1 Introduction

In the past century, there have been various health crises caused by pandemics and outbreaks that were unparalleled in human history. The majority of these pandemics were attributable to influenza viruses, including H1N1, H2N2, and H3N2 [1]. For instance, Spanish Flu and swine flu pandemics were caused by the H1N1 virus [2] and Asian flu and Hong Kong flu were caused by the H2N2 and H3N2 viruses. Various coronavirus outbreaks have also occurred in the last 20 years, such as the outbreak of SARS-CoV in 2002 and MERS-CoV in 2012 [3]. At present, the world is grappling with a severe global pandemic known as Covid-19, which is caused by a novel and highly contagious strain of the coronavirus called severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). This disease poses a significant threat to human health and well-being [4]. The disease has infected people globally in many countries. As the number of infections rose exponentially, health workers’ primary concern was to end the spread of coronavirus. Some preventive measures were taken such as social distancing, wearing masks, and washing hands regularly. Many countries took substantial steps such as imposing curfews and lock-downs, calling for work from home, and social distancing to combat the spread of Covid-19 infection [5]. Till July 2022, an estimate of 6,369,510 deaths has been reported globally [6, 7]. The pandemic also caused huge global economic losses with the enormous disruptions in many areas such as agriculture, transport, supply chain, tourism, and industry, compelling governments and owners to shut down most operations globally [8]. An overview of Covid-19 symptoms, preventive measures, worldwide effect and mitigation efforts is shown in Table 1.

Table 1 Overview of Covid-19symptoms, preventive measures, worldwide effect and mitigation efforts [9]
Full size table

The infection levels may vary from asymptomatic and mild to severe or fatal. General categorization of Covid-19 cases is shown in Table 2. A Covid-19 patient’s health deteriorates rapidly, leading to patients being admitted to the hospital’s Intensive Care Unit (ICU) [9]. According to statistics from China, almost 15–20% of Covid-19 cases were hospitalized, including 15% of cases of severe illness and 5% of patients needing intensive care [10]. In Spain and Italy, 40–50% of Covid-19 cases were hospitalized, and 7–12% were admitted to ICUs [11].

Table 2 Classification of COVID-9 patients [12]
Full size table

Even though medical professionals worked day and night to save lives, the rapid growth in the cases of Covid-19 hard-pressed the existing healthcare systems to their limits. A traditional health care systems manage multi-party workflows such as diagnosis, patient care, testing, prescriptions and remote monitoring. It is worth mentioning that patients’ data between different centers are not shared on a common platform. Hence, when patients are referred from one place to another, sometimes reexamination of patients is observed. This promotes tedious manual work and makes treatment costly and time-consuming [12]. During this, the treatment of critical patients may be delayed, which risks the life of the patients as well. Moreover, the healthcare data management system is not coordinated between healthcare facilitators. The absence of an unfailing data surveillance system that would disseminate relevant information to healthcare institutions was also noticed [13]. Main sources of Covid-19 information include government websites, hospitals, news portals, and clinical labs. However, getting the correct and reliable information was very difficult and a real challenge for the Covid treatment system. Another challenge was to process huge volumes of Covid-19 data using human-dependent medicine tools [14]. Furthermore, as the disease is highly infectious, the patients were quarantined to disrupt the spread of the disease. Isolation keeps the spread of the virus under control. Constant monitoring of patients’ vital signs is essential in highly infectious diseases like Covid-19. It is required for both low or medium condition patients in home quarantine or severe patients in the hospital.

Technological advancements can aid in overcoming these challenges. Recently, the technologies such as the Wireless Body Area Network (WBAN), Internet of Things (IoT), Blockchain, and Machine Learning (ML) have been used widely in the medical domain [15]. These technologies facilitate in develo** a smart healthcare system that can help in integration of diagnosis, treatment surgeries, clinical management, maintenance of supplies, remote monitoring of patients, smart and cost-effective supply chain in pandemic situations. In contrast to traditional medical systems, smart healthcare systems create a network where healthcare workers and patients can easily connect to exchange essential information whenever required. The smart healthcare system would assist us in being prepared for unprecedented pandemics and emergencies like Covid-19 [16].

Remote patient monitoring using IoT, Blockchain and ML is an essential part of smart healthcare systems. During the ongoing COVID-19 pandemic, the importance and need of remote monitoring has increased drastically. It enables healthcare professionals to record the vitals of a patient remotely in continuous manner. It also helps to improve the safety of the healthcare professionals as there would be minimal patient-doctor interaction. In remote monitoring process, a patient is admitted to a hospital or home quarantined and is assigned a set of sensors that measure the patient’s vitals like heart rate and blood pressure, from time to time. The vast amount of data collected from the WBAN and the previous health records of the patients are stored securely in a database [17]. Healthcare data is considered highly sensitive and private and is primarily stored in centralized servers or cloud servers [18]. These storage methods suffer from single-point failure. Healthcare facilities can be discontinued and disrupted by attacks such as Denial of Services (DoS) and Ransomware attacks. Third-party organizations generally own cloud data centers; hence the issues related to privacy, tractability and accountability of patient data is inevitable [19]. Blockchain technology has huge potential to solve the security and privacy issues in a healthcare system. Blockchain stores the data securely using cryptographic functions in a decentralized way. The stored data on the blockchain is immutable; hence it cannot be changed by malicious users [20]. The stored patient data can be processed/analyzed using well-trained ML models for prognosis and diagnosis. Machine learning algorithm uses this monitored data as well as the previous health records of the patients for training the model. It can provide valuable insights into the patient’s condition, hence assisting the healthcare workers for better decision making [18]. Machine Learning aids scientists, doctors, and researchers in aggregating, analyzing and processing the patients’ data for several valuable purposes like vaccine development, supply chain management and predicting patients’ health [15].

Hence, in Covid-19-like pandemic situations, emerging technologies may be utilized to build remote vital signs monitoring system. The sudden surge in the cases of Covid-19 globally has forced different people and communities to search for prompt solutions to weaken the effects of the Covid-19 outbreak. This paper aims to survey the recent literature on remote monitoring systems for pandemics like Covid-19.

1.1 Research Methodology

We performed a sco** literature review related to remote monitoring of Covid-19 Patients. We searched 9 databases in May 2022: Google Scholar, IEEE Xplore, Springer, Taylor and Francis, Scopus, Science Direct, ACM Digital Library, Web of Science, and Wiley. These databases were explored using search terms related to the remote monitoring (e.g., ‘Remote Monitoring’ and ‘Covid’, ‘Patient Health’, ‘IoT’), blockchain storage systems (‘Covid’ and ‘Health’, ‘Blockchain Storage’ or ‘Blockchain Sharing’) and processing/analyzing of medical data of Covid-19 like pandemics only (e.g., ‘Processing Analyzing’, ‘Prediction’, ‘Covid Patient Sensed Health Data’, ‘Machine Learning’). We also examined the references of the included studies (i.e., checking references backward) to find more related works that could be reviewed for this survey. We also analyzed the works that referenced the included studies (i.e., checking references forward). We excluded studies that were related to other diseases. Our search was restricted to English works published after the advent of COVID-19 (i.e., December 2019). We have included only peer- reviewed and good quality articles, journals and conference proceedings.

1.2 Scope and Contributions

Recently, several research studies on combating pandemics using technology have been conducted. However, no comprehensive survey on remote monitoring of pandemic patients exists. Hence, unlike existing surveys on Covid-19 management, this work focuses on reviewing the research related to remote monitoring in pandemic-like situations.

In [21], the authors have reviewed various technologies, including IoT, drones, 5G, AI, and blockchain, for the Covid-19 pandemic. They have provided a detailed overview of Covid-19 and its impact in various sectors. Similarly, in [14], the authors have discussed the applications of artificial intelligence and blockchain in the Covid-19. In [22], the authors have demonstrated the prospects of vital signs monitoring for quarantined patients using image and signal-processing techniques and deep learning methods. Authors in [23] discussed some applications of blockchain and AI in the health care sector, such as early detection of outbreaks, supply chain management of drugs and other equipment and contract tracing. The authors systematically reviewed the applications, including contract tracing, vaccine monitoring, and pandemic control and surveillance. A systematic literature review has been conducted in [24] to analyze the role of blockchain technology in combating the coronavirus pandemic. The authors show that the blockchain can significantly aid in tackling the Covid-19 pandemic, focusing on the main features of the blockchain. Similarly, a complete review on the adoption of blockchain for handling the coronavirus pandemic has been discussed in [25]. The authors have also detailed a case study on blockchain-based digital vaccine passports. Another study in [26] discusses the advantages of blockchain for combating the Covid-19 pandemic. Authors in [27] review various data analysis methods for health monitoring sensors. They have classified the related research works based on the types of sensors used, i.e., contactless and contact sensors.

In [28], the authors have presented the architecture of blockchain-enabled IoMT and discussed the various solutions offered by blockchain-enabled IoMT to fight Covid-19. The solutions discussed were based on five perspectives; social distancing and quarantine, medical data provenance, smart hospital, tracing pandemic origin, remote healthcare, and telemedicine. Authors in [29] reviewed smart health monitoring systems based on IoT that can efficiently monitor multiple patients remotely in a cost-effective manner. The roles of AI, Robotics, IoT, and blockchain in combating the Covid-19 pandemic were investigated in [30]. It reviews the potential of these technologies to handle Covid-19. The authors in [31] explored the use of AI and its applications in managing the Covid-19 outbreak. Authors in [32] discuss enabling technologies and systems to handle the Covid-19 pandemic. The work mainly focuses on three areas. Firstly, wearable devices suitable for monitoring quarantine and Covid-19 suspects; secondly, sensing systems for detecting the infected and monitored patients; and lastly, telehealth technologies for diagnosing Covid-19 and remote monitoring.

After reviewing state-of-the-art and going through the literature, the requirement of a detailed survey on pandemic patient monitoring system is felt. A complete survey that covers all the aspects related to remote monitoring, storing/sharing platforms and analyzing the monitored data is missing. Thus, we have conducted a comprehensive literature survey to fill the research gap. The proposed study is a complete survey of the work done on pandemic patient monitoring system. To the best of our knowledge, this is the first survey paper that reviews all the works specifically related to remote monitoring system of patients during Covid or similar pandemic-like situations. This paper may play a vital role in assisting professionals and academicians in identifying how emerging technologies such as blockchain, machine learning, and IoT can be used to develop a remote monitoring system. Table 3 summarizes the relative comparison of our work with the existing survey works. Figure 1 shows the scope of our work by highlighting the broader categorization of work.

Table 3 A Comparison of our work with the existing survey works
Full size table
Fig. 1
figure 1

Scope of the survey work

Full size image

In summary, the main contribution of this work is as follows:

  • The article provides an in-depth analysis of the architecture of a remote monitoring system for patients during a pandemic, with a particular focus on its layered structure.

  • A description of IoT, blockchain, and machine learning with their relevance and applicability for remote monitoring systems is provided.

  • A comprehensive survey of various works related to pandemic patient monitoring system is presented.

  • Finally, open issues and challenges in the realization of a pandemic patient monitoring system are discussed along with future research possibilities.

1.3 Organization of the Paper

The structure of the rest of the paper is outlined as follows. Section 2 details a brief background on patient monitoring and enabling technologies, including IoT, WBAN, blockchain, and machine learning. Section 3 reviews all the research studies related to remote monitoring of pandemic patients. Section 4 presents a detailed assessment of the works based on various parameters. Section 5 discusses the issues and challenges. Lastly, in Sect. 6 we conclude our work. Table 4 list out all the abbreviations used in the work.

Table 4 Abbreviations
Full size table

2 Background

This section discusses the overview of pandemic remote monitoring system. Detailed background knowledge of enabling technologies such as IoT, WBAN, blockchain, and machine learning is also given.

2.1 An Overview of Pandemic Patient Monitoring System

Continuous measurement of the Covid patients’ vitals is essential as this gives first-hand information about abnormal conditions [22]. During hospitalization of severe cases or home quarantined patients, these systems provide a tool for healthcare givers to monitor the patient’s health. Primarily, five vitals are monitored, including oxygen saturation, pulse rate, blood pressure, respiratory rate and temperature. The normal range of vital parameters for an adult is shown in Table 5 [34].

Table 5 The normal range for vital parameters
Full size table

Vital signs are the first indicators of patients’ deteriorating conditions in most diseases including Covid-19. The coronavirus has an incubation period of 1–14 days in the human body. Upon incubation, disruption of vitals is observed [34]. One of the first disruptions includes an increase in the body temperature, i.e., greater than 100.4 F. As coronavirus disease is related to lung inflammation, the attenuation of oxygen uptake capacity is also observed, leading to a decrease in the level of saturation, resulting in a reduced supply of oxygen to the body organs. To meet the requirements of oxygen in body cells, the rate at which the heart pumps blood increases; thus, an increase in heart rate is observed. Simultaneously, the body increases the number of breaths per minute to restore oxygen levels to normal, leading to an increased respiratory rate [35]. Thus, the vitals of the body are disrupted upon Covid-19 infection.

Real-time monitoring of pandemic patients through wearable sensor devices [36] is one of the most popular applications of IoMT in medical field. IoMT and Wireless Body Area network collects medical data from biosensors continuously. These IoT devices in healthcare generate a huge amount of sensitive data, whose unauthorized access can lead to dire consequences. Though IoT has shown massive potential in the healthcare sector, ensuring data security is an important aspect. Hence, a secured gateway-based architecture is needed for utilizing IoT in medical fields with a proper authentication system [37]. Many researchers have attempted to develop a secure IoT network between healthcare IoT equipment and cloud services. However, they suffer from challenges like data privacy, single-point failure, system vulnerability, and centralized supervision [38].

Since medical data is highly susceptible, it needs to be stored securely. Blockchain provides an novel way to store and disseminates data in a secured manner [39]. It offers a robust network to form a distributed and secure method for sharing and storing data owing to its properties of immutability, traceability and decentralization. Blockchain enhances the overall security by ensuring accountability and data integrity [40]. First, the blockchain security schemes, including asymmetric encryption/decryption schemes and the digital signature, gives the enhanced protection to the collected data [41]. Second, blockchain has other security mechanisms, including access control and authentication, to provide added security. Moreover, the decentralized nature of blockchain removes the problem of single-point failures and DDoS kind of attacks. Immutability achieved by cryptographic hash functions does not allow data to be changed by malicious users, hence increasing the reliability. The data stored on the blockchain is traceable throughout. Data traceability and non-repudiation properties are ensured by the decentralized consensus algorithms and asymmetric cryptographic mechanisms, i.e., digital signature. The medical data stored in the blockchain should be analyzed so as to understand the symptoms and further decision making regarding the prognosis and diagnosis.

Machine learning along with image processing techniques can be used to find patterns from the patients’ data stored on the blockchain network. This analysis can further assist healthcare workers in their decision-making process. The usage of machine learning also aids in minimal interaction between healthcare professionals and patients as these systems give immediate medical outputs to problems based on historical data, hence accelerating the recovery process and reducing the costs of treatment. ML can also be employed in diagnosis, prognosis evaluation, drug discovery, and epidemic prediction for Covid-19. The learning capability of ML could assist in forecasting future cases and the impact of a potential outbreak. The system can also create alerts for healthcare facilitators, authorities, and families during an emergency. A generalized patient remote monitoring system is shown in Fig. 2. It integrates IoMT, blockchain and machine learning in a remote monitoring system. The architecture is conceptually divided into three layers including, IoT data or data collection layer, blockchain or data storage layer and machine learning or data processing layer. In the next section, scope and contributions of our work is discussed.

Fig. 2
figure 2

A generalized architecture of a pandemic patient remote monitoring system

Full size image

2.2 Enabling Technologies

The rapid advancements in science, technology, medicine, and the usage of smart medical devices in the medical domain have transformed the medical field entirely. This subsection gives a brief background of the technologies, i.e., the IoT, blockchain, and machine learning, that play a vital role in develo** of a pandemic patient remote monitoring system.

2.2.1 Internet of Things

Internet of Medical Things (IoMT) systems are diverse and widely used in the medical domain. The Internet of Things is a network of things or physical devices communicating with each other [42]. IBM defines IoT as, “The Internet of Things is the concept of connecting any device (so long as it has an on/off switch) to the internet and other connected devices. The IoT is a giant network of connected things and people – each of which collects and shares data about how they are used and about their environment”. IoT has bought an immense revolution in various domains, including healthcare, supply chain management and industries. The technology aims to integrate technologies like cloud computing, electronic devices, sensor networks, and mobile services. This technology connects sensors and actuators that are integrated into digital devices linked to the internet, utilizing unique Internet Protocol (IP) addresses to identify these devices through both Transmission Control Protocol (TCP) and non-TCP methods.

Many transmission protocols, like ZigBee, Z-Wave, Long-Range Wide Area Network (LoRaWAN), Wi-Fi and Bluetooth Low Energy (BLE) are used to transmit IoT data [43]. In IoT, no human intervention is required. It means devices automatically transfer and store the sensing information to cloud severs or local servers or blockchain. The collected data is generally processed by fog or cloud servers. Smart gateways can act as Fog or edge nodes. Some functions of smart gateways include collecting and filtering data, preprocessing, and reconstructing data to a proper format. A standard framework for IoT systems is required to overcome the devices’ interoperability, heterogeneity, and diversity issues. Some typical applications of IoT include wireless inventory trackers, smart healthcare, biometric scanners, remote monitoring of patients and smart home security [44]. Remote monitoring of patients in hospitals or at home is an important use case of IoT.

The currently used remote monitoring systems are complicated and bulky as wires are used to establish a connection between the sensors and a wearable wireless transmitter [45]. This system suffers from disadvantages such as patient movement restriction and comfort. Therefore, a monitoring system composed of sensor nodes with wireless capability is required to overcome these issues, as shown in Fig. 3. A special-purpose wireless sensor network, Wireless Body-Area Network (WBAN), which consists of various wireless devices and networks can be used for this purpose [46]. One of the crucial applications of WBAN is in the medical domain for real-time monitoring of several patients. In WBAN, two sensor signals may collide or can interfere with other external wireless devices. WBAN is more easily managed than a traditional wired network [46].

Fig. 3
figure 3

The general architecture of A WBAN [46]

Full size image

The sensors used in WBAN should be low powered and small in size. The architecture adopted in WBAN comprises wireless sensor nodes that transmit a patient’s vital parameters such as heart rate, ECG, and blood pressure via a wireless network. Sensor nodes must meet essential requirements of miniaturization and low-power operation [46].

2.2.2 Blockchain

S. Nakamoto first highlighted blockchain in 2009 by introducing the world’s first cryptocurrency, bitcoin [47]. Blockchain consists of consensus protocol, immutable ledger, mining, hash cryptography, and distributed peer-to-peer networking [9.

Table 9 Comparison of blockchain-based data storage and data sharing platforms
Full size table

3.3 Processing Stored Patients’ Data

Machine Learning has several applications in combating the effects of Covid-19 like pandemics such as detection of Covid-19, prediction of severity level of patients, prognosis of disease and prediction of oxygen requirement. Once the patients’ data is collected using IoT sensors, this data needs to be analyzed for useful purposes. A framework for edge-centric intelligent healthcare, which analyzes measured parameters using smart wearable sensors, has been proposed by the authors in [70]. They employed an advanced machine learning technique called Bag-of-Neural Network (BoNN) for this purpose. The dataset was prepared using various preprocessing techniques and then deployed on the edge network for analyzing the symptoms to predict the severity of the disease. The proposed algorithm ensures minimum training time and higher accuracy of 99.8% on the benchmark Brazil dataset.

Testing is one of the critical steps to fight against coronavirus by identifying the infected persons. The primary difficulty in identifying the Covid-19 patients was the shortage and dependability of testing kits. Hence, healthcare workers faced difficulties in identifying and diagnosing positive cases. This results in an urgent need to develop an autonomous system for diagnosing suspected patients. A user-friendly, time and cost-effective solution is needed to identify Covid-19 patients quickly. Several research works have been proposed in this direction.

A model for automatic diagnosis based on a smart contract node is presented in [109]. The model takes in online-monitored health parameters as input. Coughing is considered a significant factor in identifying whether a patient is infected with Covid-19 or not. To extract features from the coughing audio of suspected patients, Mel Frequency Cepstral Coefficients (MFCCs) are utilized in the algorithm. The detection model is built on the Convolution Neural Network-Long Short-Term Memory (CNN-LSTM) architecture, consisting of convolution and pooling layers for extracting spatial features and LSTM layers for temporal feature extraction. The proposed algorithms ensure user privacy and achieve 90% detection accuracy using coughing signals.

The study described in [110] proposes a model that can detect Covid-19 within a few minutes by using various patient parameters such as temperature, pulse rate, age, and contact with high-risk individuals. The proposed framework utilizes nine classifiers, including Support Vector Machine, Naive Bayes, Random Forest, Quadratic Linear Discriminant Analysis, Decision Tree, Linear Discriminant Analysis, Gradient Boost, K-Nearest Neighbor, and Extreme Gradient Boost. The XGB algorithm exhibits the highest tenfold cross-validation accuracy and the best performance when considering the recall rate vs. decision threshold boundary. The study employs SHapely Adaptive exPlanations analysis to identify the most influential features, which include “cough,” “fever,” “loss of smell,” and “high risk exposure occupation.” To optimize the hyperparameters of the classifier and balance the classes in the dataset, the study utilizes a Bayesian optimizer.

In [98], the issue of preserving patient data privacy on a data-sharing platform during the pandemic is addressed. To solve this problem, the authors proposed a blockchain-based framework that employs federated learning to collect data from multiple hospitals for training a deep learning model. The use of blockchain provides data authentication, while federated learning ensures the privacy of the organizations. To deal with the heterogeneity of the data from different Computed Tomography (CT) scanners at various hospitals, a normalization technique is initially applied. Capsule Network-based segmentation and classification are then utilized to identify Covid-19-positive patients. Lastly, a federated learning-based model is proposed that guarantees privacy and security. The dataset used in this study comprises CT scan slices for 89 subjects. The architecture is depicted in Fig. 11.

Fig. 11
figure 11

Blockchain based federated learning [111]

Full size image

In [102], the authors propose a method for identifying Covid-positive patients using CT-scans and clinical reports, employing an Artificial Neural Network (ANN). The patient data is stored on the blockchain, but only for positive patients. The ANN algorithm takes the CT-scan and clinical reports as input to categorize the patients as positive or negative. The clinical reports are given 30% weightage, while the CT-scan report prediction results are given 70% weightage. In [112], the authors propose an online data monitoring framework that predicts the risk level of Covid patients. The framework aggregates data using edge computing, where local gateways and edge nodes are used for aggregating the monitored data. The authors utilize Random Forest on synthetic data for prediction, which achieves 97% accuracy.

Authors in [111] predicts the patients’ prognosis using the radiomics features of CT scans. In radiomics, features are extracted from the medical images using data-characterization algorithms. More specifically, duration of a patient’s stay in hospital is predicted using the radiomics features of initial CT scans. They have classified the patient into two categories, i.e., short-term stay and long-term stay. This is done so as to give more attention to infected patients with long-term stay. Two machine learning algorithms, Logistic Regression and Random Forest are employed for prediction of patients’ stay in the hospital. Authors in [113] have proposed an federated learning based model for processing medical records. The model uses data from 20 institutes worldwide to predict oxygen demands of Covid-19 infected patients. Some of the criteria for prediction were chest X-rays, vital signs and lab reports.

Among the Covid-19 infected people, acute respiratory distress syndrome (ARDS) is more prone to a life-threatening severe respiratory system failure. Monitoring Vital signs (e.g., heart rate) has been proven very effective for the early detection of different respiratory diseases. Authors in [114] have conducted the study on ARDS patients by monitoring their vital parameters. They observed the daily log of Heart Rate (HR) and Blood Pressure (BP) of 150 ARDS patients for long-term use. For extracting the features and statistical analysis, deep learning is utilized. Federated learning is used to maintain data anonymity where data was collected from multiple sources.

Most of the previous research works have focused on assisting healthcare workers in predicting the severity level of Covid-19. However, there is a lack of research on develo** efficient models for predicting the recovery time of Covid-19 patient. A deep learning model ‘iCovid’ based on multimodal clinical data to predict the recovery time of Covid-19 infected patients was proposed in [115]. The prognostic model development of ‘iCovid’, as shown in Fig. 12, is a regression model which generates the recovery probability distribution for a patient at admission within 48 h. The research demonstrates that parameters like treatment schemes, symptoms, comorbidities, age and biomarkers have high impact in recovery-time predictions. The authors have utilized a large-scale dataset containing data collected from 2530 Covid-19 patients. All the patients were admitted in Huoshenshan Hospital in Wuhan, China. The following parameters were collected from each of the patients; treatment schemes used, clinical features, CT scans, outcome (recovered, censored or decreased), severity-level and time (in days) after which outcome is observed after admission in hospital. iCOVID utilizes treatment schemes, clinical features and lung CT images as inputs. VGG-16 network30 is used to extract convolutional features from the lung CT scans, which are inputs to fully connected layers along with clinical features schemes for prediction of recovery-time of patients.

Fig. 12
figure 12

Prognostic model development [115]

Full size image

To develop an accurate machine learning based model, a huge amount of good quality training data is required. Medical data is present at various locations globally. Transferring of patients’ data across medical centers is a serious challenge as it causes privacy and security issues. A solution to mitigate this issue is to fine-tune the machine learning models locally with the annotations and local data. Usually, the quality and availability of local annotations differ due to the usage of heterogeneous types of equipment and the availability of medical resources at various locations worldwide. In [116], a solution is proposed using federated machine learning and semi-supervised learning. The authors developed an innovative federated semi-supervised learning technique that fully utilizes all the data, including annotations and without annotations. Federated learning deals with the privacy concern of various organizations while sharing sensitive data. Furthermore, semi-supervised learning reduces the pressure of annotating the data in a distributed environment.

H1N1 and Covid-19 infections have similar symptoms and are the most spread pandemic diseases globally. The researchers in [117] have utilized machine learning algorithms to classify H1N1 and Covid-19 patients. The system uses data from 1467 patients, including 70% of the patients infected with H1N1 and 30% of the patients infected with Covid-19 having 42 attributes e.g. temperature, diarrhea, coughing, Sore throat. Authors have demonstrated that machine learning algorithms gave promising results in classifying Covid-19 and H1N1 patients.

Authors in [118] propose a framework that integrates cloud computing, IoT, machine learning, and fog computing to develop a novel intelligent system for Covid-19 disease prognosis and monitoring. IoT devices gather streaming data from medical devices such as lung ultrasound machines and non-medical devices such as smartwatch devices. The authors have proposed a framework that utilizes federated ML as a service (federated MLaaS), a distributed batch MLaaS maintained on the cloud, and a distributed stream MLaaS implemented on a hybrid fog-cloud. The usage of Fog provides reduced latency and enhanced security to the framework.

Various researchers have proposed many machine learning models to diagnose and predict Covid. The existing models are not capable of deciding for the Covid-19 patient immediately and are also not capable of processing multiple sensor data for diagnosis. Thus, a framework for smart health monitoring and prediction called ‘iCovidCare’ is proposed in [119] to solve these challenges. The architecture of ‘iCovidCare’ is shown in Fig. 13. It uses the rule-based approach for quick decision making. Local edge devices are also used for minimizing the delay. For prediction, Ensemble Random Forest (eRF) algorithm is proposed. The trained model is also deployed on edge devices for minimizing the delay and latency. For evaluation synthetic data is used in which most dominant features are extracted using the proposed data fusion and feature selection methods.

Fig. 13
figure 13

Architecture of iCovidCare [119]

Full size image

The Covid-19 pandemic has heightened the sensitivity of public health information. The absence of controlled and reliable media information coupled with the evolving nature of Covid-19 has led to a proliferation of unverified news sources, which has overburdened call centers. To tackle this problem, a privacy-preserving model based on blockchain and federated learning is proposed by the authors in [120]. The proposed model aims to enhance the authenticity and reliability of Covid-19-related news dissemination. The framework is composed of four components, namely individuals, consortium blockchain, Center for Disease Control, and Federated Learning.

3.3.1 Summary and Discussion

Machine learning has enormous potential to help combat coronavirus-like pandemics. Machine learning applications include suspected Covid-19 patients’ detection, classification of disease severity, prediction of the amount of oxygen requirement of patients and prognosis of the duration of hospital stay of the infected patients. Most studies focused on develo** models for detecting Covid-19 and classifying the severity of Covid-19 patients. However, studies on other applications are still lacking. Lack of training data is a challenge for develo** a machine learning model for smart healthcare system.

Further, the authors have not focused on the privacy of data. If data is spread across various hospitals and healthcare centers, it may cause privacy and security issues. Hence federated learning is suggested in the case of Covid-19 data, which ensures the privacy of data and shares only the trained model, not the actual data, with the central server. Also, the performance of a machine learning algorithm depends on the amount and quality of data provided during training. Most of the works consider the storage of patients’ data on a centralized database. Therefore, there a need to focus on utilizing blockchain to store the monitored data. Table 10 lists a comparative study of all the works discussed in this section related to analyzing/processing of Covid-19 patients’ monitored patients’ data.

Table 10 Comparison of data analyzing/processing systems using machine learning
Full size table

4 Discussion by Comparison

This section compares all the works referred in this survey related to remote monitoring of pandemic patients using technologies like the IoT, blockchain, and machine learning. The comparison and analysis are based on various factors, including applications and platforms. This statistical comparison is based on the research works considered in this survey and is shown in Figs. 14, 15 and 16.

Fig. 14
figure 14

Percentage of various blockchain platforms used in storage layer

Full size image
Fig. 15
figure 15

Distribution of applications of machine learning for mitigating Covid-19 in the data analysis layer

Full size image
Fig. 16
figure 16

Distribution of database platforms to store IoT data in the data collection layer

Full size image

Figure 14 shows the percentage of various blockchain platforms used for storing the patient data. Here, notable is that equal number of research studies have utilized Ethereum and Hyperledger to store and share the monitored data. Some research studies have also used their own customized blockchain. As IoT devices collect a massive amount of data during remote monitoring, continuous data is generated as long as the patient is monitored. Efforts must be made to develop blockchain solutions that can handle the over flooding of generated data. The processing and transactional speed of the blockchain system should be is in accordance with the rate at which the data is generated.

The generated data can also be stored and shared using other technologies such as cloud storage [70, 71] or hybrid storage consisting of both cloud and edge servers [70, 76, 126,127,128]. However, the storage of patient data in these systems must be done securely to protect patients’ privacy and prevent unauthorized access.

Cloud storage is a convenient option for storing large amounts of patient data as proposed in several research works such as [70, 70,71,72, 75, 76, 79, 80]. For example the work proposed in [70] stores the data in a central location through the WSN and distributed edge devices. Here, the transition of the data will be from WSN nodes to edge devices and then from the distributed edge devices to remote cloud servers. This means that data is encrypted when it is being transmitted over the internet and also when it is stored in the cloud. In [71], a cloud manager manages the data flow from and to the servers. Cloud manager also handles efficient communication, data storage, and other data related queries. A low-cost and lightweight cloud-based mobile health monitoring system to measure heart rate, oxygen saturation and electrocardiogram of Covid patients was proposed in [74]. In yet another work, authors in [75] proposed an architecture that transfers the patient’s monitored data to cloud storage. The cloud server runs the software to provide useful information about the patient’s conditions. Hybrid storage combines cloud and edge storage to provide scalability and cost-effectiveness while maintaining control over sensitive patient data. By storing sensitive data on local edge servers and less sensitive data in the cloud, hybrid storage offers tailored solutions for organizations and patients [70, 76].

However, cloud based approaches have their own limitations. Cloud storage with encryption can still be vulnerable to cyber-attacks and unauthorized access. On-premise storage using edge servers may be costly and may not be scalable. Hybrid storage requires careful planning and coordination to ensure data is appropriately partitioned and stored. Blockchain technology can address these limitations by providing a decentralized and tamper-proof platform for storing patient data. It enables secure and transparent data sharing between multiple parties while maintaining patient privacy and consent. It also provides an immutable record of all transactions, ensuring data integrity and auditability. These features make Blockchain a promising technology for securely storing pandemic patient data and other healthcare-related data.

  • Decentralization and immutability: The data is stored in a distributed network of nodes, making it extremely difficult for any unauthorized party to tamper with the data. For example authors in [91] stores the Personal Health Records using the Ethereum blockchain and IPFS which stores and synchronizes the data on multiple distributed nodes.

  • Transparency and accountability: Every transaction on the blockchain is recorded and can be accessed by authorized parties, ensuring that the data is trustworthy and accurate. Additionally, the patient can maintain control over their data and grant permission to access it. For example, the work proposed in [93] proposed a blockchain-based secure framework to access and share the medical images in a transparent manner among the patients. A consortium blockchain-based system “HonestChain,” was also proposed in [101] that ensures auto-assurance and auto-auditability.

  • Interoperability: Blockchain allows different organizations to share data easily and securely. This is particularly important during a pandemic, where different organizations and healthcare providers may need to access a patient’s data to provide necessary care. For example works such as [100] and [102] proposed decentralized blockchain-based platforms for sharing covid electronic medical records among medical research institutions, hospitals, patients, government organizations ensuring patients’ privacy is not violated.

  • Privacy: Blockchain can enhance privacy by allowing the patient to store their data anonymously, ensuring that their sensitive information is protected while still being accessible to healthcare providers. For example, in [77], a novel remote patient monitoring approach has been proposed that uses a private blockchain to safeguard patients’ privacy. In a similar work in [94], authors utilizes the blockchain for ensuring privacy of patients.

Figure 15 shows the problem domain in which machine learning has been applied on the remotely monitored data of Covid-19 pandemic. The considered applications are as follows: detection of Covid-19, prediction of recovery time, prediction of patient’s hospital stay, prediction of the oxygen requirement of the patient and classification of disease severity. Here notable is that half of the work is on detecting Covid-19 using medical images like CT scans, x-rays, or clinical data collected from IoT devices. Low coverage attention has been given to other applications, which need to be explored further.

We infer from the survey that research works mainly used central cloud servers or local servers to store the data collected from the IoT devices, as shown in Fig. 16. Training the machine learning algorithms has also utilized the centralized approach, leading to privacy concerns. Hence researchers should focus on building decentralized systems. Blockchain and federated learning promote the decentralization of the whole workflow. These techniques should be utilized to effectively store and share the Covid patients’ data. Federated learning is more suitable as Covid-19 data is generally not present at one central location. Further, these two technologies together have the capability to protect the privacy of the data, which is of top concern in the case of healthcare data.

5 Issues and Future Directions

Healthcare workers are vulnerable to getting infected by the disease during the treatment of the infected persons. Hence the demand for smart remote monitoring systems has increased tremendously after the Covid-19 pandemic. Even though technologies like IoT, blockchain, and machine learning are quite useful and have huge potential in realizing a patient monitoring system, there are several issues and challenges that should be addressed. This section discusses these issues and challenges in detail along with future research directions.

5.1 Scalability of Blockchain Based Platforms

The rapid growth of blockchain technology has acquired huge attention around the globe. Current blockchain systems suffer from the problem of poor scalability [129], which acts as a huge hurdle in the wide-scale adoption of blockchain in healthcare systems. In a blockchain system, transaction throughput is used to measure the scalability of blockchain systems which is the number of transactions per second (tps). Blockchain has very low tps as compared to other matured commercial payment systems and conventional centralized databases. For instance, Bitcoin has a maximum throughput of only 7 tps while PayPal and Visa have a throughput of 170 tps and 200 tps respectively [33]. There should be a significant improvement in throughput and the scalability of blockchains to maintain the massive number of transactions. The different ways to improve the scalability are as follows:

  • Develo** more scalable consensus algorithms: One of the challenges in blockchain technology is develo** more scalable consensus algorithms. Public blockchains tend to have lower throughput than private blockchains when it comes to confirming transactions. Consortium blockchains perform better than public blockchains, but worse than private ones. To overcome this challenge, researchers must explore newer distributed ledgers such as Tangle. Tangle is a chainless ledger implemented using Directed Acyclic Graph (DAG) that is ideal for storing IoT data. Tangle offers high network scalability, transactional throughput, and low transactional cost, as highlighted in [48].

  • Sharding: The processing of transactions in blockchain is sequential which results in significant computation overhead as the network grows in size. To address this, sharding is a technique that divides a blockchain network into smaller partitions or shards. The ledger is also divided into multiple parts, each assigned to a different shard. This allows for faster validation of transactions, as each shard can process its assigned transactions in parallel. By dividing the blockchain into smaller partitions, sharding increases the number of transactions that can be processed by the network simultaneously.

5.2 Dataset Availability

The sudden and vast outbreak of the Covid-19 pandemic overwhelmed the current record-kee** application facility of the healthcare systems. Proper records were not maintained as the frontline workers were more focused on saving lives. Also, the records have been saved in various formats at different locations. Accumulating, aggregating, annotating and formatting enormous healthcare data is a huge challenge [130]. The available data is also not adequate for large-scale machine learning algorithms. Countries are highly reluctant to share their medical data. This phenomenon puts a challenge in evaluating the measures for combating the pandemic. Future research possibilities regarding this issue are as follows:

  • Automated data augmentation and aggregation model can be deployed on edge locations to pre-process the patient monitored data.

  • Efforts should be made to collect data from various locations and build datasets kee** in mind the ethical and privacy issues

  • Interoperability standards need to be designed for data collected from multiple sources.

5.3 Data Security and Privacy

The security and privacy of the patient’s data are of top concern and ensuring it is a big challenge. As vast volumes of data are aggregated, data security becomes a big concern as there is a risk of data breaching and manipulation. Hence a secure authentication system must be developed to secure sensitive data [106]. Data privacy can also be highly affected by transparent and distributed data storage. Privacy laws must always be adhered during data processing, storage, and visualization. Although blockchain has tremendous potential, it is still a relatively new technology. It can preserve certain amount of privacy as users operate their accounts using private and public key, their real identity is hidden but blockchain cannot provide transactional privacy as values of all transactions are publicly visible on the network. Future research possibilities regarding this issue are as follows:

  • Develo** well-defined access control mechanisms: Access control protocols need to be designed so that patients can give access of their health data to various stakeholders such as relatives and doctors.

  • Designing encryption mechanisms: Cryptographic encryption algorithm can be utilized to provide further privacy to the data stored on the blockchain.

  • Mixing: Although transactions in blockchain occur without exposing the identity of the users, it can still be guessed by malicious nodes in the network by analyzing the transactions stored publicly on the network. Mixing provides anonymity by executing transactions from multiple account addresses to multiple output addresses.

5.4 Storage and Processing of Large Healthcare Data

In a remote monitoring system, IoT devices produce a huge amount of heterogeneous health data [131]. Hence, a robust and efficient framework is needed for storing and processing the voluminous data produced at high speed. Preprocessing such as removal of redundant data, filtering, and data pruning also becomes challenging as data size keeps increasing. IoMT data also pose challenges in data analysis as data is heterogeneous. This data analysis is done by machine learning which requires extensive feature extraction. Future research possibilities regarding this issue are as follows:

  • Optimizing the blockchain storage: As the transactions keeps on increasing it becomes harder for a blockchain node to manage the full copy of the ledger. Hence, the solutions should be storage efficient as persistent data storage is involved. Utilizing the off-chain storage systems like IPFS and HDFS can be one of the promising way to tackle the above issue.

  • Usage of Machine Learning Models: Different ML models can be used for different types of data such as deep CNNs for X-rays and MRI, and RNN for time-series and EEG data.

5.5 Lack of Training and Required Skills

The doctors and healthcare workers may be hesitant to adopt these new technologies over the traditional paperwork system in a remote monitoring system. Many technical difficulties would be inevitable and thus hamper the development of such systems due to lack of knowledge and required skills. Future research possibilities regarding this issue are as follows:

  • Efficient and User-friendly technologies can be developed with well-defined access control mechanisms and mobile/web interfaces.

  • Proper demonstrations and user manuals can be prepared to implement the technologies.

6 Conclusion

The coronavirus pandemic has exerted extreme pressure on the existing healthcare systems. Scientists and researchers are striving continuously to provide solutions for humanity to be better equipped for such emergencies in the future. Smart remote healthcare monitoring is essential for combating pandemics like Covid-19. This survey discusses the role of IoT, blockchain and machine learning in the remote monitoring of pandemic patients. It also details a comprehensive survey on how these technologies are utilized for conceptualizing and designing such systems. The relevant works in the domain are categorized into three sub-domains; remote monitoring of pandemic patients using IoT, storing or sharing patients’ data securely using blockchain, and processing stored patients’ data using machine learning. A comparative study of all the related research works in the respective sub-domains has also been done. The comparative study considers various parameters, viz. objective, merits and demerits of the work, dataset used and applied scheme. Finally, various challenges and future research directions have been given. As there is an urgent requirement for a smart remote monitoring system, this survey work will be helpful in assisting researchers to develop such a system.