World Health Organization (WHO) acknowledged the coronavirus disease 2019 (COVID-19) as one of the global diseases. The novel coronavirus disease is named Severe Acute Respiratory Syndrome Coronavirus (SARS-COV2) is a mild to severe respiratory disease having fever, cough, and shortness of breath as initial symptoms. This virus disseminates through contact with infectious persons, touching contaminated surfaces and infectious air droplets. The virus invades into healthy cells of the body especially the lungs causing the respiratory problems and sometimes causes organ failure by killing healthy cells which eventually leads to death. The origin of coronavirus disease was zoonotic, as the initial cases had been reported from animals in Wuhan. This paper presents the literature review on the identification COVID-19 patients. Furthermore, the treatment and remedies of COVID-19 patients are discussed utilising machine learning and prediction models. The datasets related to the cases of COVID-19 patients are also discussed in this paper. Finally, the COVID-19 confirme d cases are becoming exponentially
COVID-19 is an epidemic virus that was initially spread within China and later it was declared by the WHO as a pandemic disease that was spread globally. On 31st December 2019, the first case of the threatening disease in human beings was disclosed in Wuhan; a city of China. SARS-COV-2 is the name given to this coronavirus disease which is mild to serious disease with the initiative symptoms retaining cough, fever and dificulty in breathing. This virus disseminates through contact with infectious persons, touching contaminated surfaces and infectious air droplets. The virus invades into healthy cells of the body especially the lungs causing the respiratory problems and sometimes causes organ failure by killing healthy cells which eventually leads to death. The origin of coronavirus disease was zoonotic, as the iniChaudhary)
Figure 1 shows the WHO statistics on COVID-19 related to the total number of active cases, cured cases, and 37,289 confirme d cases ,which included 813 deaths in china and 302 confirme d cases in multiple countries. The foremost epidemic named Severe Acute Respiratory tial cases had been reported from animals in Wuhan [1]. the WHO report. As of January 6th, 2020, 44 confirme d cases were reported in the Americas region. AccordSyndrome (SARS-COVID-19) was reported in 2002-2003. ing to WHO month-wise report from January to July Let us assume that the model yields the droplet lifetime is , hence the infection rate constant is calculated as follows: 1 = ( ), 2 = ( 1), where , are the functions. • Stage 4: It is an uncontrollable stage. This is the worst stage as the transmission becomes pervasive and the number of cases crossed the threshold limit. China was reported as the first union WHO faced all these four stages [3]. The reaction based pandemic model equation is calculated on the basis of infection growth rate as ( 2 × ), where is the time interval during the diagnosis phase.
WHO classifie d novel coronavirus into four transmission stages as presented in Figure 2.
• Stage 1: This stage begins as the initial spread of the disease with the travel history of a person from infected areas. Such people were kept quarantined for a time period of 15 days in hospitals.
The infection would be controllable at this stage because of the easy tracing of infection sources. • Stage 2: It is the local transmission in which cases were reported in the peoples who came into contact with infected persons, hence till this stage, sources were traceable. Lockdown became necessary to prevent further spread of this viral.
To break the chain of viral transmission, PM Modi announced ‘JANTA CURFEW’ on March 22nd, 2020 in India.
The global situation was unbalancing very dangerously at • Stage 3: It is community transmission having nei- that time due to the extremely high risk of this pandemic. ther travel history nor in contact with infected 266,073 confirme d cases and 11,184 deaths were reported persons. Sources became untraceable due to its at the global level at that time. India was struggling hard widespread. The reason behind its transmission from COVID-19 with a total of 360 confirme d cases and was the presence of air droplets due to exhalation deaths [4]. In India, there is an immediate requirement of by the infected person and inhalation of those hospitals all around the country so temporary hospitals droplets by a normal person. Some necessary are constructed to overcome loads of COVID-19 patients. steps were taken to control the disease by ap- Moreover, there are a huge demands of around 1 milplying social distancing, masks, and sanitizers. lion ventilators at the peak time of epidemic in India but Confirmed cases occurred due to travel
history, droplets crystalized in air Measure: Hand hygiene, cleaning, early detection, isolation, avoid traveling Infection growth rate: exp(r2 t)
Measure: Hand hygiene, cleaning, social distancing Stage 1: Imported cases and little transmission
Stage 4: Full Epidemic (Uncontrollable
stage) Clusters Transmission
Transmission Stages Sporadic Cases
Late
Epidemic CORONAVIRUS (COVID-19)
Respiratory Droplets Evaporation,
Precipitation Stage 2: Imported cases and new cases in local clusters
Interaction
Main Wave
Stage 3: Community transmission Local Transmission Measure: Hand hygiene, cleaning, early detection, isolation, avoid traveling
General Population
Droplet lifetime:ω
Infection rate constant: r1 = f(ω), r2 = g(r1) Measure: Hand hygiene, cleaning, social
distancing at present, there are approximate 30K-50K ventilators, tion for the elimination of the coronavirus by applying similarly it is estimated that the United States has 160K difer ent prediction models. For remote healthcare manventilator which is very short as per the patient demands. agement of COVID-19, Machine Learning models are mainly focused in this paper with research challenges 1.3. Comparison with the existing of COVID-19 pandemic and their key solutions. As the surveys cases were reporting from all continents of the world and the viral was going on spreading desperately, it becomes The comparison of the existing state-of-the-art schemes a threatening alarm for the whole of mankind. The news on the COVID-19 is discussed in Table 1. was coming from difer ent continents with the adverse 1.4. Motivation & Scope of the Survey efe cts of this viral disease. The whole world was facing this critical situation and the economy of the world came The motive of this study is not only limited up to the down due to instant lockdown which ultimately leads estimation of COVID-19 data but also to provide a solu- to a struggle full life for mankind at the global level. It Review: Identification and Diagnosis of Clinical Characteristics of COVID-19 Virus
Section I: Introduction Section 1.1 Data Statistics on COVID-19 Section 1.2 Transmission Stages of COVID Infection Section 1.3 Comparison with existing survey
Section 1.4 Motivation and Scope of the survey structured into mind to collect COVID-19 data and to find out solutions using technical applications that may prove helpful in eradication COVID-19. Data collected with the help of WHO reports and worldometer. Motivated from the proposals of the existing work done by the researchers on COVID-19, we presented a literature review on early detection of the symptoms of COVID virus and discuss the performance of the various prediction models based on the machine learning models.
Figure 3 presents the taxonomy of the literature review on COVID-19. Section 1 includes the definition and history of COVID-19 followed by the statistics and transmission stages. Then, the comparison with the existing survey, motivation, and scope of the review paper are discussed. Section 2 includes the clinical characteristics of COVID-19 disease. Section 3 considers the diagnosis of COVID-19 patients followed by Section 4 that presents the treatment and remedies adopted for COVID-19 based on machine learning. Section 5 discusses the datasets related to COVID-19 patients. Section 6 focused on the research gaps and Section 7 analyzed the open issues and future directions. Finally, Section 8 concludes the paper.
Coronaviruses are enfold viruses with a assurance, singlewrecked RNA genome. Coronaviruses possess the largest genomes for RNA viruses having a length ranging between 26 to 32 base pairs (kb). For replication, spike formation, and nucleocapsid of viral, all types of coronaviruses contain specific genes in ORF next regions that conceal proteins. For the attachment and entry of host cells, some glycoprotein spikes are present on the outer surface of coronavirus. The entry receptor for SARS-COV-2 spike glycoprotein is a Human AngiotensinConverting Enzyme-2 (ACE-2) which is expressed in epithelial cells of the heart, small intestines, and kidneys [5]. In humans and animals, coronaviruses cause intestinal and respiratory infections. While studying on difer ent cases of COVID-19 patients, difer ent infectious categories were reported having decreased volume of the virus in the nasopharyngeal tract and cause organ failure, decreased rate of the virus in an upper and lower tract of the respiratory tract with detection of the virus in plasma, nasopharyngeal shedding of SARAS-COVID-2 in the patients having normal symptoms by early diagnosis.
Table 2 presents the clinical and biological characteristics of the COVID-19 patients. We consider the COVID19 three patients data such as patients X, Y, and Z, where all these three patients belong to China having a sickness period of 15 days, 26 days, and 23 days, respectively. First patient X was a female having age 30, while the second patient Y is age 48 and the third patient Z is age 80 and both the Y and Z were male. To diagnose COVID-19 patients, the following parameters were taken into account: Age, Chronic Sickness, Province, Travel History, Sickness Period, Diagnosed date, COVID Symptoms, Chest X-ray, Haemoglobin, Platelet Count, Sodium, WBC Count, Urea, Creatinine, Nature Sickness Type, Test Conducted. Symptoms of Patient X and patient Y were detected which include: Fever, Conjunctivitis, Influenza, Cough (mild symptoms) and after the final treatment, both the patients were fully recovered and discharged. Third patient Z having fever, cough, diarrhea, shortness of breath symptoms (severe symptoms), was not able to recover and died on Feb 14, 2020, after a time period of 16 days of hospitalization [5].
In this section, we diagnose the symptoms of COVID-19 patients. The patient sufering from the COVID-19 virus has the following symptoms such as– fever, cough, dificulty in breathing, repeated shaking with chills, fatigue, muscle pain, headache, sore throat, taste congestion or runny nose nausea, vomiting, and diarrhea. The COVID Extract the features (reduce complexity and variance of the
training data) patient has to follow the incubation time ranges between 1-15 days after the detection of the virus. Many studies identify the prediction models for valediction of COVID19 in the broad population. Age, sex, previous hospital history incorporates the predictor factors. Many prognosticator models are applied to difer ent COVID-19 cases to diagnose those. Computed Tomography (CT) images, x-rays, RT-PCR, and many others are used to diagnose the patients in most of the studies. Difer ent prognostic models are used for the patients recognized with COVID19 by using the many factors which include sign and symptoms, CT features, laboratory symbols, LDH level, and so many others. A more detailed view of diagnostic and prognostic models is specifie d in Table 3 where all the elaborated features and tests used make it clear. ing (DL) techniques are used to optimize imaging and
This segment gives an outline of technologies which textual dataset. The last layer is the output Layer which are recently in use. The topmost attributes of the tech- evaluates and predicts the final diagnosis using the best nologies and their consequences to face the pandemic prediction model. and to merge the use of these technologies to audit the People from all fields i.e. engineers, doctors, researchers, spreading virus are highlighted in this section. The con- students, and all others are working on the eradication cerned authorities are informed to poise these technolo- of COVID-19. Numerous ideas are presented and many gies which are helpful in eliminating the disease. The innovations are done by them which contains testing work done is described under eight modules such as– pre- from wastewater, molecular point of care testing, autodiction analysis, detection analysis, discovery analysis, matic mask creator machine, jeeva setu ventilators, qura telemedicine analysis, containment analysis, infodemic ring, PPEs, microwave sanitizer atulya, UV sanitizer deanalysis, economy disruption analysis, and social con- vice and safe swab. Table 5 presents the treatment and trol analysis. The strength, weakness, opportunities, and solutions that covers the following attributes such as– threat (SWOT) analysis in diagnosing the symptoms of features, accuracy, platform used for analysis and expenCOVID-19 is summarized in Table 4. diture of these innovations. For future forecasting of positive confirme d cases, death cases and recovery cases, four prediction models – LR (Linear Regression), LASSO 4. Treatment and Remedies of (Least Absolute Shrinkage & Selection Operator), SVM COVID-19 Infection based on (Support Vector Machine), ES (Exponential Smoothing) Machine Learning are used. Table 6 presents the difer ent prediction models of COVID-19 and their applications. Table 6 covers diferFigure 4 presents a system model of a machine learning- ent applications used by difer ent researchers using varibased scheme for COVID-19 patients. The layered archi- ous medical standards and approaches. The succeeding tecture covers 5 layers to diagnose clinical characteristics applications are used for early detection and diagnosis usof COVID-19 patients. The first input Layer is designed ing radiology images, tracking the outbreak, prediction for the COVID patients database that can deliver vast outcome of COVID-19 patients, protein structure prepackets of data to the main server. The second layer is the dictions, biomedicine perceptive, pharmacotherapy and training & feature selection, it picks best-suited imaging drug discovery, and social awareness and control. Chest techniques according to the previous experience of data. X-ray images, CT images, the temperature of infected The third layer is the imaging-based techniques. CT scan, people, infected lung cells, classification of respiratory MRI, PET, X-Ray, Optical, and Digital Microscopic Imag- patterns are the difer ent medical standards used by the ing Techniques are suggested to detect disease by imag- researchers using various machine learning models. The ing tests. The fourth layer is Optimization Techniques performance of these machine learning models are sumin which Artificial Intelligence (AI), Machine Learning marized in Table 7. The summarized outcome of the final (ML), Convolution Neural Networks (CNN), Deep learn- diagnosis is that the exponential smoothing (ES) machine learning approach performs the best outcome among all the prediction models [33].
The machine learning model is a mathematical representation of the real-world data the output has been familiar with the certain type of trained designs. In order to perform improved outcomes and performance, developers re-train the existing model. Machine Learning techniques are further divided into three categories i.e., (i) supervised machine learning, (ii) unsupervised machine learning, (iii) reinforcement learning, and (iv) predictive modeling techniques.
The model will be trained with loaded labeled input data. In the labelled dataset, solutions have been given for each dataset. The machine learns from it and makes available solutions for each problem. It predicts the data according to the trained dataset. Supervised Machine Learning is used Classification and Regression methods.
An extensive range of textual classification methods have been used to estimate social media sentiment and based on their likenesses; these classifiers are assembled into various categories. In this study, two classifiers are mainly focused named as Naïve Bayes Classifier (NBC) and kNearest Neighbors (k-NN) [34].
• Naïve Bayes Classifier (NBC): With advanced accuracy to classify text, products, and documents, Naïve Bayes Classifier is a modest and operative method based on Bayes Theorem. To demonstrate this technique, binary input values are used. Bernoulli Naïve Bayes and Multinomial Naïve Bayes are the two models used in Naïve Bayes Classifier which characterizes the frequency and binary features, respectively. To estimate the required parameters, NBCs can be used with restricted size training data. This classifier is operative with real-world data and can deal with dimensionality. Naïve Bayes Classifier probability rule is computed by using the equation.
(| ) = ( |) ()/ ( ), (1) • k-nearest Neighbors (k-NN): k-Nearest Neighbors, a common non-parametric classifier is easy to implement and appropriate for multi-class complications. Based on the similarity dimension, the k-NN method classifies text and documents very excellently. With slight data, K-NN is a totally efifcient and useful algorithm. This classifier works on parallel measurement standards where the approximation of distance and proximity are the aspects used to measure the similarity between two data points. K-NN computes classification based on the bulk of nearest neighbors and their number is resolute within a static radius of an individual point.
One basic type of Machine Learning is the regression model, it consists of two variables one predictor variable and another is the dependent variable. The regression model analyse the relationship between dependent and independent variables. It predicts the best fit data. Some regression models are listed below.
• Linear Regression (LR): Linear Regression is a valuable and constant classifier , primarily used to predict relationships between continuous variables and used to organize text and documents. The least-squares algorithm method is used to minimize squared alteration between the predicted results and exact classes. To accomplish a regression task, linear regression is used which is a machine learning algorithm based on supervised learning. For future forecasting of positive confirme d cases, death cases, and recovery cases, four prediction models – LR (Linear Regression), LASSO (Least Absolute Shrinkage & amp; Selection Operator), SVM (Support Vector Machine), ES (Exponential Smoothing) are used. The performance of these models is given in Table 7. ES performs the best among all the prediction models. • The model necessity alterations if the structure is suitable due to non-continuous quantitative variables of predictors or replies. Here, the mathematical model is A, and B are two substitution variables that signify features of documents and class to be evaluated, respectively.
= 0 + 1 1 + 2 2 + 3 3 + … + . (2) Estimated value will find ∧ for a mathematical value 0∧, 0∧, … ∧ for 0, … . • Support Vector Machine (SVM): Support Vector Machine model is used for the regression and classification model of Machine Learning. Due to its increased performance in real-life applications, SVM assists as a foremost tool for data regression and classification. It is a supervised Machine Learning algorithm. SVM can perform proficiently on non-linear classification while executing linear classification. It performs excellently when dimensional spaces are greater than the number of samples.
Clustering possess acquires the complementary features This section mainly covers the dataset of COVID-19 paas of classification excluding the base. In such a method, tients. Table 8 summarized the list of datasets related to we don’t have any knowledge about the clusters or seg- COVID-19 patients. The datasets are categorized into two ments which we have to spot in our data and what we data type i.e., (i) medical imaging and, (ii) textual. The have to explore. Some efe cts like clusters, structures and medical imaging data type is further categorized on the compositions can be unexpectedly emerge in our dataset basis of CT scans and chest x-ray, while the textual data while using a clustering algorithm. Clustering is used for type is categorized on the basis of COVID-19 case reports, data compression and generalization of data. social media data, and scholarly articles. In addition, the features and applications alongwith the trained model 4.2. Unsupervised Machine Learning used in datasets are elaborated. Difer ent technologies like AI, ML, NLP are used for practical purposes.
Unsupervised Machine Learning finds an unknown pattern in a dataset without pre-existing markers, it is a self-organized technique. There is no need to train output variables only the input dataset is given previously.
Unsupervised Machine Learning used Clustering and Reduction Methods for prediction.
Reinforcement Learning is an environment interactionbased method. It works on two stages: Start and End stage, there may be difer ent ways to reach an end state. Some examples are driverless cars, robot navigation, selfnavigating vacuum cleaner, etc.
SEIR stands for Susceptible, Exposed, Infectious, and Removed model. Susceptible are individuals if it exposed infection it becomes hosts, Exposed are individuals that are infected with asymptomatic, Infectious is already infected and it can transmit infection, Removed are individuals who are no lengthier infected and recovered.
After analyzing the existing work by the researchers on COVID-19, the following research gaps have been identiife d.
• Less focus on early detection and diagnosis of COVID-19 patients: High number of patients did not detect at early stages and hence not diagnosed timely. With the rapidly increasing new confirme d cases of COVID day by day, it is necessary to timely diagnose the suspected cases to control the crises of COVID-19. • Insuficient utilization of resources like physicians, medical equipment, hospitals: There is no proper use of resources related to the treatment of COVID-19 patients. It includes many reasons like improperly or less guided doctors who are not able to treat the patients, less amount or sometimes no availability of medical equipment, lack of hospitals up to long distances, and less arrangement of ventilators and proper space in hospitals.
• Less focus on funding towards drugs and vac- 7.
cines: Many researchers and physicians explore that drugs will be a quicker way to tackle the coronavirus pandemic. By drugs and vaccines, protein expansion speeds up in our body for eficient treatment of coronavirus. So the government should be funded towards drugs and vaccination to deal with this deadly disease.
The open issues, future Directions, and impact of COVID19 on social life is summarized in Table 9. While presenting difer ent future directions, diverse open issues are challenged which are- 1) In the Transportation sector where vehicles did not move from one place to another • No focus on practical and cheap diagnostic: due to lockdown. Services were not available due to In this pandemic, proficiency of treatment should non-continuous production. 2) In Medical Sector where be improved and reduce the cost of diagnostic Emergency health centers and mask creator machine setests. The tests should be available in an afor d- tups and sanitizer machine setups were used. The rest of able range so that every citizen can make himself the issues are described in the table. checked and contribute to eliminating the pan- The future directions include touchless hi-technology demic permanently by staying healthy and fol- using various approaches, a tracking system for sample lowing the instructions given by the government. collection, the progress of medical services through AI, etc. Social life is afe cted very badly by a coronavirus.
People are facing a lowering of economy, lack of hospital facilities, climate variation, dense population, etc. • Insuficient sample size: Quality of sample size for further treatment and new research (for better solutions) is not adequate. For better treatment and to stop the crisis, it should be considered as a necessary step for a valuable solution.
• Less focus on ventilators available for COVID In this paper, we focused on the identification and
diagpatients sufering from breathing: There is nosis of the clinical characteristics of COVID-19. Initially,
no suficient availability of ventilators in the hos- we discuss the data statistics related to COVID-19
alongpitals till now. There is a need for the proper with their clinical characteristics. Then, we diagnosis
arrangement of ventilators in the high range for the symptoms and prognosis models of COVID-19
paCOVID patients by keeping social distancing. With tients. Based on the finding of the research gaps and
a proper supply of oxygen, hygienic conditions challenges of COVID-19, we presented the remedies with
should be maintained. the help of prediction models such as machine learning,
and prediction models. In addition, the datasets model of
• Less focus on the remote treatment of COVID- COVID-19 patients is discussed. Finally, the social impact
19 patients at the time of limited healthcare of pandemic disease, open issues, and future challenges
resources: The COVID-19 patients have to face a are analyzed. In the future, we implement the forecasting
lot of problems due to scarcity of doctors, nursing model for the diagnosis of COVID-19 and suggest the
staf, and other clinical resources. To overcome methodology for the treatment of the COVID patient.
such issues, some remote controlling techniques
should be used which include robots, automatic
machines, ventilators accessed with automatic References
solutions of problems faced by patients such as
the supply of oxygen when it gets lowering down, [1] ”WHO Coronavirus Disease (COVID-19) Dashboard.”
and so on. Such remote-controlled resources can (Online Available: https://covid19.who.int/).
satisfy the needs of patients in the absence of [
125 COVID-19 images related to MERS- deep and https://github.com/ieee8023/ CoV, SARS-CoV, and ARDS , 16 attributes transfer like ID of patient, age of patient, date of learning admission, and location of patient. Applications: COVID diagnosis. 275 CT scans extracted from 760 medRxiv and bioRxiv preprints. Application: Diagnosis of COVID deep CNN with accuracy of around 85% CNN and SVM, COVIDCAPS model https://github.com/UCSD-AI4H/COVID-CT https://github.com/ieee8023/covid-chestxray-dataset https://www.kaggle.com/lachmann12/correcting-underreported-covid-19-case-numbers
19 Case
19 Case
MERS-CoV: Middle East Respiratory Syndrome CoronaVirus, SARS-CoV: Severe Acute Respiratory Syndrome, ARDS: Acute Respiratory Distress Syndrome (ARDS), CNN: Convolutional Neural Network, SVM: Support Vector Machine, COVID-CAPS: Capsule Network Model, NLP: Natural Language Processing.
https://www.kaggle.com/tawsifurrahman/covid19radiography-database
Impact of COVID-19 on Social Life Open Issues Future Directions Troubled economy, agriculture, organization, and so- Privacy Laws for operator protection and public pri- Touchless hi-technology by using AI, IoT, ML, DL, Data cial life vacy Analytics No availability of hospital facilities, alternate solu- Medical database Laws for treatment and proper rec- AI advance for smart healthcare to progress medical tion followed by peoples like Social distancing, self- ommendations services quarantine, self-hygiene Technology emphasis on value making rather than hu- Security Laws for proposal and development of techno- Enhance blockchain for a fast response without netman administration logical solutions work delay and with upgraded security Mostly obstructed measures floating worldwide dispar- Legal Authority and Copyright Laws for distribution Essential more secure and apparent tracking system for ity, climate variation, tense population content & Personal data sample collection, drug delivery, and telemedicine.