figure 12 we can see radar signals used in signals were nicely canceled. Significantly more distor-
the second tumor extraction technique: rotation subtrac- tions can be seen in the rotation subtraction signal
tion. Signals are shown for the same multistatic channel (dashed line). The tumor response was reasonably recov-
as used in background subtraction. Both signals from ered from measured data. However, signals at other
Figure 12 contain tumor response, they were obtained ranges also appeared. Especially the skin reflection
when array was in two different positions due to array signal is visible. It has actually higher magnitude that
rotation (AP1 and AP2, as described in previous para- the tumor response. The remaining of the sin reflection
graph). Again we can distinguish signals originating signal is due to measurement imperfections during array
from antenna coupling, skin reflection and the tumor. rotation. Because in raw measured data the skin reflec-
[33] In Figure 13 we present signals obtained from tion signal is few orders of magnitude stronger than
background and rotation subtraction methods. We can tumor backscattered response, even small changes in
see that the ‘ideal’ background subtraction method (solid distance between antennas and the skin layer, or changes

9 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 15. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P1: x = 10, y = 0, z = 20: (a) 3-D focused image, (b) 2-D image through the
horizontal plane z = 15, (c and d) channel data at the location of the detected tumor (x = 9, y = 0,
z = 15). Images obtained using background subtraction for array position AP2.

in a skin thickness will result in skin reflection signal at


4.3. Comparison of Tumor Response Extraction
two array positions not being the same. This can be
Methods: Experimental Imaging Results
observed in Figure 12 on raw signals (before applying
subtraction; only small difference in amplitudes for both [35] In this section we compare the techniques pre-
skin reflection signals, hardly visible). It is worth adding, sented above for tumor response extraction and their
that the antenna coupling signal was also well canceled impact on imaging quality. Using the system described in
using rotation subtraction technique. section 2, we performed a number of measurements of
[34] The problem of remaining signals when using tumors in different locations within the breast phantom.
rotation subtraction although clearly visible on individual [36] We present an example of a 8 mm (diameter)
multistatic radar signals, is not very critical overall. spherical tumor phantom at two different locations in the
These clutter signals will usually add incoherently during breast phantom. For each location, we compared the
focusing processing. But the tumor response will be three focused images obtained using the tumor response
combined coherently, as presented in a next section. In extraction methods described above. Additionally, we
conclusion, we can clearly see the difference between evaluated the quality of coherent radar operation by
ideal (not practically useful) background subtraction and examining the radar signals obtained after focusing at
the practically useful rotation subtraction method. the focal location where the tumor was detected.

10 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 16. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P1: x = 10, y = 0, z = 20: (a) 3-D focused image, (b) 2-D image through the
horizontal plane z = 15, (c) channel data at the location of the detected tumor (x = 9, y = 3, z =
21), (d) as Figure 16c but normalized. Images obtained using rotation subtraction for array
positions AP1 and AP2.

4.3.1. Results for Tumor Location P1: x = 10, y = 0, image (in gray). There is a slight shift in the location of
z = 220 mm the detected tumor, but otherwise the image is clear and
[37] Below we present experimental tumor detection of good quality (there is no clutter present in the image).
results for a 8 mm spherical tumor phantom located in The small spatial offset is most likely due to nonideally
the breast at a position P1: x = 10, y = 0, z = 20 (all coherent summation of received pulse (which differ
positions quoted in mm). Following the measurement slightly in shape and duration).
procedure described in section 4.1, three focused images [39] When looking at the horizontal plane where tumor
of the detected tumor were created, two using background was detected (z = 15), shown in Figure 14b, we can
subtraction method and one using rotation subtraction. easily identify the tumor. Moreover, there is no visible
[38] In Figure 14 we present results for the background clutter in this 2-D image, demonstrating the good quality
subtraction method with an array in position AP1 (as of our imaging process. Two-dimensional contour plots
described in section 4.1). Figure 14a shows a three- show signal energy on a linear scale, normalized to
dimensional (3-D) focused energy image (3 dB maximum in the 3-D volume, values below 0.1 rendered
energy contour), with the tumor detected at position: as white.
x = 9, y = 3, z = 15. The skin is also shown in the

11 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 17. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P2: x = 0, y = 30, z = 20: (a) 3-D focused image, (b) 2-D image through
the horizontal plane z = 15, (c) channel data at the location of the detected tumor (x = 6, y = 24,
z = 12), (d) as Figure 17c but normalized. Images obtained using background subtraction for
array position AP1.

[40] To further evaluate the quality of our results, we [41] In Figures 14c and 14d we can see the array
investigated the coherence quality of our radar system. channel data at the focal point where the tumor was
As described in section 3.2, using a synthetic postrecep- detected (x = 9, y = 3, z = 15). Figure 14c shows the
tion focusing algorithm, by proper time-shift and align- signal with absolute amplitude values, to investigate
ment of received signals, coherent operation of the radar time-alignment but also the amplitude spread of the
should be achieved. Therefore, at the focal location signals after preprocessing. The figure shows about
where tumor is located, after ideal focusing and prepro- 800 time samples (x axis) of the received channel data
cessing of all radar channels the received pulses should (with 16 antennas we record 120 channels, each dis-
have the same amplitude and be time-aligned. In reality played as a line parallel to the x axis). The black lines on
we cannot expect all pulses to be the same, because we the figure represent the extent of the integration window
do not account for frequency-dependent factors (e.g. t used in our modified DAS algorithm. We can observe
antenna characteristics, losses). Also, the tumor scatter- that pulses have a similar amplitude across all channels,
ing is angle-dependent, resulting in slightly different demonstrating their good equalization. In Figure 14d the
pulse shapes received by different antennas. However, same data are presented but with the amplitudes normal-
we do expect the pulses to be well time-aligned. ized to the maximum in each channel, we can see more

12 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 18. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P2: x = 0, y = 30, z = 20: (a) 3-D focused image, (b) 2-D image through the
horizontal plane z = 15, (c) channel data at the location of the detected tumor (x = 3, y = 24,
z = 12), (d) as Figure 18c but normalized. Images obtained using background subtraction for
array position AP2.

easily the good time-alignment of the received pulses. the tumor response from raw measured data. Focused
The slight differences in time-shifts at individual chan- energy images (Figures 16a and 16b for 3-D and 2-D
nels are most likely due to the effects discussed above. results, respectively) show that although tumor was again
[42] In Figure 15 we present imaging results and detected without any problems, detection quality is
channel data for the second background measurement, slightly degraded. Beside the tumor response, there is
after the array was rotated by 10 degrees to position AP2. also some clutter present in images. The reason for this
We can observe in the focused images (3-D in Figure 15a can be seen by looking at channel data in Figures 16c
and 2-D in Figure 15b), that the tumor phantom is again and 16d. Here the received channel data have a much
successfully detected. However, due to the array rotation larger amplitude spread, compared with results for back-
it is now seen at a slightly different location (x = 9, y = 0, ground subtraction. Moreover, the coherence of pulses is
z = 15) by the array. The channel data presented in also not as good.
Figures 15c and 15d show again that good coherent radar [44] The degraded performance can be explained by
operation is achieved. differences in time-shifts for different array channels, due
[43] Finally, in Figure 16, results are shown for the case to the difference in the physical displacement of each
when rotation subtraction method was used to de-embed antenna during rotation: when rotating the array around

13 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

Figure 19. Experimental imaging and antenna array evaluation results for 8 mm spherical tumor
located at position P2: x = 0, y = 30, z = 20: (a) 3-D focused image, (b) 2-D image through the
horizontal plane z = 15, (c) channel data at the location of the detected tumor (x = 3, y = 27,
z = 12), (d) as Figure 19c but normalized. Images obtained using rotation subtraction for
array positions AP1 and AP2.

its center, antennas close to the center of rotation and 17b for position AP1, and Figures 18a and 18b for
experience smaller displacement than antennas located position AP2, we can observe that tumor was in both
further away. Nevertheless, this rotation subtraction still cases easily detected. However, the difference in the
provides performance sufficient to detect small tumors in location of the detected tumor is larger than for the first
the phantom. tumor location, as would be expected (due to larger
4.3.2. Results for Tumor Location P2: x = 0, y = 30, physical displacement for the same angle of rotation).
z = 220 mm The tumor was detected at position: x = 6, y = 24, z = 12
[45] We present here results similar to those shown in for AP1, and position: x = 3, y = 24, z = 12 for AP2.
the previous subsection, but for a tumor located at a Focused images are clear with no clutter visible. This
different position: P2: x = 0, y = 30, z = 20. An 8 mm good performance is again confirmed by the channel data
spherical tumor was again used in the measurements. (Figures 17c and 17d and Figures 18c and 18d) at the
[46] In Figures 17 and 18 we present results for two focal points where the tumor was detected: signals are
background subtraction measurements, with the array in relatively well-aligned and with comparable amplitudes.
position AP1 and AP2. The array was rotated by 10 degrees, [47] When using rotation subtraction, the tumor was
as before. By looking at focused images in Figures 17a also detected, as seen in the focused images Figures 19a

14 of 15
RS6S06 KLEMM ET AL.: BREAST CANCER IMAGING RS6S06

and 19b. Although a small amount of clutter exists in Bialkowski, M. E., and C. K. Wee (2007), Investigations into an
these images, their overall quality is more than satisfac- uwb microwave radar system for breast cancer detection,
tory. The channel data in Figures 19c and 19d show IEEE Antennas Propag. Int. Symp., 2160 – 2163.
again that the signals vary more in amplitude after Craddock, I. J., R. Nilavalan, J. Leendertz, A. Preece, and
rotation subtraction, compared to the ideal background R. Benjamin (2005), Experimental investigation of real aper-
subtraction, and they are also less aligned. But the good ture synthetically organised radar for breast cancer detection,
quality of focused images shows that this effect of IEEE Antennas Propag. Soc. Int. Symp., 1B, 179 – 182.
smaller coherence is equally translated onto the entire Fear, E. C. (2005), Microwave imaging of the breast, Technol.
focusing domain. This is an encouraging conclusion of Cancer Res. Treatment, 4, 69 – 82.
our evaluations, since possibly only the rotation subtrac- Fear, E. C., P. M. Meaney, and M. A. Stuchly (2003), Micro-
tion method could be used in realistic breast cancer waves for breast cancer detection?, IEEE Potentials, 22(1),
detection scenarios. 12 – 18.
Hagness, S. C., A. Taflove, and J. E. Bridges (1998), Two-
dimensional FDTD analysis of a pulsed microwave confocal
5. Conclusions and Future Work system for breast cancer detection: Fixed-focus and antenna-
[48] In this contribution we presented a microwave array sensors, IEEE Trans. Biomed. Eng., 45(12), 1470 –
system for breast cancer detection, employing a confor- 1479.
mal antenna array. The antenna elements populate the Hines, M. E., and H. E. Stinehelfer (1974), Time-domain oscillo-
inside of a section of a hemisphere, this being a suitable graphic microwave network analysis using frequency-domain
geometry for clinical application. A 3-D physical breast data, IEEE Trans. Microwave Theory Tech., 22, 276 – 282.
phantom was also presented including a geometrically Klemm, M., I. J. Craddock, J. A. Leendertz, A. Preece, and
and electrically realistic skin (this being the dominant R. Benjamin (2008), Improved delay-and-sum beamforming
source of clutter). algorithm for breast cancer detection, Int. J. Antennas Pro-
[49] The new investigations presented herein have pag., 761402, doi:10.1155/2008/761402.
focused on the evaluation of two methods of de- Kosmas, P., C. M. Rappaport, and E. Bishop (2004), Modeling
embedding tumor response from raw measured data. with the FDTD method for microwave breast cancer detec-
We compared quality of detection as well as coherence tion, IEEE Trans. Microwave Theory Tech., 52(8), Part 2,
quality of these methods. Results show that both de- 1890 – 1897.
embedding techniques provide good quality images and Lazebnik, M., et al. (2007), A large-scale study of the ultrawi-
there is no difficulty in detecting 8 mm spherical tumor at deband microwave dielectric properties of normal breast tis-
different locations within the breast phantom. However sue obtained from reduction surgeries, Phys. Med. Biol., 52,
using a rotational subtraction the coherent radar opera- 2637 – 2656.
tion is degraded, compared to the ideal (but unachiev- Leendertz, J., A. Preece, R. Nilavalan, I. J. Craddock, and
able) case of background subtraction. R. Benjamin (2003), A liquid phantom medium for micro-
[50] This effect is due to the different physical dis- wave breast imaging, paper presented at 6th International
placement of each antennas after array rotation arising Congress of the European Bioelectromagnetics Association,
from the different antenna locations in the array. We Budapest, Hungary, Nov.
believe that this effect can be lessened to a certain degree Nilavalan, R., I. J. Craddock, A. Preece, J. Leendertz, and
and this is the subject of ongoing research. R. Benjamin (2007), Wideband microstrip patch antenna
design for breast cancer tumour detection, IET Microwaves
References Antennas Propag., 1(2), 277 – 281.
Sill, J. M., and E. C. Fear (2005), Tissue sensing adaptive radar
Abas, S., X. Meilian, N. Sima, T. Parimala, and S. Pistorius
for breast cancer detection - Experimental investigation of
(2007), Efficient microwave breast imaging technique using
simple tumor models, IEEE Trans. Microwave Theory Tech.,
parallel finite difference time domain and parallel genetic
53(11), 3312 – 3319.
algorithms, IEEE Antennas Propag. Int. Symp., 2176 – 2179.
Benjamin, R. (1996), Synthetic, post-reception focusing in
near-field radar, paper presented at EUREL International 
R. Benjamin, 13 Bellhouse Walk, Kingsweston, Bristol, UK.
Conference, The Detection of Abandoned Land Mines: A
I. J. Craddock, M. Klemm, J. Leendertz, and A. Preece,
Humanitarian Imperative Seeking a Technical Solution, 7 – 9
Centre for Communications Research, Department of Electrical
Oct.
and Electronic Engineering, University of Bristol, Merchant
Benjamin, R., I. J. Craddock, G. S. Hilton, S. Litobarski,
Venturers Building, Woodland Road, Bristol BS8 1UB, UK.
E. McCutcheon, R. Nilavalan, and G. N. Crisp (2001), Micro-
(m.klemm@bristol.ac.uk)
wave detection of buried mines using non-contact, synthetic
near-field focusing, IEE Proc. Radar Sonar Navig., 148(4),
233 – 240.
15 of 15

Você também pode gostar