The primary goal of this study was to determine the full scope of AI's impact on COVID-19 containment. Scopus was chosen as the first source because it contains papers from prestigious journals and conference proceedings, providing a diverse sample of the importance of AI in COVID-19 containment. Scopus search returned 332 articles for first search “artificial intelligence for medical imaging to combat coronavirus disease” keywords used are, (“ai” OR “artificial AND intelligence” OR “artificial AND intelligent” OR “machine AND learning” OR “deep AND learning”) AND (“medical” AND ima^*) AND (“COVID-19” OR “novel AND coronavirus” OR “COVID-19 AND pandemic” OR “coronavirus” OR “SARS-CoV-2”).

Scopus search for second search “AI in managing social distancing” in the article title, abstract, or keywords, (“ai” OR “artificial AND intelligence” OR “artificial AND intelligent” OR “machine AND learning” OR “deep AND learning”) AND (social AND dist^*) AND (“COVID-19” OR “novel AND coronavirus” OR “COVID-19 AND pandemic” OR “coronavirus” OR “SARS-CoV-2”), returned 159 articles. Scopus returned in total 491 articles; those were then manually reviewed for relevance to our review concept. All publications that met the following requirements were deemed relevant.

COVID-19 is a transferable disease, as we all know. Minimum human interaction is required to stop this viral disease. COVID-19 data is collected, a monitoring system is developed, and COVID-19 data is visualized using AI-based technologies.

BlueDot is a powerful analytics tool developed in Canada wherein ML and Natural Language Processing (NLP) techniques are employed (Panda, 2020). It was able to detect the onset of COVID-19 and send out watchful warnings. It was also used to send out preventive notices to those cities wherein individuals arrived after January 2020 from Wuhan. The confluence of the user's traces with the traces of infected people was checked using the Global Positioning System (GPS) on smartphones. To keep the data safe, they utilized cryptographic procedures (Arif et al., 2021). After encountering sick people, the app creates early warning indications to assess the danger of infection spreading. Instagram has created “RT. live.”, a COVID-19 tracker which displays state-wise data on virus transmission in the United States. The virus infection's multiplication rate was calculated using data analysis methods (Comba, 2020).

AI was used to track and forecast COVID-19 infection using data visualization techniques/dashboards. They gave a complete picture of vulnerable patients from every part of the world. Local and global dashboards are the two primary categories of data visualization systems. For a wider population of COVID-19, Johns Hopkins University's Center for System Science and Engineering (JHU-CSSE) created the COVID-19 dashboard. The New York Times provided a virus infection dashboard for several parts of the world as well (Anastassopoulou et al., 2020). The AI tracker of Bing was trained to display diagnosed, cured, and deceased cases from across the world. This filter was used on a specific country for displaying infected people from a certain province. Big data was required for further analyzing medication discovery, risk assessment, infection spread, therapy, and cure of COVID-19 infected patients using Artificial Intelligence approaches. Text, social media, biomedical, speech, and case studies are some of the broad categories for the datasets. The COVID-19 datasets are broadly classified as shown in Figure 7 (Anastassopoulou et al., 2020).

Text datasets included risk variables, non-pharmaceutical interventions, virus infection propagation, incubation period, and environmental stability. Most of the tweets about COVID-19 contained myths and incorrectly classified information about coronaviruses (Abd-Alrazaq et al., 2020; Minaee et al., 2022). To address this issue, various fake news and rumor detection techniques have been proposed. Due to the open-source nature of the COVID-19 chest X-ray dataset, medical images were used to train deep learning or machine learning models (Hughes et al., 2020). According to another study, COVID-19 can also be distinguished by its vocal sounds. The University of Cambridge scholars requested that participants' voice samples be collected to create an open-source cough sample repository. The COVID-19 databases were issued by the World Health Organization and the National Centers for Disease Control and Prevention to track the progression of infection and analyze the influence of preventive measures. COVID-19 contaminated locations can be seen on Johns Hopkins and Git-hub (Andrejeva et al., 2020).

The diagnosis of a COVID-19 patient is challenging. Each patient is tested individually, which takes a substantial amount of time. For that reason, the medical report must be generated quickly and at a reduced cost. Therefore, COVID-19 patients were also screened using AI-based approaches. There were three main AI-based strategies for screening patients namely face recognition, wearable devices, and virtual healthcare assistant as shown in Figure 8. Applying face scans to determine symptoms, was a potential research field in machine learning to aid in diagnosing COVID-19. Wearable technology, like smart watches, could be used for checking patterns in a patient's resting heart rate. Using chatbots driven by ML to assess patients depending on self-reported complaints (Robitzski, 2020).

Fever in patients was detected using drones/robots equipped with thermal scanners from a safe distance. Baidu, a Chinese company, created infrared cameras to scan crowds for temperature readings (Robitzski, 2020). In a single minute, they can scan hundreds of people. Apple created an AI-powered watch that can detect a patient's temperature along with the heart rate to diagnose their symptoms (Wadhera et al., 2020). Oura Ring, which is designed by OURA enterprise solutions, helps in tracking activity that analyze the patient's onset patterns, progress, and recovery (Nepogodiev et al., 2020).

Similarly, AI-powered Stallion developed a chatbot using natural language processing (NLP) skills as a virtual healthcare agent (Agrebi, 2020). It encourages people to take precautions, monitors their symptoms, and makes recommendations for home quarantine or hospitalization. COVID-19 information was also provided via a virtual assistant developed by Google Cloud. This virtual assistant reacts swiftly to customer questions and provides the best information possible (Sheerman et al., 2020). AskDoc, an AI-enabled doctor video bot, was also developed to offer voice and text replies to COVID-19 questions. Likewise, Facebook created a WhatsApp bot to help people stay up to date on the COVID-19 pandemic (Ardakani et al., 2020).

To detect COVID-19 from X-ray and CT images many AI models like, using imaging algorithms have been proposed. X-ray machineries are used to scan the infected body nearly identical to how pneumonia and malignancies are treated. Researchers have put forward a deep convolutional neural network design which was termed as COVID-net where chest X-ray (CXR) images have helped in diagnosing COVID-19. It was developed earlier, as studies which observed those cases, showed irregularities in chest radiography images that were symbolic to those individuals who got infected with COVID-19 (Mahase, 2020). Emara et al. (2021) provided two solutions for an automated COVID-19 diagnostic system. The first was based on transfer learning models that have been pre-trained with chest X-ray scans as input. To achieve this, they employed pre-trained models such as AlexNet, GoogleNet, Inceptionv3, InceptionresNetv2, SqueezeNet, DenseNet201, ResNet18, ResNet50, ResNet101, VGG16, and VGG19. The second method focuses on creating a CNN termed CONV-COVID-net from scratch to identify COVID-19. ResNet models namely ResNet18, ResNet50, and ResNet101 achieved the maximum classification accuracy on the two datasets. According to the results, they achieved 97.67, 98.81, and 100% accuracy using the first dataset, respectively. On the other hand, the reported accuracies for the second dataset were 99, 99.12, and 99.29%, respectively (Chen et al., 2020; Kanne, 2020; Emara et al., 2021).

Using a combination of deep convolutional neural network (CNN) and discrete wavelet transform (DWT) attributes, Mostafiz et al. (2020) suggested an innovative way to identify COVID-19 from a chest X-ray picture. The detection process was completed using a random forest-based bagging strategy. A lengthy experiment was conducted, and the findings depicted that their methodology outperformed existing approaches by more than 98.5% (Mostafiz et al., 2020). The Random Forest (RF) classifier was reported by Sunil et al. for COVID-19 detection. They employed resampling mechanisms to correct the data's imbalance problem. The results of the analysis revealed that accuracy, geometric mean, and f-measure are 0.94, 0.93, and 0.92, respectively. The proposed method offered comparable sensitivity and greater accuracy to deep and transfer learning models, which are computationally costly and less accessible (Kumar and Ratnoo, 2022).

Algorithms applied to the demographic data of multiple cases helped in identifying and tracing COVID-19 patients. Contact tracing ensured identifying individuals with symptoms of the disease and monitored the cluster of cases from spreading the virus further. For this, various mobile applications based on artificial intelligence were developed for tracing the patients as virtual contact tracing is quicker than non-digital methods of tracing. These applications collect the personal data of user and analyze that through AI algorithms to trace if the user is in contact with any COVID-19 patient. Hence, the real-time forecasting model was introduced. It helped in generating predictions of positive cases and deaths in the future by analyzing personal data of the users. The model also checked whether the user was likely to spread the disease in particular regions which in turn helped in taking safety measures in those areas (Beck et al., 2020).

For example, the British government tested an app that used Bluetooth protocol to locate other smartphone owners that are in close vicinity to one another. As a result, a person who was not affected but was in close proximity to someone who had COVID-19 symptoms could be notified. In addition to contact tracing, Nieva, Melek, and a group of Waterloo Engineering students are developing software algorithms based on smartphone's Bluetooth data for grouping persons within a community. They're also working on wearable Bluetooth devices to increase the app's accuracy even more. Through the assessment of specific vital signs, the AI algorithms will increase contact-tracing accuracy and real-time observation of the recovery progress (Richardson et al., 2020).

Table 2 shows several infected countries that were efficient in developing contact tracing applications based on machine learning and artificial intelligence. They created a digital contact tracing method with a mobile application, integrating various technologies like Bluetooth, GPS, contact information, network-based API, system physical address and mobile tracking data. According to studies, over 36 countries have effectively implemented digital contact tracing using centralized, decentralized, or a mixture of both systems to reduce work and enhance the efficacy of traditional healthcare diagnostics procedures (Tang et al., 2021). All these apps are intended to capture personal data, which will be analyzed by machine learning and artificial intelligence (AI) algorithms to track down an individual person who is susceptible to the virus because their latest contact sequence (Chakraborty and Ghosh, 2020; Ubaid et al., 2021).

It is evident that AI aids in increasing the speed of the drug testing process which involved a great deal of time initially. Since it boosts diagnosis prediction and drug development, as a result, it helped in producing the vaccine much faster. A molecule transformer-drug target communication design was introduced by scientists that can help in finding an anti-viral drug to treat coronavirus (Laudanski et al., 2020). The model compared with an application named Autodok Vina and applied algorithms on some existing drugs approved by the FDA. It was found that antiretroviral drug Antazanavir and Remdesivir were beneficial for COVID-19 patients. There is a low-risk factor to use existing drugs and moreover it is a less expensive process comparatively. Consequently, AI was noted as one of the most significant technologies for enhancing the drug development process and finding accurate results (Yigitcanlar et al., 2020).

Several research laboratories worked continuously to create COVID-19 vaccines and medicines during this epidemic. They utilized artificial intelligence to find new vaccinations or repurpose old ones (Kühl et al., 2020). The most potential virus-host protein-protein interactions for HIV as well as for H1N1 have been effectively predicted using ML models trained on datasets of protein, considerably decreasing the work necessary to create a comprehensive map of the virus-host interactome. AI may be used to screen existing medication compounds and determine whether or not they are effective in combatting coronavirus. South Korea and USA employed a system based on AI for discovering the “Atazanavir” medicine that was repurposed to treat COVID-19 (Liu et al., 2020). Benevolent AI researchers discovered the medications “Baricitinib” and “Myelofibrosis” for curing COVID-19 (Hossain et al., 2020).

As we cast our gaze beyond the ongoing COVID-19 pandemic, it becomes increasingly critical to explore how Artificial Intelligence (AI) can play a proactive role in preventing and mitigating future pandemics, including emerging threats like Dengue, Salmonella, and Monkeypox. AI offers significant potential to strengthen our preparedness and response strategies. Trying to predict if a variant of influenza will spread from person to person via zoonotic transmission can assist doctors and other experts to plan of time and prepare for pandemics that could occur. Influenza A, for instance, is predominantly found in birds, but it has the potential to spread to humans (Louten, 2016).

The researchers detected possible zoonotic influenza strain with great precision using machine learning. Many AI/ML-based models have been successful in determining the future of pandemics, such as infection or transmission rate, according to the literature. As a global ML-based prediction method, Ramesh et al. proposed the probability of a COVID-19 recurrence. Exponential Smoothing (ES), Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine (SVM), and four typical prospective predictions, such as Linear Regression, were all identified as risk factors for COVID-19 in this study. The study estimated the number of newly infected COVID-19 patients, death rates, and COVID-19 recovered cases in the next 10 days (Mojjada et al., 2020). Similarly, other researchers have also proposed different AI/ML based modeling techniques to forecast a pandemic such as COVID-19. A comparison of existing works on COVID-19 prediction methodologies is depicted in Table 3.

Although more research is needed to develop models for predicting direct transmission, the knowledge of which strains of influenza are most probable to leap is a vital first step in planning for the next pandemic. In the future, AI systems could be developed for pandemic prevention and response, with its capabilities spanning early disease detection, predictive modeling, rapid vaccine development, efficient resource allocation, contact tracing, targeted public health communication, and genomic analysis. More patient data exchange, as well as ML approaches that would enable models to be built whilst limited data is accessible, would help. Few-shot learning, in which AI can understand patterns from a limited number of findings, along with transfer learning, in which an AI that has successfully received training to achieve one task could be swiftly changed to perform some another similar task, are examples of potential gains. Such proactive approaches align well with the evolving landscape of infectious disease management and underscores AI's pivotal role in safeguarding global public health in the future. However, there is still more work to be accomplished. In addition, we can also explore the use of large language models, namely ChatGPT or Google Bard in case of any future pandemic outburst (Sohail et al., 2023a).

AI language models like ChatGPT hold significant potential for future pandemics. They can play a vital role in disseminating accurate information to the public, answering queries about prevention, symptoms, testing, and vaccination, thus aiding in education. Additionally, AI can be leveraged to power crisis helplines and mental health support, offering much-needed emotional assistance and guidance (Sohail et al., 2023a,b). ChatGPT's language translation capabilities can ensure that critical information reaches non-English speakers and diverse populations. Monitoring public sentiment through AI can guide authorities in tailoring responses effectively. Furthermore, AI can assist in detecting and correcting misinformation, aiding in research and drug discovery by sifting through vast amounts of scientific literature, supporting remote patient monitoring, engaging citizens in policy discussions, designing awareness campaigns, and enhancing healthcare worker training.