Detection of COVID-19: A Smartphone-Based Machine-Learning-
Assisted ECL Immunoassay Approach with the Ability of RT-PCR CT
Value Prediction
Ali Firoozbakhtian, Morteza Hosseini,* Mahsa Naghavi Sheikholeslami, Foad Salehnia, Guobao Xu,
Hodjattallah Rabbani, and Ebtesam Sobhanie
Cite This: Anal. Chem. 2022, 94, 16361−16368
Read Online
ACCESS
Metrics & More Article Recommendations
*
sı
Supporting Information
ABSTRACT: The unstoppable spread of severe acute respiratory
syndrome coronavirus 2 (SARS-CoV-2) has severely threatened
public health over the past 2 years. The current ubiquitously
accepted method for its diagnosis provides sensitive detection of
the virus; however, it is relatively time-consuming and costly, not to
mention the need for highly skilled personnel. There is a clear need
to develop novel computer-based diagnostic tools to provide rapid,
cost-ecient, and time-saving detection in places where massive
traditional testing is not practical. Here, we develop an electro-
chemiluminescence (ECL)-based detection system whose results
are quantified as reverse transcriptase polymerase chain reaction
(RT-PCR) cyclic threshold (CT) values. A concentration-depend-
ent signal is generated upon the introduction of the virus to the
electrode and is recorded with a smartphone camera. The ECL images are used to train machine learning algorithms, and a model
using artificial neural networks (ANNs) for 45 samples was developed. The model demonstrated more than 90% accuracy in the
diagnosis of 50 unknown real samples, detecting up to a CT value of 32 and a limit of detection (LOD) of 10
−12
g mL
−1
in the
testing of artificial samples.
■
INTRODUCTION
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-
2) is a positive-sense single-stranded RNA (+ssRNA) virus
from the family of Coronaviridae.
1,2
Rapid human-to-human
transmission with the high infectious capacity of the virus has
been responsible for the increasing number of infected people
worldwide. As of September 2022, there were more than 616
million reported cases of the disease, of which over 6.54
million deaths have been reported.
3
From the very start,
debates have been opened on how to control the spread of the
virus, and social distancing and mass population screening have
been suggested as practical ways of controlling the disease. As
the virus has a large incubation period
4
and symptoms similar
to other common diseases (e.g., cold), routine population
screening seems a highly rational and practical way of
hampering the growing number of infected people. Dierent
detection strategies have been reported for SARS-CoV-2 like
the detection of the nucleocapsid protein gene (N gene),
RNA-dependent RNA polymerase gene (RdRp gene), spike
protein gene (S gene), and envelope protein gene (E gene), all
with the help of nucleic acid amplification techniques.
5−7
However, these methods require nucleic acid extraction,
amplification, and detection of the end product, making their
feasibility and portability a challenge. The lateral flow
immunoassay rapid tests are another alternative for the highly
expensive and time-consuming reverse transcriptase polymer-
ase chain reaction (RT-PCR); however, they suer from a lack
of sensitivity and can be of use after a certain period of being
infected since they detect the IgG/IgM antibodies. Thus,
assays that are based on the detection of the whole virus
without pretreating the samples seem more feasible, cost-
ecient, and time-saving. The surface proteins of SARS-CoV-2
are spike protein (S), envelope protein (E), nucleocapsid
protein (N), and membrane protein (M).
8,9
Employing
detection strategies based on the detection of SARS-CoV-2
surface proteins is easier to adapt, less expensive, and does not
require skilled personnel since the sample taking is just the
nasopharyngeal swab immersed in a buer solution without
any further treatments. These immunoassays benefit from a
high level of sensitivity and specificity due to the employment
Received: August 11, 2022
Accepted: November 4, 2022
Published: November 16, 2022
Articlepubs.acs.org/ac
© 2022 American Chemical Society
16361
https://doi.org/10.1021/acs.analchem.2c03502
Anal. Chem. 2022, 94, 16361−16368