Breast Tumor Detection Using UWB Circular-SAR Tomographic
Microwave Imaging
Daniel Oloumi, Student Member, IEEE, Pierre Boulanger, Member, IEEE, Atefeh Kordzadeh,
Karumudi Rambabu, Member, IEEE

Abstract— This paper describes the possibility of
detecting tumors in human breast using ultra-wideband
(UWB) circular synthetic aperture radar (CSAR). CSAR
is a subset of SAR which is a radar imaging technique
using a circular data acquisition pattern. Tomographic
image reconstruction is done using a time domain global
back projection technique adapted to CSAR. Experiments
are conducted on a breast phantoms made of pork fat
emulating normal and cancerous conditions. Preliminary
experimental results show that microwave imaging of a
breast phantom using UWB-CSAR is a simple and low-
cost method, efficiently capable of detecting the presence
of tumors.
I. INTRODUCTION
Early detection of breast cancer reduces the risk of
mortality. For that reason, different imaging modalities
have been developed for detecting tumors in human
breast. X-ray mammography is the most commonly used
method for early screening of breast cancer as it is
relatively cheap [1], [2]. However it is limited in
sensitivity and fails to detect breast cancerous tissues in
many cases [3]. It also uses ionizing radiation that can
be detrimental to the patient over a long period [4].
Computed tomography scan (CT-scan) and magnetic
resonance imaging (MRI) are other methods to image
and detect cancerous tissues in human breast based on
high resolution images [5], [6]. Nevertheless, both of
these methods are very expensive.
Considering the high cost and complexity of current
methods, microwave imaging may turn out to be a
simple and efficient method to perform breast imaging.
Microwave imaging is a simple and cheap method
capable of providing adequate image resolutions.
Extensive research has been done on microwave imaging
of breast and have resulted in promising progresses [7]–
D. Oloumi, and K. Rambabu, are with the Electrical and Computer
Engineering Department, University of Alberta, Edmonton, AB T6G
2V4 Canada (e-mail: oloumi@ualberta.ca, rambabu@ualberta.ca)
A. Kordzadeh is with Biomedical Engineering Department and joint
with Electrical and Computer Engineering Department, University of
Alberta, Edmonton, AB T6G 2V2 (Canada),
(email:kordzade@ualberta.ca)
P. Boulanger is with the Computing Science Department, University
of Alberta, Edmonton, AB T6G 2V4 Canada (e-mail:
pierreb@ualberta.ca).
. data is acquired over the radar trajectory at every 5ᵒ
978-1-4244-9270-1/15/$31.00 ©2015 IEEE 7063
[11]. The semi-spherical shape of human breast is well-
suited to projection measurements. Microwave
tomography, can produce high-resolution images
especially for circular data acquisition system. However,
a complex inverse problem needs to be solved in order to
reconstruct images [12].
In this work, an Ultra wide-band (UWB) circular
synthetic aperture radar (CSAR) is used for the human
breast imaging. The raw data is acquired by sending and
receiving UWB pulses into the breast tissues from
various viewpoints. A tomographic image reconstruction
is then performed using a global back projection (GBP)
algorithm, in time domain, which is very suitable for
UWB pulses [13]. GBP based on tomographic
reconstruction with CSAR is very fast and easy to
implement. Image resolution provided by linear
synthetic aperture radar (LSAR) depends on pulse shape,
width, and integration angle [14]. However, using CSAR
imaging technique, radar image resolution depends only
on the pulse width and shape.
In this paper, Section II describes the UWB CSAR
algorithm and its implementation in Matlab. In Section
III, various measurements, on normal and cancerous
tissue inside a breast phantom, were performed to
evaluate the ability of UWB-CSAR for breast imaging
applications. In Section IV, we will demonstrate the
capability of UWB-CSAR algorithm to provide images
with adequate resolution to image the internal structure
of the breast and to detect the presence of tumors.
II. UWB CIRCULAR SAR
In this section, the CSAR configuration,
mathematical implementations, and image
reconstruction algorithm are described.
A. Raw data acquisition
The circular SAR system scheme is shown in Fig.
1(a). Radar transceiver rotates around the area to
acquire the raw data from multiple angles. A second
derivative Gaussian pulse, with 10 GHz bandwidth and
center frequency of 6.5 GHz, is used as the transmitted
pulse. The raw data is generated by two point targets as
shown in Fig. 1(a). Targets are located at locations (-
1.75mm, 1.75mm) and (1.75mm, -1.75mm) in a
medium with relative dielectric constant of 5. The raw
 0 0
1 1
Angle [Radians]
[degree]
2 2
3 3
4 4
5 5
6 6
0 1
Time [ns]
(a)
(b) (c)
Fig.1. (a) Radar scanning style. (b) Raw data (c) envelope of
detected raw data
increment and is shown in Fig.1 (b). The measured data
is in the form of a traditional sinogram as the
measurement path is circular. Envelope of the received
pulses are also generated as may eventually provide the
ability to resolve targets more clearly [14]. As shown in
Fig.1 (c), envelope detection of the received pulse is
performed by applying the Hilbert transform to the raw
data as shown in Fig.1 (c).
B. Image reconstruction
Image reconstruction is performed using a time
domain global back projection (GBP) algorithm [13].
The GBP is a coherent summation of the raw data,
measured at different angles over the radar trajectory, at
each pixel of the reconstructed image. The measured
data is in the form of N-theta and Nr-data which is the
number of aperture positions over the radar trajectory
and time samples, respectively. The image is
reconstructed as a M  N pixels image in the x-y
directions. Here M and N are selected to be 500 in each
direction. To find the appropriate sample data in time,
the round trip time measurement from each aperture
position to each image pixel, t (M , N ) , is calculated as
follows:
DM , N  (1)
t (M , N ) 
Vs
Where the distance, D(M , N ) , is given by:
D( M , N )  X pf Tr  XM 2
 Y pf Tr  YM
X pf  Rx X 
M
2
 Y pf  Rx  YM
Here V s is the signal speed for SAR processing, which
is same as pulse group velocity in the medium, and XM
and YM are pixels’ coordinates in the reconstructed
image. X pf Tr / Rx , Ypf Tr / Rx are the positions of the
transmitting (Tx) and receiving (Rx) antennas and are
defined by:
X pf Tr / Rx  r. cos( pf Tr / Rx )
Ypf Tr / Rx  r. sin( pf Tr / Rx ).
1
The round trip time is found by dividing round trip
0.8 distance by signal speed as shown in (1). Dividing the
Angle [Radians]
round trip time by the time resolution, t , provides the
0
[degree]
0.6
0.4
index of appropriate sample of the received pulse as
-0.5
follows:
0.2
 t (M , N ) 
IndexM , N   round  . (4)
2
-1
0 1
Time [ns]
2  t 
The precise sample for each image pixel is picked by
the nearest neighbor method. The image is reconstructed
through:
2
I M , N    rawdata( , index( M , N ).) (5)
 0
In which I and θ are reconstructed image and radar
measurement angles, respectively. A volumetric 3D
image can be produced by scanning the breast in the z-
direction, in different slices.
The reconstructed images in both real and absolute
value are shown in Fig.2 (a) and Fig.2 (b). One can see
that targets are detected and resolved. Side lobes
appeared in the reconstructed images are due to the
GBP and can be removed by generating a positive
image defined in [14] by:
Positive( I )  abs( I )  real( I ). (6)
The positive image in Fig.2 (c) shows clear boundaries
of the targets. Fig.2 (d) shows the reconstructed image
using the envelope detected raw data. As one can see
that the targets are detected and imaged more clearly.
The blurring effects appeared in the envelop image is
due the smoother curve of the envelope detected raw
data. The blurring effect can be removed by applying a
high-pass filter to the image similar to the back
projection filter used in traditional CT reconstruction.
C. CSAR image resolution
Since the target is scanned from all angles in a
circular pattern, range and cross-range are the same.
Therefore, a general term “resolution” is assigned to
both. Unlike LSAR [14], image resolution provided by
CSAR only depends on the pulse width. The image
2
 (2) resolution by a CSAR configuration based on using a
 2
. second derivative of Gaussian pulse as the transmitted
pulse is defined as:
c
r = (6)
3BW  r
where, c, BW , and  r are the speed of light, pulse
bandwidth, and relative permittivity of the medium,
respectively. Hence, narrower pulse width or higher
value of medium permittivity results in higher CSAR
image resolution. In our simulation, CSAR images with
4.5 mm resolution were achieved as a result of using 10
(3) GHz bandwidth pulse and medium of propagation with
a relative dielectric constant of 5.
7064
 Real(image) Positive(image) 1
Sr(t) Samb(t) Sc(t) Sh(t)
2

Normolized amplitude
-0.02 0 -0.02 0
1.5

X[m]
X[m]

0 0 1
-0.5 -1 First reflection to Second reflection to
penetrate into the
0.5 phantom
leave the phantom
0.02 0.02
-2
-1 0
-0.02 0 0.02 -0.02 0 0.02
Y[m] Y[m]
-3
(a) (c) 0 1 2 3 4
Time [ns]
Envelop(image) (a)
Absolute(image)
1 1 00 11 00 11
-0.02 -0.02 0.8
0.8 50
50 50
50
0.8
0.8 0.8
0.8
0.6
X[m]
0.6
X[m]

0 100
100 100
100
0

[degree]

[degree]
Angle [degree]

Angle [degree]
0.4 0.4 0.6
0.6 0.6
0.6
150
150 150
150
0.02 0.2 0.02 0.2
200
200 200
200

Angle

Angle
-0.02 0 0.02 0.4
0.4 0.4
0.4
-0.02 0 0.02
Y[m] Y[m] 250
250 250
250
(d) 0.2
0.2 0.2
0.2
(b) 300
300 300
300
Fig.2. Reconstructed images: (a) real image, (b) absolute image, (c) 350
350 00 350
350 00
positive image, (d) envelop image. 11 22 33 11 22 33
Time
Time[ns]
[ns] Time
Time[ns]
[ns]
(b) (c)
III. EXPERIMENTAL MEASUREMENTS Fig.4. Measurements, (a) Received pulse at a radar position, (b)
sinogram raw data (c) envelope detected sinogram raw data
A. Measurement setup
The measurement setup consists of a UWB radar calibrated before image reconstruction. To calibrate the
system, a breast phantom, and a matching liquid raw data, the ambient pulse is removed from the
container. The radar system is setup using a pulse received pulse. The ambient pulse is measured by
generator, a sampling oscilloscope, and a transceiver. sending and receiving a pulse into the matching liquid
The radar transceiver consists of a pair of UWB container in the absence of the breast phantom. By
antennas to transmit and receive a pulse with 10 GHz - subtracting ambient signal from the received pulse, all
10dB bandwidth. The antenna is customized to radiate effects from nearby objects and also antenna mutual
efficiently inside the matching liquid. A vegetable oil coupling are removed. The calibration procedure for
with a dielectric constant of 3 is used as the matching normal breast model is shown in Fig.4 (a). As one can
liquid. The normal breast phantom is built using pork fat see up to 1.5 ns, the received and ambient pules are
inside a plastic mold as shown in Fig.3 (a). In order to almost the same which is pure mutual coupling.
model cancerous breast tissue, a section of an orange, However after this point there are changes, between
which has approximately the same dielectric properties these two pulses, which are caused by the presence of
of cancerous tumor, is placed inside the phantom (Fig.3 the breast phantom. The solid blue line is the target
(b)). The breast model is located in the middle of a signature and solid black line which is the detected
plastic container inside the matching liquid. The radar envelop of the pulse. The calibrated raw data and its
transceiver is inside the matching liquid as shown in envelope are shown in Fig.4 (b) and Fig.4 (c),
Fig.3 (c). The imaged area is approximately 14 cm x 14 respectively. The same dataset is also generated for
cm. cancerous breast model.

B. Raw data acquisition IV. IMAGE RECONSTRUCTION AND DISCUSSIONS

The raw data is collected by rotating the phantom and Similar to the simulations results, both positive
recording the received pulse at every 5ᵒ, covering 360ᵒ, images and envelop images are reconstructed. For the
for normal and cancerous scenarios. The raw data for CSAR image reconstruction algorithm, it is necessary to
normal breast are shown in Fig.4 which contains 4050 have the exact dimensions of the scanned area. If the
samples in each angle. The raw data has to be dimensions used for image reconstruction is not the
same as the measurement setup, the image will not be
Orange lobe TRX
emulating tumor
focused. Since the medium of propagation is a
combination of matching liquid and breast phantom, to
Breast calculate the signal speed for CSAR processing, the
phantom
effective dielectric constant has to be considered. This
Rotating table
value can be estimated based on materials in the
(a) (b) (c)
medium, using dielectric probe measurement. For
Fig.3. Breast phantom (a) Normal, (b) with tumor model (d)
measurement setup
instance, for this measurement setup, we have pork fat

7065
 -0.06
Positive (Image)
-0.06
Positive (Image)
processing techniques such as envelope detection and
60
-0.04
50 -0.04 60
positive image subtraction are applied to improve image
-0.02
40 -0.02 quality. Error in group velocity results in image
Y[m]

degradation. These preliminary results from

Y[m]
0 0 Tumor 40
30
location
0.02 20 0.02
20 measurements demonstrate the ability of UWB-CSAR
0.04 Breast
Boundary
10 0.04
Breast to effectively image human breast with and without
0.06 0 0.06Boundary 0
-0.05 0
X[m]
0.05 -0.05 0
X[m]
0.05 cancer. Further work needs to be done on a more
(a)
(b) accurate breast phantom and eventually a pilot study
-0.06
Envelope (Image)
Envelop (Image)
Envelope (Image) with pigs will be performed.
100 -0.06
-0.06
100
-0.04 -0.04 100
90 -0.04
-0.02
80 -0.02
-0.02
90
REFERENCES
80
80
Y[m]

Tumor
Y[m]

0
70
ast
0.02
Bre dary
un
60
Bo
0.04 50
0.06
-0.05 0 0.05
X[m]
(c) (d)
Fig.5. Reconstructed real images: (a) normal (b) tumorous,
reconstructed envelope images: (c) normal (d) tumorous.
with a dielectric constant that ranges between 5 and 6
and also vegetable oil with a dielectric constant of 3.
The average dielectric constant should be between 4 and
5. Then by using these values into the image
reconstruction algorithm, the exact number for the
average permittivity can be found. A medium electric
permittivity of 4.5 is found to be optimal for image
reconstruction.
The reconstructed positive and envelop images are
shown in Fig.5. Obviously having these two types of
reconstructed images provides more information about
the scanned object. The image of the normal breast
phantom is shown in Fig.5 (a) and Fig.5 (c). The
positive image, Fig.5 (a), has some reflections near the
center, which is due to the discontinuities in the breast
phantom or signal processing artifacts. However, the
envelop image shows a smooth region with lower
reflected power associated to the inside of the breast
phantom and a hot ring around, showing the breast
boundary. This image is considered to be a reference for
comparison. The images for the cancerous breast
phantom are shown in Fig.5 (b) and Fig.5 (d). By
comparing the positive images Fig.5 (a) and Fig.5 (b),
one can see that the only difference is the appearance of
reflections due to tumor model in the breast phantom.
This fact can also be observed by comparing the
envelop images, Fig.5 (c) and Fig.5 (d), which clearly
shows the presence and location of the simulated tumor.
These measurement results demonstrate the ability of
UWB-CSAR to be used for breast cancer imaging
applications.
V. CONCLUSION
Breast microwave imaging using ultra wideband
(UWB) circular synthetic aperture radar (CSAR) is
presented. Breast images are reconstructed with the time
domain global back projection algorithm. Signal
00 location 70 [1] J. G. Elmore, K. Armstrong, C. D. Lehman, and S. W.
0.02
0.02
ast
Bre dary
60
60 Fletcher, “Screening for Breast Cancer,” Clin. corner, vol.
0.04
0.04 Bo
un 50 293, no. 10, pp. 1245–1256, 2014.
40
0.06
0.06
-0.05
-0.05 0 0.05 [2] C. H. Lee and et al., “Breast cancer screening with imaging:
X[m]
recommendations from the Society of Breast Imaging and
the ACR on the use of mammography, breast MRI, breast
ultrasound, and other technologies for the detection of
clinically occult breast cancer.,” J. Am. Coll. Radiol., vol. 7,
no. 1, pp. 18–27, Jan. 2010.
[3] A. Nnewihe, E. Staroswiecki, N. Bangerter, and B.
Hargreaves, “Dual tuned helmholtz coil for breast cancer
imaging,” in Proceedings 16th Scientific Meeting,
International Society for Magnetic Resonance in Medicine,
2008, vol. Toronto, p. 1114.
[4] E. J. Hall and D. J. Brenner, “Cancer risks from diagnostic
radiology.,” Br. J. Radiol., vol. 81, no. 965, pp. 362–78,
May 2008.
[5] P. T. Weatherall, G. F. Evans, G. J. Metzger, M. H.
Saborrian, and A. M. Leitch, “MRI vs . histologic
measurement of breast cancer following chemotherapy :
comparison with X-Ray mammography and palpation,” J.
Magn. Reson. Imaging, vol. 875, pp. 868–875, 2001.
[6] B. Zangheri, C. Messa, M. Picchio, L. Gianolli, C. Landoni,
and F. Fazio, “PET/CT and breast cancer.,” Eur. J. Nucl.
Med. Mol. Imaging, vol. 31 Suppl 1, no. May, pp. S135–42,
Jun. 2004.
[7] E. C. Fear, S. C. Hagness, P. M. Meaney, M. Okoniewski,
and M. A. Stuchly, “enhancing breat tumore detection with
near-field imaging,” IEEE Microw. Mag., no. March, pp.
48–56, 2002.
[8] A. E. Bulyshev, S. Y. Semenov, A. E. Souvorov, R. H.
Svenson, A. G. Nazarov, Y. E. Sizov, and G. P. Tatsis,
“Computational modeling of three-dimensional microwave
tomography of breast cancer,” IEEE Trans. Biomed. Eng.,
vol. 48, no. 9, pp. 1053–1056, 2001.
[9] E. J. Bond, S. Member, X. Li, S. C. Hagness, and B. D. Van
Veen, “Microwave imaging via space-time beamforming for
early detection of breast cancer,” IEEE Trans. Antennas
Propag., vol. 51, no. 8, pp. 1690–1705, 2003.
[10] A. H. Golnabi, P. M. Meaney, S. Geimer, and K. D.
Paulsen, “Microwave imaging for breast cancer detection
and therapy monitoring,” in BioWireleSS, 2011, pp. 59–62.
[11] A. H. Golnabi, P. M. Meaney, and K. D. Paulsen,
“Tomographic microwave imaging with incorporated prior
spatial information,” IEEE Trans. Microw. Theory Tech.,
vol. 61, no. 5, pp. 2129–2136, 2013.
[12] M. Ostadrahimi, P. Mojabi, and A. Zakaria, “Enhancement
of gauss – newton inversion method for biological tissue
imaging,” IEEE Trans. Microw. Theory Tech., vol. 61, no.
9, pp. 3424–3434, 2013.
[13] D. Oloumi, “Oil well monitoring by ultra-wideband ground
penetrating synthetic aperture radar,” M.Sc. thesis, Blekinge
Institute of technology, 2012.
[14] D. Oloumi, J. Ting, and K. Rambabu, “Design of pulse
characteristics for near-field UWB-SAR imaging,” Under_
Revis. IEEE Trans. Microw. Theory Tech., pp. 1–7, 2014.

7066