2021 2nd International Conference on Robotics,Electrical and Signal Processing Techniques (ICREST)

UWB Microwave Imaging for Non-Invasive

Anomaly Detection in Human Lung and Possible
2021 2nd International Conference on Robotics, Electrical and Signal Processing Techniques (ICREST) | 978-0-7381-3042-2/20/$31.00 ©2021 IEEE | DOI: 10.1109/ICREST51555.2021.9331250

Application in COVID-19 Diagnosis: A Review

Now shin Alam, Abir Ahmed, M d A shif Islam Oni, Tahseen A sm a Meem, Md. Abdur Rahman
Department of EEE, Faculty of Engineering
American International University-Bangladesh
Dhaka, Bangladesh
nowshin.alam@ aiub.edu, abir.ahm ed@ aiub.edu, ashif.oni@ aiub.edu, tahseen@ aiub.edu, arahman@ aiub.edu

Abstract— In this paper an overview o f the general process improvements to the existing work and future research
and im portant design factors o f UW B M icrowave Im aging is directions for applying the concepts for COVID-19 detection
given for m edical purposes, and the feasibility of its application in Section VII.
in the context o f COVID-19 detection is discussed in brief. The
recent research into COVID-19 detection using other im aging
technologies are reviewed for the sake o f comparison, and the
research lim itations for em ploying UW B im aging for the same II. R e s e a r ch o n COVID-19 D e t e c t io n

goal with acceptable results are identified. Currently, the most widespread method for COVID-19
detection is a Reverse Transcriptase Polymerase Chain
Keywords— UWB, ultra wide band, microwave imaging, reaction (RT-PCR) test, where swab samples are taken from
COVID-19 detection. the respiratory tract to be tested. But the amplification of a
specific part of the transcribed viral DNA during RT-PCR test
I. I n t r o d u c t io n
is a cumbersome and laborious process, the duration o f which
The illness ascribed to the new coronavirus was originally is prolonged even further in laboratories under heavy
called Novel Coronavirus-Infected Pneumonia (NCIP), which workload. A faster and more reliable automated system is
was later renamed by the World Health Organization (WHO) necessary for the pandemic stricken countries to deal with the
to COVID-19, an abbreviated form of Coronavirus Disease current crisis in a timely manner.
2019. [1] According to doctors, cloudy areas are seen at some
locations on the medical scans o f COVID-19 patients who A. Computerized Tomography (CT) scan
develop pneumonia. As the lung infection worsens, the air Modern CT scans are believed to be more accurate than a
sacs get full of the fluid leaking from the blood vessels in the blood test such as CRP (C-Reactive Protein Level). Song, et
lungs and are severely inflamed. At this stage the clouds form al. [7] developed an automated deep learning diagnosis system
rounded clusters and slowly turn white in color, and the based on CT scan datasets with a 95% accuracy claim. Similar
patients’ ability to absorb oxygen decreases, causing shortness CT scan based detection approaches using deep learning and
of breath, cough and other symptoms. In the worst cases convolutional neural networks were also explored in [8],[9].
patients may develop acute respiratory distress syndrome Hasan, et al. implemented an additional Q-deformed entropy
(ARDS), which is a type of lung failure and may render the algorithm and long short-term memory (LSTM) neural
patient unable to breath on their own or even die. network classifier to further increase the accuracy [10].
RT-PCR, blood test, X-ray, and computed tomography B. X-ray
(CT) scans [2] are the existing common methods for Several AI based automated detection systems based on
diagnosing COVID-19, and doctors peruse these imaging X-ray images [11] - [14] have also been developed, and may
scans to look for swelling, inflammation, or fluid in the lungs. be preferable to CT scan based approaches due to higher
While choosing a technology for medical imaging, several availability and accessibility while maintaining similar
characteristics are desired: attaining adequate penetration accuracy. An open source convolutional neural network
depth (better at low frequencies), imaging resolution (better at design called COVID-Net by Wang, et al. [12] aimed at
high frequencies), minimization o f health risk to the patient, COVID-19 detection from chest X-ray (CXR) scans. Later a
and implementation ease and cost. model called COVIDetectioNet was presented by Turkoglu in
In this paper, the feasibility of using Ultra Wideband [15] that used a pre-learned deep features ensemble and
(UWB) microwave imaging is considered as a possible low feature selection for detecting COVID-19, where the previous
risk, portable and inexpensive method for the early detection studies on the topic were also briefly reviewed and compared
of lung tissue abnormality in COVID-19 patients. There has in terms o f accuracy.
already been multiple instances o f UWB medical imaging Despite the high accuracy of the COVID-19 detection
application in the field of detecting and monitoring of water methods involving X-ray and CT scans, both technologies are
accumulation in the human body, as detecting the presence of associated with health risk problems from exposure to
water in human organs makes diagnosis of different diseases ionizing rays. That, coupled with the unavailability of the
(pulmonary edema, congestive heart failure, liver cirrhosis, screening equipment and higher costs o f operation, prevents
etc.) possible [3]-[6]. This paper briefly reviews the current their use as a long-term monitoring tool. On the other hand,
research in the field o f COVID-19 detection in Section II and microwave imaging is safer as it does not expose the patient
in the field of UWB imaging for fluid accumulation detection to ionizing rays. It is also an inexpensive and easier to operate
in Section IV, provides an overview o f UWB imaging technology with relatively higher scope of development when
methods in Section III, V, VI, and suggests possible it comes to design portability.

© IEEE 2021. This article is free to access and download, along with rights for full text and data mining, re-use
and analysis.
Page 772

Authorized licensed use limited to: Mapua University. Downloaded on March 31,2021 at 03:47:27 UTC from IEEE Xplore. Restrictions apply.
 2021 2nd International Conference on Robotics,Electrical and Signal Processing Techniques (ICREST)

III. M i c r o w a ve Im a g in g in an array or an antenna is moved to take

measurements from multiple locations.. Backscatter
Microwave imaging uses electromagnetic (EM) waves in
techniques for imaging using a single antenna keeps
the microwave range (i.e., from 300 MHz to 300 GHz) to map
calculation simple while the multistatic approach
the electrical property distributions of an object into an image
displaying the internal structure of the object material. provides the possibility of spatially focusing the signal
for a better image.
A. Justification fo r using UWB
• M odelling: While collecting data, the imaging
The UWB technology uses a very large bandwidth (>500 mechanism can either use physical phantoms modeled
MHz) and little power [4]. The selection o f the UWB pulses to mimic the properties o f human body parts, or design
for M W imaging can be justified with several factors: a) very the model in a simulated environment (e.g. Finite-
high resolution (i.e., capability of differentiating between two difference time-domain or FDTD simulation using
objects in close proximity), b) moderately high penetration in electromagnetic solver software like CST).
the human body, and c) lack of harmful physical repercussions
due to its non-ionizing properties [3]. The frequency band safe • Signal processing domain: Depending on the
for human body suggested by [16] was from 3 GHz to 5 GHz. equipment being used, the received signals can be time
domain signal (e.g. from an oscilloscope) or frequency
B. General Overview o f Imaging Process domain S-parameters (e.g. from vector network
Identifying lung abnormality with microwave imaging analyzer).
depends on the contrast in dielectric properties o f damaged
lung tissues full o f fluid compared to healthy tissues at • Algorithm Parameters: A multitude of algorithms
microwave frequencies. These differences are have been developed for microwave imaging that
characterized by relative permittivity and conductivity [4]. derive the intensity of points on the target image based
Usually conductivity increases and permittivity decreases on specific parameters from the backscattered data,
with frequency increase, so this dependence needs to be which can be amplitude, frequency, or phase o f the
taken into account when designing an imaging system. received signal. This topic has been discussed in more
detail in Section VI.
A near-field microwave imaging system attempts to
characterize the dielectric property distribution by
transmitting microwave pulse signals from an antenna
IV. R e s e a r ch on UWB I m a g i n g for Fl u id
placed close to the human body. The antennas can be
Ac c u m u l a t io n
directly mounted on the body itself or be immersed in a
coupling medium between air and skin to reduce the signal While there has been substantial amount of research on
reflection caused by the difference between the dielectric different types o f cancerous tumor detection using UWB
constant o f the two mediums [4]. The pulse signal is then imaging, there exist few works that utilize UWB imaging to
partially reflected by the air-skin interface and partially detect fluid accumulation in human body. One notable
transmitted through the human body and backscattered by example o f a time domain method that yielded experimental
the interfaces separating tissues with varying dielectric results is [3] where Pancera, et. al used two Vivaldi antennas,
properties. The next step is measuring the scattered signals a simple 3-layer phantom (muscle-water-muscle) and a time-
from the receiving antenna placed at multiple positions domain algorithm to observe how the received signal at
around the body. Finally, images are formed from the oscilloscope correlates with the transmitted signal from pulse
energy information o f the detected signal. generator to detect the presence of urine in human bladder.
The distance between peaks in correlation result allowed
The quality of the image can be judged by the contrast, calculation o f water layer for a specific cross section o f the
which refers to the ratio between the brightest and darkest bladder, and the process could be repeated from different
points (i.e. pixels) of an image. A point is bright when it is angles to assess the total water volume (left as future work).
caused by strong reflections from anomalies with
considerably different permittivity or conductivity values The most noteworthy application o f UWB imaging in
compared to the surroundings. On the other hand, small detecting fluid accumulation in lungs can be found in [5],
objects with low permittivity/conductivity contrast with where Rezaeieh, et al. employed UWB imaging methods to
respect to their environment result in weaker scattering detect fluid accumulation in lungs, also known as pulmonary
and thus darker pixels. The amount o f contrast that can be oedema. This work is especially relevant to COVID-19
achieved depends on multiple factors: the sidelobes of the detection as it achieved valuable experimental results where
system impulse response, multipath signals, device UWB imaging of an artificial human torso was performed and
internal clutter from measurement equipment, receiver water accumulation as low as 1 milliliter (Fig. 1) was clearly
noise, propagation loss, etc. identifiable in a pair of lumb lungs placed in it. In the
experimental setup, a portable VNA and an antenna array
C. Important Design Factors combined with a switching network was used for multistatic
The following factors must be taken into account when data collection from the back of the torso. This work proposed
building an imaging system. an original frequency-domain imaging algorithm using the
concepts o f Bessel function and Poynting vector, but there is
• Data acquisition: Observation signal is usually room for possible improvements since the artifact removal
collected from receiver antenna using either a method involved simple average subtraction and an average
monostatic or a multistatic approach. In the monostatic estimate o f relative permittivity. The proposed clinical system
case, a single antenna or antenna array element acts also yielded promising results when conducted with healthy
both as a transmitter and a receiver, whereas in the volunteers, so testing the reproducibility of the work and its
multistatic scenario, either multiple antennas are used possible usage in COVID-19 detection is worth investigating.

Page 773

Authorized licensed use limited to: Mapua University. Downloaded on March 31,2021 at 03:47:27 UTC from IEEE Xplore. Restrictions apply.
 2021 2nd International Conference on Robotics,Electrical and Signal Processing Techniques (ICREST)

[17] and [18] developed multilayer lung models based on

CT or MRI scan data o f the horizontal cross section of the
human chest, where each layer corresponding to a tissue type
is characterized by its relative permittivity, conductivity,
density and thickness. Several empirical models were
developed to explain the frequency dependence (i.e.
dispersion) o f conductivity and permittivity, some notable
examples being Debye model and Cole-Cole model [19],
[20]. More accurate models can be derived from detailed
anatomical representations o f the body. National Library of
M edicine’s Visible Human Project data set was used by
researchers in the past [21], and recently the Computational
Electromagnetics Group at The University o f Texas at Austin
developed open source electromagnetic voxel models o f the
human body named AustinMan and AustinWoman [22].
(a) (b)
Fig. 1. Images from torso phantom using (a) healthy lamb lungs and (b)
lamb lungs with 1 mL injected water.[5] VI. D a t a Co l l e c t io n & Al g o r it h m s

The electrical properties, shape, size, and mass of the

TABLE I. Co mpa r is o n a n a l y s is o f e x is t in g w o r k target obj ect (lung) are all factors o f the target’s characteristics
Study Inversion Method Experimental Result (impulse response) embedded in the backscattered signal. We
Qualitative, Synthetic Simple 3 layer bladder observe a cause-effect relationship: e =/ (c) where e is the
Pancera et al. [3] aperture focusing phantom, no imaging observed scattered signal, the “effect”, c indicates the physical
(Time domain) result scattering scenario (i.e. target position, size, shape) , the
Qualitative, Synthetic Torso phantom with lamb “cause”, and f is the forward solution of the electromagnetic
Rezaeieh et al. [5] aperture focusing lung (with/without water) scattering problem (i.e. the function relating cause and effect).
(Frequency domain) + healthy volunteers
Finding f is a nonlinear problem with no closed-form solution,
Only simulated phantom
Quantitative, Born and it is hard to develop a mathematical model with a simple
Dogu et al. [6] based on human lung X-
iterative method transfer function. Imaging requires solving this inverse
ray images
problem as the resultant image characterizes the scattering
scenario c, and there are two approaches for it: model-based
Later Dogu, et al. [6] proposed a simulation only detection and data-driven. Most o f the recent research has been data-
method of pulmonary oedema using Born approximation, but driven due to the mathematical complexity of the model-based
considering the level o f experimental results, extrapolating method requiring an exact model describing the environment.
Rezaeieh, et al.’s work with the help of other existing literature
on UWB imaging appears to hold greater premise. Solving the inverse problem in real space (i.e. using
position variables x, y, and z in real space) directly can be done
with both frequency domain data (e.g., S-parameters) and time
domain data (e.g., voltage waveforms). The goal is to collect
V. M o d e l l in g T h e Im a g in g Sy s t em
the signals over as many spatial points as possible and increase
The refractive index determining how much the path of an the antenna aperture, a measure o f the effectiveness of an
electromagnetic wave is reflected and refracted when entering antenna at receiving electromagnetic power. This is done by
a material depends on relative permittivity. If two adjacent using multistatic systems, and the method where the antenna
media have the same relative permittivity then there is no collects data over a line or a surface, one spatial sample at a
reflection or backscattering at the boundary between them, but time, is referred to as synthetic aperture radar (SAR).
there would be some reflection at the boundary otherwise.
Because the various human tissues/organs differ in relative A. Fourier-domain methods
permittivity, an UWB signal penetrating the human body can In these methods, the data sets o f the observed received
provide information o f the tissue properties and identify signals are subjected to Fourier transformation and the image
anomalies that sport different reflection characteristics with construction is then performed using Fourier domain
respect to the surrounding tissues [3]. Additionally, the algorithms. In the most common scenario, the responses are
portion o f the transmitted UWB absorbed by the tissues varies the S-parameters o f the microwave (MW) network formed by
across regions, and the amount of attenuation the signal the imaged object and the acquisition setup, including the
experiences in each tissue layer partially depends on the transmitting (Tx) and receiving (Rx) antennas.
conductivity of the layer.
The relationship o f voltages and currents with network
performance becomes complex at high frequencies. S-
parameters are related to power waves and offer simpler
explanations. At each frequency we have a set o f S-
parameters, which form the scattering matrix or S-matrix of
the network. Vector network analyzers (VNA) are used to
accurately and efficiently measure the S-parameters o f M W
networks. For imaging purposes, the transmission line
formalism is used to estimate the reflection and transmission
coefficients at the interfaces between different layers and the
Fig. 2. Multilayer planar model derived from the Duke’s anatomy.[20]

Page 774

Authorized licensed use limited to: Mapua University. Downloaded on March 31,2021 at 03:47:27 UTC from IEEE Xplore. Restrictions apply.
 2021 2nd International Conference on Robotics,Electrical and Signal Processing Techniques (ICREST)

behavior of the penetrating pulse. Every layer of tissue is After this calibration, the backscattered signals at all
considered a transmission line of the same length, and the antennas are time-shifted by the appropriate amount of delay
system is treated as a multiport network where Sm is corresponding to the focus point to be imaged. The time-
interpreted as the reflection of the signal from tissue layers and shifted signals are then all added together, passed through a
S„m as the transmission of the signal. window function to remove parasitic signals, and finally
integrated over time to produce the intensity of the point in
VNA sweeps over a specified frequency range by question. Later researchers introduced improvements to the
measuring the response transmission and reflection relative to classical delay-and-sum (DAS) algorithm [19] by
the applied incident signal at each frequency and represents implementing additional processing steps that involved cross
the received signal in the form of these S-parameters, return multiplications, adaptive weighting and qualify factors. The
loss, or reflection coefficients. It is also possible to construct MWDAS[23], DMAS [20], and RAR [24] algorithms are all
time domain signals by applying inverse Fourier transform similar to the DAS approach described in Fig. 3, with the steps
since the measurement includes both phase and amplitude. 14-16 involving windowing/weighting being different.
B. Time-domain methods
VII. Ch a l l e n g e s i n Re s e a r c h
The time-domain imaging systems send out pulse signals
and collect the scattered signals as waveforms over time. The In order to build upon the existing UWB imaging methods
basic concept is the same as the frequency-sweep S-parameter for water accumulation described in Section III in the context
measurements, except that all the signals are functions of time of COVID-19 detection, certain factors that impact the
and calculations for the algorithm is performed in time accuracy of the imaging result need to be investigated.
domain. Delay-and-sum (DAS) algorithms are the simplest 1) Artifact removal: In order to eliminate the artifact
and the most basic approach to image generation with pulsed noise due to the fact that the torso is a heterogeneous medium
signals. The total response of an antenna array is calculated by comprising of tissues with different dielectric properties,
forming an appropriately delayed and linear weighted majority of past research work opted to use an average
combination of the sensor outputs. As an example, the step by subtraction method, which may not be an accurate portrayal
step process of the DAS algorithm is explained in Figure 2,
of the environmental noise, especially when the noise
where pre-processing techniques are first used to eliminate the
noise and background reflections from the data received by includes sudden spikes or the scattered wave at each antenna
the antenna array. experiences different patterns of noise. While many methods
aim at noise removal with the assumption that the noise has a
Algorithm 1 DAS algorithm Gaussian distribution of zero mean, it may be advisable to use
1 : Transmitter Antenna Signal s(t) prior knowledge about the noise to identify different noise
2: Receiver Antenna Signal rrn(t) > due to the contribution types and subtract them from the unprocessed signal. The
o f all (x,y) points in environment different size and shape of patient bodies and the nearby
M
environment also need to be taken into account for more
3: c.(t) = j j ^2 rm(t) > calibration/reference signal that
*=1 universal algorithms that do not suffer from loss in generality.
approximates scattered signal from environment 2) Antenna design limitations: The dilemma of
4: va - average velocity o f propagation in the environment
transmitting a strong, directional signal that can penetrate
at the center frequency o f input pulse
obstacles like bone without harming the human body is a
5: For each o f the M antennas: significant limitation in signal detection. While UWB has low
6: for i = 1 to M do dissipation (i.e, the conversion of EM energy into heat) for
x 'm ( t ) = r m (t ) — c{t) > environment noise removal short distance propagation through air, the medium losses due
T
x„{t) = ^ £ x>m(t) > moving average of x'm(t) to human body still may be high enough that the scattered
t= o signal goes below the detectable level of the receiver antenna.
x m (t) = x'm (i) — xa(t) > scattered signal only from
The frequency band the antenna needs to operate on to
imaged object
7: For each (x,y) point in N X N grid:
provide acceptable results corresponds to a large anenna size,
8: for j = 0 to N —1 do which complicates the design process. The radiation pattern
9: for /.: = 0 to JV I do also needs to as unidirectional as possible for efficient signal
tm (x,y) = ^ JL > delay needed utilization and minimalizing interference. Also, non­
for signal to travel from transmitter o f mth antenna to the idealities in antenna impulse response due to frequency
(x.y) point and back to the receiver o f mth antenna dependency of antenna can cause distortions in the
10: end for transmit/receive signal. The frequency dependency of the
11 : Calculate x m( t-tm(x, y)) t> delay shifting position of the phase center of a UWB antenna is also a factor
12: end for
not considered in most approaches, which can impact the
13: end for
delay calculation in time-shifting algorithms.
14: For each (x.y) point in N X N grid: 3) Permittivity assumptions: Another common issue
15: for x = 0 to JV —1 do with MW imaging algorithms are the assumptions about
16: for y = 0 to N — I do permittivity when computing signal propagation speed
M
y ( t,x .y ) = £ x,n(t-tm{x,y)) through a medium. Time delay calculation in RAR [24] and
z(t, x, y) = w(t ) * y(t, x, y) > Windowing DMAS [20] involves assuming that the permittivity values of
I{x. y ) = f z(t)2dt. > Calculate pixel intensity different body tissue layers are constant over the whole
spectrum at the center frequency, which may lead to
Fig. 3. Pseudocode for DAS Algorithm. distortions in the image. Most works also opt to choose the

Page 775

Authorized licensed use limited to: Mapua University. Downloaded on March 31,2021 at 03:47:27 UTC from IEEE Xplore. Restrictions apply.
 2021 2nd International Conference on Robotics,Electrical and Signal Processing Techniques (ICREST)

largest or the average permittivity when calculating time 2017 10th International Conference on Electrical and Electronics
Engineering (ELECO), Nov. 2017, pp. 1054-1057.
delays o f signal passing through multi-layered body models,
[7] Y. Song et al., “Deep learning Enables Accurate Diagnosis of Novel
w hich can result in degradation o f image quality. Coronavirus (COVID-19) with CT images,” medRxiv, p.
4) Effect o f respiration: As the lungs sport different 2020.02.23.20026930, Feb. 2020, doi: 10.1101/2020.02.23.20026930.
dielectric properties during respiration, inhalation and [8] S. Wang et al., “A deep learning algorithm using CT images to screen
exhalation, the differences between these stages need to be for Corona Virus Disease (COVID-19),” medRxiv, p.
2020.02.14.20023028, Apr. 2020, doi: 10.1101/2020.02.14.20023028.
considered in the context o f modelling the lung for COVID-
[9] X. Xu et al., “Deep Learning System to Screen Coronavirus Disease
19 induced pneumonia detection. Babarinde, et al [17] 2019 Pneumonia,” arXiv:2002.09334 [physics], Feb. 2020, Accessed:
included the effect of respiration by applying different tissue Oct. 30, 2020. [Online]. Available: http://arxiv.org/abs/2002.09334.
simulating liquids in their experimental phantom for lung [10] A. M. Hasan, M. M. AL-Jawad, H. A. Jalab, H. Shaiba, R. W. Ibrahim,
and A. R. AL-Shamasneh, “Classification of Covid-19 Coronavirus,
tumor detection, but simulation based modelling tends not
Pneumonia and Healthy Lungs in CT Scans Using Q-Deformed
take this factor into account. Entropy and Deep Learning Features,” Entropy, vol. 22, no. 5, Art. no.
5) Specific absorption rate (SAR): While 5, May 2020, doi: 10.3390/e22050517.
microwave imaging is a safer imaging technique than others, [11] J. Zhang et al., “Viral Pneumonia Screening on Chest X-ray Images
Using Confidence-Aware Anomaly Detection,” arXiv:2003.12338 [cs,
care should be taken when designing the system to ensure that eess], Sep. 2020, Accessed: Oct. 30, 2020. [Online]. Available:
the SAR value, the rate of energy absorption in human body http://arxiv.org/abs/2003.12338.
tissues, does not exceed beyond the safety level o f exposure [12] L. Wang and A. Wong, “COVID-Net: A Tailored Deep Convolutional
as defined by the official safety standards. The maximum Neural Network Design for Detection of COVID-19 Cases from Chest
X-Ray Images,” arXiv:2003.09871 [cs, eess], May 2020, Accessed:
permissible SAR is 1.6 W/kg for an average mass of 1 g Oct. 30, 2020. [Online]. Available: http://arxiv.org/abs/2003.09871.
according to IEEE C 95.1:1999 guidelines and 2 W/kg for an [13] I. D. Apostolopoulos and T. A. Mpesiana, “Covid-19: automatic
average mass o f 10 g as per International Commission on detection from X-ray images utilizing transfer learning with
Non-Ionizing Radiation Protection (ICNIRP) standards. convolutional neural networks,” Phys Eng Sci Med, vol. 43, no. 2, pp.
635-640, Jun. 2020, doi: 10.1007/s13246-020-00865-4.
Cavagnaro, et al. [25] derived the maximum Effective
[14] A. Narin, C. Kaya, and Z. Pamuk, “Automatic Detection of
Isotropic Radiated Power (EIRP) o f an UWB antenna with a Coronavirus Disease (COVID-19) Using X-ray Images and Deep
3.1 - 10.6 GHz spectrum to be 0.556 mW in order to satisfy Convolutional Neural Networks,” arXiv:2003.10849 [cs, eess], Oct.
the FCC limits. The SAR value calculated by Rezaeieh, et al. 2020, Accessed: Oct. 30, 2020. [Online]. Available:
http://arxiv.org/abs/2003.10849.
[5] was reported to be 0.03 W/Kg w hich is within the
[15] M. Turkoglu, “COVIDetectioNet: COVID-19 diagnosis system based
acceptable limits. on X-ray images using features selected from pre-learned deep features
ensemble,” Appl Intell, Sep. 2020, doi: 10.1007/s10489-020-01888-w.
VIII. C o n c l u s io n [16] L. M. Camacho and S. Tjuatja, “FDTD simulation of microwave
The review of currently existing COVID-19 detection scattering from a lung tumor,” in 2005 IEEE Antennas and Propagation
Society International Symposium, Jul. 2005, vol. 3A, pp. 815-818 vol.
methods reveals the research gap in the field o f developing a 3A, doi: 10.1109/APS.2005.1552382.
low-cost and risk-free detection technique, and to the author’s
[17] O. J. Babarinde, M. F. Jamlos, P. J. Soh, D. M. M.-P. Schreurs, and A.
knowledge, no UWB microwave imaging based approach has Beyer, “Microwave imaging technique for lung tumour detection,” in
been developed so far. Additionally, literature on UWB 2016 German Microwave Conference (GeMiC), Bochum, Mar. 2016,
imaging techniques have been reviewed. From the works pp. 100-103, doi: 10.1109/GEMIC.2016.7461566.
conducted by the past researchers, especially Rezaeieh et al.’s [18] M. Cavagnaro, E. Pittella, and S. Pisa, “UWB pulse propagation into
[5], it has been found that UWB imaging holds notable human tissues,” Phys. Med. Biol., vol. 58, no. 24, pp. 8689-8707, Dec.
2013, doi: 10.1088/0031-9155/58/24/8689.
premise in detecting anomalies in COVID-infected lungs as
[19] S. C. Hagness, A. Taflove, and J. E. Bridges, “Two-dimensional FDTD
long as the existing limitations in antenna design and image
analysis of a pulsed microwave confocal system for breast cancer
processing are properly addressed. detection: fixed-focus and antenna-array sensors,” IEEE Transactions
on Biomedical Engineering, vol. 45, no. 12, pp. 1470-1479, Dec. 1998,
doi: 10.1109/10.730440.
Re f e r ences
[20] H. Been Lim, N. Thi Tuyet Nhung, E.-P. Li, and N. Duc Thang,
“Confocal Microwave Imaging for Breast Cancer Detection: Delay-
[1] N. Zhu et al., “A Novel Coronavirus from Patients with Pneumonia in Multiply-and-Sum Image Reconstruction Algorithm,” IEEE
China, 2019,” New England Journal of Medicine, vol. 382, no. 8, pp. Transactions on Biomedical Engineering, vol. 55, no. 6, pp. 1697­
727-733, Feb. 2020, doi: 10.1056/NEJMoa2001017. 1704, Jun. 2008, doi: 10.1109/TBME.2008.919716.
[2] L. Fan et al., “Progress and prospect on imaging diagnosis of COVID- [21] V. Spitzer, M. J. Ackerman, A. L. Scherzinger, and D. Whitlock, “The
19,” Chin J Acad Radiol, vol. 3, no. 1, pp. 4-13, Mar. 2020, doi: visible human male: a technical report,” J Am Med Inform Assoc, vol.
10.1007/s42058-020-00031-5. 3, no. 2, pp. 118-130, Apr. 1996, doi: 10.1136/jamia.1996.96236280.
[3] E. Pancera, T. Zwick, and W. Wiesbeck, “Ultra wideband radar [22] “AustinMan Electromagnetic Voxels.”
imaging: An approach to monitor the water accumulation in the human https://web.corral.tacc.utexas.edu/AustinManEMVoxels/AustinMan/i
body,” in 2010 IEEE International Conference on Wireless Information ndex.html (accessed Oct. 30, 2020).
Technology and Systems, Honolulu, HI, USA, Aug. 2010, pp. 1 ^ , doi: [23] F. Yang and A. S. Mohan, “Detection of malignant breast tissues using
10.1109/ICWITS.2010.5611899. microwave imaging,” in 2009 Asia Pacific Microwave Conference,
[4] E. Pancera, “Medical applications of the Ultra Wideband technology,” Singapore, Singapore, Dec. 2009, pp. 397^00, doi:
in 2010 Loughborough Antennas Propagation Conference, Nov. 2010, 10.1109/APMC.2009.5384517.
pp. 52-56, doi: 10.1109/LAPC.2010.5666802. [24] T. Yin, F. H. Ali, and C. C. Reyes-Aldasoro, “A Robust and Artifact
[5] S. Ahdi Rezaeieh, “Wideband Microwave Imaging Systems for the Resistant Algorithm of Ultrawideband Imaging System for Breast
Diagnosis of Fluid Accumulation in the Human Torso,” PhD Thesis, Cancer Detection,” IEEE Trans. Biomed. Eng., vol. 62, no. 6, pp.
The University of Queensland, 2016. 1514-1525, Jun. 2015, doi: 10.1109/TBME.2015.2393256.
[6] S. Dogu, i. Dilman, M. Qayoren, and i. Akduman, “Imaging of [25] M. Cavagnaro, S. Pisa, and E. Pittella, “Safety aspects of human
pulmonary edema with microwaves — Preliminary investigation,” in exposure to ultra wideband radar fields,” Sep. 2012, pp. 1-5, doi:
10.1109/EMCEurope.2012.6396885.

Page 776

Authorized licensed use limited to: Mapua University. Downloaded on March 31,2021 at 03:47:27 UTC from IEEE Xplore. Restrictions apply.