**3.1 Computational intelligence methods for detection and diagnosis of (SARS-CoV-2): Use cases examples**

One of the issues confronting all nations, including the USA during this coronavirus pandemic was inadequate testing tools for detecting and diagnosis COVID-19. There is a need for other alternative tools for diagnosis and detection of COVID-19 different from Real-Time Polymerase Chain Reaction (RT-PCR) [34]. Lack of diagnostic tools and efficient tests continue to cause a major problem in controlling the spread of the disease [2]. Research shows lack of RTPCR test units was enormous and it takes 4–6 Hours to acquire results. Thus, results to many infected patients cannot be distinguished from healthy individuals and keep on infecting the other healthy folks. Therefore, to halt the spread of the disease, there is a need for fast diagnosis and detection of COVID-19. Since the results of diagnosis of COVID-19 show symptoms associated with pneumonia symptoms which identity in the image and genetic test.

However, researchers all over the globe working tirelessly to control the spread of the disease using the medical image to explored computational intelligence approaches on digitized images. Several CI techniques play a significant role in the diagnoses of COVID-19 using Chest X-ray (CXR) and Computed Tomography (CT). Recently, several pieces of research have been conducted from the digitized image using neural network (CNN) to detect and diagnose COVID-19 [35, 36]. For instance, a study by [37] using the digitized image of computed tomography (*CT*) to detect COVID-19 based on Convolutional neural network (CNN) approaches. Also, in [38] classification of *CT* images into three classes: healthy, COVID-19 and bacterial pneumonia have been experimented with using a modified version of the ResNet-50 pre-trained network. in [39] Chest X-ray images (*CXR*) were utilized by a CNN to extract the high-level features based on various ImageNet pre-trained models. To detect COVID-19 those features extracted were pass as input into SVM as a machine learning classifier. Furthermore, in [40] based on transfer learning approaches, a proposed model on CNN algorithms called COVID-Net applied to classify the *CXR* images into four classes: COVID-19 viral infection, non-COVID, bacterial infection, and normal.

Moreover, a study by [20], aimed to diagnose COVID-19 using deep learning techniques and a transfer learning system. The system utilized a combination of convolutional neural network (CNN) architecture (one convolutional layer with 16 filters followed by batch normalization, rectified linear unit (ReLU), two fullyconnected layers), and a modified AlexNet [21]. Their proposed model shows an accuracy result of 94.00%. In addition, an investigation by [22] to ascertain the uncertainty and interpretability of deep learning-based techniques for COVID-19 diagnoses in X-ray images, in other to provides the diagnostic confidence for a clinician, a Bayesian Convolutional Neural Networks (BCNN) was utilized to estimate the uncertainty on their proposed model. The results for detection accuracy of 92.86% on X-ray images were obtained by the proposed model.

#### **3.2 The use of computational intelligence approaches for COVID-19 prediction**

Prediction [41] refers to the output of an algorithm after it has been trained on a historical dataset and applied to new data when forecasting the likelihood of a particular outcome. The algorithm generates probabilistic values for an unknown variable for each record in the new data, allowing the model builder to identify

#### *The Power of Computational Intelligence Methods in the Containment of COVID-19 Pandemic… DOI: http://dx.doi.org/10.5772/intechopen.98931*

what that value will most likely be. This is heavily used in computational intelligence methods and we shall see how it has been implemented to mitigate the spread of COVID-19.

COVID-19 data was explored based on a proposed model [42] from Hikvision's temperature screening thermographic and hotspot non-contact infrared device using an acoustic device for collecting and analyzing COVID-19 data embedded with the pervasive computing devices. The principal component analysis (PCA) model was used to pre-process the collected data while Mode and Mean Missing data imputation (MMM-DI) method for removing outliers and filling missing data. To reduce noise and prevent false alarms, an Artificial Intelligence detector is also embedded with the sensor device. Using coding in MATLAB, the Susceptible, Infected, and Recovered (SIR) epidemic model is implemented to classify the cases as suspected, infected, and recovered generated for classification of the demographic data. With the previous history stored in the hidden layer, the data is then fed into a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) model to forecast the coronavirus disease cases. The proposed model helps in further treatment by exploring pervasive computing technologies in coronavirus disease prediction and detection. The issue of trust and privacy has to be handled to find a favorable rectification. Redundancy and noise challenges can be overcome by a new algorithm to assist in the area of data conversion.

For the diagnosis and prediction of coronavirus disease, prediction models such as autoregressive integrated moving average (ARIMA), LSTM, and prophet algorithm (PA) were utilized over the next 7 days to predict the number of coronavirus disease confirmations, recoveries, and death in a Computational Intelligence based technique that was proposed [43]. The algorithm with the best performance was PA. It gave a prediction of the number of coronavirus disease confirmations, recoveries, and deaths in Australia and gave accuracies of 99.94%, 90.29%, and 94.18%, consecutively.

In Jordan, the PA technique obtained the number of coronavirus disease confirmations, recoveries, and deaths with accuracy in prediction of 99.08%, 79.39%, and 86.82%, consecutively. More advanced prediction models are expected in future work. Using X-ray images of the chest, a diagnosis model which implemented VGG16 was proposed to find coronavirus disease. Being capable of obtaining an F-measure of 99% the technique allowed quick and reliable coronavirus disease detection, using a dataset that is augmented. The researchers believe that future studies will aid in diagnosing coronavirus disease using the VGG-XX versions in chest CT scans and compare their performances using larger datasets. The analysis of the spread of coronavirus disease and its related statistical data based on worldwide regional distributions was a further contribution of their study. Using their Artificial Intelligence-based analysis; two major conclusions were arrived at: (1) similar characteristics are observed in the most highly infected areas (2) in coastal environments, the spread of COVID-19 is tremendously higher than in other noncoastal environments. Henceforth, extra attention and care ought to be rendered to coastal cities. Effects of terrain, humidity, and temperature on the coronavirus disease and its spread in countries and cities would be good to be investigated in upcoming work.

## **3.3 Contact tracing based applications for COVID-19**

The act of identifying all people that a coronavirus disease patient has come in contact with in the last fortnight is known as contact tracing. The infection is known to spread to people through coughing, sneezing, saliva, droplets, or discharge from the nose through contact transmission. Various applications, methodology, and

tools have been considered for use to curtail the spread of the virus and in this section, we shall discuss some computational intelligence approaches.

One of the major strategies for the containment of COVID-19 is contact tracing. In ordinary contact tracing, medical doctors interview the infected patients to trace and find others who may be contaminated through contact with the patient. The major problem of the above methods was the difficulty for the individual to recalled all his contact. In addition, other strategies may require experience clinicians and other resources. However, recent technology innovation enhanced contact tracing methods by reducing human intervention in the process, using a smart methodology known as digital proximity (DP) contact tracing. The DP method uses network technologies to recognize and find people who could be conceivably contaminated through contact.

With the boundless accessibility of computing networks and mobile applications - and their related technologies including cell phones, smartwatches, and others - the majority of the innovation-based contact tracing frameworks are based on mobile platforms [44, 45]. However, computational intelligence is right now used through the whole life cycle of COVID-19 starting from identification to mitigation [46]. A virtual computational intelligence Agent is an option in contrast to a medical doctor on account of traditional contact tracing. In digital contact tracing (DCT) frameworks, Bluetooth innovation is generally utilized as a vicinity identifier for COVID cases. Notwithstanding, the presentation of Bluetooth-based contact tracing applications might be influenced by changing sign power, which can be shown by various cell phones, versatile positions, body positions, and actual boundaries [47].

COVI was a Computational intelligence-based contact tracing application created in Canada that uses probabilistic risk levels to profile a person's contamination hazard level [48]. COVI utilizes the advantages of CI algorithms to improve and automate the mix of pseudonymized client information in surveying the danger levels. A deduced variant of an epidemiological model-based reproduced dataset is utilized to pre-train the CI models. Upon assortment of genuine data through an application, the test system boundaries are tuned to coordinate with genuine data. The effect of CI in the COVI application is seen by utilizing the CI predictor inside the test system to impact the conduct of the specialist in suggesting the danger levels. The contact tracing application can be utilized to foresee the lockdown territory dependent on places visited by a contaminated patient. In [45] the researchers proposed a K-Means clustering algorithm with DASV seeding to foresee the lockdown region. The proposed technique has been tried in Denver, USA, and effectively distinguished the territory to be locked down as clients strolling around there approach each other regularly.

#### **3.4 The used of CI methods in case of COVID-19 treatment**

Treatment in the context of COVID-19 is the approach that could be harnessed to help a patient get back on his feet. This could be by the use of medications or other methods. Computational intelligence techniques help with decision-making tools amongst other things to bring this to pass.

The introduced [49] platforms and conceptual structures in the research area of artificial intelligence-based methods suitable for fighting coronavirus disease were addressed. Extreme Learning Machine (ELM), Generative Adversarial Network (GAN), LSTM, and RNN are varying techniques that have been developed, incorporating coronavirus disease diagnostic systems. The major issues with coronavirus disease included geographical problems, radiology, recognizing, and high-risk people according to their studies. A mechanism was revealed that helped

#### *The Power of Computational Intelligence Methods in the Containment of COVID-19 Pandemic… DOI: http://dx.doi.org/10.5772/intechopen.98931*

in selecting the right models for predicting and estimating desired parameters using several nonclinical and clinical datasets. Considering these platforms help artificial intelligence specialists to analyze large datasets and assist physicians train machines, set algorithms, or optimize the data being analyzed for fighting COVID-19 with greater accuracy and speed. They are desirable because of their potential for creating a workspace while physicians could work side by side with artificial intelligence specialists as discussed. However, while artificial intelligence speeds up the methods to defeat coronavirus disease, real experiments ought to occur because a comprehensive knowledge of the advantages and limitations of computational intelligencebased methods for coronavirus disease is yet to be achieved, and new approaches need to be in place for challenges of this level of complexity. Building an arsenal of methods, platforms, approaches, and tools that converge to solve the sought goals and help in saving more lives is going to greatly assist in the combat against the coronavirus disease and its eventual annihilation.

Fingerprint and differentially expressed genes (DEGs), two types of drug data were clustered by a multimodal restricted Boltzmann machine (mm-RBM) according to a study [50]. Showing the chemical structures, the first type of data is binary data. From drug-induced perturbations in cell lines, the second one was extracted. First, the intrinsic correlations within each input modality were encoded using the modality-specific hidden variables in the proposed multimodal RBM model. By merging unknown variables, the intra-modality features were fused next and a typical representation of cross-platform features was formed. Data integration yields significant clusters based on the indications of the proposed approach. Henceforth, to discover medications that may prove useful in treating COVID-19, the clusters consisting of drugs used for curing coronavirus disease were chosen. Having antiviral properties, the introduced drugs are similar to sophisticated drugs that have been used to control coronavirus disease. Although the outcomes seem to yield a satisfactory explanation and are significant, further clinical research such as in vivo or in vitro tests needs to be carried out.

However, COVID-19 treatment is categorized into two drug discovery and vaccine development. As we all know, without drug discovery and vaccination there will no be any treatment of COVID-19 patients which indicated its high importance and urgent need. Computational intelligence methods have been utilized in search of new chemical combinations that can lead to effective medicine, provided integrated characteristics predictions, behavior prediction, reaction prediction, and ligand-protein interactions. Proteomics and genomics investigation have been suggested on the development of mDiverse drugs and vaccines for SARS-CoV-2. CI approaches in the development of new drugs and vaccines contributed immensely to the battle against COVID-19. Integrating CI methods in the pharmaceutical arena has proven both cost-effective and less time-consuming.

Many pharmaceutical companies embraced the use of computational intelligence techniques such as artificial neural networks, Support Vector Machines (SVM), deep learning, and many others to develop various drugs and vaccines [36]. A review of recently developed algorithms in [36] to design drug development pipelines consisting of drug discovery, drug testing, and drug re-purposing. Generative Adversarial Networks (GAN) were utilized to identify DNA sequences associated with specific functions, and proteins of interest produced with lower costs using Bayesian Optimization (BO) during drug discovery. To determine the best treatment, Bayesian-based Multi-Armed Bandit (MAB) algorithms which is a sequential decision-making algorithm are utilized in drug testing to test several drug candidates. Text mining methods and graph-based recommender systems were used in repurposing to identify correlations and predict drug-disease interactions.

Several pharmaceutical companies have employed ML-based algorithms such as artificial neural networks, Support Vector Machines (SVM), deep learning, and many others to develop various drugs and vaccines [51]. The authors in [51] provide a review of recently developed algorithms to design automated drug development pipelines consisting of drug discovery, drug testing, and drug re-purposing. In drug discovery, the deep learning algorithm Generative Adversarial Networks (GAN) is used to identify DNA sequences associated with specific functions, and Bayesian Optimization (BO) is used to produce proteins of interest with lower costs. In drug testing, sequential decision-making algorithms such as the Bayesian-based Multi-Armed Bandit (MAB) algorithms are used to test several drug candidates and determine the best treatments. In drug re-purposing, text mining methods and graph-based recommender systems are used to identify correlations and predict drug-disease interactions. The authors compiled a list of relevant data sets for drug development pipeline studies.

In an attempt to identify probable vaccine candidates and constructing an epitope-based vaccine against COVID-19 authors in [52] developed a computational intelligence system that incorporated reverse vaccinology, bioinformatics, immunoinformatic and deep learning techniques. Also, in a study by [53] to predict and evaluate potential vaccine candidates for COVID-19, the authors utilized Vaxign Reserve Vaccinology (VRV) tool and Vaxign-ML, a computational intelligencebased prediction and analysis framework. The results in their research showed the second-highest protective antigenicity as a non-structural protein (nsp3), in addition to the commonly used S protein.

## **3.5 COVID-19 recovery methods based on the use of CI techniques**

COVID-19 recovery could be evaluated as the phase when we can say a patient has gotten back to his feet after being infected. Computational intelligence offers models that could assist reach this phase.

Data mining models were developed for forecasting coronavirus disease infected patients' recovery using the epidemiological dataset of coronavirus disease patients of South Korea in a study by [54]. Using a python programming language, support vector machine (SVM), logistic regression (LR), k-nearest neighbor (K-NN), decision tree (DT), naïve Bayes (NB), random forest (RF) algorithms were applied directly to the dataset. The most efficient was found to be the model developed by DT with the highest percentage of accuracy of 99.85%, followed by RF with 99.60% accuracy, then SVM with 98.85% accuracy, then K-NN with 98.06% accuracy, then NB with 97.52% accuracy and LR with 97.49% accuracy. The developed models would be very helpful in healthcare for the combat against COVID-19.

As people's way of life has changed amongst many other things due to the coronavirus disease pandemic, many losses have also been incurred and means of sustenance of a lot of people [55]. It greatly affected economic and commercial activities due to the suspension of both at certain intervals of time to control the spread of COVID-19.

Through technology management, accelerated COVID-19 recovery is emphasized as an approach to utilize with the advancements in healthcare and expansion in the access to electronic data. The area of healthcare can apply AI to address problems in the area, using substantial computation power, especially during an ongoing pandemic. Many of these machine learning systems ultimately present the most substantial transformative role in healthcare governance though many of them remain experimental [55]. Machine learning modeling has evaluated multiple scenarios to focus on the COVID-19 recovery index with the proposed research. To identify specific patterns and help the masses overcome the

#### *The Power of Computational Intelligence Methods in the Containment of COVID-19 Pandemic… DOI: http://dx.doi.org/10.5772/intechopen.98931*

impending outcome of coronavirus disease, the research presents a strong case where machine learning models can be used.

The generalization of developed machine learning models is possible as the study [55] feeds on near-time data and comprehensive academic underpinning. Developed and developing countries can use insights from this work as they apply to national and global levels for developing strategies. Machine learning should consider the limitations on algorithm development and understanding its appropriateness to apply like other revolutionary technologies.

Machine learning has the potential to play a key role in the advancement of healthcare and societal health enhancement as researchers are continuously attracted by predictive modeling techniques. The presented work [55] could offer counsel to make policy recommendations to help authorities develop well-informed health policies and accelerate the COVID-19 recovery.

## **3.6 Computational intelligence (CI) based quest for COVID-19 drug discovery**

Recently, Computational intelligence approaches have revolutionized many fields in medical sciences and beyond. It has generally changed our everyday lives, from speech and face recognition [56]. Two of the most affected areas influenced by CI techniques are drug and vaccine discovery [57], in which CI methods have offered compound property prediction [58], activity prediction [59], response expectation [60], and ligand-protein cooperation. Graph Convolutional Neural Network (GCNN) has been the front runner on the prediction of drug discovery applications [61, 62]. Several studies show, drug property prediction can handle by (GCNN) and extract features through encoding the adjacency information within the features [63–65]. Protein interface assessment [66], reactivity forecast [67], and drug–target connections etc. [68, 69].

Noteworthy, CI methods have additionally improved in the field of vaccine design recently. VaxiJen was the first implementation of CI techniques in RV approaches and has shown promising outcomes for antigen forecast [70, 71]. Also, drug candidate created during the process of drug discovery needs to be safe for human utilization. This means confirmation that the drug is non-poisonous is required during the drug side effect observation. To achieve the above, it requires the creation of a database that can be utilized to facilitate modeling toxicology. Several investigations, based on CI techniques were implemented to identify the cardiotoxicity of a candidate drug, hydroxychloroquine, using ECG data from smartwatches [72].

In the case of COVID-19 drug discovery, several studies used CI approaches for both repurposed drug candidates and new chemical entities. The former aimed to exploit and predict interconnected biological pathways or the offtarget biology of existing medicines that are proven safe and can thus be readily tested in new clinical trials. However, studies by Gordon et al. experimentally identifying 66 human proteins linked with 26 SARS-CoV-2 proteins, paved the way for the repurposing of candidate drugs [73]. Furthermore, for analyzing the virus-host interactome network-based model simulation has been the main computational approach used over wet-lab approaches [74]. Li et al. analyze the genome sequence of three main viral family members of the coronavirus and then relating them to the human disease-based pathways lead to the discovery of 30 drugs for repurposing [75]. Using an alternative approach by Zhou et al. offered a combination of network-based methodologies for repurposed drug combination [76]. Research has shown that the experimental evaluation of all drug and vaccine candidates was extremely challenging. However, researchers believed that leveraging computational intelligence approaches will speed up the discovery effort,

and capable of filtering generating therapy. Utilizing artificial neural networks and supervised learning methods has proven to be a vital game-changer when used for virtual filtering and *de novo* design. Large-scale training datasets and relevant bio targets are required in other to achieve desired performance using computational approaches.

## **3.7 CI methods used in COVID-19 19 surveillance**

Surveillance is the art of monitoring people or things via various techniques like directly looking at them or using tools such as binoculars, sensing them via sensors, and generally keeping track of them often in relationship to time. Since the outbreak of COVID-19, it became imperative to monitor those that were infected especially after putting them in hospitals and isolation centers, and closely watch them as treatment was rendered to them and as an effective remedy was researched by the scientific and medical community. Those that were infected that came close to other individuals contributed to initiating the method referred to as contact tracing which is a form of monitoring to trace all those who are likely to be infected by the virus. When they are found via contact tracing, they are often put under isolation for several days so they do not infect others if they contact the virus. Surveillance was also carried out on the general public by ensuring they maintain social distancing, wear a mask, and use sanitizers by enforcers. In this section, we are primarily concerned about computational intelligence techniques that were used and can be used with surveillance to mitigate the spread of COVID-19.

To fight and overcome coronavirus disease like pandemics, a beyond 5G (B5G) enabled smart health care framework was proposed [77]. A cloud layer, an edge layer, and a stakeholder layer are all contained in the framework. Into the system was the integration of a mass surveillance system in terms of mask-wearing, social distancing, and body temperature detection. Analysis at the edge utilizing the latest generation of high-power edge computers was done on human vital signs and hospital test data. This diagnostic method for coronavirus disease could be extended to any infectious disease. Protecting sensitive personal data at the edge to protect anonymity, verifying non-coronavirus disease patients, and reducing overcrowding in health centers will all be helpful. Other protease sequence analyses and deep learning models will be tested in the framework in the upcoming work. A timeseries analysis model and a prediction model could be embedded in the framework also in future work. For low latency and better security, pervasive edge computing could also be added.

To assist in reducing the coronavirus disease outbreak, an embedded surveillance system was presented [78] which detects the elderly ones who are more affected by COVID-19 in the recent pandemic. To determine the age of an individual an age estimation is used. To enhance the results of pre-trained deep networks, an enhancement age estimation method is used by utilizing face alignment. To refer to the presence of the elderly in an environment, a notification is sent to mobile or any other device systems using the Internet of Things. Using a public database, the proposed system was evaluated and the results obtained show that the system was satisfactory in its performance. Two types of comparison were additionally used to compare the accuracy of the proposed system. Pre-trained deep networks and face alignment were implemented in the first one for the enhancement of the deep learning model. The combination of face alignment and pre-trained deep networks proved age estimation performance from the obtained results. Implementation using two kinds of hardware and comparison between them was further done in the proposed system.

*The Power of Computational Intelligence Methods in the Containment of COVID-19 Pandemic… DOI: http://dx.doi.org/10.5772/intechopen.98931*
