This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 1

Microwave Breast Screening Prototype:

System Miniaturization with IC Pulse Radio
Lena Kranold*, Student Member, IEEE, Mohammad Taherzadeh,
Frederic Nabki, Member, IEEE, Mark Coates, Senior Member, IEEE,
Milica Popović, Senior Member, IEEE

Microwave Can IC pulse radio be System 1: Picosecond and GigaBERT

Breast integrated in the • 44 cm/41.5 cm/25 cm
Screening system for microwave • US$ 23,000
breast cancer System 2: IC Pulse Radio
Prototype
screening? • 8 cm/7.5 cm/3 cm
• US$ 100

2-D cross-sectional views of the reconstructed scatterers

Maximum scatterer plane Tumor depth plane

IC Advantages System 1 Time-domain

✓ Lower cost Ph B microwave
✓ Smaller 3.0 GHz radar scan
dimensions
✓ Implementable in
System 2
wearable
Ph B
prototype 3.0 GHz Delay-multiply-
✓ Similar and-sum image
performance reconstruction
✓ Novel application System 2
to MWR Breast Ph A
screening 3.6 GHz

Microwave Breast Screening Prototype: System Miniaturization with IC Pulse Radio.

Take-Home Messages
• The reported advancement is motivated by the objective to design an integrated wearable system for breast
health monitoring.
• We show that the IC pulse radio can replace the off-shelf components without compromising the breast tumor
detection capabilities of the clinical prototype.
• The objective application of the system is to design an integrated wearable prototype for breast health
monitoring.
• We have successfully integrated the IC pulse radio into the microwave radar system; the results with phantom
measurements indicate that this would be a viable alternative to bulky and expensive off-the-shelf
components, bringing us a step closer to the envisioned compact, low-cost wearable device for radar-based
microwave breast screening.

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

Microwave Breast Screening Prototype:

System Miniaturization with IC Pulse Radio
Lena Kranold*, Graduate Student Member, IEEE, Mohammad Taherzadeh,
Frederic Nabki, Member, IEEE, Mark Coates, Senior Member, IEEE,
Milica Popović, Senior Member, IEEE

Abstract In this study, we report on system advancements for an ultrawideband (UWB) microwave radar prototype for early
breast cancer detection. We introduce an integrated circuit (IC) pulse radio as a suitable replacement for both a bulky and
expensive off-the-shelf pulse generator and a clock. To assess the system’s suitability as a breast screening prototype, we
compare measurements on two different experimental breast models (phantoms) with the established and previously reported
prototype using off-the-shelf components and the here-introduced IC. We test both systems on a homogeneous fat-mimicking
phantom with a 2-mm skin layer as well as a phantom with a 2-mm skin layer and glandular insertions, while the antennas,
antenna housing, and sampling oscilloscope are the same for both systems. Additionally, we advance with the IC to higher
frequencies, aiming to comply with the band intended for microwave imaging devices with medical applications. Furthermore,
we compare the economic requirements of the IC and of the previously reported system by evaluating their cost and
compactness. The objective of this study is to investigate if the IC pulse radio can replace bulky off-shelf components to allow us
to implement the pulse generation circuitry in one flexible circuit board with the switching and antennas.

Keywords — Microwave imaging, integrated circuits, medical diagnostic imaging, multistatic radar.

[4, 5, 6] maps the dielectric scatterers within the breast. The

I. INTRODUCTION1 microwave radar can be designed in a monostatic setup,
where a motor rotates one antenna around the breast, and
B REAST cancer is the second most common cause of
cancer deaths in women [1], and currently it is
diagnosed using established screening technologies as X-
the said antenna transmits and receives the signal [4].
Alternatively, a multistatic setup uses an array of antennas
that are strategically positioned around the breast [5, 6].
ray mammography, ultrasound, MRI or biopsy. These
Additionally, [7] introduced a multistatic radar prototype
established technologies suffer from certain drawbacks, for using CMOS technology in a clinical setup for breast
example: the use of ionizing radiation, expensive cancer screening. The system introduced and tested in an
operational cost, limited accessibility or invasiveness. initial patient study in [7] uses mechanical rotation of the
Hence, researchers worldwide are investigating the antenna array to create a baseline estimate for imaging of
development of new alternative technologies. the breast, while the system is handheld, and the patient lies
An option for these alternative technologies are in the supine position. The UWB pulses are generated using
microwave devices, which rely on the reported inherent CMOS technology and the pulse generation is integrated in
dielectric contrast within the breast tissue [2]. There are the handheld device. The center frequency of the system in
different approaches to use microwave technologies. While [7-9] is at 6 GHz with a bandwidth of 6.7 GHz, and the
microwave tomography (MWT) [3] aims to recover the pulse width is 160 ps. The skin artifact removal is
dielectric profile of the breast, a microwave radar (MWR) performed by using a two-step method based on signal
selection and filtering [8,9].
1
This work was supported by the following grants of the Natural On the other hand, the reports in [3,10] introduce a MWT
Sciences and Engineering Research Council of Canada (NSERC):
prototype that solves an iterative inverse problem.
Discovery (NSERC RGPIN-2019-05850) and Collaborative Research and
Development (CRDPJ 521870 – 17, with Analog Devices, Inc). Therefore, a discrete dipole approximation is applied to
L. Kranold is with the Department of Electrical and Computer solve the electromagnetic fields.
Engineering at McGill University, Montreal, QC, H3A 0E9, Canada (e- The work in [11] reports on a cost-effective solution of
mail: [email protected]). an ultrawideband radar for biomedical application such as
M. Taherzadeh is with the Electrical Engineering department, Ferdowsi
University of Mashhad, Iran (e-mail: [email protected]). breast cancer detection. The system can be implemented in
F. Nabki is with the Department of Electrical Engineering at École a portable device, and picosecond pulses are generated
Technologie Supérieure, Montreal, QC, H3C 1K3, Canada (e-mail: using step-recovery diodes (SRD). The receiver is realized
[email protected]). with a synchronized transmitter and receiver, based on time
M. Coates is with the Department of Electrical and Computer
Engineering at McGill University, Montreal, QC, H3A 0E9, Canada (e- expansion.
mail: [email protected]). Furthermore, [12] introduces the use of prior information
M. Popović is with the Department of Electrical and Computer based on ultrasound for the system in [4]. The ultrasound
Engineering at McGill University, Montreal, QC, H3A 0E9, Canada (e-
scan information is used to incorporate a structural estimate
mail: [email protected]).

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 3

whether a lower-cost component with much more desirable

form factor can generate similar results for breast tumor
detection using phantoms. Therefore, we first perform
measurements with both systems on a homogeneous fat
phantom with a 2-mm skin layer and assess the tumor
detection capability using two different detection
algorithms. Next, we test both systems with a second
heterogeneous phantom that is composed of fat, with a 2-
(a) (b) mm thin skin layer and glandular insertions. Since several
international regulations [15] require microwave medical
devices to operate above 3 GHz, we introduce the option to
vary the IC center frequency to comply with these
regulations. Finally, we identify the different spatial
requirements of both system options as well as their current
costs. This provides an economic assessment of replacing
the current pulse generator and clock with the IC. Please
note that the absolute costs are provided for the purpose of
comparison only. The novelty to this research is the
application of the IC pulse radio to a wearable microwave
radar breast screening prototype, allowing the integration of
the pulse generating circuitry with the antennas into one
(c) flexible prototype in the future.

Fig. 1. (a) Envisioned wearable bra microwave radar (image by Clinton II. MATERIALS AND METHODS
Ford Illustrations for McGill University). (b) Current antenna arrangement
in prosthetic bra for volunteer testing [6]. (c) Current antenna arrangement A. Original System: Ultrawideband Microwave Radar
in 3D-printed hemisphere for phantom testing.
Prototype
based on three tissue types to divide the 3D-scanned region The UWB radar prototype [6] investigated in this study
into homogeneous volumes. This prior information is employs several off-the-shelf components to perform pulse
processed to enhance the MWR imaging results. generation. A Picosecond Labs Model 3600 Impulse
The ultrawideband (UWB) time-domain microwave Generator [16] is triggered at 25 MHz by a Tektronix
radar introduced in [6] operates at 2 – 4 GHz and is GigaBERT 1400 clock generator [17] and generates a
designed to be wearable, with the vision of having an Gaussian pulse. The pulse is then shaped by a custom
affordable bra to detect breast cancer at an early stage. microwave filter from [18] to the desired range of 2 –
Fig. 1 (a) shows the envisioned wearable design. The
proposed system targets a wearable, time-domain, cost- Computer Clock
efficient device for breast health monitoring. While the ADC
systems in [3-5, 7-12] report on microwave technologies on Acquisition 25 MHz
breast cancer detection, little work has been done on the
efforts of developing a wearable solution. Our system is
Pulse
Picoscope Splitter
unique with the application of frequent breast health Generator
Trigger
monitoring in the form of an affordable bra.
Although the current system under investigation still LNA +24dB
includes bulky off-the-shelf-components, the 16 monopole
antennas [13] are flexible and embedded in a wearable
Pulse Shaper
prosthetic bra as in Fig. 1 (b). This prototype system has 2-4GHz Gaussian
been used for testing on volunteers and patients. However, Pulse
Received Signal
to repeatably test the system performance, experimental
breast models (phantoms) are essential. A long-term stable HP Pulse
Switching
phantom composed of carbon black, graphite and Amp Shaper
polyurethane was introduced in [14], and fabricated for the Circuit +35 dB UWB 2-4GHz
UWB radar system in this study. The phantoms are of Input
hemispherical shape, and to establish proper contact TX RX Pulse
between the antennas and the phantom surface, a 3D-
printed hemisphere in Fig. 1 (c) was fabricated, with
approximately the same antenna arrangement as in the bra 16
for volunteers.
In this study, we evaluate the current prototype system Antennas
using an off-the-shelf pulse generator and clock, and Fig. 2. System diagram of the original system.
compare performance with an IC pulse radio, to establish

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

4 GHz, amplified at +35 dB before it is passed through a filter on the receiver side of the system as shown in the
custom switching circuit and delivered to one of the 16 schematic in Fig. 2. On the receiving line is a custom
transceiving antennas [13]. The 16 antennas are set to microwave filter that reshapes the received and amplified
transmitting and receiving mode one at a time, and thus 240 signal to 2 – 4 GHz.
signals are recorded in total. The data is collected with a The antennas of the prototype are housed in a 3D-printed
Picoscope digital sampling oscilloscope PS9201a at hemisphere (Fig. 1 (c)) for phantom experiments to ensure
160 GSa/s, taking 16 averages per sample and 4096 proper contact between the antennas and the phantom
samples per signal. Thus, the Picoscope records each surface. The Picosecond Labs Model 3600 has dimensions
sample 16 times and forms the average before proceeding of 41.5 cm  41.5 cm  9 cm and costs ca. US$ 19,000
to the next sample. One time-domain signal, corresponding [19], but it is from a discontinued line and is hence no
to the signal passed between one antenna pair, contains longer purchasable. The Tektronix GigaBERT [17]
4096 averaged samples, and 240 signals in total are measures 44 cm  36 cm  15.5 cm and is available for
recorded for one complete phantom or breast scan. For the approximately US$ 3,500. The custom designed microwave
more complex experimental breast models, we add another filter has a fabrication cost of ca. US$ 500 and the filter is a
low-noise amplifier (LNA) of +24 dB and a microwave 2-layered printed circuit board on Isola 340HR substrate. It
measures 5 cm  5 cm  0.8 mm. This filter is used either
on the transmitting line before amplification or, for more
complex phantoms, on the receiving line. In the latter case,
the transmitting line filter is a microstrip line from [18].
The total cost for the system components is approximately
US$ 23,000. This system has been approved by the McGill
Research Ethics Board for trials on volunteers and patients.
B. IC System: Integrated Circuit Pulse Generator
In this paper, we present a system modification that
introduces an integrated circuit (IC) pulse radio to replace
the Picosecond Labs Model 3600, the Tektronix GigaBERT
clock and the custom microwave filter. The IC generates
UWB pulses of a bandwidth of approximately 1 GHz with
(a) adjustable center frequency. The IC contains one oscillator,
to generate the trigger signal, one oscillator to generate a
Computer IC Pulse 3 GHz – 6 GHz radio-frequency (RF) signal, a narrow-
pulse generator that is controlled by the trigger signal, and
ADC Generator an RF mixer to multiply the RF signal and the narrow pulse
Acquisition in order to generate an UWB impulse signal.
Trigger For this application, we choose the center frequency of
Picoscope Splitter 3.0 GHz to compare to our original system. We then
investigate if the new system can successfully detect the
tumors in experimental breast models if the center
LNA +24dB frequency is changed to 3.6 GHz. The IC is fabricated in a
UWB CMOS technology and packaged in a QFN48, measuring
Pulse Shaper Input 7 mm  7 mm for the total package size. Including the
2-4GHz Pulse driving circuitry and FPGA it measures
8 cm  7.5 cm  3 cm. The IC generates an UWB signal
Received Signal with an amplitude of 2 Vpp and a reconfigurable pulse width
of 1 ns to 2 ns. Additionally, it has an integrated clock that
HP
Switching can trigger the pulse generator as well as the Picoscope
Amp sampling oscilloscope. An FPGA is used to control the
Circuit +35 dB output frequency, the pulse width, and the trigger clock
frequency of the IC. Figure 3 (a) shows the printed-circuit
board (PCB) that contains both IC and the FPGA. The PCB
TX RX
module includes several other parts that are not related to
this work. The PCB module is powered through the USB
connection to the computer that controls the FPGA.
16 The UWB pulse that is generated by the IC is amplified
Antennas by the same power amplifier at +35 dB and then passed
through the same switching circuit to the 16 antennas in the
(b) 3-D printed hemisphere. The received pulse passes through
Fig. 3. (a) The pulse generator module, containing the fabricated IC and the the receiving side of the switching circuit and afterwards is
FPGA. (b) System diagram of IC system.

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 5

amplified by a low noise amplifier (LNA) of +24 dB gain tumor-bearing plug with a 1 cm diameter spherical tumor
and filtered using a 2 – 4 GHz band-pass filter implemented embedded at its far end. We chose this small tumor size to
by the pulse shaper from the original system. The Picoscope cover the description of breast cancer stages 0 and IA which
then records the signal. A system diagram is shown in is considered as a measure for early breast cancer and hence
Fig. 3 (b). has a significantly higher 5-year survival rate of 99 % [1,
The non-subsidized IC fabrication cost is US$ 3,700 [20] 21]. In Tab. II the plug properties are summarized. Note
for 100 chips. Considering the packaging cost, each IC here that plugs 1 and 3 are used with Phantom A and plugs
costs approximately US$ 100. For a higher production 2 and 4 with Phantom B to match the fat-mimicking
volume, the cost per IC is likely to be significantly lower. properties.
C. Breast Phantoms
Breast phantoms are used to test microwave radar D. Methods
prototypes, because they allow for system testing with To compare the two systems, we conducted several
controlled laboratory experiments. The tests are repeatable, measurements with the different phantoms. For each
so it is possible to isolate the effects of specific hardware phantom, we collected two consecutive baseline scans
changes. We fabricated carbon-polyurethane-based (healthy plug 1 or 2) followed by two tumorous scans (plug
phantoms with interchangeable tight-fitting insertions, here 3 or 4). The signals were collected with minimum
referred to as “plugs”, to mimic different scenarios, disturbance of the system, hence only the plug was changed
following the procedure described in [14]. An example between healthy and tumorous scans. We repeated the
phantom is shown in Fig. 4 (a) and a plug in Fig. 4 (b). We experiment three to six times on different days to ensure the
used two different phantoms to compare the performance of measurements we report are repeatable. After comparing
the original system with the new system that incorporates both systems, we repeated the experiment with Phantom A
the IC (IC system). Initial tests with a homogeneous fat and the IC at a higher center frequency to comply with
phantom without a skin layer were promising, but to draw several international regulations for microwave radars in
meaningful conclusions, it is necessary to include a skin the medical field [15].
layer in the comparisons. Therefore, Phantom A consists of We analyzed the data using two methods. First, we
homogenous fat with a 2-mm skin layer, while Phantom B compare the mean energy of signal differences (MED), and
is heterogeneous, consisting mainly of fat and 8 %vol gland afterwards we apply the delay-multiply-and-sum (DMAS)
with a 2-mm skin layer. Table I summarizes the phantom algorithm from [22] to create a 3-D image of the phantoms
properties at the center frequency of 3 GHz. While these with the two systems. For both analysis methods, the
phantoms are not as complex as the breast anatomy, they signals are windowed to the time around the main pulse.
offer good representation of the average electrical This window is chosen to be 6.25 ns starting with the main
properties of breast tissue in the frequency range of interest pulse, which is different for the two systems and different
and allow us to conduct a comparative study of our systems phantoms. For the original system the starting point with
in a straight-forward fashion. The tumor locations are phantom A is 8 ns and with phantom B 10 ns while for the
specified using a Cartesian coordinate system where the IC system the starting point for phantom A and B is 2 ns.
origin is at the center of the hemispherical base The MED calculates the mean energy that is contained in
(representing the chest wall) and the z-axis points from the signal differences of two complete scans. Therefore, first,
chest wall towards the nipple location (the center of the for each antenna pair, the signals of two separate scans are
curved surface of the hemisphere). time-aligned and then subtracted from each other. Then, the
Both phantoms are measured with a homogenous fat- time-integral of this difference is calculated to determine
mimicking plug that matches the phantom and with a the energy of this difference. Lastly, the mean is calculated
TABLE I over all 240 differences from all antenna pairs between two
PHANTOMS AND THEIR PROPERTIES scans. We expect that the MED of two scans of the same
Tumor location nature is significantly smaller than the MED of two scans of
Phantom Contents εr at 3 GHz
(x/y/z) in cm different nature. For example, a baseline-baseline
A
Skin 35
(-1.3/2.7/3.0)
comparison is expected to show a smaller MED than a
Fat 11 baseline-tumor comparison. For these MEDs, we determine
Skin 35 the mean and standard deviation using the MATLAB built-
B Fat 8 (-3.0/0.0/3.0)
Gland 25 in function normfit() [23], as well as the 95 % confidence

TABLE II
PLUGS AND THEIR PROPERTIES
Applicable
Plug Contents εr at 3 GHz
phantom
1 Fat 11 A
2 Fat 8 B
Fat 11
3 A (a) (b)
tumor, 1 cm diameter 70
Fat 8 Fig. 4. Carbon-polyurethane phantom (a) with hole to insert various plugs
4 B (b).
tumor, 1 cm diameter 70

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

interval of the mean for each comparison. The mean is determines the distance of the highest scatterer in the 3-D
calculated as the mean of the 240 energy differences that volume to the center of the known tumor location. The
resulted in subtracting and integrating one time-domain signal-to-mean-ratio (SMR) is the ratio of the highest
signal (baseline) from another (tumor or baseline), while intensity within the box around the tumor location to the
the standard deviation represents the square root of the mean intensity of the 3D volume outside the box around the
unbiased variance estimator. It represents the 95 % tumor. Hence, it quantifies the contrast of the scattering at
confidence interval of the mean. Hence, we consider the the tumor location relative to the surrounding volume.
system to be successful in detecting the tumor if the lower
bound of the baseline-tumor comparison is higher than the III. RESULTS
upper bound of the baseline-baseline comparison.
The DMAS algorithm from [22] calculates a 3D image of A. Generated Input Pulses
the scatterers within the phantom using a differential scan. To assess if the generated pulses of the IC compare to the
Therefore, we subtract signals of a baseline scan from a pulses generated by the Picosecond Labs Pulse Generator
tumor scan after time-aligning each antenna pair signal. and microstrip filter, we show here a basic comparison of
These signals are then processed by determining their delay the time-domain pulses and the Fourier transform in the
to each voxel in the 3-D volume, using the knowledge of frequency domain.
the antenna locations, and intensities for each voxel are Fig. 5 (a) shows the recorded time domain signals of the
determined. The DMAS images in this paper are all original system as well as the IC system at center frequency
normalized to the maximum scatterer within the 3-D of 3.0 GHz and 3.6 GHz. All three pulses have a pulse
volume, with dark blue representing the lowest scattering width of under 1 ns, and the original system UWB pulse has
and dark red the maximum intensity 1. The x-y-coordinates an amplitude of 0.6 V. Both IC pulses, at 3.0 GHz and
of the known tumor locations are marked by a red “x” in the 3.6 GHz have an amplitude of 1.2 V. Although we expect
images. the pulses to not be the exact same, the amplitude
We can determine quantitative values from the DMAS discrepancy is accounted for with the input restrictions of
image to assess whether tumor detection is successful. The the high-power amplifier by placing attenuators to not
signal-to-clutter-ratio (SCR) determines the ratio of the exceed the maximum voltage input. Furthermore, the IC
intensity of the highest scatterer within a generated signals show more peaks than the original
3 cm  3 cm  3 cm box around the known tumor location system signal.
to that of the highest scatterer outside this box in dB. The frequency domain responses of the pulses in
Hence, a positive SCR implies that the highest scatterer is Fig. 5 (b) were generated with a discrete Fourier transform
near the tumor location, while a negative SCR means that and they confirm that the frequency content of the original
the tumor detection is unsuccessful. The localization error system UWB pulse is from 2 – 4 GHz. The IC pulse
centered at 3.0 GHz has a bandwidth of 1 GHz. The second
IC pulse is at its expected center frequency of 3.6 GHz with
a bandwidth of 1 GHz.
B. Phantom A
In this section, we compare the results of the original
system with the IC system by looking at an example of
unprocessed recorded signals as well as the MED and the
DMAS results for the scans using Phantom A
TABLE III
MED FOR PHANTOM A WITH THE ORIGINAL SYSTEM
(a)
Standard Confidence
Scans Mean
deviation Interval
Baseline1-Baseline2 0.0243 0.0259 [0.0210;0.0276]
Tumor1-Tumor2 0.0177 0.0216 [0.0150;0.0205]
Baseline1-Tumor1 0.0760 0.1341 [0.0590;0.0931]
Baseline1-Tumor2 0.0763 0.1365 [0.0589;0.0936]
Baseline2-Tumor1 0.0531 0.0939 [0.0412;0.0651]
Baseline2-Tumor2 0.0556 0.1053 [0.0422;0.0690]

TABLE IV
MED FOR PHANTOM A WITH THE IC SYSTEM
Standard Confidence
Scans Mean
deviation Interval
(b) Baseline1-Baseline2 0.0039 0.0069 [0.0030;0.0047]
Fig. 5. Raw recorded signals of generated pulses: (a) Time-domain signal Tumor1-Tumor2 0.0039 0.0108 [0.0026;0.0053]
and (b) frequency content normalized to 0 dB. The pulse generated by the Baseline1-Tumor1 0.0227 0.0278 [0.0192;0.0263]
Picosecond Labs and microstrip filter (original system) is shown in dashed Baseline1-Tumor2 0.0234 0.0300 [0.0196;0.0273]
black, the IC pulse at center frequency of 3.0 GHz in solid blue and at Baseline2-Tumor1 0.0181 0.0225 [0.0152;0.0209]
center frequency of 3.6 GHz in dot-dashed red. Baseline2-Tumor2 0.0183 0.0232 [0.0154;0.0213]

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 7

(homogeneous fat, skin layer). is expected since the IC has a bandwidth of 1.0 GHz, while
In Fig. 6, we display one example of the received signals. the microwave filter has a higher bandwidth of 2.0 GHz and
In Fig 6 (a), we show the time-domain signals of a the signals of the higher frequency content experience more
reflecting antenna pair, i.e., two antennas that are placed attenuation than the lower-frequency ones due to the
next to each other in the wider ring of the arrangement dispersive and lossy nature of the tissue [24]. Furthermore,
(transmitting antenna 1 and receiving antenna 2), for the the IC system shows more low-frequency noise. Hence,
original system and in (b) for the IC system. In Fig. 6 (c), before processing this data with the MED or DMAS
we display the frequency-domain content, obtained through algorithms, we digitally filter all signals with a bandpass at
the Fourier transform, of the same antenna pair. Both 1.7 – 4 GHz, to avoid including the low frequency noise in
signals are from the respective baseline scans. We observe our analysis, as it is not part of the input signal that has
that the original system has a longer time delay compared to passed through the breast phantom.
the IC pulse generator system with respect to the starting In Table III, we present the MED results for the original
point of the signal. This is caused by more and longer system, and in Table IV for the IC system. We observe that,
cables that are used with the Picosecond Labs Model 3600 for both systems, we consider the tumor detection based on
and the GigaBERT clock generator. In the frequency the MED analysis as successful. For the original system, the
domain, we observe that the original system has a higher lower bounds of the different-case-comparisons (baseline-
content in the range of 1.5 – 2.5 GHz, while the IC system tumor) are a factor of 1.5 larger than the upper bounds of
shows more content between 2.5 – 3.5 GHz. Although both the same-case-comparisons. For the IC system, we observe
systems have a center frequency of 3.0 GHz, this behavior a factor of 2.8 between the same-case and different-case
comparisons.
The DMAS slices for measurements of the original
system of the maximum scatterer location and the tumor
depth location are displayed in Fig. 7 and for the IC system
in Fig. 8. The quantitative DMAS results are summarized in
Table V. We observe that both systems detect one
significant scatterer in the region around the tumor, but the
original system shows more low scattering in the 3D
volume. However, the localization error of the IC system is
slightly larger than that of the original system. Both systems
TABLE V
SCR, SMR AND LOCALIZATION ERROR FOR PHANTOM A
(a)
Localization Succe
System SCR SMR
Error ss?
Original System 10.9 dB 38.3 dB 0.9 cm Yes
IC System 14.1 dB 48.9 dB 1.2 cm Yes

(b)

(a) (b)
Fig. 7. DMAS image slices of Phantom A for measurements with the
original system: (a) Maximum scatterer slice and (b) tumor depth slice.

(c)
Fig. 6. Raw recorded signals of homogeneous Phantom A (skin and fat)
with plug 1 (fat): (a) Time-domain signal of the original system, (b) of the (a) (b)
IC system of reflecting antenna pair of two antennas next to each other and Fig. 8. DMAS image slices of Phantom A for measurements with the IC
(c) frequency content normalized to 0 dB. system: (a) Maximum scatterer slice and (b) tumor depth slice.

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

model, antennas 1 and 2. Fig. 9 (a) shows the recorded

time-domain signals for both systems, and Fig. 9 (b)
displays the frequency domain content for both systems.
We notice here that the signal recorded using the original
system is later in time again, as expected due to cable
delays, than the IC system. Looking at the frequency
content of the original system, it has less attenuation
between 2 – 2.5 GHz compared to 2.5 – 4 GHz, but it
passes the signal in the desired frequency range of 2 –
4 GHz. In contrast, the frequency content of the IC system
appears to be contained in the 2 – 3.5 GHz range.
(a)
The results of the MED analysis are presented in Table
VI for the original system and in Table VII for the IC
system. We observe that the original system has a factor of
10 difference between same-case-comparisons and
baseline-tumor comparisons. Hence the tumor detection
using the MED method is considered successful for the
original system on Phantom B. Furthermore, the IC system
shows a factor of 1.1 difference between comparisons of
scans of the same nature and different nature. The MED
also successfully detects the tumors for this system on
Phantom B.
(b) Additionally, in Fig. 10 (a) we show the slice of the
Fig. 9. Raw recorded signals of heterogeneous Phantom B (skin, fat and maximum scatterer of the signals recorded with the
gland) with plug 2 (fat): (a) Time-domain of reflecting antenna pair of two original system, and in (b) the slice of the tumor depth.
antennas next to each other and (b) frequency content normalized to 0 dB. We observe a single intense scatterer very close to the
tumor location, but the contrast in the image is not as
show a high SMR, but the IC system has a higher SCR with high, with an SMR of 36.1 dB. The SCR for this case is
14.1 dB than the original system with 10.9 dB. positive at 8.0 dB, and hence the original system can
C. Phantom B successfully detect the tumor in Phantom B with the
Since both systems show comparable results for the DMAS algorithm, with the localization error of 1.8 cm.
Fig. 11 shows the maximum scatterer location slice in (a)
homogeneous phantom with skin, in this section we present
and the tumor depth slice in (b) for the scans of
comparative measurements of the heterogeneous
Phantom B with the IC system. The image results in an
Phantom B, which consists of fat and 8 %vol glandular
SCR of 2.9 dB and we notice one distinct scatterer in the
insertions with a 2-mm thin skin layer around it. In Fig. 9,
we present the raw signals of the same antenna pair as in
the previous section. As a reminder, these correspond to
two antennas in proximity to each other on the wider ring of
the arrangement, closer to the base of the hemispherical

TABLE VI
MED FOR PHANTOM B WITH THE ORIGINAL SYSTEM
Standard Confidence
Scans Mean
deviation Interval
Baseline1-Baseline2 0.0021 0.0039 [0.0016;0.0026] (a) (b)
Tumor1-Tumor2 0.0017 0.0038 [0.0012;0.0022] Fig. 10. DMAS image slices of Phantom B for measurements with the
Baseline1-Tumor1 0.0447 0.0984 [0.0321;0.0572] original system: (a) Maximum scatterer slice and (b) tumor depth slice.
Baseline1-Tumor2 0.0487 0.1063 [0.0351;0.0622]
Baseline2-Tumor1 0.0396 0.0897 [0.0282;0.0510]
Baseline2-Tumor2 0.0428 0.0952 [0.0307;0.0550]

TABLE VII
MED FOR PHANTOM B WITH THE IC SYSTEM
Standard Confidence
Scans Mean
deviation Interval
Baseline1-Baseline2 0.0101 0.0226 [0.0073;0.0130]
Tumor1-Tumor2 0.0072 0.0167 [0.0051;0.0094]
Baseline1-Tumor1 0.0227 0.0391 [0.0177;0.0277]
(a) (b)
Baseline1-Tumor2 0.0205 0.0363 [0.0159;0.0252]
Fig. 11. DMAS image slices of Phantom B for measurements with the IC
Baseline2-Tumor1 0.0192 0.0339 [0.0149;0.0235]
system: (a) Maximum scatterer slice and (b) tumor depth slice.
Baseline2-Tumor2 0.0183 0.0341 [0.0140;0.0227]

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 9

TABLE VIII center frequency and if, in that case, the tumor detection
SCR, SMR AND LOCALIZATION ERROR FOR PHANTOM B
remains successful. Fig. 12 shows the simulated S11 for all
Localization antennas distributed around a hemispherical breast model
System SCR SMR Success?
Error
with 2-mm skin layer. Simulations are performed using
Original System 8.0 dB 36.1 dB 1.8 cm Yes
IC System 2.9 dB 25.0 dB 1.5 cm Yes Ansys HFSS [25]. The S11 graph demonstrates that, when
positioned in contact with the skin, the antennas can
tumor region (localization error of 1.5 cm) with operate successfully in the 2 – 5.5 GHz range. With the 1-
surrounding low scattering. The contrast in this image is GHz bandwidth in mind, we set the new center frequency
lower than for the original system, with an SMR of at 3.6 GHz, with the frequency band ranging from 3.1 –
25.0 dB. Comparing Fig. 10 and Fig. 11, we observe that 4.1 GHz. The antenna S11 has been verified using a VNA
the original system results in a clearer image of the in [26] and it has been tested extensively on phantoms and
Phantom B, but both systems are able to detect the tumor in direct contact with human skin in volunteer trials [27].
in the correct region using the DMAS algorithm. The First, we present in Fig. 13 (a) the time-domain
quantitative results are summarized in Table VIII. recorded signals of two different antenna pairs. The first
antenna pair, TX1 RX2, is the same antenna pair chosen
D. Advancements to Higher Center Frequency for the results in the previous sections. The second
In addition to comparing the IC system with the original antenna pair, TX1 RX5, consists of two antennas located
system, we investigate if the IC can operate at a higher across from each other in the upper ring of the
hemisphere. Hence the signal for this antenna pair has
passed through the entire phantom. In Fig. 13 (b), we
display the frequency content of the same two signals,
together with the frequency response of the custom 2 –
4 GHz microwave filter on the receiving line. We observe
that the microwave filter restricts the signal to 1.5 –
4.1 GHz, and hence still allows the higher frequency band
at center frequency 3.6 GHz.
We present the MED analysis results in Table IX. We
observe that the lower bound of the confidence interval of
the different-case-comparison is a factor of 2 larger than the
Fig. 12. Ansys HFSS simulated S11 of all 16 antennas with a hemispherical upper bound of the comparison between two scans of the
breast model, with antennas placed at the same locations as for the 3D-
printed hemisphere.
same nature. Therefore, we consider the MED analysis to
be successful in differentiating a tumor scan from a healthy
scan.
In Fig. 14, we present the DMAS images for the higher
frequency scan with the IC system on Phantom A. In
Fig. 14 (a) we show the maximum scatterer slice and in (b)
the slice of the tumor depth. We notice here that there is
TABLE IX
MED FOR PHANTOM A WITH THE IC SYSTEM AT CENTER FREQUENCY 3.6
GHZ
Standard Confidence
Scans Mean
deviation Interval
Baseline1-Baseline2 0.0022 0.0050 [0.0015;0.0028]
(a) Tumor1-Tumor2 0.0019 0.0056 [0.0012;0.0026]
Baseline1-Tumor1 0.0074 0.0114 [0.0060;0.0089]
Baseline1-Tumor2 0.0073 0.0110 [0.0059;0.0087]
Baseline2-Tumor1 0.0072 0.0113 [0.0058;0.0087]
Baseline2-Tumor2 0.0068 0.0096 [0.0056;0.0080]

(b)
Fig. 13. Raw recorded signals of Phantom A with plug 1 with IC at center
frequency of 3.6 GHz: (a) Time-domain signal of reflecting antenna pair of (a) (b)
two antennas next to each other and (b) frequency content normalized to Fig. 14. DMAS image slices of Phantom A for measurements with the IC
0 dB, displayed with response of the microwave filter used on the receiving system at center frequency 3.6 GHz: (a) Maximum scatterer slice and (b)
line. tumor depth slice.

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

TABLE X IV. CONCLUSION

SCR, SMR AND LOCALIZATION ERROR OF IC SYSTEM FOR PHANTOM A AT
DIFFERENT CENTER FREQUENCIES In this study, we presented a system advancement using an
Localization IC pulse radio with integrated clock to replace the
System SCR SMR Success?
Error Picosecond Labs Model 3600 pulse generator, Tektronix
fc = 3.0 GHz 14.1 dB 48.9 dB 1.2 cm Yes GigaBERT 1400 clock generator and microwave filter in a
fc = 3.6 GHz 9.6 dB 28.9 dB 1.2 cm Yes UWB radar prototype for early breast cancer detection. We
one distinct scatterer in the tumor region, and only a few first compared scans of two different breast phantoms of the
smaller scattering points in the 3D volume. The DMAS original system with the introduced IC system and evaluate
analysis results in a positive SCR of 9.6 dB and hence their abilities to detect breast cancer using analysis of the
detects the tumor in the correct region with a localization mean energy of signal differences as well as the DMAS
error of 1.2 cm. The DMAS results compared for both algorithm. All tumor-bearing phantom scans included a
investigated center frequencies of the IC are shown in spherical tumor with 1-cm diameter, which corresponds to
Table X, and we notice that both center frequencies result breast cancer stages 0 and IA, and thus is considered an early
in the same localization error. Therefore, with the IC, we detection of the cancer. Hence, we show that the IC system
are able to move the center frequency of our microwave can detect the 1-cm diameter tumor within the different
radar for breast health monitoring to a higher range in order breast phantoms with a similar accuracy as the original
to comply with most international regulations and without system. Future investigations might explore more
the need for a new custom microwave filter. complicated phantoms with smaller tumors.
Additionally, we explore the options to move to a higher
E. Economic Comparison center frequency using a different setting of the IC to comply
So far, the presented results show that the IC pulse radio with international regulations for medical devices. Therefore,
is a suitable replacement for the Picosecond Labs Model we conclude that with a center frequency of 3.6 GHz, the
3600 and the Tektronix GigaBERT 1400 with respect to system has effectively the same capability to detect tumors.
data analysis and successful tumor detection in our In addition, we perform an economic comparison of both
microwave breast health monitoring prototype. In our work, systems to assess if we can not only miniaturize the system
we strive to design a low-cost microwave breast screening with the IC, but also design it to be more affordable. With the
system that can be made available to small family clinics IC at the prototype stage, we can save approximately
and possibly to economically disadvantaged environments. US$ 22,900 compared to the bulky off-the-shelf equipment.
For this reason, it is vital to investigate if the replacement of These advancements are introduced to bring our system a
original system by its IC counterpart would be economical step closer to a cost-efficient and wearable device, providing
and beneficial from the standpoint of system the opportunity to integrate a small pulse generating circuitry
miniaturization. into one flexible circuit board instead of using bulky off-shelf
The original system uses the Picosecond Labs Model components, bringing us closer to a wearable device in the
3600 and the Tektronix GigaBERT 1400, which have future.
approximate dimensions, when stacked on top of each The novelty of this proposed system lies in the successful
other, of 44 cm  41.5 cm  25 cm. Although the integration of an IC pulse radio to a wearable microwave
Picosecond Labs Model 3600 is discontinued, together the breast screening tool for early cancer detection. In the
components have a total purchase cost of approximately literature [7-9], a handheld device has been proposed that is
US$ 23,000. In contrast, the IC system only measures applied to a patient in the supine position. This device used
8 cm  7.5 cm  3 cm including mounted components and CMOS pulse generating technology and creates an image of
an FPGA. The IC together with the driving circuit are in the the breast by subtracting signals of a physical rotation of the
prototype phase, costing approximately US$ 100. The cost antenna array. On the other hand, the here-proposed system
of the IC, however, will decrease drastically if mass- is designed to be integrated in one wearable device that
produced. The comparison of dimensions and costs of both tracks changes in the breast. The goal of our study is to
systems are summarized in Table XI. With our vision to provide an affordable early warning system that detects
design an affordable wearable device for early breast cancer unhealthy changes at an early stage and can be applied on a
detection, the IC proves to be a suitable replacement for the frequent basis without posing a health risk.
current pulse generator, clock and microwave filter with
respect to tumor detection as well as advantages of lower ACKNOWLEDGMENT
costs and significantly smaller dimensions. The IC system
reduces the prototype costs by approximately US$ 22,900. The authors would like to acknowledge SPARK
TABLE XI Microsystems International Inc. (sparkmicro.com) for
COMPARISONS OF ORIGINAL SYSTEM AND IC SYSTEM providing the IC and the module.
Dimensions Additionally, the authors would like to thank Analog
System Components Appr. Costs
w/l/h Devices for the collaboration and their support.
Original
Tektronix GigaBERT 44 cm/ Specifically, these measurements utilized the switching
Picosecond Labs Model 3600 41.5 cm/ US$ 23,000 circuit and custom filter designed and fabricated by their
System
Microwave filter 25 cm
8 cm/
team.
IC IC Furthermore, the authors would like to thank Mr. Don
7.5 cm/ US$ 100
System Driving Circuit
3 cm

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology
JERM-2020-05-0066 11

Pavlasek and his workshop for his support in designing and algorithms,” Master’s Thesis, McGill University, Montreal,
Canada, 2017.
fabricating the 3D-printed hemisphere.
[20] CMC Microsystems, “TSMC 65-nanometer GP CMOS Process
Technology”, 2019, [Online], Available: https://www.cmc.ca/tsmc-
65-nanometer-cmos-process-technology-2-2/. [Accessed 17.
REFERENCES February 2020].
[1] American Cancer Society, “Cancer Statistics Center,” 2018, [21] Breastcancer.org, “Breast Cancer Stages,” 2020, [Online],
[Online], Available: Available:
https://cancerstatisticscenter.cancer.org/#!/cancer-site/Breast. https://www.breastcancer.org/symptoms/diagnosis/staging.
[Accessed 12. February 2020]. [Accessed 16. March 2020].
[2] N. Nikolova, “Microwave Imaging for Breast Cancer,” IEEE [22]H. Been Lim, N. Thi Tuyet Nhung, E. P. Li and N. Duc Thang,
Microw. Mag., vol. 12, no. 7, p. 78–94, 2011. “Confocal Microwave Imaging for Breast Cancer Detection: Delay-
[3] N. R. Epstein, P. M. Meaney and K. D. Paulsen, “3D parallel- Multiply-and-Sum Image Reconstruction Algorithm,” IEEE Trans.
detection microwave tomography for clinical breast imaging,” Biomed. Eng., vol. 55, no. 6, pp. 1697-1704, June 2008.
Review of Scientific Instruments, vol. 85, no. 12, p. 124704, 2014. [23] MATLAB, “Mathworks MATLAB,” 2020, [Online], Available:
[4] E. C. Fear, J. Bourqui, C. Curtis, D. Mew, B. Docktor and C. https://www.mathworks.com/products/matlab.html. [Accessed 16.
Romano, “Microwave breast imaging with a monostatic radar-based March 2020].
system: A study of application to patients,” IEEE Trans. on [24] M. Lazebnik et al., “A Large-Scale Study of the Ultrawideband
Microw. Theory Techn., vol. 61, no. 5, pp. 2119–2128, 2013. Microwave Dielectric Properties of Normal, Benign and Malignant
[5] D. Byrne, M. Sarafianou and I. Craddock, “Compound Radar Breast Tissues Obtained from Cancer Surgeries,” Phys. Med., vol.
Approach for Breast Imaging,” IEEE Trans. on Biomed. Eng., vol. 52, pp. 6093-6115, 2007.
64, no. 1, pp. 40–51, 2016. [25] ANSYS Inc., “ANSYS HFSS”, 2020, [Online]. Available:
[6] L. Kranold, C. Quintyne, M. Coates, M. Popović, “Clinical Study http://www.ansys.com/products/electronics/ANSYS-HFSS.
with a Time-Domain Microwave Breast Monitor: Analysis of the [Accessed 16. March 2020]
System Response and Patient Attributes,” in Proc. Eur. Conf. [26] E. Porter, “A microwave time-domain radar for monitoring breast
Antennas and Propagation, Krakow, Poland, Mar. 2019. health: development and testing of an early prototype”, PhD thesis,
[7] Song, H., et al,.”Detectability of Breast Tumor by a Hand-held McGill University, Montreal, QC, Canada, p. 118, 2015.
Impulse-Radar Detector: Performance Evaluation and Pilot Clinical [27] E. Porter, H Bahrami, A. Santorelli, B Gosselin, L. Rusch, and M.
Study”, Sci Rep 7, 16353 (2017). Popović, "A Wearable Microwave Antenna Array for Time-
[8] Song, H., et al,.”A two-stage rotational surface clutter suppression Domain Breast Tumor Screening," IEEE Trans. Med. Imaging, vol.
method for microwave breast imaging with multistatic impulse- 63, no. 3, pp. 530-539, 2016.
radar detector”, IEEE Trans. Instrum. Meas., 2020.
[9] Song, H., H. Watanabe, X. Xiao, T. Kikkawa. ”Artifact removal for
impulse-radio radar-based breast cancer detection”, in Proc ICEAA-
IEEE APWC, Granada, Spain, Sep. 2019 Lena Kranold received the B.Eng. (2013) degree
[10] S. Hosseinzadegan, A. Fhager, M. Persson, P. Meaney. ”A discrete in Technology Management and Marketing from
dipole approximation solver based on the COCG-FFT algorithm the University of Applied Sciences Kiel,
and its application to microwave breast imaging”, Int. J. Antenn. Germany, and the M.Sc. degree (2015) in
Propag., 2019. Electrical and Information Engineering from the
[11] D. Oloumi, E. Fear.”An inexpensive ultrawideband radar for high- Christian-Albrechts-University of Kiel, Germany,
resolution applications”, in Proc. 18th Int. Symp. on Antenna specializing in computational electromagnetics.
Technology and Applied Electromagnetics, 2018. She was the recipient of the prestigious
[12] N. Abdollahi, D. Kurrant, P. Mojabi, M. Omer, E. Fear, J. LoVetri. scholarship of the Stiftung der Deutschen
”Incorporation of ultrasonic prior information for improving Wirtschaft e. V. (Foundation of German
quantitative microwave imaging of breast”, IEEE J. Multiscale Business), 2013 – 2015. She is a PhD candidate in
Multiphys. Comput. Tech., vol 4, 2019. Electrical Engineering at McGill University,
[13] H. Bahramiabarghouei, E. Porter, A. Santorelli, B. Gosselin, M. Montreal, Canada, where she received the McGill Engineering Doctoral
Popović and L. A. Rusch, “Flexible 16 Antenna Array for Award (2016-2018), the STARaCom Collaborative Grant Supplement
Microwave Breast Cancer Detection,” IEEE Trans. Biomed. Eng., (2019-2020) and the STARaCom Scholarship (2020-2021). Her doctoral
vol. 62, no. 10, pp. 2516-2525, Oct. 2015. work focuses on the design of a prototype for breast health monitoring
[14] J. Garrett, “Average dielectric property analysis of non-uniform with low-power microwaves and conducting clinical trials. From 2017-
structures: Tissue phantom development, ultra-wideband 2018, Lena was the treasurer of the IEEE Montreal Women in Engineering
transmission measurements, and signal processing techniques,” Affinity Group, and since 2019, she has been serving as the Vice-Chair of
Master’s Thesis, University of Calgary, Calgary, Canada, 2014. the IEEE Engineering in Medicine and Biology Society, Montreal Chapter.
[15] Federal Communications Commission., “FCC-02-48”, 2002,
[Online], Available: Mohammad Taherzadeh-Sani received the B.Sc.
https://transition.fcc.gov/Bureaus/Engineering_Technology/ Orders/ degree from the Ferdowsi University of Mashhad,
2002/fcc02048.pdf. [Accessed 19. February 2020]. Iran, in 2001, the M.Sc. degree from the
[16] IEEE GlobalSpec Engineering 360, “Picosecond Labs Impulse University of Tehran, Iran, in 2004, and the Ph.D.
Generator Model 3600”, 2020, [Online], Available: degree from McGill University, Montreal,
https://datasheets.globalspec.com/ds/1463/PicosecondPulseLabs/89 Canada, in 2011. He was a recipient of a J. W.
616C72-4B52-40B2-95DF-C546EA8BF88B. [Accessed 12. McConnell Memorial Fellowship from McGill
February 2020]. University in 2007 and 2008 for his doctoral
[17] Toronto Surplus & Scientific Inc., “Tektronix GigaBERT 1400”, research, and a Post-Doctoral Fellowship from
2020, [Online], Available: the Le Fonds Québécois de la Recherche sur la
https://www.torontosurplus.com/microwave-logic-tektronix- Nature et les Technologies for 2012 and 2013
gigabert-1400-gb1400-drx-analyzer-pattern-generator-and-error- (declined). In 2012, he joined Ferdowsi
detector-1400mb-s-receiver-unit-in-stock-z1.html. [Accessed 12. University of Mashhad as an Assistant Professor. He authored several
February 2020]. journal publications in distinguished journals (e.g., JSSC, TCAS-I, TCAS-
[18] A. Santorelli, et al., “Experimental Demonstration of Pulse Shaping II, and T-VLSI). His research interests focus on biomedical circuits and
for Time-Domain Microwave Breast Imaging,“ in PIER, vol. 133, systems, high-quality and highspeed data converters, and radio frequency
pp. 309–329, 2013. integrated circuits.
[19] K. El Hallaoui, “Microwave time-domain radar for monitoring
breast health: Miniaturized hardware and anomaly detection

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JERM.2020.3029214, IEEE Journal
of Electromagnetics, RF and Microwaves in Medicine and Biology

Frederic Nabki (S’99–M’10) received the B.Eng.

degree (Hons.) in electrical engineering and the
Ph.D. degree in electrical engineering from
McGill University, Montreal, QC, Canada, in
2003 and 2010, respectively. In 2008, he joined
the Université du Québec à Montréal (UQAM),
Montreal, QC, Canada, where he was an
Associate Professor in microelectronics
engineering. In 2016, he joined the École de
technologie supérieure, Montreal, QC, Canada, a
constituent of the University of Quebec, as an
Associate Professor with the department of
Electrical Engineering. He has published two book chapters and over 130
publications, and holds 24 issued patents and 25 pending patent
applications related to MEMS and CMOS/MEMS monolithic integration.
His research interests include MEMS and RF/analog microelectronics. He
was a recipient of the Governor General of Canada’s Academic Bronze
Medal, the J.J. Archambault IEEE Canada Medal and the UQAM Faculty
of Science Early Career Research Award. He holds or has held financial
support from the Microsystems Strategic Alliance of Quebec (ReSMiQ),
the Quebec Fund for Research in Nature and Technology (FRQNT), the
Ministry of Economy, Science and Innovation (MESI) of Quebec, the
Natural Sciences and Engineering Research Council of Canada (NSERC),
and the Canada Foundation for Innovation (CFI).

Mark Coates (M. 2000, S.M. 2004) received

the B.E. degree in computer systems
engineering from the University of Adelaide,
Australia, in 1995, and a Ph.D. degree in
information engineering from the University of
Cambridge, U.K., in 1999. He joined McGill
University (Montreal, Canada) in 2002, where
he is currently a Professor in the Department of
Electrical and Computer Engineering. He was
a research associate and lecturer at Rice
University, Texas, from 1999-2001. In 2012-
2013, he worked as a Senior Scientist at Winton Capital Management,
Oxford, UK. He was an Associate Editor of IEEE Transactions on Signal
Processing from 2007-2011 and a Senior Area Editor for IEEE Signal
Processing Letters from 2012-2015. In 2006, his research team received
the NSERC Synergy Award in recognition of their successful collaboration
with Canadian industry, which has resulted in the licensing of software for
anomaly detection and Video-on-Demand network optimization. Coates'
research interests include statistical signal processing and machine
learning, communication and sensor networks, and Bayesian and Monte
Carlo inference. His most influential and widely cited contributions have
been on the topics of network tomography and distributed particle filtering.
His contributions on the latter topic received awards at the International
Conference on Information Fusion in 2008 and 2010.

Prof. Milica Popović received her M.Eng. (EE,

1994) and PhD (EE, 2001) at Northwestern
University. Since 2001, she has been with the
Department of Electrical and Computer Engineering
at McGill University, where she currently holds
associate professorship. She teaches courses on
electromagnetic fields, electrical measurements and
antennas and propagation. In 2020, she received
Samuel and Ida Fromson Outstanding Teaching
Award in Engineering. Her research revolves, in
large part, around biomedical applications of
electromagnetic theory. Prof. Popović is a Fellow of the World Innovation
Foundation, a Senior Member of the IEEE and a member of the
Professional Engineers of Ontario. For the IEEE Montreal Chapter, she
has served as the as the Chair of IEEE Women in Engineering (WIE), and
is presently the chapter’s Chair for IEEE Engineering of Medicine and
Biology Society (EMBS). Prof. Popović is on the technical program
committee of IEEE Microwave Theory and Techniques MTT-28
Biological Effects and Medical Applications, and an Associate Editor for
the IEEE Journal of Electromagnetics, RF and Microwaves in Medicine
and Biology (IEEE J-ERM).

2469-7249 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Middlesex University. Downloaded on November 03,2020 at 08:21:32 UTC from IEEE Xplore. Restrictions apply.