Convolutional Neural Network Approach in
Covid-19 Screening in Asymptomatic Individuals
Mohammed-Amine Zyad
Polydisciplinary Faculty
Sultan Moulay Slimane University
Beni Mellal, Morocco
zyad@usms.ma
Belaid Bouikhalene
Polydisciplinary Faculty
Sultan Moulay Slimane University
Beni Mellal, Morocco
B.BOUIKHALENE@usms.ma
Abdelmajid Zyad
Faculty of Science and Techniques
Sultan Moulay Slimane University
Beni Mellal, Morocco
A.ZYAD@usms.ma
Abstract—The new coronavirus (COVID-19) spread rapidly
around the globe and hits around six million of cases as reported
by the World Health Organization (WHO) at the time of writing.
This growth in number of cases made it hard for doctors and
radiologists to correctly diagnose the COVID-19 disease. The use
of automated methods can be helpful to efficiently and accurately
detect the disease, and can also be used to assist doctors who
lack the proper tools for diagnosis. In the present study, we
provide a novel method using Convolutional Neural Networks
(CNN) to accurately classify COVID-19 cases from raw chest
X-ray images by 95%. The proposed model is lightweight and
can be easily deployed to the Cloud or mobile devices with little
compute power.
Index Terms—classification, convolutional neural networks,
COVID-19, coronavirus
I. I NTRODUCTION
In December 2019, a new coronavirus disease (COVID-19)
was discovered in Wuhan, China [1]. The virus that caused it
is known as severe acute respiratory syndrome coronavirus
2 (SARS-CoV-2). This virus has the capacity to transfer
from animals to humans due to its zoonotic nature [2]. The
symptoms of this disease are divided into two groups, as Zhang
[3] stated: ”Those with gastrointestinal symptoms (G group)
and those without gastrointestinal symptoms (NG group).
Common gastrointestinal symptoms included inappetence, di-
arrhea, nausea, abdominal pain, and vomiting. Significantly
higher proportions of patients with fever, dizziness, myalgia,
and fatigue were noted in group G than in group NG.”
The coronavirus started originally in China and spread to
Japan, South Korea and Europe after that, Italy and Spain in
particular, where it hit very hard [4]. In Morocco the first
known case traced back to 2
nd
of March 2020 and it went
up to 5856 cases in 11
th
of May 2020. According to World
Health Organization (WHO), the total confirmed cases in the
world is around 6.2 million.
Doctors and radiologists’ important task is to diagnose
cases, whether they have COVID-19 or not, to a certain accu-
racy. The problem is that it is easy to get overloaded looking
at the huge number of cases. This can lead us to think about
automatic methods, mainly machine learning methods which
gained a huge success in recent years to assist radiologists in
this mission.
In the present study, we propose an automatic method using
Convolutional Neural Networks (CNN) to detect and classify
X-ray chest images wether the patient suffers from COVID-19
or not. In Section III, we will present a brief history about the
neural networks and their usage in the medical imaging field
and present some alternative methods using automatic methods
to classify COVID-19. In Section IV, we present the dataset
and the model used to achieve the results cited in this paper.
Finally, in Section V we present the results after successfully
applying the model on the dataset and an interpretation of what
those results mean, besides a limitation to this method.
II. DIAGNOSTIC APPROACHES OF SARS-COV-2
The demonstration of the positivity of patients to Covid19
is carried out by a molecular biology test (PCR) and a
serological test looking for antibodies (IgM and IgG) against
the SARSCoV2 virus.
A. Evaluation by a Quantitative RT-PCR Assay
Generally, serial samples (throat swabs or sputum) from
patients were collected and tested for SARS-CoV-2 [5] [6].
The presence of SARS-CoV-2 in respiratory specimens was
detected by quantitative real-time RT-PCR of SARS-CoV-2
[5]. The Total RNA was extracted using the respiratory sample
RNA isolation kit. Then, vortex and transfer of 50 mL of cell
lysates into another collection tube were made. After that,
the tubes were centrifuged at 1000 rpm for 5 min. After 10
min incubation at room, 5 mL of RNA sample was prepared
to be used for RT-PCR amplification. Real time RT-PCR was
performed using the nucleic acid testing kit specific for SARS-
CoV-2 detection. The open reading frame 1 ab (ORF1ab) and
nucleocapsid protein (N) were simultaneously selected as the
two target genes. The human GAPDH gene was used as an
internal control [7]. The specific primers and probes set for
ORF1ab and N were as follows: ORF1ab-forward primer 50-
ACCTTCTCTTGCCACTGTAGC-30, ORF1ab-reverse primer
50-AGTATCAACCATATCCAACCATGTC-30, probe 50-
FAM ACGCATCACCCAACTAGCAGGCATAT- BHQ1-30,
N-forward primer 50-TTCAAGAAATTCAACTCCAG-
3, N-reverse primer 50-AGCAGCAAAGCAAG 978-1-6654-4103-2/21/$31.00 ©2021 IEEE
2021 7th International Conference on Optimization and Applications (ICOA) | 978-1-6654-4103-2/20/$31.00 ©2021 IEEE | DOI: 10.1109/ICOA51614.2021.9442641