Sage Journals: Discover world-class research

Abstract

Objectives

Telehealth monitoring applications are latency-sensitive. The current fog-based telehealth monitoring models are mainly focused on the role of the fog computing in improving response time and latency. In this paper, we have introduced a new service called “priority queue” in fog layer, which is programmed to prioritize the events sent by different sources in different environments to assist the cloud layer with reducing response time and latency.

Material and Methods

We analyzed the performance of the proposed model in a fog-enabled cloud environment with the IFogSim toolkit. To provide a comparison of cloud and fog computing environments, three parameters namely response time, latency, and network usage were used. We used the Pima Indian diabetes dataset to evaluate the model.

Result

The fog layer proved to be very effective in improving the response time while handling emergencies using priority queues. The proposed model reduces response time by 25.8%, latency by 36.18%, bandwidth by 28.17%, and network usage time by 41.4% as compared to the cloud.

Conclusion

By combining priority queues, and fog computing in this study, the network usage, latency time, bandwidth, and response time were significantly reduced as compared to cloud computing.

Keywords

Cloud computing fog computing telehealth priority queue clinical decision support system

Introduction

The internet of things (IoT) is a network pattern that includes many interconnected tools in this network. These tools are called things. Wearable wireless sensors, and mobile devices are examples of these things.¹ It is estimated that by 2030 the number of IoT devices will exceed the number of people on earth.² Due to the increasing number of medical devices, sensors (ingestible sensors, invasive sensors, strip sensors, implantable sensors), and mobile devices that connect via the internet, the IoT has turned its attention to the field of health care.³ The advent of IoT with low-cost, high-performance sensors has accelerated the use of home healthcare monitoring.⁴ Because IoT allows patients and healthcare providers to transfer the treatment process to the patients’ side, it enables patients to manage the disease and receive help from the healthcare team remotely via mobile devices. This technology has been considered a promising solution for the healthcare industry to improve the quality of care.² IoT-based healthcare systems improve monitoring efficiency and provide adequate emergency services to patients.⁵ Sensors and smart devices used to monitor the patient have computational and storage capacity limitations, so they cannot manage the large amount of data generated. The cloud computing approach has been adopted to address such IoT challenges.⁶ The cloud computing model provides the ability to store, process, and manage large volumes of data.^7,8 Despite the above advantages, due to the centralized nature of this technology, it is deprived of services with minimal delay, location awareness, and geographical distribution, which is critical for IoT applications.^9,10 Because the cloud computing approach requires a constant internet connection¹¹ at reliable speeds with sufficient bandwidth,¹² in cases where bandwidth is insufficient and network interruptions occur, cloud delays increase. It is not easy to provide health care services.¹³ To provide health care services where it is necessary to make decisions and respond to events with minimal delay, another processing model is called fog computing, a generalization of cloud computing.¹⁴ It is introduced by the CISCO company.¹⁵ In this model, sensitive data is analyzed at endpoint IoT devices, and based on the defined policies, part of the data is sent to the cloud layer. This leads to improved use of bandwidth and low latency.¹⁶ Fog computing as a virtual platform built into the edge of the network that provides computing, storage space, and network-based services between endpoint devices and the cloud server.¹⁷ Providing health care services is highly dependent on network connectivity. Network failure may cause interruptions and delays in providing health care services, leading to unpleasant consequences for the patient’s quality of life and even death.

Telemedicine is defined as providing telemedicine services on telecommunications infrastructure.¹⁸ Telemonitoring is a subset of telemedicine. Designing advanced telehealth monitoring systems is an open challenge and still an active research topic.¹⁹ Due to the rapid growth and development of IoT web services, users are facing an information explosion.²⁰ Therefore, we are witnessing the growth of the development of decision support systems (DSSs) in various fields, such as health care which is named clinical DSSs (CDSSs).²¹ CDSSs can improve and enhance the quality of care.²²

Cloud-based CDSS architectures have been developed to process data collected by sensors and telehealth monitoring devices to predict and diagnose acute clinical conditions. These architectures allow real-time data collection analysis minimally and with low latency.²³ Cloud-based DSS architectures suffer from a lack of location awareness, mobility support, and low latency. By performing data processing on the edge of the network, fog computing architectures have improved and enhanced the traditional cloud-based DSS architectures to overcome these limitations.²⁴

In telehealth monitoring systems, response time is a critical measurement. Fog nodes send all information to the cloud server for processing based on the input time. In the standard models, there is no difference between emergency and non-emergency requests to receive services from the cloud server. In the current fog models, much information is sent to the cloud server for processing since only some of this information is in an emergency and has priority for processing.

In this paper, a telehealth monitoring model based on fog computing with the ability to support CDSS on the cloud layer is developed and evaluated. To maintain the efficiency of the model, we used priority queues as a service on the fog layer, which was considered very effective in improving the response time to emergency requests.

The key contributions of this paper are listed below:

• Proposing an innovative model for developing priority queues on fog computing.

• Evaluation of the proposed model using CDSS of diabetes disease.

• Evaluation of the proposed model regarding network usage, latency time, response time, and bandwidth.

Background

Recently, we have witnessed the development of IoT-based telehealth monitoring architectures and fog computing.^2,5,6 These architectures are often divided into three layers: cloud, fog, and edge.²⁵ The clinical decision-making computing section in these architectures is divided into lightweight computing and heavy computing tasks.²⁶ According to the servers’ processing capability, lightweight computing task is deployed on the fog server, and heavy computing task is deployed on the cloud server.

Lightweight calculations consist of comparing data collected and sent from sensors with a threshold value to detect abnormal conditions and send an alert about the person being monitored, or the lightweight data mining method is used on the fog server to monitor the health status of patients in real time.

Among the processing and preprocessing used in the fog server in various studies can be cited the encryption of patient data to increase data privacy, feature extraction, normalization of data received from sensors, deletion of noise data received from bio-signals,²⁷ diagnosis of abnormal conditions, clinical decision making, health status monitoring, disease prediction,²⁸ data filtering, trimming, data analyzing and reconstruction.²⁹ According to the results of the study,³⁰ there are three general categories related to the classification of applications of fog computing in the healthcare industry. These three categories are patient monitoring, environmental monitoring, and education and reference, of which the most common application of fog computing has been identified in patient monitoring. This study also categorizes the features of health solutions based on fog computing into low latency, security, privacy, streaming, and real-time application, reducing the dispatch of information on the network, location awareness, heterogeneity, interoperability, storage, and computing, geographical mobility distribution, energy efficiency, and a large number of nodes wireless access. Among these, low latency and a strong presence of streaming and real-time applications have been identified as the main reasons for using fog calculations in health care. Because patient and environmental monitoring applications are sensitive to latency compared to other domains, the growth of fog computing has occurred in the monitoring domain.³¹ Several studies have been conducted to use fog computing to develop efficient health monitoring systems.³¹ Home quarantine during a disease pandemic is one of the applications of this technology. Doctors can guide patients with the right decisions by monitoring a person with Coronavirus disease 2019 (COVID-19)in real-time.^32,33 For example, El- Rashidy et al. proposed an end-to-end deep-learning framework for monitoring COVID-19 patients in real-time. The proposed system has been declared effective regarding quarantine and remote care of patients with COVID-19.³⁴ Smart e-Health gateway can be presented with higher-level services such as data processing, local storage, embedded machine learning, and real-time local data processing at the edge of the network. In this regard, Rahmani et al. present an intelligent electronic health gateway based on fog computing. They also implemented an IoT-based early warning score health monitor to demonstrate the effectiveness and relevance of the proposed system in conducting a medical case study. The results of this study show an increase in intelligence, energy efficiency, mobility, performance, interoperability, security, and reliability in the use of fog computing.³⁵ Verma and Sood used data mining techniques to classify patients’ status, distributed storage and informed on the edge of the network to smart monitor patients at home. Event triggering based data transmission methodology has been used to process patient data in real-time.³⁶ As another application, machine learning techniques are used in the fog layer to process physiological data. Barik et al. present a mist computing-based architecture that uses deep-learning techniques to analyze physiological data collected by diabetic patients’ smartwatches. The proposed model is based on big medical data processing, and the results of this study showed that mist computing is useful for analyzing big medical data, especially in telehealth monitoring.³⁷ Liu et al. present the HPCS framework for privacy protection in fog computing decision support systems. In this architecture, a lightweight data mining method is used on the fog server to monitor the health status of patients in real-time. Newly discovered abnormal signals can be sent to the cloud server for more accurate predictions while maintaining privacy. Specifically, they developed a new secure internal outsourcing protocol for fog servers to achieve a lightweight, secure monolayer neural network. Also, a privacy-protection polynomial calculation protocol allows the cloud server to perform activation functions on a multilayer neural network confidently. The evaluation of the study confirms the achievement of real-time and high-accuracy health monitoring goals without compromising privacy.²⁸ The applications of fog computing are not limited to monitoring chronic diseases. For example, Bandopadhaya et al. have proposed a healthcare monitoring architecture for soldiers deployed in adverse environmental conditions. This architecture implements time-series pattern analysis based on data filtering at the fog layer.³⁸ Various wireless connectivity solutions are used for health monitoring systems based on fog computing. Kharel et al. presented an intelligent health monitoring system based on fog computing that used low-power wide-area network (LPWAN) technology. This model is suitable for monitoring the condition of patients in places that do not have access to the internet.³⁹

Heart patient telemonitoring is one of the areas that benefited from the benefits of fog computing to monitor the condition of patients. In Scirè et al. study, data processing and analysis tasks were transferred to the edges of the network to increase the lifespan of wearable devices. In this paper, new, lightweight algorithms are developed to analyze and interpret electrocardiogram (ECG) sensor data that can be implemented in a processor embedded in a wearable device to detect heart rate and arrhythmia classification.¹⁹ Gia et al. proposed a low-cost health monitoring model to collect the health information of heart patients. Additionally, sensor nodes monitor and analyze the ECG in real-time to analyze the heart patient’s data professionally.²⁷ Gia et al. proposed an IoT-based health monitoring system that has been used in fog computing for the ECG feature extraction system as a lightweight process. The evaluations show that using fog computing in intelligent gateways compared to cloud computing will minimize the transmission delay.⁴⁰

Table 1 shows the strengths, weaknesses, and descriptions of the architecture and objectives of previous studies. In most of these studies, only some performance evaluation measurements are evaluated, and some necessary measures, such as response time, latency, and bandwidth, are not assessed.

Table 1.

Limitations and strengths of related studies.

Authors	Objectives	Design methods	Limitations and strengths
Rahmani et al.³⁵	Presented a proof of concept implementation of an IoT-based remote health monitoring system which includes a demonstration of an intelligent e-Health gateway called UT-GATE	By exploiting several UT-GATEs, an intermediary processing layer was formed to demonstrate the fog computing concept for IoT-based healthcare systems.	⁃ Lack of evaluation of response time and latency time. ⁃ The system demonstration includes all the data flow processes from data acquisition at sensor nodes to the cloud and end users.
Scirè et al.¹⁹	Presented an end-to-end system that is suitable for execution within a resource-constrained wearable device	The system processes the ECG sensor data automatically detects the heartbeats, and conducts an arrhythmia classification using a combination of supervised-learning approaches	⁃ Does not evaluate the quality of service criteria and performance evaluation. ⁃ Evaluate the performance of the system in comparison with various existing approaches in terms of achieved accuracy in the detection of abnormal events.
Gia et al.²⁷	Proposed a low-cost health monitoring system that provides continuous remote monitoring of ECG together with automatic analysis and notification.	The sensor nodes collect and wirelessly transmit ECG, respiration rate, and body temperature to an intelligent gateway that appropriate caregivers can access.	⁃ System does not evaluate by performance evaluation. ⁃ Systems represent the collected data in valuable ways, perform automatic decision making and provide real-time notifications.
Liu et al.²⁸	Proposed a framework for a hybrid privacy-preserving clinical decision support system in fog–cloud computing.	Design a new privacy-preserving clinical decision support system that can be used for health status monitoring and disease prediction in the fog layer.	⁃ System does not evaluate by any performance evaluation parameters. ⁃ Use a lightweight data mining method on the fog server to monitor patients' health conditions in real-time.
El-Rashidy et al.³⁴	Proposed an end-to-end framework for monitoring COVID-19 patients	Convolutional neural network-based deep learning model and fog computing	⁃ Lack of evaluation of the quality of services. ⁃ Proposed model with high detection accuracy.
Verma et al.³⁶	Proposed layered framework for fog-assisted intelligent home patient monitoring	use data mining techniques to classify patients' status and inform on the edge of the network	⁃ Lack of evaluation using main parameters of quality of service (QoS). ⁃ Proposed model for classifying patients based on condition statutes.
Bandopadhaya et al.³⁸	Proposed a healthcare monitoring architecture for the soldiers deployed in adverse environmental conditions	The step-wise filtration of the redundant information at the WSBAN node and fog node improved channel utilization and reduced the computational burden on the cloud.	⁃ The article does not evaluate the quality of service criteria. ⁃ The layer-wise filtering approach removes unnecessary data, resulting in a fast system response time.
Gia et al.⁴⁰	Proposed an efficient IoT-enabled healthcare system architecture that demonstrates the effectiveness of fog computing in IoT-based healthcare systems in terms of bandwidth utilization, QoS assurance, and emergency notification.	Designed a lightweight algorithm that is suitable for requirements of fast computation and low hardware resources consumption to extract P wave, T wave, and heart rate from original ECG signals for keeping track of the heart’s activities and verifying the smart gateway’s warning service	⁃ The article does not evaluate response time and latency measures. ⁃ Introduced a flexible, lightweight template for ECG feature extraction
Kharel et al.³⁹	The proposed intelligent health monitoring system architecture is based on fog computing exploiting LoRa as connectivity.	Using fog computing alongside LoRa as connectivity remotely monitors users’ biometric information kilometers away.	⁃ In this article does not evaluate the quality of service criteria. ⁃ The remote monitoring requires no internet connectivity and less power for data transmission.

Methodology

In this study, the main event simulation capabilities in CloudSim are used to execute iFogSim architectural functions. Therefore, events between the fog computational components in iFogSim are performed by the CloudSim core layer, and the main CloudSim layer is responsible for managing the various functions among the fog computational components in iFogSim. It is created by simulating services and entities. The configuration (GHz CPU, RAM size, and Power) of the various fog devices are described in Table 2.

Table 2.

Configuration details of different fog nodes.

Device type	CPU GHz	RAM (GB)	Power (W)
ISP gateway	4.0	6	117.445
Smartphone	2.6	2	88.64
Wi-Fi gateway	4.0	6	117.445
Cloud VM (virtual machine)	4.0	6	117.445

Cloudlets were also used to understand patient data and execute requests. To run fog programs, only one host is provided for each node to provide resources. With the help of fog computing, the cloud environment processes the patient’s data on the fog nodes. Due to the closeness of the IoT devices to the fog nodes, this process causes the network bandwidth to be saved and the delay time and response time to be reduced.

Dataset

The dataset used in this study is a subset of the Indian data set called PIMA, which was collected from the UCI University machine learning repository.⁴¹ The data set consists of several independent prediction variables and a dependent target variable. Independent variables include frequency of pregnancy, body mass index (BMI), insulin level, age, etc. PIMA consists of 768 samples, of which 268 are diabetic, and 500 are non-diabetic. The valuable features in predicting diabetes are shown in Table 3.

Table 3.

PIMA data set description.

Feature name	Type	Values range	Description
Times pregnant	Integer	0–17	Frequency of pregnancy (FP)
Plasma glucose	Integer	0–199	Values obtained from oral glucose tolerance test (OGTT)
Diastolic BP	Integer	0–122	Diastolic blood pressure (DBP)
TRI skin fold	Integer	0–99	Triceps skin fold thickness (TSFT)
Insulin	Integer	0–846	Values obtained from 2-h serum insulin (2HSI)
BMI	Number	0–67.1	Body mass index (BMI)
D_PED	Number	0.078–2.42	Diabetes pedigree function (DPF)
Age	Integer	21–81	age

Proposed model

Figure 1 shows the flow of different entities in the home environment using fog computing. The fog layer can retrieve information about the patient’s health history from the cloud in the intelligent communication mechanism. On the other hand, in conventional communications, updates related to the actions taken by the node are transferred to the cloud along with the patient’s data for future actions. To improve the understanding of the dimensions of the proposed model, we have provided a layering approach to the smart home remote control model based on the fog shown in Figure 1. In this model, by combining different IoT devices, sensors, and other internet-connected hardware devices, a sensor network is formed that can record various patient-centered events efficiently. The main goal of this model is to monitor patients who need special remote care using IoT-based technology. The fog layer consists of fog nodes located at the edge of the network.

Figure 1.

Proposed model based on fog computing for patient monitoring in smart home.

In addition, fog characteristics such as real-time interactive services and mobility and scalability support can be an optimal choice in the IoT-based telehealth monitoring environment. The layering approach in the proposed model for patient telehealth monitoring based on fog consists of three layers, data acquisition, fog, and cloud. The three layers offer seven services:

(1) Data collection service (2) data pre-processing service (3) event classification service (4) event prioritization service (5) decision-making service (6) cloud storage service (7) update of the fog servers. Each service performs its own defined essential function, which is used as an input to adjacent services. All three layers are described below:

Data acquisition layer

In this layer, data from IoT devices related to the patient’s health status and home environment are widely obtained from various wireless hardware devices embedded in various locations in the home and the sensor network embedded in the patient’s body or wearable sensors. Each sensor node is integrated with biosensors and other medical sensors. Physiological and environmental parameters of a patient are sent to the fog layer in text, graphic, and numerical form.

Fog layer

The data sent by the sensors will be stored in this layer, which is referred to as local storage. Storage is necessary for local processing in the fog layer. Other operations performed on the data stored in the fog layer are data encryption based on size, type, and sensitivity. This service aims to increase the confidentiality of data. Another fog layer service is data standardization based on various health informatics standards. The data filtering service in this layer clears the data before analysis based on various filtering and noise removal techniques. Data fusion service is used to remove additional data, replace new data, and reduce the volume of sensor data. The data analysis service in the fog layer processes the sensor data locally, which reduces response time, sends data to the cloud, and improves system performance.

Compression methods without data loss are used in the data compression service to reduce the amount of data transmitted over the network from the fog layer to the cloud. The physical proximity of the fog layer to the body area network (BAN), sensors, and stimuli allow us to meet the needs of IoT health care. Some of these services are public and useable for any other than the healthcare industry. Figure 2 shows the fog layer services in the proposed model.

Figure 2.

Proposed model layer services.

In addition to the services mentioned in the fog layer to increase the data transfer speed between the fog and cloud layers, reduce the bandwidth used to send data, and reduce response time, three other services have been proposed.

Pre-processing service

One of the fog node’s tasks is to convert the received data into the appropriate format before sending it to the cloud layer. Also, fog nodes are designed to synchronize all data collected from heterogeneous devices over a universal time stamp. This service provides computing, networking, storage, encryption, compression, filtering, data standardization, and fusion.

Events classification service

The data sent by the sensors is stored in real-time in the fog nodes. The fog layer performs the data evaluation process by interacting with other fog nodes and services in each fog node for analysis. In this process, the data are compared with the normal values range predetermined by physicians. Heterogeneous data collected from different medical devices before classification are converted to the appropriate format. The normal values range of different parameters for classifying events is defined in the fog node. During the classification, the event is placed in one of the two classes: the normal event or abnormal event. The class of abnormal events includes those events that can exacerbate the patient’s health. A sensitive or abnormal event indicates that one or more parameters of a patient’s collected data were not within the normal range. It is a normal event that all the parameters collected are in the normal range and do not harm the patient’s health. Each range of normal values is considered a threshold to classify patient data into two abnormal or normal classes. In this way, only if the threshold is violated will the data be sent to the cloud for review, thus reducing the data sent to the cloud and saving bandwidth consumption.

Queuing service

The priority queue is a specific type of queue. Each queue item has additional information that indicates the priority of that item. Unlike a regular queue, values in the priority queue are listed on a priority basis instead of the first-in-first-out (FIFO).

In Figure 3, A and B, etc. denotes the value of items while 1, 2, etc. denote the priority of items. A priority queue is an extension of a queue that contains the following characteristics:

- Every element in a priority queue has a priority value associated with it.

- The element with the higher priority will be moved to the top and removed first.

- If two elements in a priority queue have the same priority value, they will be arranged using the FIFO principle.

Figure 3.

Priority queue basics.

Using priority queues allows higher-priority requests to be received and processed faster than lower-priority requests. Priority queues are helpful in applications that offer different service level agreements to customers. Priority queues are widely used in the real world:

- Algorithms: Certain foundational algorithms rely on priority queues, such as Dijkstra’s shortest path algorithm, prim’s algorithm, heap sort algorithm, etc.⁴²

- Data compression: It is used in data compression techniques like Huffman code.⁴³

- Operating Systems: Priority queues select the following process, ensuring high-priority tasks run before low-priority ones. It is also applied for load balancing and interrupt handling.⁴⁴

- Bandwidth Management: Priority queues are utilized to prioritize the critical data packet, so the network can make sure that those packets reach the destination as quickly as possible.⁴⁵

Our study used a priority queue to manage bandwidth and improve patient response time.

According to Figure 3, if a patient’s condition is detected as abnormal by the event classification service, it should be sent to the cloud for decision-making by CDSS, and relevant results should be sent to the target groups (health care team, patient, and patient family) with minimal delay. In some diseases, such as heart disease, responding to the event sent in the shortest time or real-time is necessary.

Response time will be affected when multiple events are sent to the fog node from different patients and environments. Queuing service is responsible for providing events in which one or more patient-related parameters are detected outside the normal range for sending to the cloud layer. However, when the number of these events increases, a queue of ready-to-send events will be formed. The queuing service must prioritize the events based on the critical level. Events in which a patient’s critical condition is detected will be a priority for sending to the cloud layer.

The patient’s submitted data in the proposed model will be mentioned as an event. Each event will contain different patient data that will be consistently collected and sent to the fog node during the periods specified by the sensors. The data of each event will be referred to as the parameter. For each parameter, normal, high, very high, low, and medical professionals will determine very low ranges and each range according to the importance and risk of that range for the patient by the specialist physician using Table 4 in the range [0–1] will be valued. When one or more parameters related to an event are in an abnormal range, this event should be prioritized using Table 4 for abnormal parameters.

Table 4.

The valuation of abnormal parameters.

Parameter range	Value
Low	ε to 0.4
High	ε to 0.4
Very low	0.4 to 1
Very high	0.4 to 1

For this purpose, by using equation (1), the total value of the abnormal event parameters will be calculated.

w_{i} = \sum_{k = 0}^{n} {a b n o r m a l p a r a m e t e r}_{k}

(1)

Equation (1): Calculating the value of each event based on abnormal parameters

In equation (2), to unify the weight of events containing abnormal parameters, by using equation (2), we will normalize the obtained weights to the range [0, 1].

w e i g h t ({e v e n t}_{i}) = \frac{(w_{i} - \min ({a b n o r m a l p a r a m e t e r}_{k})) \times (1 - ε)}{(w_{i} - \min ({a b n o r m a l p a r a m e t e r}_{k})) + ε}

(2)

Equation (2): Calculating the weight of abnormal events

In the following, an example of the application of the proposed approach is explained.

In the fog layer, each of the features is compared to normal values. In the above event (Table 5), the values of plasma _glucose, diastolic_ BP, and values obtained from 2-h serum insulin parameters, which normal range is 140–200, less than 90, and less than 166, respectively, were detected as abnormal. Due to sending several events to the fog layer, emergency events should be identified and prioritized for sending to the cloud. Based on the proposed model, abnormal symptoms are evaluated using uncertainty theory. This valuation is based on the opinion of experts that has already been prepared (Table 4). According to Table 4, plasma _glucose was 0.25, diastolic_ BP was 0.33, and values from 2-h serum insulin were 0.39. Using equation (2),

w e i g h t ({e v e n t}_{34}) = 0.79

was calculated.

Table 5.

Instance data sent to the fog layer (record #34).

Feature name	Value
Times pregnant	6
Plasma glucose	92
Diastolic BP	92
TRI skin fold	0
Insulin	0
BMI	19.9
D_PED	0.188
Age	28

Cloud layer

In the fog layer, the patient’s health status is evaluated. This evaluation shows that the patient is in normal or abnormal status. When a patient’s health status is classified as normal by the fog server, his/her historical data is stored in an electronic health record (EHR) on the cloud layer at regular intervals to control network traffic. On the contrary, when the patient’s health status is assumed to be abnormal, the patient’s data will be sent to the cloud server for decision-making.

Decision-making services

This service is located on the cloud layer. When the patient’s received data is abnormal on the fog server, it is sent to the cloud layer for decision-making. In this layer, the data is delivered to the decision-making service. By using the presented model in this study and placing CDSS of diabetic disease in the decision-making service, it is possible to monitor patients' health status. The CDSS examines the data and reports on the patient’s health status. The report is sent to the healthcare team, the patient, and the patient’s family.

Cloud storage service

This service is important in receiving and storing patients' health data from different fog nodes. The cloud storage service provides the required data when the CDSS requires historical data stored in an EHR. Also, the health care team for the final prescription, in addition to the evaluation results of the CDSS and data received from the patient, the historical data stored in EHR are required, which is provided by this service.

Fog update service

This service is responsible for adding new parameters and editing the value of the parameters set by the medical experts. When a new parameter, such as glucose, needs to be added, the process of defining the parameter and determining its normal range occurs on the cloud layer. This new information should also be applied to fog nodes for use. The fog update service is responsible for updating the fog nodes and the value of the intervals of each parameter. By determining the value of each parameter range by medical experts, it will place in the fog layer for further exploitation.

Results

Model evaluation

Different requests were used to evaluate the performance of the proposed model (Table 6). These requests are a combination of normal and abnormal class requests. Abnormal classes have higher priority and must be treated urgently.

Table 6.

Some parts of the patient’s requests.

Number of request	Request class
Number of request	Abnormal	Normal
10	6	4
50	25	25
130	48	82
210	78	132
290	109	181
370	140	230
450	171	279
530	187	343
610	210	400
690	236	454
768	268	500

Figure 4 shows the calculation of network usage time for a different number of patient requests. This figure shows that the amount of network usage time in the fog-based model is significantly lower than in the cloud-based model because the fog-based model reduces patient requests to the cloud layer.

Figure 4.

Network usage time versus number of patient requests.

According to this result, the fog-based model reduced the network usage time by an average of 41.4% and 29.05 s compared with the cloud-based model. Based on the proposed model, only requests that have one or more patient parameters out of the normal range will be classified in the abnormal class and will be with one or more patient parameters out of the normal range will be classified in the abnormal class and recognized as eligible for network usage by the fog server.

Figure 5 compares latency time in a fog-based model to a cloud-based model with a different number of patient requests.

Figure 5.

Latency versus number of patient’s requests.

The fog-based model performs better than the cloud-based model in terms of latency. Compared to the cloud-based model, the fog-based model reduced the average latency by 9.1 s (36.18%).

Figure 6 shows the response time comparison in the fog-based model with the cloud-based model with a different number of patient requests.

Figure 6.

Response time versus number of patient requests.

Since only requests in which the patient’s condition is diagnosed as urgent are sent to the cloud layer, the response time is significantly improved by 9 s (25.8%).

Figure 7 shows a comparison of the bandwidth used in the proposed model and the cloud-based model. The proposed model reduced the use of bandwidth by 28.17%. Ten requests were sent in the first evaluation, from which six were detected with abnormal symptoms. Due to the 29-byte volume of each request, the proposed model used 174 bytes, and the cloud model used 290 bytes of bandwidth to send requests.

Figure 7.

Bandwidth versus number of patient requests.

This study used CDSS extracted based on the PIMA Indian diabetes dataset in study⁴⁶ to evaluate the proposed model. Some of the extracted rules are shown as follows:

- R1: If (FP between 4 to 6) AND (OGTT between 78 to 98) AND (DBP between 56 to 76) AND (TSFT between 11 to 31) AND (2HSI between 13 to 33) AND (BMI between 21 to 27) AND (DPF between 0.31 to 0.35) AND (AGE between 28 to 31) Then Non-Diabetes

-R2: If (FP between 11 to 14) AND (OGTT between 135 to 155) AND (DBP between 72 to 92) AND (TSFT between 10 to 20) AND (2HSI between 100 to 120) AND (BMI between 20 to 26) AND (DPF between 0.23 to 0225) AND (AGE between 55 to 59) Then Non-Diabetes

After uploading the rules of this CDSS to evaluate the performance, the requests of patients diagnosed abnormal by fog nodes will be sent to the cloud server for review based on priority.

In this evaluation, out of 7798 submitted requests, 2801 were detected as abnormal by fog nodes and sent to the cloud layer for a more detailed review. Based on this evaluation, CDSS on the cloud layer successfully identified the condition of 658 patients and sent the correct recommendations and warnings to the healthcare team, the patient, and the patient’s family. In this evaluation, the accuracy of cloud-based CDSS was 85.7% (Table 7).

Table 7.

Evaluation of the CDSS accuracy on the cloud layer.

CDSS	Wrong answer	Correct answer	Accuracy, %
PIMA	110	658	85.7

Discussion

In this paper, we have introduced a new service called “priority queue” in the fog layer, which is programmed to prioritize the events sent by different sources in different environments to assist the cloud layer with reducing response time and latency. Priority queue service was considered very effective in improving the response time to emergency requests. Using the proposed model, the incoming events are weighted; those with higher weights indicate a critical health status and require emergency health services to be delivered immediately. Therefore, higher-weight events will have a higher priority in the delivery queue and will receive faster access to medical services. By combining CDSS, priority queues, and fog computing in this study, the network usage time, latency, and response time were reduced by 37.8%, 38.2%, and 25.8% compared to cloud computing.

With the spread of chronic diseases in the present century, the need to develop advanced infrastructure to manage these diseases as an open challenge remains. One of the trends in recent research in this field is fog-based architecture. CDSS is a service of these architectures for processing data collected by wearable sensors and monitoring tools for predicting and diagnosing acute clinical conditions. The current fog-based telehealth monitoring models are mainly focused on the role of fog computing in improving latency. Recent works have provided models for reducing the latency impact and executing applications that use fog and cloud resources more efficiently. Using priority queues for analyzing patients' status and their effect on response time in fog computing has not been studied until now (Table 8). According to Table 8, although the response time criterion is essential in providing telehealth monitoring services, most existing research has not reported a result for this evaluation.

Table 8.

Comparison of existing fog computing models with the proposed model.

Study	Prediction system	Priority queue	Performance parameters
Study	Prediction system	Priority queue	Network usage time	Power consumption	Execution time	Latency	Network bandwidth	Response time
Rahmani et al.³⁵	—	—	—	55.7%	—	80.8%	74.1%	—
Scirè et al.¹⁹	95.2%	—	—	—	—	—	—	—
Gia et al.²⁷	—	—	—	20%–50%	—	10%–50%	40%–80%	—
Liu et al.²⁸	99.5%	—	—	—	—	—	—	—
El-Rashidy et al.³⁴	97.95%	—	—	—	—	—	—	—
Verma et al.³⁶	85%	—	—	—	—	—	—	—
Gia et al.⁴⁰	—	—	—	—	—	—	90%	—
Azimi et al.⁵	—	—	—	—	—	—	—	16.45%
Azimi et al.⁴⁷	96%	—	—	—	72.8%	—	—	19.04%
Gill et al.⁴⁸	—	—	22%–26.7%	23.56%	—	19.5%–29.4%	—	—
Moosavi et al.⁴⁹	—	—	—	26%	—	16%	—	—
Our study	85.7%	✔	41.4%	—	—	36.18%	28.17%	25.8%

The use of priority queues to identify patients with emergency status has made it possible to provide minimum service delay. In the present study, leveraging priority queues in the fog layer and allocating the required bandwidth between the fog and cloud layers significantly improved response time. As a result, requests of patients with acute conditions were more quickly and effectively handled.

Conclusion

The provision of telehealth services is highly dependent on the network connection. Network failure may lead to interruptions and delays in the provision of health care services and have unpleasant consequences for the patient’s quality of life and even death. Because fog computing develops at the edges of the network, it can provide many features, such as support for user mobility, geographic density distribution, and minimum latency and response time. This technology is essential for providing emergency services sensitive to delay. However, fog computing cannot substitute for the cloud, and interaction and cooperation between them are necessary to analyze large amounts of data and implement distributed health care. The results of this study showed that the use of priority queues improved response time and latency in telehealth monitoring systems. The evaluation of the proposed model is based on file-based input data as a limitation that can be received directly from sensors in the natural environment. Also, the performance of the proposed model was evaluated using the iFogSim tool, which can be evaluated in a natural cloud environment in future works. In this study, the proposed model was evaluated with diabetes data. It is recommended that this model be used for other diseases, such as heart disease which response time is crucial for decision-making.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Ethical approval

The study protocol was approved by the Shiraz University of Medical Sciences (SUMS) Ethics Board (IR.SUMS.NUMIMG.REC.1400.073). The research ethics committee certificate is available at: https://ethics.research.ac.ir/ProposalCertificateEn.php?id=247842&Print=true&NoPrintHeader=true&NoPrintFooter=true&NoPrintPageBorder=true&LetterPrint=true.

ORCID iDs

Azita Yazdani

Seyedeh Fatemeh Dashti

Yeganeh Safdari

References

Adhinugraha

Rahayu

Hara

, et al. On internet-of-things (IoT) gateway coverage expansion. Future Gener Comput Syst 2020; 107: 578–587.

Hassen

Ayari

Hamdi

. A home hospitalization system based on the internet of things, fog computing and cloud computing. Inform Med Unlocked 2020; 20: 100368.

Morales-Botello

Gachet

de Buenaga

, et al. Chronic patient remote monitoring through the application of big data and internet of things. Health Informatics J 2021; 27(3): 14604582211030956.

Hjelm

Hedlund

. Internet-of-Things (IoT) in healthcare and social services–experiences of a sensor system for notifications of deviant behaviours in the home from the users’ perspective. Health Informatics J 2022; 28(1): 14604582221075562.

Azimi

Anzanpour

Rahmani

, et al. HiCH: hierarchical fog-assisted computing architecture for healthcare IoT. ACM Trans Embed Comput Syst 2017; 16(5s): 1–20.

Sood

Mahajan

, et al. Fog–cloud assisted IoT-based hierarchical approach for controlling dengue infection. The Computer Journal 2022; 65(1): 67–79.

Sadoughi

Ali

Erfannia

. Evaluating the factors that influence cloud technology adoption—comparative case analysis of health and non-health sectors: a systematic review. Health Informatics J 2020; 26(2): 1363–1391.

Ibtihal

Hassan

. Homomorphic encryption as a service for outsourced images in mobile cloud computing environment. Cryptography: breakthroughs in research and practice. IGI Global 2020; 7(2): 316–330.

Hou

Ren

Wang

, et al. Distributed fog computing for latency and reliability guaranteed swarm of drones. IEEE Access 2020; 8: 7117–7130.

10.

Javadzadeh

Rahmani

. Fog computing applications in smart cities: a systematic survey. Wireless Networks 2020; 26(2): 1433–1457.

11.

Stergiou

Plageras

Psannis

KE,

et al. Secure machine learning scenario from big data in cloud computing via internet of things network. In: Handbook of computer networks and cyber security. Cham: Springer, 2020, pp. 525–554.

12.

D’Agostino

Morganti

Corni

, et al. Combining edge and cloud computing for low-power, cost-effective metagenomics analysis. Future Gener Comput Syst 2019; 90: 79–85.

13.

de Moura Costa

da Costa

da Rosa Righi

, et al. Fog computing in health: a systematic literature review. Health and Technology 2020; 10(5): 1025–1044.

14.

Bellendorf

Mann

ZÁ

. Classification of optimization problems in fog computing. Future Gener Comput Syst 2020; 107: 158–176.

15.

Khanagha

Ansari

Paroutis

, et al. Mutualism and the dynamics of new platform creation: a study of cisco and fog computing. Strat Mgmt J 2022; 43(3): 476–506.

16.

Dhingra

Madda

Patan

, et al. Internet of things-based fog and cloud computing technology for smart traffic monitoring. Internet of Things 2021; 14: 100175.

17.

Farahani

Barzegari

Aliee

, et al. Towards collaborative intelligent IoT eHealth: from device to fog, and cloud. Microprocess Microsyst 2020; 72: 102938.

18.

Sadoughi

Erfannia

Sancholi

, et al. Teleradiology in southeast Iran: evaluating the views of senior executives and radiologists. Health Care Manag 2017; 36(3): 301–307.

19.

Scirè

Tropeano

Anagnostopoulos

, et al. Fog-computing-based heartbeat detection and arrhythmia classification using machine learning. Algorithms 2019; 12(2): 32.

20.

Sung

W-T

C-Y

(eds). Smart warehouse management based on iot architecture. In: International Symposium on Computer, Consumer and Control (IS3C), Taichung, Taiwan, 6–8 December 2018. IEEE.

21.

Sutton

Pincock

Baumgart

, et al. An overview of clinical decision support systems: benefits, risks, and strategies for success. NPJ Digit Med 2020; 3(1): 1–10.

22.

Olakotan

Mohd Yusof

. The appropriateness of clinical decision support systems alerts in supporting clinical workflows: a systematic review. Health Informatics J 2021; 27(2): 14604582211007536.

23.

Mahmoud

El-Sappagh

El-Bakry

, et al. A real-time framework for patient monitoring systems based on a wireless body area network. Int J Comput Appl 2020; 176: 12–21.

24.

Roy

Bose

Sarddar

. A fog-based dss model for driving rule violation monitoring framework on the internet of things. International Journal of Advanced Science and Technology 2015; 82: 23–32.

25.

Habibi

Farhoudi

Kazemian

, et al. Fog computing: a comprehensive architectural survey. IEEE Access 2020; 8: 69105–69133.

26.

Tuli

Basumatary

Gill

, et al. HealthFog: an ensemble deep learning based smart healthcare system for automatic diagnosis of heart diseases in integrated IoT and fog computing environments. Future Gener Comput Syst 2020; 104: 187–200.

27.

Gia

Jiang

Sarker

, et al. (eds). Low-cost fog-assisted health-care IoT system with energy-efficient sensor nodes. In: 13th international wireless communications and mobile computing conference. IWCMC), Valencia, Spain, 26–30 June 2017. IEEE.

28.

Liu

Deng

Yang

, et al. Hybrid privacy-preserving clinical decision support system in fog–cloud computing. Future Gener Comput Syst 2018; 78: 825–837.

29.

Abdel-Basset

Manogaran

Gamal

, et al. A novel intelligent medical decision support model based on soft computing and IoT. IEEE Internet Things J 2019; 7(5): 4160–4170.

30.

Kraemer

Braten

Tamkittikhun

, et al. Fog computing in healthcare–a review and discussion. IEEE Access 2017; 5: 9206–9222.

31.

Farahani

Firouzi

Chang

, et al. Towards fog-driven IoT eHealth: promises and challenges of IoT in medicine and healthcare. Future Gener Comput Syst 2018; 78: 659–676.

32.

Erfannia

Amraei

Arji

, et al. Reviewing and content analysis of persian language mobile health apps for COVID-19 management. Stud Health Technol Inform 2022; 289: 106–109.

33.

Abdali

T-AN

Hassan

Aman

AHM

(eds). A new feature in mysejahtera application to monitoring the spread of COVID-19 using fog computing. In: 3rd International Cyber Resilience Conference (CRC), Langkawi Island, Malaysia, 29–31 January 2021. IEEE.

34.

El-Rashidy

El-Sappagh

Islam

, et al. End-to-end deep learning framework for coronavirus (COVID-19) detection and monitoring. Electronics 2020; 9(9): 1439.

35.

Rahmani

Gia

Negash

, et al. Exploiting smart e-Health gateways at the edge of healthcare internet-of-things: a fog computing approach. Future Gener Comput Syst 2018; 78: 641–658.

36.

Verma

Sood

. Fog assisted-IoT enabled patient health monitoring in smart homes. IEEE Internet Things J 2018; 5(3): 1789–1796.

37.

Barik

Priyadarshini

Dubey

, et al. Leveraging machine learning in mist computing telemonitoring system for diabetes prediction InAdvances in data and information sciences 2018; 30, pp. 95–104.

38.

Bandopadhaya

Dey

Suhag

. Integrated healthcare monitoring solutions for soldier using the internet of things with distributed computing. Sustainable Computing: Informatics and Systems 2020; 26: 100378.

39.

Kharel

Reda

Shin

. Fog computing-based smart health monitoring system deploying lora wireless communication. IETE Technical Review 2019; 36(1): 69–82.

40.

Gia

Jiang

Rahmani

A-M

, et al. (eds). Fog computing in healthcare internet of things: a case study on ecg feature extraction. In: IEEE international conference on computer and information technology; ubiquitous computing and communications; dependable, autonomic and secure computing; pervasive intelligence and computing, Liverpool, UK, 26–28 October 2015. IEEE.

41.

UCI machine learning repository . In: University of California I, School of Information and Computer Sciences, editor. https://archive.ics.uci.edu/ml/datasets/Pima+Indians+Diabetes (2017, accessed 10 July 2020).

42.

Chen

Chowdhury

Ramachandran

, et al. Priority queues and dijkstra’s algorithm. TX: Computer Science Department, University of Texas at Austin, 2007.

43.

Moffat

. Huffman coding. ACM Computing Surveys (CSUR) 2019; 52(4): 1–35.

44.

Derbala

. Priority queuing in an operating system. Computers & Operations Research 2005; 32(2): 229–238.

45.

Snehalatha

Julia

Rodrigues

. Survey of bandwidth management techniques. IJISME 2013; 1(8): 12–15.

46.

Hayashi

Yukita

. Rule extraction using recursive-rule extraction algorithm with J48graft combined with sampling selection techniques for the diagnosis of type 2 diabetes mellitus in the Pima Indian dataset. Inform Med Unlocked 2016; 2: 92–104.

47.

Azimi

Takalo-Mattila

Anzanpour

, et al. (eds). Empowering healthcare IoT systems with hierarchical edge-based deep learning. In: Proceedings of the 2018 IEEE/ACM International Conference on Connected Health: Applications, Washington, DC, USA, 26–28 September 2018. Systems and Engineering Technologies.

48.

Gill

Arya

Wander

, et al. (eds). Fog-based smart healthcare as a big data and cloud service for heart patients using IoT. In: International Conference on Intelligent Data Communication Technologies and Internet of Things. Berlin, Germany: Springer, 2018.

49.

Moosavi

Gia

Nigussie

, et al. End-to-end security scheme for mobility enabled healthcare internet of things. Future Gener Comput Syst 2016; 64: 108–124.

A fog-assisted information model based on priority queue and clinical decision support systems

Abstract

Objectives

Material and Methods

Result

Conclusion

Keywords

Introduction

Background

Methodology

Dataset

Proposed model

Data acquisition layer

Fog layer

Pre-processing service

Events classification service

Queuing service

Cloud layer

Decision-making services

Cloud storage service

Fog update service

Results

Model evaluation

Discussion

Conclusion

Footnotes

Declaration of conflicting interests

Funding

Ethical approval

ORCID iDs

References