Internet of Things with Deep Learning Techniques for Pandemic Detection: A Comprehensive Review of Current Trends and Open Issues

Abstract

Technological advancements for diverse aspects of life have been made possible by the swift development and application of Internet of Things (IoT) based technologies. IoT technologies are primarily intended to streamline various processes, guarantee system (technology or process) efficiency, and ultimately enhance the quality of life. An effective method for pandemic detection is the combination of deep learning (DL) techniques with the IoT. IoT proved beneficial in many healthcare domains, especially during the last worldwide health crisis: the COVID-19 pandemic. Using studies published between 2019 and 2024, this review seeks to examine the various ways that IoT-DL models contribute to pandemic detection. We obtained the titles, keywords, and abstracts of the chosen papers by using the Scopus and Web of Science (WoS) databases. This study offers a comprehensive review of the literature and unresolved problems in applying IoT and DL to pandemic detection in 19 papers that were eligible to be read from start to finish out of 2878 papers that were initially accessed. To provide practitioners, policymakers, and researchers with useful information, we examine a range of previous study goals, approaches used, and the contributions made in those studies. Furthermore, by considering the numerous contributions of IoT technologies and DL as they help in pandemic preparedness and control, we provide a structured overview of the current scientific trends and open issues in this field. This review provides a thorough overview of the state-of-the-art routing approaches currently in use, as well as their limits and potential future developments, making it an invaluable resource for DL researchers and practitioners and it is a useful tool for multidisciplinary research.

1. Introduction

The COVID-19 pandemic has made life more difficult for people everywhere [1,2]. The most severe effect, a rise in deaths and injuries (globally), has demonstrated the necessity of social and commercial limitations [3,4]. Recent developments in Internet of Things (IoT) and Internet of Medical Things (IoMT) technology have resulted in the growth of numerous IoT-based applications for pandemic preparedness and control. It has recently been demonstrated that distinct pandemic signal data can be gathered from IoT devices and processed using deep learning (DL) or machine learning (ML) approaches to identify various diseases of this system, including COVID-19, which is thought to be a pandemic that is still going strong worldwide [5,6,7]. Owing to its many advantages, IoT technology has been used in a variety of medical applications, making the healthcare industry one of the IoT’s active applications. Examples of applications where the IoT is used are in remote health surveillance, blood pressure, oxygen saturation, geriatric care, chronic disease monitoring, wheelchair management and control, and exercise regimens. IoT gadgets utilized recently in the healthcare industry are shown in Figure 1. Furthermore, several healthcare diagnostic and imaging devices can be classified as smart devices and are essential to the IoT [8]. It is anticipated that IoT integration in the healthcare industry reduces healthcare service costs while also improving quality of life and user experience. Healthcare firms view IoT technology as having the potential to decrease device downtime through remote service provisioning. In response, the IoT can accurately determine when different devices should refill their supplies in order to carry out continuous and adaptable procedures. Furthermore, the IoT facilitates efficient and feasible resource scheduling to ensure that patients receive the best care and have enough rest [9].

Figure 1. Application of IoTs in healthcare [5].

Considering the importance of the healthcare industry, integrating IoT and the cloud in the healthcare space has the potential to have a wide range of applications in daily life and society [10]. By utilizing cloud and IoT technologies, these applications can grow. For patients with chronic diseases, for example, who require long-term monitoring, there are various applications [8,11] where continuous surveillance can be vital. Cloud computing can provide security while offering large amounts of storage and powerful processing [12]. Networks connected to the IoT are helpful for gathering essential data for analysis and forecasting. As a result, they have been widely incorporated into several everyday tasks to create a smart world in several crucial applications, including smart cities (SC), smart industries, smart agriculture, smart environment, and smart health [13,14]. Because these applications create large amounts of data and require real-time data transfer, they are power-hungry, bandwidth-intensive, and sensitive to data transmission delays [15].

Every technological advancement and invention must be used in the fight against pandemic outbreaks. Healthcare, like many other fields, depends on new technologies like DL and IoT for help. One of the primary goals of AI is the exploration of datasets connected to diseases, data preparation, and the prevention and control of infectious diseases. To comprehend the COVID-19 pandemic, address it, and find a COVID-19 vaccine, IoT and DL are essential. With the use of DL techniques, computers can now simulate vast amounts of data and use their intelligence to forecast the pattern and rate of disease spread. For properly screening, analyzing, forecasting, and tracking both present and possible future patients, result-oriented technology is used [16]. AI tracks data from individuals who have died, improved, and have coronary heart disease. One of the technologies used to detect the coronavirus’s spread and identify its effective parameters is DL. The DL approach is effective in identifying people who are at high risk and in anticipating the steps that would be required to address any potential infections and lessen the disease’s impact. By using historical data, DL-based techniques can calculate the risk of patient death. This method is an additional medical tool that uses data and evidence to enhance patient planning, treatment, and reduction. However, new technology also lowers the cost of diagnosis and treatment and enhances decision making. DL tools help radiologists’ ability to see patterns in pictures and improve their capacity to identify potential diseases and diagnose them early on.

In 2006, Hinton et. al. [17] proposed DL, a subset of ML and artificial intelligence (AI) [17]. DL is derived from neural networks and, in contrast to ML, is better at learning from data [18]. For big data sets, DL performs more efficiently than ML. In particular, researchers and practitioners can now expand previously labor-intensive feature extraction processes thanks to DL [19,20]. Moreover, numerous layers are used by DL to represent data abstraction and construct computational models. These benefits have helped DL become much more popular and have made it possible for DL to be successfully used in a variety of industries, including computer vision, natural language processing (NLP), and the development of computational models in the healthcare industry, including pandemic investigation. In the same vein, the remarkable learning capabilities of DL have led to a notable increase in scientific research in recent years [21,22]. The versatility of DL makes it useful in a wide range of applications, including image identification, object recognition, and classification issues. This section examines how IoT and DL might be used in medical research to employ tactics to counteract the COVID-19 pandemic. Significant advancements in healthcare over the past few years have created new chances to enhance people’s lives. Through the analysis of pertinent datasets and medical imagery, DL has been applied in studies to diagnose a variety of disorders. In the past, medical personnel had to manually go through reports, examine photos, and analyze various pandemic datasets. This took a lot of time. Due to their improved outcomes and growing significance, ML and DL have sped up this process. They are transforming healthcare, particularly in times of epidemic.

The use of novel sensors that generate vast volumes of data is one of the growing trends that is further propelling the application of DL models and solving societal problems including pandemic. New opportunities have arisen as a result of society’s growing technological advancements, and these could make life easier for us all while also offering more effective services or manufacturing methods [23]. IoT refers to a clear system of digital, mechanical, and networked computing technologies that can transfer data over a designated network without requiring any kind of human involvement. Every one of the previously mentioned devices/sensors is identified by a code or unique number. Nowadays, the IoT is a well-proven and established technology that connects several strategies, real-time analytics, ML, DL, and sensory products.

This comprehensive review seeks to identify the challenges facing existing IoT-DL-based techniques in achieving optimal COVID-19 detection. Therefore, the goal of this study was achieved to provide key contributions in Section 1.1 and it was articulated with the study organization in Figure 2.

Figure 2. Review contribution and organisation.

1.1. Some of the Key Contributions of this Paper Are Listed as Follows

• A thorough review backs up the role of IoT applications in prevalence pandemic controls and prevention, which motivates and emphasizes the need for a reliable detection strategy to combat the pandemic
• A thorough review backs up the role of DL techniques usable for pandemic diagnosis and detection
• Current trend of IoT-DL contribution in pandemic detection and control
• The provision of extensive discussion and open issues for developing novel IoT-DL approaches pandemic detection.

1.2. Study Organization

The rest of the paper is structured as follows. The contributions of an IoT-based pandemic detection system and control is detailed in Section 2. Background information on DL taxonomy and architecture as well as data analysis with respect to the COVID-19 pandemic from recent publications is described in Section 3. Current trends in IoT-DL pandemic detection are given in Section 4. State-of-the-art review and discussion of findings and open issues are presented in Section 5. Section 6 contains the summary of the study, conclusions, and suggestions for future study towards overcoming the obstacles and unresolved issues of IoT-DL in pandemic detection.

2. Internet of Things (IoT)-Based Pandemic Detection Systems

The COVID-19 pandemic will not be defeated until the vast majority of people worldwide have received vaccinations or until herd immunity is established. Numerous scientists have taken advantage of IoT architecture based on AI techniques to monitor and identify possible COVID-19 instances to stop the virus’s spread [24]. IoT is a rapidly expanding topic in computer science that is exciting to work in. The need for efficiency and automation has also fuelled technological developments in this area. The increased connectivity of the internet has led to breakthroughs in wireless networking technologies, which have coincided with the growth of IoT devices. The increasing need for efficiency and automation is reflected in the fact that almost every common object may now be connected to the internet. The development of IoT-based systems for pandemic analysis and detection, together with related wearable technology and data collection, is reviewed in this section [25].

2.1. Review of Existing IoT-Based Systems for Pandemic Control

IoT technology utilization in healthcare has advanced significantly and is still a viable field for expansion. Important inventions that use IoT sensors, like the AliveCor heart monitor, demonstrate the value of technology in healthcare and its potential to save lives [26]. Technology has always been a vital factor in the healthcare sector, and IoT devices are widely used in healthcare environments. Using remote health monitoring to keep an eye on patients at home instead of in hospitals is one way that IoT devices can be helpful in the healthcare industry [27]. IoT device data are useful in medical contexts because they can be examined and applied in ways like early disease prediction [28].

IoT technology, which makes it easier to link common products to the internet and create a massive network of interconnected gadgets, has quickly become a groundbreaking topic in computer science. The proliferation of IoT devices can be attributed to two factors: the progress made in wireless networking technology and the growing need for efficiency and automation in various industries. There are many different industries using this technology, and more industries are predicted to use it in the future. IoT is gradually permeating society and is now especially important in domains like security, smart homes, SC, healthcare, and agriculture, among others. The development of networking and technological capacities highlights how important developing technologies are to the global spread of IoT. Though it has promise, there are several obstacles, such as privacy and security issues, that researchers need to overcome with creative thinking. With networking and technology developments fuelling the IoT’s growth, its uses will expand and become more deeply ingrained in society. The emergence of IoT devices, their common uses in healthcare, agriculture, and SC, and the future possibilities of this exciting sector have all been covered in their study [10]. During the COVID-19 pandemic, IoT sensors proved invaluable in assisting medical professionals in keeping a closer eye on vital indicators that, if changed promptly, might potentially save lives [29]. Researchers can identify more methods to progress this field of study by looking at these IoT device applications in the healthcare sector [25]. In the recent COVID-19 pandemic, IoT technologies’ contributions included, but were not limited to, prompt identification and diagnosis of the infection, prevention and management of its spread, monitoring of patients under quarantine, contact tracing of affected individuals, support for medical personnel in obtaining supplies of food, medicine, and medical equipment, remote patient monitoring, and data collection [10,30], as shown in Figure 3.

Figure 3. Contribution of IoT technologies in pandemic control.

2.2. Applications and Descriptions of IoT Technologies, Including Wearables, and Sensors in COVID-19

Wearables and sensors, among other IoT technologies, have proven crucial in managing the COVID-19 pandemic in a number of ways. They have been crucial in tackling the COVID-19 pandemic’s many issues, from tracking the virus’s progress to keeping an eye on health indicators and guaranteeing people’s safety in diverse settings. The digitalization process has allowed “smart” to become the central concept of current technical developments. In fact, IoT technologies are currently regarded as one of the main pillars of the fourth industrial revolution due to their huge potential for innovations and useful benefits for the general population. This section examines current IoT initiatives as well as potential solutions for a range of COVID-19 problems, including tracking infected/quarantined individuals and pre-screening. Major applications of IoT technologies towards the control of COVID-19 pandemic description of the role played by as established from the literature are presented in Table 1.

Table 1. Major applications and description of the role played by IoT in COVID-19 pandemic.

Refs.	Applications	Description of the Roles Played in the COVID-19 Pandemic Control
[25,31,32]	Internet-connected healthcare centers and data collection	Wearable health monitoring devices in real-time surveillance. To support pandemics like COVID-19, hospital facilities require a fully integrated network for the adoption of IoT.
[25,31]	Communicate medical staff during any emergency	Patients and staff would be able to respond more swiftly and efficiently when necessary thanks to this integrated network.
[25,33]	Transparent COVID-19 treatment	The patients can avail the benefits offered without any partiality and favors
[34]	Automated treatment process	The selection of effective treatment modalities facilitates the proper management of cases. Provides healthcare support whenever and wherever needed. Telemedicine as a tool to stop infections and manage viral transmission
[25,33]	Telehealth consultation	Specifically, to use well-connected teleservices to make therapy available to those in need in rural areas.
[32,33]	Wireless healthcare network to identify COVID-19 patient	Smartphones can be equipped with a variety of genuine applications, and sophisticated tracking of affected individuals can facilitate a more efficient and successful identification process.
[33]	Smart tracing of infected patients	The impactful tracing of patients ultimately strengthened the service providers to handle the cases more smartly
[35]	Real-time information during the spread of this infection	Accurate case handling and timely information sharing are made possible by the well-informed and connected locations, channels, and other aspects of IoT-based devices created a wearable, IoT-based quarantine band that may be used to track and identify fugitives in real time.
[25]	Rapid COVID-19 screening	The correct diagnosis would be tried using smart, connected therapy equipment as soon as the case is received. In the end, this enhances the overall quality of the screening procedure.
[33]	Identify innovative solution	The ultimate objective is the overall standard of supervision. It can be accomplished by establishing innovations as ground-level successes.
[36]	Connect all medical tools and devices through the internet	IoT connects all medical equipment and gadgets via the internet to provide real-time information during COVID-19 treatment.
[37]	Accurate forecasting of virus with the help of data analytics tools	Forecasting and predicting the rate of infection and making disease diagnoses. Utilizing statistical methods can aid in forecasting the future state of affairs based on the data report that is currently accessible. It will also assist in planning for a better working environment for the government, physicians, academics, and other professionals.

3. Deep Learning Techniques for Pandemic Detection

In 2020, the coronavirus COVID-19 caused a pandemic that killed many people globally, according to the World Health Organization (WHO). The virus that causes the disease has afflicted millions of individuals, and new infections continue to arise even if it seems to have subsided. Reverse transcription-polymerase chain reaction testing (RT-PCR) or medical data analysis are needed to identify COVID-19. Because scanning and analyzing medical data is expensive and time-consuming, researchers are concentrating on automated computer-aided techniques. Researchers from various disciplines are collaborating on the use of different tools, including DL, to identify COVID-19 from medical data, including ultrasonography, magnetic resonance imaging (MRI), CT scans, X-rays, cough sounds, and clinical signs. DL has its usual workflow, which includes data collection where IoT devices are prominent, data preprocessing, features selection and model training, and COVID-19 detection/classification or prediction algorithms.

3.1. Overview of DL Techniques Suitable for Pandemic Detection

To identify COVID-19 or order related infectious diseases, RT-PCR or medical data analysis are required. Researchers are focusing on automated computer-aided procedures since scanning and interpreting medical data is costly and time-consuming. The use of DL for COVID-19 detection is examined in this section. Since the middle of the 1960s, medical imaging techniques have been utilized to help diagnose chest problems [38,39]. During the mid-1980s, one of the most common uses of computer-aided diagnosis systems (CAD) was for the diagnosis of cancer through chest X-rays [40,41,42]. Convolutional neural network (CNN) models have been utilized by researchers more recently to classify diseases based on medical imagery [43]. Another area of research and development is the COVID-19 pandemic. To detect COVID-19, ML and DL are currently widely used. The latest pandemic has provided a significant boost to DL techniques for COVID-19 infection detection from photos. The present study is unique in that it concentrates on the challenges related to DL and image processing techniques for COVID-19 detection. The normal range of data dimensionality in medical picture analysis and computer-assisted intervention is from two to five dimensions. Volumetric pictures are captured by many medical imaging methods, including as MRI, CT, PET, and SPECT. Multiple images collected over time, known as longitudinal imaging, are commonly used in interventional settings and are therapeutically beneficial for assessing the development of disease and organ function [40,44,45]. An online version of the DL model workflow in [45] is presented in Figure 4. There are input layers or nodes, the neural networks present as hidden nodes or layers, and the output nodes or layers. These nodes or layers can be modified to improve the performance of the networks. However, manual modifications or alterations are hectic and ineffective. It is worth noting that DL models typically require no feature selection due to the design of the architectures.

Figure 4. Typical workflow of Deep Learning Model [46].

The laborious procedure of manually processing features is progressively eliminated as DL techniques effectively automate the task of learning feature representations. By employing a hierarchical layer of feature representation, the three classes of DL (discriminate learning, generative learning, and hybrid learning) aim to replicate the architecture and functionality of the human visual cortical system, especially CNN [47]. By using a layered feature representation strategy, CNNs were able to outperform manually constructed feature detection approaches by automatically learning a variety of visual features. CNN models are made to compute weights and extract features during the training phase in order to evaluate multidimensional data, such as time series or picture data. They have been dubbed convolutions since they employ a convolution operator to make complex jobs simpler. CNNs are most frequently utilized in health-related applications because of their ability to automatically derive features from images [22,48]. CNNs are also able to transmit their knowledge from one task to another through fine-tuning and transfer learning (TL). Tasks involving categorization have shown this strategy’s effectiveness [49]. CNNs have several variants such as VGG-16, VGG19, ResNet, InceptionV3, InceptionV4, AlexNet, DenseNet, which are among the most common [44,50,51]. To preprocess small amounts of input data, they mix mathematical procedures with multiple variable parameters with a multilayer stack of neurons. However, multiplayer perception (MLP) and recurrent neural networks (RNN) perform well in NLP compared to image processing.

3.2. Integration of DL with IoT Devices for Real-Time Monitoring in a Time-Series Data

Time series data forecasting is an important area of study in business, finance, and economics as it relates to healthcare. The development of advanced ML algorithms and approaches, such as DL, along with recent advances in computing power have led to the creation of new techniques for evaluating and predicting time series data [19,52]. By leveraging the current information as a time series and extracting the essential components of previous data to anticipate the values of an upcoming time sequence, this work advances a DL-based time series forecasting approach and conducts a comparative analysis among three models: ARIMA, LSTM, and FB-Prophet [53]. When it comes to evaluating time-series data in a variety of industries, including banking, healthcare, weather forecasting, and industrial processes, DL techniques have demonstrated impressive results. The following examines many well-liked DL methods for time-series data analysis: CNN, RNN, transformer-based models, Autoencoders, hybrid models, attention mechanisms, transfer learning, ensemble model, and transfer learning. CNNs have traditionally been used in image processing, but they have also been adapted for time-series data analysis. One-dimensional CNNs can be used to automatically extract relevant features from time-series data, capturing local patterns effectively. They are computationally efficient and can be parallelized, making them suitable for large-scale datasets.

Making effective emergency plans and policy decisions requires accurate forecasting of the energy use of the commercial and residential sectors. Time series of energy consumed by residential and commercial sectors show notable temporal and spatial patterns that are influenced by long-term evolution and geographical location. IoT is important to these initiatives because it makes it possible to gather vast amounts of data and then use it wisely to forward the objectives of SC. The use of DL for time-series forecasting has garnered significant attention in research because of its practical applications in various industries, including banking, retail, medicine, and the environment. DL is a well-known time-series forecasting technique and its application to the sphere of SC, particularly to multivariate challenges involving many IoT time-series [54]. Climate control systems for automated environments maximize resource use and enable real-time monitoring and early identification of plant health problems, helping to minimize crop losses to guarantee premium produce.

4. Current Trend in Pandemic Detection Based on IoT-DL Approach

4.1. Data Gathering

Technical and useful literature for a particular topic can be successfully found through a comprehensive review of literature [55]. To conduct a comprehensive review of literature, relevant papers must be properly gathered. One of the primary methods for assessing the caliber of review work is dataset preparation [56]. The process of selecting papers involves three steps: first, deciding which search engines to use to find pertinent papers; second, deciding which search libraries keywords to use to address the research questions; and third, filtering the papers that come up according to how well they address the questions in the abstract. Using the right keywords and searching libraries is necessary to find the most relevant studies. Only the most pertinent studies have been gathered for this study from the WoS and Scopus libraries as we gave preference to studies published in respectable journals. IoT and DL-based detection papers published between 2019–2025 were included in this review. The primary keywords and search terms are shown in Table 2 using advanced document search menu. Figure 5 is a Literature review PRISMA framework that presents the ratio of data accessed from the WoS to Scopus libraries. Accordingly, Table 3 shows the inclusion and exclusion criteria and the explanations; only 19 studies were included in qualitative synthesis and read from start to finish.

Table 2. Searching Sources and Queries.

Search Sources	Search Queries
Scopus	“Internet of things” OR “IoT” AND “pandemic” OR “COVID-19” AND “Deep learning” OR “DL” AND “Optimisation” OR “Optimization” OR “Optimised” OR “Optimized” AND PUBYEAR > 2019 AND PUBYEAR < 2025
WoS	“Internet of things” OR “IoT” AND “pandemic” OR “COVID-19” AND “Deep learning” OR “DL” AND “Optimisation” OR “Optimization” OR “Optimised” OR “Optimized” AND PUBYEAR > 2019 AND PUBYEAR < 2025

Figure 5. Literature review PRISMA Framework.

Table 3. Inclusion and exclusion criteria and the explanations.

I/E	CRITERIA	EXPLANATION
INCLUSION	Conference Paper	A well-defined IoT-DL-based paper concentrating on pandemic especially, covering processes, data gathering methods, results and analysis approaches, as well as the conclusions that form the basis for an oral presentation.
	Research-based Chapter in Book	Study having a well-defined IoT-DL-based paper concentrating on pandemic specifically, covering procedures, data gathering strategies, analysis techniques, and findings.
	Research Paper	The paper aimed to investigate specific research problems related to IoT-DL-based paper applications on pandemic
EXCLUSION	Duplicated papers	The identical document that occurs more than once
	Non-research papers	This article is not scientific in nature. Editorial notes, remarks, comments review and related papers were eliminated
	Non-related papers	The problem being studied outside the coverage of this work.
	Non English papers	The paper was not written in English
	Implicitly related papers	The paper does not directly express the research focus pandemic and the use of DL approach.

4.2. Exploration of Recent Advancements in IoT and DL for Pandemic Detection

The review findings that offer insightful information for choosing suitable detection strategies for the given task are presented in this section. Figure 6 provides statistical results on the application of IoT-DL approach on pandemic detection. This depicts the trend of the available studies in the domain. Relevant publications in the databases accessed started in 2020 and the peak was 2023; a substantial number of studies have been published in 2024, and these may increase before the end of 2024. The types of open access studies explored in this study are presented in Figure 7. General open access studies were the highest with 43% and Bronze access was 5%. Others are Green, Hybrid Gold, and Gold. Figure 8 shows the subject area where IoT-DL studies were categorized. Computer Science was the leading subject area, in which IoT-DL approach was prominent for pandemic detection. Engineering and mathematics were the subject areas that closely studied pandemic diseases using IoT-DL. Graphical results of document types accessed are presented in Figure 9. Journal articles have been the leading source and editorial notes are the least. Amidst all languages in the world, only three languages were prominent the accessed document; they are English, Chinese, and Turkish, as shown in Language chat at Figure 10. English was the highest and Turkish was the lowest.

Figure 6. Trend of publication per year.

Figure 7. Types of open access publications.

Figure 8. Subject areas.

Figure 9. Document types.

Figure 10. Language chat.

According to Figure 7, it can be stated that in the year 2020, the COVID-19 pandemic began to manifest itself; the unexpected situation brought an unexpected threat to everyone, and the reactions were different. Subsequently, within two years, studies with data and information on the course of this pandemic began to be processed. According to Figure 8, it is clear what approaches were chosen and the permission to manipulate.

5. State-of-the-Art Review of Literature

In the section, after screening of all the papers that were accessed from the repositories, relevant papers were reviewed through keywording using abstracts. Out of 2880 that were initially accessed, only 19 were relevant and selected for reading start to finish in the state-of-the-arts review. Table 4 summarizes the study goals approaches and conclusion as the state-of-the-arts review on IoT-DL in the papers that were read from start to finish.

Table 4. State-of-the-arts Review on IoT-DL.

Ref.	Study Goals	Approach Used	Contribution Made
[57]	To present a novel model for enhancing the standard of treatment in smart healthcare systems (SHS) by integrating AI and IoT technologies.	Introducing an upgraded particle swarm optimization-long short-term memory (PSO-LSTM) algorithm to optimize the IoT-based SHS model. Comparing the performance of PSO with PSO-LSTM for classifying patient medical data. Tuning several metrics and benchmarks to achieve the highest performance in processing patient data. Evaluating the proposed model using test sets to predict patient health risks.	The study demonstrates that the PSO-LSTM algorithm provides a more satisfactory performance with higher efficiency in classifying patient medical data, achieving an accuracy of 92.5%. This indicates a more secure, reliable, and improved patient satisfaction experience. The integration of AI and IoT in smart healthcare systems offers advanced methods for managing medical records and optimizing patient data processing performance, thereby enhancing healthcare services.
[58]	To introduced a pioneering hybrid DL model for precise energy consumption prediction, aiming to optimize energy efficiency in residential and commercial buildings.	Utilizing IoT-enabled smart meter data to achieve granular energy consumption forecasts. Developing a hybrid DL model that combines CNNs and LSTM units. Conducting a comparative analysis against established DL models to evaluate performance. Focusing on accurately predicting weekly average energy usage in both residential and commercial spaces.	The study showcases a novel model architecture that demonstrates superior performance in energy consumption forecasting, particularly excelling in predicting weekly average energy usage. The hybrid model’s demonstrated capability underscores its potential to drive sustainable energy utilization and provide invaluable guidance for more energy-efficient futures. This innovative approach offers significant promise in guiding tailored energy management strategies, thereby fostering optimized energy consumption practices in buildings.
[59]	The main aim of the research was to develop an efficient real-time IoT-based COVID-19 monitoring and prediction system using a DL model. The goal is to monitor COVID-19 patients, report health issues immediately, and predict COVID-19 suspects in the early stages.	Utilizing IoT-based healthcare systems for real-time monitoring and prediction. Collecting symptomatic patient data from sensors. Selecting effective parameters using the Modified Chicken Swarm Optimization (MCSO) approach. Employing a hybrid Deep Learning model called Convolution and graph LSTM (ConvGLSTM) for COVID-19 prediction. Implementing four stages: data collection, data analysis (feature selection), diagnostic system (DL model), and cloud system (storage).	The developed model is experimented with using a dataset from Srinagar, evaluating parameters such as accuracy, precision, recall, F1 score, RMSE, and AUC. The study demonstrates that the proposed model is effective and superior to traditional approaches in early identification of COVID-19.
[60]	The research introduced a novel AI-based mechanism for optimizing threat mitigation in IoT banking systems, addressing the growing vulnerabilities in this critical sector.	Developing an AI-based mechanism leveraging a Deep Neural Architecture known as Pointer Networks. Focusing on threat identification and mitigation in IoT banking systems, ensuring high precision and recall. Conducting extensive threat-specific evaluations to test the mechanism’s performance across various scenarios. Implementing scalability testing to validate the mechanism’s practical applicability across varying sizes of IoT ecosystems.	Demonstrating a robust defense mechanism with a precision of 0.88, a balanced recall of 0.79, and an F1 score of 0.83. Proving the mechanism’s versatility and high performance in detecting and mitigating different types of cyber threats: Malware: Precision: 0.89, Recall: 0.82, F1 score: 0.85. Denial of Service (DoS) attacks: Precision: 0.87, Recall: 0.78, F1 score: 0.82. Unauthorized access attempts: Precision: 0.90, Recall: 0.81, F1 score: 0.85. Ensuring the mechanism maintains high precision and F1 score values across different sizes of IoT ecosystems, validating its scalability and practical applicability.
[61]	To develop a sophisticated and effective epidemiological surveillance system for COVID-19 that overcomes the limitations of conventional approaches by leveraging IoT and advanced data analytics.	The developed framework created the SEIR-Driven Semantic Integration Framework (SDSIF) designed to handle diverse data sources using IoT technology. COVID-19 Ontology: Develop an extensive COVID-19 ontology to enable unmatched data interoperability and semantic inference. Data Integration and Analytics: Facilitate real-time data integration and utilize RNN for advanced analytics, anomaly detection, and predictive modeling. Scalability and Flexibility: Ensure the framework is scalable and flexible to adapt to various healthcare environments and geographical regions. Evaluation: Assess the performance of SDSIF using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared score.	The SDSIF framework revolutionizes COVID-19 epidemiological surveillance by integrating and analyzing data in real-time, offering unparalleled data interoperability and semantic inference through its innovative ontology. This framework enhances predictive modeling and anomaly detection capabilities, proving highly accurate and precise in predicting COVID-19 trends. The rigorous evaluation metrics demonstrate the framework’s effectiveness, with an RMSE of 8.70, MSE of 3.03, and an exceptional R-squared score of 0.99, highlighting its robustness in explaining disease data variations. This contribution marks a significant advancement in managing and responding to the COVID-19 pandemic and potentially other epidemiological crises.
[62]	The goal of the study was to develop a remote diagnostic system, called IFCnCov, for diagnosing COVID-19 patients in real-time. The system integrates IoT, fog computing (FC), cloud computing (CC), ensemble learning (EL), and DL principles to achieve accurate diagnosis remotely.	IFCnCov was designed as a two-layered architecture, incorporating DL approaches trained on two different datasets: a symptom-based dataset and a chest X-ray imaging dataset sourced from the Kaggle repository. IoT, FC, and CC Integration: The system leverages the integration of IoT, FC, and CC principles to address latency, bandwidth, energy consumption, security, and privacy issues associated with remote diagnosis. Ensemble Learning and DL: EL and DL techniques are utilized for accurate diagnosis, with DL models trained on the symptom-based and chest x-ray imaging datasets. The performance of IFCnCov was evaluated using various evaluative measures, including accuracy, precision, sensitivity, specificity, and F1-scores. Validation was conducted on network parameters such as scalability, energy consumption, network utilization, jitter, processing time, throughput, and arbitration time.	Enhanced Accuracy: IFCnCov achieves significantly high accuracies, precision, sensitivity, specificity, and F1-scores in both stages of diagnosis, outperforming some other state-of-the-art works. Validation of Network Parameters: The study validates the performance of IFCnCov in terms of various network parameters, demonstrating its scalability, energy efficiency, processing speed, and overall effectiveness.
[63]	The goal of this study was to comprehensively explore the intersection of cloud computing and AI in education and assess their combined impact on accessibility, efficiency, and quality of learning. The study aimed to investigate how these technologies enhance personalized learning experiences, increase user capacity, reduce administrative errors, improve scalability, and enrich overall learning experiences.	The study employs a mixed-research design to investigate the convergence of cloud computing and AI in education. It identifies improvements in educational content personalization attributed to AI and enhancements in simultaneous user capacity facilitated by cloud computing. The methodology involves analyzing data to quantify the extent of improvement in accessibility, efficiency, and quality of learning resulting from the integration of these technologies.	The study contributes to the understanding of the synergistic effects of cloud computing and AI in education by providing empirical evidence of their impact on various aspects of teaching and learning. It reports a 25% improvement in educational content personalization and a 60% increase in simultaneous user capacity, along with reductions in administrative errors and improvements in scalability. By comparing these findings with previous research, the study positions itself as a critical resource for guiding future developments and improvements in the education sector in the context of a digitally advanced world.
[64]	The primary goal of this study was to design and develop an integrated system of audio and video sensors, leveraging the IoT, to recognize and monitor coughing and sneezing, which are key symptoms of COVID-19. The system aims to provide real-time detection and alerting capabilities to support early intervention and containment measures, ultimately reducing the spread of the virus.	One way to get around these setbacks was through sensor integration. Furthermore, it raises the accuracy of event recognition. We suggested a real-time integrated IoT architecture to enhance the outcomes of coughing and sneezing detection because low-cost audio and video sensors are widely available. A cloud computing infrastructure was integrated with edge computing. Edge computing involves the camera and microphone being embedded with a DL engine and being connected to the internet.	A real-time coughing and sneezing activities are detected by edge computing by feeding it audio and video data. Comparing the accuracy and recall of the cloud detector to audio-only and video-only detectors, the cloud computing technology, which is built on the Amazon Web Service (AWS), improves both on average by 43% and 15%, respectively. The F-score increased 1.24 times on average.
[65]	The paper focused on designing and development of “Smart COVIDNet”, an IoT-based framework for predicting COVID-19 disease. The framework leverages an ensemble of deep learning models with attentive and adaptive mechanisms to improve the accuracy and efficiency of COVID-19 diagnosis.	The study used IoT devices to gather real-time personal health data from participants, including temperature, heart rate, and breathing rate. Preprocessing procedures are used to collected data to guarantee quality and eliminate noise, preparing it for analysis. The system makes use of a group of DL models, each with a focus on a distinct area of the data. Transformer-based architectures, RNNs, and CNNs are some examples of these models. In order to improve the model’s capacity to recognize significant patterns connected to COVID-19, an attention mechanism was employed to concentrate on pertinent aspects in the data. By adding fresh data to the models on a regular basis, the framework adjusts to shifting patterns in the data and gradually increases the forecast accuracy of the models. An ensemble approach was used to integrate the outputs from individual models to create a final forecast that is anticipated to be more accurate than the individual predictions.	High precision and efficacy in COVID-19 infection prediction are demonstrated by the Smart COVIDNet platform. Early and accurate COVID-19 diagnosis can be achieved with a robust system that combines attentive and adaptive DL techniques with real-time data collection from IoT devices. By facilitating accurate and timely disease prediction, the study finds that Smart COVIDNet can be an effective tool for managing and containing the spread of COVID-19.
[66]	The study’s objective was to increase COVID-19 detection efficiency and accuracy by utilizing IoT data and sophisticated deep learning techniques with a recurrent neural network (RERNN) improved by recalling and optimized with the Golden Eagle Optimization (GEO) algorithm.	The system collected real-time health data from people using IoT sensors, such as temperature, oxygen saturation, heart rate, and respiration rate. The gathered data was preprocessed to guarantee quality and eliminate noise, preparing it for DL analysis. Over time, the RERNN, a specialized RNN variation, was developed to improve memory recall capacities and increase its efficacy in finding patterns connected to COVID-19. The performance of the RERNN and its hyperparameters were optimized using the GEO Algorithm. Inspired by the hunting tactics of golden eagles, this optimization technique aims to increase the neural network’s accuracy and convergence speed. Using preprocessed IoT data, the RERNN model was trained, and standard metrics including accuracy, precision, recall, and F1-score were used to assess the model’s performance.	High accuracy and efficiency in COVID-19 detection are demonstrated by the IoT-based COVID-19 detection system that uses the RERNN optimized with the GEO algorithm. The study comes to the conclusion that combining IoT data with cutting-edge deep learning and optimization methods can greatly improve COVID-19 early identification and treatment. The suggested system has demonstrated potential in giving medical practitioners a dependable and prompt diagnosis tool, improving control and reducing the spread of the illness.
[67]	Authors developed a smart IoT-based monitoring system for COVID-19 utilizing hybrid DL models aimed to enhance the monitoring, detection, and management of COVID-19 by combining various DL techniques and leveraging IoT technology.	The system used IoT devices to continuously monitor and gather people’s real-time health data, including body temperature, heart rate, and breathing rate. To assess the data gathered, the study uses a hybrid model that combines several deep learning approaches. In order to improve the focus on pertinent data aspects and increase the model’s accuracy in finding COVID-19-related patterns, this contains CNNs, RNNs, and Attention Mechanisms. Preprocessing is done on the gathered data to guarantee quality and eliminate noise. This data was processed by the hybrid model, which uses it to derive insightful conclusions and precise forecasts. Real-world scenarios are used to implement and evaluate the performance of the Internet of Things-based monitoring system. Standard performance indicators including accuracy, precision, recall, and F1-score are used to assess the model’s efficacy.	This study’s hybrid DL model-based smart IoT monitoring system shows excellent efficacy in COVID-19 detection and monitoring. IoT technologies and cutting-edge DL methods work together to offer a reliable solution for early COVID-19 symptom detection and real-time health monitoring. According to the study’s findings, a system like this can greatly help with the effective treatment and control of COVID-19 by giving medical practitioners immediate and precise health insights to help them make decisions.
[68]	The study aimed to utilize DL models’ capacity to evaluate health data gathered from IoT devices to identify COVID-19 infections precisely and promptly. Authors created an effective COVID-19 identification system by integrating DL methods with IoT technology.	The system uses IoT devices to collect people’s real-time health data. This includes variables like oxygen saturation levels, heart rate, breathing rate, and body temperature. Preprocessing procedures are used to clean and standardize the obtained data, making it ready for DL research. The preprocessed data is processed and analyzed using several DL models. These algorithms have been trained to identify characteristics and patterns that point to COVID-19 infection. A dataset of health-related factors is used to train the DL models, and optimization techniques are used to increase the models’ precision and effectiveness. The models are adjusted to enhance their capacity to recognize COVID-19 cases. Standard metrics like accuracy, precision, recall, and F1-score are used to assess the effectiveness of the DL-based identification system. To evaluate the system’s dependability and efficacy, it was put through a number of situations.	The study comes to the conclusion that the DL-based IoT system created for COVID-19 identification is incredibly accurate and efficient. Real-time monitoring and early COVID-19 infection identification are made possible by the integration of DL models with IoT technology, giving medical professionals a useful tool. Based on quick and accurate diagnosis capabilities, the results show that such a system may greatly enhance COVID-19 management and control.
[69]	The study was to provide a DL system for early COVID-19 evaluation that was based on the IoT. The system seeks to identify and evaluate COVID-19 infections early on by utilizing the capabilities of IoT devices and cutting-edge DL algorithms.	The framework uses a variety of IoT devices, including smart devices and wearable sensors, to continuously gather personal health data from users. Sophisticated DL models were utilized for the purpose of data analysis. The purpose of these models was to find trends and abnormalities that point to COVID-19 infection. Prior to feeding the input into the DL models, the quality and relevancy of the data gathered from IoT devices were checked through preprocessing. A dataset comprising COVID-19 positive and negative examples is used to train the DL models. To evaluate the models’ dependability and accuracy in identifying COVID-19, validation was carried out.	The study concluded that the DL framework based on the IoT was useful for early COVID-19 detection. IoT device integration makes it possible to monitor health in real time, and DL models use the data gathered to accurately detect COVID-19. Early intervention and control measures are essential for controlling the virus’s transmission and enhancing patient outcomes, and the framework appears to be promising in this regard. This could enhance prompt medical intervention and slow the virus’s spread.
[70]	The study provided an IoT-integrated ensemble DL framework for COVID-19 automated diagnosis. The objective was to improve COVID-19 detection efficiency and accuracy by utilizing IoT devices for real-time data gathering and analysis and merging many DL models.	Real-time health data, such as body temperature, respiration patterns, and other vital signs important for COVID-19 diagnosis, are collected by the framework using IoT sensors. Many DL models are utilized to examine the gathered information. These models comprise, among others, CNNs and RNNs. To increase the overall diagnostic accuracy, the outputs of various DL models are combined using an ensemble technique. To combine the predictions from several models, methods like weighted averaging and voting procedures are used. Preprocessing is done on the gathered IoT data to reduce noise and standardize the inputs for improved model performance. Labeled COVID-19 positive and negative samples are included in a varied dataset that is used to train the DL models. The ensemble model’s performance was verified through the use of metrics like accuracy, precision F1 score and recall.	The study came to the conclusion that the accuracy and dependability of automated COVID-19 diagnosis are greatly increased when the ensemble DL framework and IoT technology are used together. Using the advantages of several models, the ensemble technique increases the detection system’s robustness. Because of the integration with IoT devices, continuous and real-time monitoring is made possible, which makes the system useful for accurate and timely COVID-19 detection. With prompt identification and action, the suggested paradigm shows promise for helping healthcare systems manage the pandemic more skillfully.
[71]	The study created an intelligent COVID-19 monitoring system by integrating wearable IoT sensors with DL algorithms. The goal was to improve patient care and efficiently control the virus’s spread by offering ongoing, real-time surveillance and early detection of COVID-19 symptoms.	The framework uses intelligent wearable Internet of Things sensors to continuously gather physiological data from people, including heart rate, body temperature, oxygen saturation, and respiration rate. Superb DL algorithms are applied to the data collection process. The models are trained to identify abnormalities and patterns that point to COVID-19 infection. Preprocessing is done on the data collected by the wearable sensors to guarantee accuracy and consistency. This entails normalizing the data and eliminating noise. A large dataset with both COVID-19 positive and negative events is used to train the DL models. The models’ efficacy in identifying COVID-19 is then evaluated by validating them using a range of performance indicators, including accuracy, precision, recall, and F1 score. Using the continuous data stream analysis from the wearable sensors, the framework allows for real-time monitoring and alarms.	The study concluded that the intelligent monitoring framework integrating DL with smart wearable IoT sensors is effective for the early detection and monitoring of COVID-19. The continuous real-time data collection and analysis enhance the ability to detect COVID-19 symptoms promptly, enabling timely medical intervention. The proposed framework demonstrates significant potential in improving patient outcomes and aiding in the control of the COVID-19 pandemic through effective monitoring and early diagnosis.
[72]	The goal of the project was to create a framework for leveraging data from IoT-based wearable devices to remotely monitor COVID-19 patients’ health. The framework analyzes health data using CNNs and metaheuristics in order to provide accurate and ongoing patient health status monitoring.	The framework gathers real-time health data from COVID-19 patients using wearable technology. Vital indicators including heart rate, temperature, and oxygen saturation levels are included in this data. The feature selection and data pretreatment procedures are optimized through the use of metaheuristic algorithms. The effectiveness and precision of the data analysis are improved by these algorithms. The preprocessed data is processed and analyzed using CNNs. The purpose of these deep learning models is to find patterns and abnormalities in the medical data that might point to alterations in the patient’s condition. Preprocessing procedures, such as noise reduction and normalization, were applied to the gathered data. Prior to feeding the data into the CNNs, the metaheuristics optimize the selection of pertinent characteristics. Through ongoing analysis of the patient’s medical records and the provision of real-time alarms and status updates, the framework makes remote monitoring possible.	The study finds that the suggested framework—which combines CNNs and metaheuristics with IoT-based wearable devices—was useful for remotely keeping an eye on COVID-19 patients’ health. The accuracy and dependability of the health monitoring system are improved by the application of DL techniques and sophisticated optimization algorithms. For COVID-19 patients, the framework offers prompt alerts and insights into their health status, which can result in improved management and intervention techniques.
[73]	The goal of the study was to address the security concerns in electronic healthcare systems, particularly focusing on the protection of sensitive health images transmitted over networks. The study aims to propose and implement a secure lightweight key frame extraction approach and an encryption scheme for ensuring the integrity and confidentiality of COVID-19 CT-images, while also utilizing AI techniques for COVID-19 testing.	The methodology involves the identification of security concerns recognizing the need for secure transmission of health images in electronic healthcare systems and understanding the limitations of traditional encryption methods for image data. Development of Secure Lightweight Key Frame Extraction Approach: Proposing and implementing a lightweight key frame extraction approach to ensure the accuracy and privacy protection of e-health services. Encryption Scheme Development: Building an encryption scheme incorporating a hashing version of the Blum Blum Shub (BBS) generator, known as Hash-BBS (HBBS), to achieve high-grade integrity and confidentiality in the transmission of COVID-19 CT-images. Utilization of AI Techniques for COVID-19 Testing: Applying CNN as an AI technique for COVID-19 testing to enhance secure prediction. Evaluating the proposed framework’s performance compared to alternative security and transfer learning methodologies in terms of security and prediction benchmarks.	Enhanced Security: Introducing a secure lightweight key frame extraction approach and an encryption scheme (HBBS) to ensure the integrity and confidentiality of COVID-19 CT-images transmitted in electronic healthcare systems. Utilization of AI for COVID-19 Testing: Employing CNN for COVID-19 testing to improve secure prediction of COVID-19 cases. Demonstrating through evaluation that the proposed framework outperforms alternative security and transfer learning methodologies, providing reliable transmission of CT-images for COVID-19 patients while meeting strict security and prediction benchmarks.
[8]	The goal of this methodology is to enhance security and safety measures in public gathering places by utilizing IoT technology and DL concepts. Specifically, the aim is to verify the presence of safety items such as masks and gloves, as well as detect body temperature, of individuals entering public spaces.	A camera, temperature sensor, and other safety sensors are installed at the entry points of public gathering places. Image processing techniques, coupled with DL algorithms, are employed to analyze the images captured by the camera. This analysis verifies the presence of safety items like masks and gloves. The temperature sensor measures the body temperature of individuals entering the premises. Additionally, the setup includes a sanitizer sprayer that activates when hands are placed in front of it. All sensors are connected to a single-board computer (SBC), such as Raspberry Pi, which processes the sensor data and triggers actions accordingly. If safety requirements are met, the locks are opened. Otherwise, individuals are flagged for further monitoring and disciplinary actions.	Enhancing security and safety measures in public gathering places using IoT and DL technologies. Achieving over 95% object detection accuracy through DL and image processing techniques. Ensuring worker safety at public places by implementing low-cost safety precautions. Reducing the workload of supervisors and minimizing manpower required for safety monitoring.

5.1. Discussion of Findings on the Implementation of Optimal IoT-DL Techniques for Pandemic Detection

In this section, we discuss the literature survey findings regarding the IoT-DL for pandemic detection. We provide an overview of top-notch papers on IoT-DL-based pandemic detection that were released between 2019 and 2024. IoT solutions are widely used for real-time applications and monitoring in hospitals, as well as the adoption of facemasks worn in supermarkets, banks, and malls, and the changing language used online (Twitter). Additionally, sensors have been implemented into IoT devices to monitor vital signs and COVID-19 symptoms. To raise awareness and provide education on how to combat COVID-19, virtual reality was deployed. Drones and cameras have primarily been employed for social distancing. The nature of IoT technology, which goes through a realistic process and demonstrates its performance in practical applications, may be one of the primary causes of this. IoT could handle the tasks in all applications with success. In this way, IoT offered a quick and effective way to monitor the spread of the illness. The pitfalls of IoT technology combined with DL are open access and population concerns with the disclosure of medical reports and privacy, on the basis of which data on the progress of the epidemic are processed. Other pitfalls are mentioned in the study; these should still be worked on and provide the population with education within the framework of AI and these technologies.

Our research demonstrates that numerous methods and applications, each with a distinct goal and utilizing a different technology, have been put forth to deal with the COVID-19 pandemic. However, several previously developed solutions—like those found in [5,69]—proposed methods that combine IoT and DL. These approaches aim to address the problem of integrating and harmonizing the massive amounts of data generated by IoT devices. According to other studies [65], predictive analytics and preventive tailored health services in an IoT environment require the combination of machine learning and semantics. In addition to its apparent uses in pandemic prevention, DL can help with complicated decision-making at potentially higher intelligence levels that could have advantages for IoT technologies [52]. This requires gathering enough data in real-time settings outside of simulated scenarios, which is still a big difficulty [8,65]. By building a generator and a discriminator, the metaheuristic optimisation algorithms (MOA) [57] increases learning efficiency in contrast to ML, which is merely an expansion of an already-existing technique. This eliminates the need to define the system’s reward or loss function. This has applications in image synthesis, monitoring, and pattern recognition that would directly affect mining and automation. It also makes learning tasks that were before impractical viable.

The nature of IoT technology, which goes through a realistic process and demonstrates its performance in practical applications, may be one of the primary causes of this. IoT could handle the tasks in all applications with success. In this way, IoT offered a quick and effective way to monitor the spread of the illness. Performance requirements were specified in most of the research using DL-based approaches to handle COVID-19-based datasets. The accuracy factor is associated with the most widely used performance criteria. It can be used to compare different datasets with DL-based algorithms. Training DL models can be costly, time-consuming, and require large sample sizes for optimal accuracy [43]. To build more reliable models, better parameter optimization is needed because DL is prone to deceit and misclassification as well as being trapped on local minima [74].

Even if the healthcare sector is becoming more digitally advanced and uses technology more often, much more capacity is anticipated in terms of pandemic preparedness and control. Thus, recent advances can provide insight into future advancements. Some new ML techniques that are closely related to DL are starting to emerge in this field. This includes creating and developing models for pandemic data gathering and analysis that are based on real-time IoT-DL. DL techniques for COVID-19 detection pose many difficulties. DL-based COVID-19 detection from X-ray and CT scans must still overcome a number of societal and technological challenges, despite encouraging results [45]. The review’s findings are a great resource for researchers and practitioners looking to optimize DL techniques, which in turn promotes a variety of practical and dependable real-world applications for the study of pandemic data [75,76,77].

Sharing large amounts of data between IoT-enabled devices, which referred to scalability, remains a challenge now and in the future. The complexity of managing the growing amount of data and devices increases as more and more gadgets are connected to the internet [25,61]. IoT device network architecture needs to be able to manage the growing amount of data flow. Network congestion, latency, and bandwidth limitations can all cause scalability problems, particularly in settings where a lot of densely placed devices are present. Another thing to keep in mind when using IoT technology in the current pandemic scenario, from the perspective of patient health, is specific to the security and privacy of the data that are received [33]. A related concern is the caution to be exercised when integrating the data network among the participating devices and protocols [33]. A significant challenge in the analysis of COVID-19 is the scarcity of adequate and trustworthy data during the initial quarter of the pandemic. Numerous fatalities and viral illnesses go undetected due to the small number of tests (especially the one accessed through IoT devices). There is not a single nation on the planet that can offer trustworthy data about the virus’s prevalence in a sample that is representative of the whole public. Large-scale, high-quality dataset accessibility is essential for ML and DL applications [44].

5.2. Research Part and Open Issues

Installing sensor gear into an IoT system is the first step in the development process. Enhancing IoT sensor performance is the main problem that must be solved in order to enable the network’s functionality to function as intended [78]. Given that pandemic control must be rendered quickly and in real-time, particularly in light of the COVID-19 pandemic and Ebola pandemic, a wireless sensor’s lifetime and power consumption should be optimized. If not, the system will not be able to function in emergency situations. The existing work focuses on using static datasets for their investigation; static datasets that collect by data produced by human experts may also lead to infection and the spread of the disease. Undoubtedly, the security issue is difficult because people are wary of going viral and typically do not want to disclose their medical conditions to strangers. However, it is possible for someone to knowingly falsify a patient’s report, which might put the patient in danger. The COVID-19 epidemic put the whole planet on lockdown, with social separation emerging as the sole effective means of preventing its spread. Therefore, there is a growing need for IoT-based remote healthcare solutions. Not everyone, though, has access to a fast internet connection. Therefore, the service provider’s obligation to cover such areas remains open. In addition, the data are stored in the cloud via the online method, which makes it accessible to both the doctors and patients. Another fascinating phenomenon in the realm of IoT is the management of shared resources. However, due to a lack of confidence, it is unlikely that the general population will receive service from a distance. More public knowledge and outreach are required for this kind of application, and people should be encouraged to seek remote assistance [79,80].

In order to effectively control pandemics like COVID-19, future research initiatives must examine the Fourth Industrial Revolution (4IR) technologies; these include IoT technologies that have enormous potential for data collection and transmissions and big data analytics such ML and DL. The development of DL, ML, and AI will be essential in spurring innovations in pandemic preparedness and control, opening up new avenues for IoT applications due a fast data collection and analysis of these technologies. To ensure IoT technology’s continued success and expansion, it is also necessary to address concerns including the interoperability and scalability of IoT systems, improve affordability, and increase awareness among users and stakeholders. Furthermore, scholars in this domain should be concerned about the hyperparameter configuration and data dimensionality of DL-based methods [40]. Additionally, it is essential to create models that simplify different systems while simultaneously effectively representing them, striking a balance between their complexity and predictive capabilities. Exploring metaheuristic algorithms for the exploration and exploitation of data as well as hyperparameter turning towards selection of the optimum tuning is hereby suggested for future research directions and collaborations. Potential applications of IoT and DL in other healthcare domains.

A further drawback of IoT and DL-based methods for COVID-19 applications is the absence of a comprehensive high-quality dataset. This might be the result of distinct models being developed using little amounts of data for particular applications within the same data area. Utilizing IoT, AI, or DL-based approaches is intended to address a particular issue in the real world with a practical application that calls for specialized tools and technology. The cost and accessibility of creating and outfitting communication devices for IoT-based or therapeutic, diagnostic, estimating, and forecasting applications are constrained. Furthermore, IoT developers have limited access to best practices. The absence of IoT-based incident response activities as the most effective techniques has resulted in limitations on the application and passage of rules, regulations, and policies regarding the use of this technology due to the lack of IoT edge authentication and licensing requirements, ethical and privacy concerns associated with data collection and analysis, scalability and interoperable systems, and random sampling. This is in agreement with [54,81]. Because of these restrictions, efforts to figure out how to obtain situational awareness of the security of IoT assets in a medical complex are currently unfocused. Various significant aspects that affect the usage of open data are discussed in some publications, such as the scarcity of open data for study, adoption and sharing of hurdles, privacy concerns, and guidelines and recommendations for data sharing in information system [82]. The results of the survey suggest the need to permit network scalability of precise IoT-DL systems without jeopardizing information privacy. Pandemic detection in IoT-based computing environments is a problem with limited scalability that presents a significant difficulty. It is also required to more thoroughly consider all the causes of the pandemic and provide opportunities to better prepare against it based on the results found. The use of fourth industrial revolution technologies that would analyse multimodal datasets and detect patterns in real-time is vital [80,83].

From a technological standpoint, the most popular, secure, and reliable way to identify COVID-19 cases is the RT-PCR technique. The use of DL in healthcare is a relatively young and quickly expanding field of study. Because DL-PCR-based techniques are currently widely accessible, it is not possible to completely replace them. Consequently, DL-based detection systems may be used by medical practitioners as a first screening technique or as a tool to aid in their decision-making. DL-based COVID-19 detection presents many difficulties. DL-based COVID-19 detection from X-ray and CT scans must still overcome several societal and technological challenges, despite encouraging results. In recent years, several CNN architectures have been proposed. CNN architectures that are most frequently used to identify COVID-19, including DenseNet, VGG, Xception, ResNet [67], AlexNet, MobileNet [36], and Inception. To choose the best architecture for the job at hand, it is essential to look at different designs’ depth, tasks, resilience, and input size as the overview of the architectures is given [44,83,84,85]. Some of the discriminative DL models with high accuracy scores were not evaluated based on other metrics for efficiency and effectiveness of the models.

6. Conclusions

A summary of benchmarking results is provided, along with a discussion of the limits of IoT-DL-based pandemic control techniques. We present an overview of current knowledge, point out gaps and unresolved problems, and make recommendations for additional research that will be helpful to anyone hoping to advance their career in this field and become an expert in it. The review sheds light on open problems within the IoT-DL domain by offering perceptive analyses of the topics that remain unresolved and the current developments. Researchers can use this work to create more efficient IoT applications for pandemic preparation and management. This study raises new, significant research questions that should be investigated to ensure that everyone has access to healthy settings in the event of a pandemic. This survey was motivated by the need for effective and efficient technological-based pandemic preparedness; as such, we present a comprehensive review of the recent developments in IoT and DL in pandemic detections. This work presents a thorough analysis and exploration of related studies that have employed IoT-DL approaches for pandemic detection. It provides a thorough analysis of the literature on IoT-DL-based pandemic detection techniques. Based on the objectives of the studies, the methodology employed, and the contributions made by earlier researchers in the field, we provide the state-of-the-art as well as the trends in current research. We also provided information on the current difficulties and possible future paths for this field of study. It adds to the body of knowledge by pointing out that novel approaches, such the MOA IoT-DL, have not always been employed in pandemic research, especially when trying to get around issues with hyperparameter tuning and data dimensionality.

In order to address COVID-19 and other pandemic-related issues, the current study review existing IoT-DL-based studies. Monitoring, detection, identification, classification, and diagnosis are the categories into which the primary applications fall. The results showed that the discriminative DL models (CNN, RNN, and MLP) and their variance are under investigated. The discriminative DL classifiers also have pre-trained networks that perform differently based on the task and dataset. We anticipate that this unique survey will make a substantial contribution to identifying research gaps in the intricate and quickly changing field of multidisciplinary pandemic control. The evaluated publications have offered a range of discriminative DL models that were created to offer a rapid and precise automated solution for COVID-19 viral diagnosis. We concentrated on the study objectives, important elements of the IoT-DL techniques employed, and the contributions produced in this comprehensive review. It is worthy of note that none of the studies consulted have used any of the metaheuristic algorithm in their investigation.