A Comprehensive Survey of The Internet of Things I

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3047960, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
A Comprehensive Survey of the Internet of

Things (IoT) and Edge Computing in Healthcare
Fatima Alshehri and Ghulam Muhammad, Senior Member, IEEE
Department of Computer Engineering, College of Computer and Information Sciences, King Saud University, Riyadh 11543, Saudi Arabia
Corresponding authors: Ghulam Muhammad (e-mail: ghulam@ksu.edu.sa).

The authors extend their appreciation to the Deputyship for Research & Innovation, “Ministry of Education” in Saudi Arabia for funding this research work
through the project number IFKSURP-1436-023.
ABSTRACT Smart health care is an important aspect of connected living. Health care is one of the basic
pillars of human need, and smart health care is projected to produce several billion dollars in revenue in the
near future. There are several components of smart health care, including the Internet of Things (IoT), the
Internet of Medical Things (IoMT), medical sensors, edge computing, cloud computing, artificial
intelligence, and next-generation wireless communication technology. Many papers in the literature deal with
smart health care or health care in general. Here, we present a comprehensive survey of IoT- and IoMT-based
smart health care, mainly focusing on journal articles published between 2014 and 2020. We survey this
literature by answering several research areas on IoT and IoMT, edge and cloud computing, security, and
medical signals fusion. We also address current research challenges and offer some future research directions.
INDEX TERMS Internet of Things (IoT), Internet of Medical Things (IoMT), edge computing, cloud
computing, medical signals, smart health care, artificial intelligence
I. INTRODUCTION enter health documents and link people, resources, and

The rising number of chronic patients and the aging of the organizations. Intelligent medical treatment includes diverse
population render the avoidance of diseases an important actors, including physicians, staff, hospitals, and research
requirement of healthcare. Prevention is not only defined by bodies. It comprises a dynamic framework with many facets,
regular exercise, nutrition, and periodic preventive controls including disease prevention and identification, assessment
as a way to sustain a healthier environment but also as a and evaluation, management of healthcare, patient decision-
method of keeping serious conditions from becoming worse. making, and medical research. Elements of intelligent
The future health sector must tackle an increasing number of healthcare involve automated networks like the IoT, mobile
chronic problems and the scarcity of treatments to satisfy Internet, cloud networking, Big Data, 5G, microelectronics,
patient demands [1]. COVID-19 has recently highlighted the and artificial intelligence (AI), along with evolving
importance of quick, comprehensive, and accurate biotechnology.
eHealthcare and intelligent healthcare involving different Sensors have been gradually embedded into diverse
types of medical and physiological data to diagnose the virus. systems of our lives through computer technology,
The use of emerging technology in protective policies and automation, and automated signal processing. Sensor-
behavioral systems can help identify potential health produced data can enable clinicians to more quickly and
conditions early and enable the scheduling of appropriate reliably recognize critical situations and help patients
steps, such as concurrently monitoring treatments and become more informed of their symptoms and future
preparing new assessments. The world’s smart health market treatments. Intrusive and noninvasive tools—ranging from
is forecast to reach USD 143.6 billion in 2019, which will devices to read bodily temperature to dialysis control
expand by an average growth rate of 16.2% between 2020 systems—provide personal and multimedia details and
and 2027 [2]. assistance to patients and the health care sector.
Smart healthcare refers to platforms for health systems Medical signals come in the form of electrocardiograms
that leverage devices such as wearable appliances, the (ECGs), electroencephalograms (EEGs), electroglottographs
Internet of Things (IoT), and the mobile Internet to easily (EGGs), electrooculograms (EOGs), electromyograms
VOLUME XX, 2017 1
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
(EMGs), body temperature, blood pressure (BP), and heart II. METHODS
rate, among others. A health care monitoring system may use We used the systematic review process PRISMA (Preferred
these medical signals to monitor a patient. Reporting Items for Systematic Reviews and Meta-Analyses)
The IoT is slowly starting to connect both doctors and to identify studies and narrow down results for this review of
consumers through health care. Ultrasounds, BP readings, the fusion of the IoMT and medical signals in smart health
glucose receptors, EEGs, ECGs, and more continue to care, as shown in Fig. 1. The PRISMA process has four main
monitor patients’ wellness. Conditions like follow-up visits sequential steps: identification of papers, scanning of papers
to doctors are critical. Several health care facilities have by removing duplicates, eligibility filtering, and inclusion of
started to utilize smart beds, which can detect a patient’s papers to produce a final paper list. The present study followed
movement and automatically adjust the bed to the correct a three-step procedure: definition of research areas, data
angle and location. The Internet of Medical Things (IoMT) collection, and data extraction.
refers to the IoT used for medical purposes. When
developing a fully integrated health environment, the IoMT
can play an important role.
Sometimes, relying on only one type of medical signal
may not fulfill the requirements for a complete diagnosis of
a certain disease. In such cases, multimodal medical signals
can be deployed for a better diagnosis. These signals can be
fused at different levels, including the data level, the feature
level, and the classification level [3]. When fusing signals,
many challenges may be encountered. These challenges
include synchronization when acquiring signals from
different sensors, data buffering, feature normalization, and
classification fusion [4].
In order to ensure patients’ and stakeholders’ satisfaction,
intelligent health care has been revolutionized with the FIGURE 1 PRISMA study selection diagram. N represents the number
development of AI and machine learning (ML) algorithms in of papers.
the context of deep learning (DL) and wireless local area
network (wLAN) technologies [5]. The medical industry has A. Research areas
been able to manage numerous medical signals from the The research areas we used to select the articles were as
same user—simultaneously improving disease detection and follows: “state of the art regarding IoMT and medical signals
prediction precision—due to these technologies’ high for smart health care”; “the techniques of multimodal medical
computational performance, high data volume, data fusion”; “cloud- and edge-based smart health care”; and
accommodation of several terminal units, and the “security and privacy of the IoMT”. Included in these
introduction of 5G and beyond 5G wireless technology. research areas are aspects of the challenges encountered in
In this paper, we present a comprehensive survey of IoT- implementing new technologies and the proposed solutions
and IoMT-based smart health care systems. The survey to mitigate such challenges.
includes all related papers published between 2014 and 2020,
located using IEEE Xplore, ScienceDirect, SpringerLink,
B. Data collection
MDPI, Hindawi, the ACM Digital Library, and Google
After identifying the research questions, we collected data
Scholar. The survey’s focus is mainly on the IoT and IoMT from the selected fusion of medical IoT and medical signals
as used in smart health care, sensors used to acquire medical for smart health care studies. This step involved the search
signals, the fusion of different medical signals, and privacy strategy and selection of studies.
and security issues related to the integration of sensors into
these systems. At the end of the paper, we discuss some
challenges in IoMT-based health care and suggest related C. Search strategy
future research directions. Our survey of articles used a combination of keywords and
involved formulating a search strategy and selecting data
The paper is organized as follows. Section II describes the
sources. We used the following combination of keywords: a)
methodology adopted to select the papers. Section III
“Internet of Medical Things”; b) “Fusion medical signals”;
presents a comprehensive survey of the literature and
c) “Multimodal medical data”; d) “Cloud/edge based smart
answers several research questions. Section IV mentions health care”; and e) “Security and privacy Internet of
some challenges and offers future research directions in this Medical Things.” The number of papers elicited by each
field. Finally, Section V concludes the paper. search strategy (item) after searching is shown in Fig. 2.
VOLUME XX, 2017 9
E. Data extraction
The following data categories were collected from
articles:
a. Task information
b. IoMT/modality
c. Features
d. Classifier
e. Dataset
f. Accuracy
The data categories that were collected from certain

reviews are:
FIGURE 2. Number of papers by item. a. Task information
b. Fusion type
c. Application area
d. Years covered
e. Publications covered
III. RESEARCH AREAS

The survey is divided into four areas: IoT or IoMT and
medical signals; IoMT or medical signals fusion; edge- and
cloud-based smart health care; and security and privacy in
IoMT-based health care.
A. IoT or IoMT and medical signals
The research in [3] used a multi-sensor platform with two-
channel pressure pulse wave (PPW) signals and one-channel
ECG to estimate BP. A total of 24 physiological and 11
FIGURE 3. Number of papers by year. informative features were derived from the signals gathered.
After extracting these features, a weakly supervised feature
(WSF) selection method was proposed to identify key aspects
in order to develop a BP model using spectral analysis. A
The search strategy was implemented based on the content multi-instance regression algorithm was then used to fuse the
of the main research questions. We restricted our selection to selected features. This model was validated on a private
papers written between 2014 and 2020, as shown in Fig. 3. dataset which included 85 patients (hypertensive and
To locate appropriate papers, we scanned for related hypotensive). The results of the proposed model showed high
publications in major online research repositories, including robustness and reliability for cuffless BP measurement.
IEEE Xplore, ScienceDirect, SpringerLink, MDPI, Hindawi, Authors in [4] presented a technique for emotion
the ACM Digital Library, Google Scholar, .and other health recognition and classification across subjects. It integrated the
and engineering journals. The number of publications significance test and sequential backward selection with a
collected from the IEEE archive was greater than the number support vector machine (ST-SBSSVM) to enhance the
of those obtained from other repositories. precision of emotion recognition. The input modalities used
included 32-channel EEG signals; four-channel EOG signals;
D. Selection of studies four-channel EMG signals; and vital signals measuring
Our initial search identified 168 papers. The “Internet of respiration, plethysmography, galvanic skin response, and
Medical Things” keyword obtained the largest number of body temperature. Ten types of linear and non-linear EEG,
papers. After removing duplicate and irrelevant articles, the EOG, and EMG features were extracted and fused with the
search was reduced to 100 articles. We used Zotero as a vital signals to produce a high-dimensional feature vector. The
reference management software. To optimize our search, we features were fused and selected using significance tests and a
added requirements for inclusion of screened papers: (1) backward selection search. The selected features were then fed
papers must be in the English language; (2) papers should be into a support vector machine (SVM) classifier. The
related to the IoT; and (3) articles should be in the medical experiments were performed using two publicly available
and health care fields. datasets, namely DEAP and SEED. The proposed method
achieved 72% accuracy on the DEAP dataset and 89%
accuracy on the SEED dataset.
VOLUME XX, 2017 9
FIGURE 4. Taxonomy of the survey.
Gu et al. [5] presented an outline for a situation awareness showed that the proposed method was superior to other
system to guarantee the accuracy of signal transmission and approaches at the time.
protect workers in mines. Multi-sensor data were Lin et al. [8] developed a hybrid BSN architecture based on
preprocessed, and information entropy theory was used to multi-sensor fusion (HBMF) to enable smart medical services.
weight the data based on various characteristics. The study Their goal was to address the lack of conventional multi-
employed a random forest (RF) SVM-based model to fuse the sensor fusion approaches in medical applications. The study
data and identify situation levels. To ensure forecast accuracy, presented in detail the specific functions of the four layers in
RF-SVM output was viewed as extreme learning machine the HBMF architecture. To enhance the performance of fusion
(ELM) input data. The results showed a root mean square error decision-making, a multi-sensor fusion method was proposed
(RMSE) below 0.2 and a TSQ no greater than 1.691 after 200 based on an interpretable neural network (MFIN) and using AI
iterations. technologies (see Fig. 5). The proposed design showed an
Muzammal et al. [6] presented a discussion of the improvement in both reliability and flexibility over existing
generation of unified activity data which could be used for multi-sensor fusion approaches.
medical purposes and proposed a multi-sensor data fusion In [9], seven channels from functional near-infrared
framework. The data obtained from the body sensor networks spectroscopy (fNIRS) were fused with seven EEG electrodes
(BSNs) were fused and inserted into an ensemble classifier, to improve the detection of mental stress. Simultaneous
which was placed in the fog computing (FC) environment. measurements of fNIRS and EEG signals were carried out on
After extensive research, the empirical study considered novel 12 subjects. These measurements were conducted while
kernel RF the best option to evaluate the proposed system. The subjects solved arithmetic problems under two different
results showed a prediction accuracy of 98% when the number conditions (control and stress). The performance of the fusion
of estimators was set to 40 at a tree depth of 15. of fNIRS and EEG signals was superior to the performance of
Steenkiste et al. [7] presented a reliable model based on each separately.
sensor fusion methods to improve performance of predicting In [10], a fusion of EEG and ECG videos was proposed
sleep apnea. The proposed method used backward shortcut using three different transforms to improve video resolution:
connections to collect and combine multi-sensor data, discrete cosine transform (DCT), discrete wavelet transform
including oxygen saturation, heart rate, thoracic respiratory (DWT), and hybrid transforms. Both peak signal-to-noise
belt, and abdominal respiratory belt. Convolutional neural ratio (PSNR) and mean squared error (MSE) parameters were
network (CNN) and long short-term memory (LSTM) models used to measure the fusion effect. This empirical study found
were used to assess robustness and performance. The results that hybrid transforms improved image reconstruction.
VOLUME XX, 2017 9
FIGURE 5. Overview of multi-sensor fusion framework
FIGURE 6. Fusion model for to predict blood pressure from ECG data.
potential field clustering (PFC), and FOR. The proposed

Authors in [11] suggested a method of medical image method was based on the physics notion of potential field and
fusion using rolling guidance filtering (RGF). The study used viewed the intensity of a pixel in an MRI scan as a “mass”
an RGF to filter input images into either low-frequency or which produces a potential field. The performance was
high-frequency components. First, the RGF separated the validated on a publicly available MRI benchmark database
input images into low-frequency and high-frequency called Brain Tumor Image Segmentation (BRATS) and
components, each of which had its own fusion role. A showed that both PFS and FOR were similar methods.
Laplacian Pyramid (LP)–based fusion rule was used with the However, PFS was an exclusive segmentation algorithm and
structural component and a sum-modified-Laplacian method required fewer parameters.
was utilized for the detailed component. The last step was Nathan and Jafari [13] proposed an approach using particle
image reconstruction. The proposed method was compared filtering to enhance heart rate monitoring with the existence
with another six methods, and the results showed that the of motion artifacts and using wearable sensors. In this
proposed fusion method had an advantage in processing time approach, heart rate was formulated as the only state to be
and the best high-frequency information. estimated apart from multiple specific signal features. This
Cabria and Gondra [12] used a potential field segmentation led to the fusion of information from different sensors and
(PFS) algorithm to segment brain tumors in magnetic signal modalities to increase monitoring accuracy. The
resonance imaging (MRI) scans. They presented the use of efficiency of this approach was tested on actual motion
ensemble tactics which gathered the output produced by PFS, objects caused by ECG and PPG data with corresponding
VOLUME XX, 2017 9
accelerometer observations, and results showed encouraging IID+IIC, and IID+HIS were superior to other existing
average error levels of less than 2 beats per minute. methods.
Authors in [14] presented a method based on multi-level Guanqiu [19] proposed a medical image fusion framework
information fusion to develop a predictive model for which combined two methods: dictionary learning and
measuring BP from ECG sensor data. In this method, the data entropy-based clustering. The presented framework utilized
were fused in five levels (see Fig. 6). In level 1, data were a Gaussian filter to split source images into high-frequency
fused from multiple ECG sensors. In level 2, the features and low-frequency components. Dictionary learning and
from the input data were extracted using various techniques. weighted average algorithms were used to fuse high-
The fusion of output information from seven different frequency and low-frequency components, respectively.
classifiers was input into the meta-classifier in level 3. Furthermore, to obtain the final fused image, the algorithm
Knowledge from multi-target regression models for each BP merged low-frequency and high-frequency components. The
type was integrated into level 4, and a single predictor for result showed enhanced performance compared with other
systolic BP (SBP), diastolic BP (DBP), and mean arterial current image fusion methods.
pressure (MAP) was obtained in level 5. Baloch et al. [20] presented a layered context-aware data
In [15], the author presented a method based on combination tactic for IoT health care applications. It
physiological signals fusion to improve the accuracy of included three phases: situation building, filtering and
emotion recognition. Its performance was validated by context acquisition, and intelligent inference. Reliable,
comparing both fused and non-fused physiological signals on accurate, and timely data were gathered from various sources.
two publicly available datasets. A feedforward neural The aim of the analysis was to resolve issues such as
network classifier was trained using both fused and unfused uncertainty, irregularity, restricted range, and sensor
signals. The result of the proposed method showed an deficiency. The drawback of this analysis was that no
improvement in performance on the DEAP and BP4D+ particular method was used to evaluate the suggested
datasets compared with other current methods. solution.
Chen et al. [16] modified an existing real-time system to In [21], a distributed hierarchical data fusion architecture
produce a recognition system for human action. The device at various levels was employed using complex event
obtained data from various sensor types, such as depth processing (CEP) technology to improve decision accuracy
cameras and wearable inertial sensors. Low-computation and timely. It divided the task of data fusion into three-level
effective depth perception features and inertial signal features processing models (low, middle, and high levels of data
were inserted into two computationally powerful shared fusion). A smart health care scenario was prepared with
collaborative representation classifiers (CRCs). The appropriate IoT network topologies to prove the effectiveness
proposed method was tested on a publicly available dataset of the proposed architecture. This empirical research found
called UTD-MHAD, and the results showed an improvement that the proposed solution allowed for effective decision-
in overall classification rate (> 97%) compared to using each making at various stages of data fusion and showed an overall
sensor separately. increase in the efficiency and response time of primary health
Zhang et al. [17] developed a data fusion cluster-tree services.
construction algorithm based on event-driven (DFCTA) and
Survey papers on IoMT and medical signals
further discussed data fusion and associated routing
technology. The minimum fusion delay path was presented Herrera et al. [22] discussed the latest in sensor fusion for
by computing the fusion waiting time of the nodes, and the hand rehabilitation applications. They concluded that only
fusion delay issue within the network was analyzed. The limited research had focused on sensor fusion for hand
DFCTA was tested and compared with traditional methods, rehabilitation and could be categorized into three application
with the results showing that the proposed method was areas: hand movements, exoskeletons, and serious games for
reliable, fast, and timely. hand rehabilitation. In most studies, the position and strength
In multi-scale transform fusion methods, there is of the user’s limb were measured during the rehabilitation
complexity in extracting structural and functional process. Of the types of sensors used, sensors based on EMG
information from both MRIs and positron emission signals were the most common.
tomography (PET) images utilizing the same decomposition Wearable devices play a vital role in long-term health
scheme. For this reason, authors in [18] proposed two monitoring systems and are currently at the heart of IoMT
procedures built on intrinsic image decomposition (IID). The [23]. Authors in [23] presented a comprehensive review that
presented methods were used to decompose both MRIs and aimed to present the most important wearable health care
PET images into two components in the spatial domain. The devices, including biophysiological signs, motion trackers,
authors used the rank of image coefficients to output the final and devices to measure EEGs, ECGs, BSCs, and so on. It
fused image by combining the decomposed two-scale proposed an extensive comparison between IoMT devices
components. Three fusion methods were employed based on and considered both motion trackers and vital signs the most
IID models. When the planned method was tested, IID+PCA, important elements in health monitoring.
VOLUME XX, 2017 9
In [24], the authors argued that it was complicated to detect study aimed to decrease costs and improve treatment
and resolve obstructive sleep apnea (OSA), although it is one outcomes by employing an interconnected network for
of the most common diseases. The paper highlighted IoT efficient flow and exchange of data. Singh et al. [31] also
systems that had supportive technologies and were utilized to presented an IoMT concept based on ML approaches to
diagnose OSA, including FC, smart devices, ML, the cloud, tackle the COVID-19 health crisis. It provided treatments and
and Big Data. It further considered the improvement in the solutions to issues related to orthopedic patients.
monitoring of sleep quality and other remote monitoring in Kaleem et al. [32] discussed ways to actively apply the IoT
AI-based health systems. In addition to the survey, a novel in the medical and smart health care sectors and provided a
IoMT optimization paradigm was proposed to improve the method named k-Healthcare in IoT. The proposed method
quality of remote OSA diagnosis. The model showed an used smartphone sensors to collect and transmit data to the
enhancement in the sensitivity, accuracy, energy cloud for processing and then to stakeholders.
consumption, and specificity of the system of remote OSA In [33], an event-driven data fusion tree routing algorithm
diagnosis. was presented. The paper discussed the theory of health
Gravina et al. [25] provided a comprehensive and information and the sports information gathering system,
systematic review of existing multi-sensor fusion which is divided into terminal nodes and client management
technologies for BSNs by formally categorizing these systems. The proposed algorithm designed communication
integrations in the BSN domain. They also provided an in- mechanisms according to the characteristics of IoT
depth study and evaluation of data fusion in the context of communication and used visual methods for modeling. The
physical activity. Their research pinpointed specific outcomes showed an enhancement in accuracy and timeliness
properties and parameters which impacted the choice of compared with other methods.
fusion design at the data, feature, and decision levels. They Chiuchisan et al. [34] provided the design for a health care
also addressed these design choices and fusion strategies for network to track at-risk patients in smart intensive care units
general health applications and the recognition of emotion. (ICUs) based on the IoT model. It used a series of sensors and
Sumithra and Malathi [26] provided a comprehensive the Xbox Kinect to track patient motions and any required
study of various modalities, such as MRI, PET imaging, adjustments in environmental parameters to notify physicians
computed tomography (CT), X-ray, and ultrasound. They in real time.
compared CT and MRI according to criteria such as cost, Sharipudin and Ismail [35] proposed a health care
risks, benefits, advantages, and limitations. The study monitoring system to manage and process data in the patient
highlighted types of multimodal fusion and further noted that, monitoring system. The proposed system was combined with
by merging both CT frames and MRI slices, the exact health care sensors that measured health parameters. The
boundary of the tumor in the brain could be detected. extracted parameters were then sent to cloud storage for
[27] presented a thorough overview of the application of medical staff’s reference.
image fusion technology in tumor treatments and diagnosis, Dimitrov [36] presented a discussion of IoMT applications
in particular liver tumors. It highlighted the key values of and Big Data in the health care field which permitted
image fusion techniques by considering their limitations and innovative commercial models and allowed for variations in
prospects. It further presented an extensive review of the work progression, customer experiences, and output
procedures and algorithms used in medical image fusion and enhancements. Wearable sensors and mobile applications
concluded with a discussion of the research challenges and were used to fulfill numerous health needs and to collect Big
trends in medical image fusion. Table 1 presents a summary Data from patients to advance health education.
of the papers described above on the IoT or IoMT and Authors in [37] established early warning score systems
medical signals. based on the characteristics of vital signs. The proposed
system supported the estimation of a health state by providing
B. IoMT and medical signals fusion
a helpful decision and cause for critical care interference. It
Swayamsiddha and Mohanty [28] discussed different investigated the most appropriate ML technique to predict the
applications of the cognitive IoMT (CIoMT) to tackle the risk associated with input medical signals.
COVID-19 pandemic. Their review showed that the CIoMT Sanyal et al. [38] proposed a federated filtering framework
was a successful tool for fast detection, decreasing the (FFF) based on the forecast of data at the central fog server
workload of the health industry, dynamic monitoring, and using aggregated model from IoMT devices. This framework
time tracking. used models provided by local IoMT devices and then shared
Yang et al. [29] proposed a combination of point-of-care with the fog server. It presented a solution for many common
diagnostics and the IoMT to assist patients in receiving issues, such as energy efficiency, privacy, and latency for
proper health care at home. The proposed platform might resource-constrained IoMT devices.
reduce national health costs and monitor disease spread. Luna-delRisco et al. [39] addressed recognition, obstacles
to implementation, and threats to the usage of wearable
Singh et al. [30] highlighted the overall applications of the
technology in the Latin American health care system. Major
IoT philosophy in tackling the COVID-19 health crisis. This
VOLUME XX, 2017 9
problems that the authors noted included the training and was performed using the Infomax and entropy bound
allocation of human capital in health care, the connectivity of minimization (EBM) algorithms. The experiment revealed
public care, funding arrangements for health programs, and that the joint ICA model could be superior to the transposed
inequality in health. They considered smart wearable sensors IVA model. In the case of joint ICA, a robust ICA algorithm
in health care to be part of the solution. such as EBM was superior to the Infomax algorithm.
Adali et al. [40] used a system where joint independent Authors in [41] presented a deep CNN model for seizure
component analysis (ICA) and transposed independent detection utilizing an excellent cross-patient seizure
vector analysis (IVA) were employed to fuse functional MRI, classifier. The visualization method demonstrates the spatial
structural MRI, and EEG data. Results were obtained from distribution of the characteristics learned by the CNN in
healthy controls and schizophrenia patients using an audible various frequency bands when studying the seizure and non-
oddball (AOD) function. The presented system was validated seizure classes.
on a private dataset which included 36 subjects. The analysis
Table 1 Summary of papers regarding IoT/IoMT and medical signals.
Ref. Task IoMTs Classifier Database Accuracy
[6] Cuff-Less One ECG, two multi-instance Private; total 85 participants Estimation error:
Blood Pressure photoplethysmog regression including 17 hypertensive and 12 around 1.50
Measurement ram signals (pulse algorithm hypotensive
pressure wave
sensors)
[7] Emotion 32-ch EEG signals, SVM Physiological signals (DEAP) 72% (DEAP); 89%
recognition 4-ch EOG, 4-ch dataset and the SJTU Emotion (SEED)
EMG, respiration, EEG Dataset (SEED)
plethysmograph,
galvanic skin
response, body
temperature.
Health and risk Multi sensors RF-SVM Private RMSE=0.017
[8] assessment for (Regression
miners Forecast -SVM)
and ELM
[9] Heart disease Multi sensors Random Forest NA 98%
prediction and Kernel
Random Forest
ensemble
[10] Sleep Apnea Abdominal CNN, LSTM Sleep-Heart-Health-Study-1 AUPR = 0.67 (CNN);
Detection respiratory belt, database AUPR = 0.71
thoracic (LSTM)
respiratory belt,
heart rate and
oxygen saturation
[11] Medical BSNs Cross-domain, NA NA
human–robot Incremental
interaction classifier and
scenario multi-sensor
fusion
[12] Rate of mental Functional Near SVM 12 healthy subjects with no history Mean detection
stress infrared of psychiatric, neurological illness rate 98%
Spectroscopy or psychotropic drug use
(fNIRS) and
Electroencephalo
graph (EEG)
VOLUME XX, 2017 9
[13] Medical video ECG and EEG DCT, DWT and Each video comprises 36 frames NA
fusion videos Hybrid with 18 frames per second
Transforms
[14] Medical image MRI and CT LP-based The test image pairs are from public The Avg is:
fusion images fusion rule, website [31] 𝑄 AB/F =0.669
Sum-Modified- MI=4.249
Laplacian SML VIF=77.178
and LP-based
fusion rule
[15] brain tumor MRIs Intersection Brain Tumor Image Segmentation Avg=0.62
detection and union (BRATS) MRI benchmark database. Std=0.211
Med=0.68
[16] Heart rate Wearable sensors A particle filter Database was taken from 2015 IEEE Error < 2 beats/min
estimates ECG, PPG Signal Processing Cup (SP Cup) [32]
The MIT-BIH Noise Stress Test
Database was also used
[17] Blood pressure Multiple ECG Bagging, Private database and the Physionet MAE: 7.93, 6.41,
predictive sensors Boosting, SVM, database. and 5.72 for SBP,
K-means, RF, DBP, and MAP (the
Naive Bayes, traditional
J48, META approach). 16.60,
9.24, and 9.80 for
SBP, DBP, and MAP
(custom approach)
[18] Emotion Physiological Feedforward BP4D+ and DEAP 95.81% on DEAP
Recognition signals Neural 91.51% on BP4D+
Network
[19] A human A depth camera CRC University of Texas at Dallas >97%
action and wearable Multimodal Human Action Dataset
recognition inertial sensor (UTD-MHAD)
[20] Medical data Multi sensors Cluster tree Private dataset having 120 nodes NA
fusion data fusion
[21] Intrinsic image MRI and PET Image 30 pairs of abnormal brain from Avg running time
decomposition images coefficients Harvard University and clinical IID+PCA=0.5
(IIC), PCA, IHS cases IID+IIC=0.5
6 pairs of images with resolution IID+IHS=0.7
changes.
[22] Medical image CT, MRI, PET, and Gaussian filter, Data were collected from two 𝑄 AB/F =0.6291
fusion SPECT images weighted public repositories [31],[33] MI=2.1045
average VIF=0.3426
algorithm,
dictionary-
learning based
algorithm
cameras of eight subjects. When the performance of the

Bernel et al. [42] presented a DL method for the fusion of fusion method was evaluated, the proposed method was
multimodal data to assist and monitor a user in performing superior to other state-of-the-art fusion approaches.
multi-step tasks. They further presented a deep temporal Xu et al. [43] presented a modified pulse-coupled neural
fusion scheme to extract deep features from individual data network (PCNN) model using the quantum-behaved particle
sources before the fusion took place. The Insulin Self- swarm optimization (QPSO) algorithm to solve the issue of
Injection (ISI) dataset consists of motion data captured with PCNN’s parameters and to improve the effectiveness and
a wrist sensor and video data obtained from the wearable correctness of medical image fusion. Various criteria were
VOLUME XX, 2017 9
used to judge the performance of different methods, Irfan and Ahmad [50] reviewed current architectural
including mutual information, standard deviation (SD), models and produced a new one for the IoMT. They
spatial frequency (SF), and structural similarity (SSIM). The pinpointed the motivations that would lead medical
proposed method was validated on five pairs of multimodal practitioners to decide to adopt the IoMT and further
medical images from a publicly available dataset [44] and demonstrated privacy and security problems in the IoMT.
showed an improvement in performance over other current Authors in [51] presented a comprehensive review of the
methods. current architecture for IoMT devices and discussed different
In [45], an approach based on weighted principal aspects of the IoMT, including communication modules and
component analysis (PCA) for multimodal medical fusion in major sensing technologies. The paper further discussed the
the contourlet domain was presented. One of the contourlet challenges and opportunities related to using the IoMT in the
transform’s limitations was capturing limited directional health care industry. Communication gateways, data
information. In this study, the contourlet transform was acquisition, and cloud servers were the main components of
combined with PCA to overcome this limitation and improve the IoMT framework.
the fusion of medical images. It used max and min fusion In [52], the author presented a comprehensive overview of
rules to merge the decomposed coefficients, and the results multimodal fusion of brain imaging data. This survey
showed improvement. discussed in detail the merits of multimodal data fusion and
Torres et al. [46] proposed a formulation that merged two summarized various multivariate voxel-wise data fusion
features from three different modalities to categorize human methods. Numerous of fusion studies in multimodal medical
sleep poses in an ICU atmosphere. Unlike other methods that data, particularly related to psychosis, were reviewed. The
extract one feature by merging data from various sensors, this author summarized this analysis by highlighting the
method extracted features independently and then utilized importance of multimodal convergence in minimizing
them to estimate labels. Various properties and scenes were misdirection and perhaps discovering links between the brain
obtained from different modalities, cameras, and RGB (red, and mental illness.
green, and blue) and depth sensors. Both shape and Table 2 presents a summary of the papers described above
appearance features were extracted and used to train single regarding IoMT and medical signals fusion.
modal classifiers and generate an estimation of the trust level
of each modality.
Using a hybrid technique combining non-subsampled C. Edge- and cloud-based smart health care
contourlet transform (NSCT) and stationary wavelet
transform (SWT), Ramlal et al. [47] produced an enhanced An edge- or cloud-based privacy-preserved automatic
multimodal medical image fusion scheme. NSCT was used emotion recognition system utilizing a CNN was proposed in
to decompose the source image into various sub-bands, and [53]. In [54], the authors suggested an appropriate training
SWT was used to decompose the NSCT approximation system for a deep neural network named ETS-DNN in an
coefficients into sub-bands. The efficiency of the proposed edge-computing environment. In order to change DNN
procedure was assessed through four sets of experiments. The parameters, ETS-DNN was combined with a hybrid
suggested system was compared to other existing fusion algorithm for hybrid modified water wave optimization
schemes and showed improvement in brightness, clarity, and (HMWWO) In order to minimize data traffic and latency,
edge information in the merged image. data preprocessing and classification was carried out at the
Manchandaa and Sharmab [48] proposed an improved edge of computation. The results showed that ETS-DNN was
algorithm based on a fuzzy transform for multimodal medical superior to the compared approaches.
image fusion. They produced highly informative fused Authors in [55] developed a clustering model for medical
medical images with consideration paid to the limitation of applications (CMMA) to increase the reliability and lifetime
losing information in the error images when using fuzzy of communication and reduce the energy consumption of
transform (FTR) pairs. Different datasets were used to edge-based IoMT systems. The remaining energy, delay,
validate the proposed algorithm, and the result was compared distance from the base station, capacity, and queue of the
with existing multimodal medical image fusion algorithms. IoMT devices were considered when choosing a cluster head.
The outcome indicated an increase in the standard of smooth The practical analysis showed that CMMA performed better
edges; sharp, fine texture; and high clarity in the presented in terms of energy-efficient communication compared with
algorithm. other approaches.
Authors in [56] presented a cognitive IoT (CIoT) cloud-
Survey papers on IoMT and medical signals based smart health care framework with an EEG seizure
detection method using DL. Authors in [57] proposed a voice
Joyia et al. [49] presented the contributions of IoT in the pathology monitoring system integrating IoT and cloud
medical field and their major challenges in the IoMT. technology.
Numerous applications and research in IoMT were discussed
in terms of how they solved issues faced by the global health
care industry.
VOLUME XX, 2017 9
Table 2 Summary of papers regarding IoMT or medical signals fusion.
Ref. Task Modality Fusion / classifier Database Accuracy

[46] Auditory fMRI, sMRI, Joint ICA and Private; 22 healthy and 14 Mutual information = 0.59
oddball task EEG transposed IVA patients
for control
and
schizophrenia
patients
[48] Human action Google CNN-LSTM Insulin 90% (clip)

and activity glass Self-Injection (ISI) Dataset; 4
recognition wearable male and 4 female participants
for health camera
monitoring (video) and
Invensense
motion
wrist
sensor
(motion)
[49] Similarity CT-MR, CT– PCNN, QPSO- Group 1 from Group 1:
measures of MR T2, MR PCNN (http://imagefusion.org), Groups STD= 65.1832
different CT T1–MR T2, 2-5 from [31] SF= 22.8200
scan images MR T1–FDG MI_AF= 3.2100
and MR T2– Entropy for Group 2-5:
SPET G2= 4.5362
images. G3= 5.4726
G4= 5.4726
G5= 3.7875
[50] Fusion of CT and MRI Min–Max fusion 3 datasets downloaded from E= 6.3364
obligatory rule Brain Atlas [31] SSIM= 0.9957
anatomical 𝑄 𝐴𝐵/𝐹 =0.6511
minutiae
images to
progress
medical
diagnosis
[51] Sleep Pose A Carmine Linear 26,400 images from five actors 100% accuracy in bright
Recognition camera, Discriminant Available at and clear scenes;
RGB and Analysis (LDA) http://vision.ece.ucsb/research. 70% in poorly illuminated
depth and SVM scenes;
sensors 90% in occluded scenes
[52] Fusion of CT, MRI Entropy of -38 CT and MRI images of 14 MI= 4.3780
brain images square, weighted patients. 𝑄 𝐴𝐵/𝐹 = 0.7780
obtained from sum-modified -Harvard Medical School website SD= 58.0671
CT scan and Laplacian (http://www.med.harvard.edu/
MRI. AANLIB/home.html)
[53] Generate a MRI/CT, Fuzzy transform 8 datasets Fusion factor=5.946
fused medical MRI the same size 512 51 × 2 with 256 SSIM=0.8871
image from T1/MRI T2, grayscale levels. Table 7 Feature similarity
error images CT/PET, index measure
MRI/SPECT (FSIM)=0.8581
VOLUME XX, 2017 9
In [58], Olokodana et al. used the ordinary kriging method two related systems. The study found that the proposed
to present a real-time seizure detection model in an edge technique improved performance in terms of detection and
computing paradigm. Fractal dimension features were classification with 98.5% accuracy.
extracted from EEG signals, and an ordinary kriging model Oueida et al. [70] provided a resource preservation net
was then used for classification. Computational time (RPN) framework which integrated a custom cloud, edge
complexity is one of the limitations of kriging. In the computing, and Petri net. The framework improved
proposed model, a previously trained ordinary kriging model reliability and efficiency and reduced both resources and time
was moved to an edge device for real-time seizure detection. consumption. The proposed system was suitable for
The empirical study achieved a training accuracy of 99.4% emergency departments and other types of queuing systems.
and a mean seizure detection latency of 0.85 seconds. In [71], Kharel et al. used Long Range (LoRa) wireless
In [59], an energy-efficient smart-health system based on communication and FC to produce an architecture for smart
fuzzy classification was proposed for seizure detection. The remote health monitoring. LoRa radio provides long-range
raw EEG data was processed at the edge before being communication and energy consumption for IoT devices and
transmitted to the mobile–health cloud (MHC). The proposed is used in the proposed system to link the edge user’s device
system minimized energy consumption by reducing the with health centers. FC preserves network bandwidth and
amount of transmitted data and provided high classification reduces latency by minimizing data exchange with the cloud.
accuracy. The result showed an extension in battery life of Tests showed that LoRa and FC had promising performance
60% and a classification accuracy above 98%. in remote health care monitoring.
A new network paradigm, CIoT, has been proposed based In [72], the author utilized several wearable sensors and a
on the application of cognitive computing technologies [60]. DL method (namely a recurrent neural network [RNN]) to
In [61], Chen et al. combined the advantages of edge introduce a human activity prediction system. Data, features,
computing and cognitive computing to create an edge- and activity prediction were processed on fast edge devices
cognitive-computing–based (ECC-based) smart health care like personal computers. To predict human activities from a
system which allocated maximum edge computing resources public dataset, the RNN was trained based on the features,
to higher-risk patients. The empirical experiments showed achieving 99.69% mean prediction performance.
that the proposed system was capable of improving energy Authors in [73] produced a task scheduling approach called
efficiency and user quality of experience (QoE). HealthEdge that assigned priority to each task based on its
Authors in [62] presented an edge-IoMT computing emergency level in order to decide whether to process the
architecture which minimized latency and improved given task remotely (i.e., in the cloud) or locally. They also
bandwidth efficiency. It consisted of two components: edge provided a priority-based task queuing method which
computing unit modules which compressed and filtered real- allowed emergency tasks to be processed earlier. The results
time video data, and cloud infrastructure modules which showed that increasing the local edge workstation reduced
securely transmitted medical information to the physician. processing time.
Akmandor et al. [63] discussed different edge-side In [74], Vasconcelos et al. proposed a new method called
computing options which were designed to address adaptive brain tissue density analysis (adaptive ABTD) to
challenges in smart health care systems. They demonstrated improve the detection and classification of strokes. Edge
an edge-side reference model comprised of three levels: computing devices provided low computation and cost and
sensor node, communication, and base station. The reduced time consumption in detection and diagnosis. The
compatibility between sensors and edge-side requirements integration of the adaptive ABTD with edge devices and the
enabled smart edge-side decision-making. IoT introduced speedy and efficient stroke diagnosis.
DL was utilized on a mobile health care platform to Authors in [75] presented a model for cloud-IoT–based
investigate a speech pathology detection method in [64] and health service applications in an integrated Industry 4.0
an EEG-based remote pathology detection system in [65]. environment by enhancing the selection of virtual machines
In [66], an automated voice disorder recognition system (VMs). They implemented their cloud-IoT model using three
was used to monitor people of all age groups and professional optimizers: particle swarm (PSO), genetic algorithm (GA),
backgrounds. By identifying the source signal from the and parallel particle swarm (PPSO). The proposed
speech using linear prediction analysis, the proposed system architecture consists of stakeholders who use IoT devices to
could determine the voice disorder. send tasks through cloud computing in order to receive
In [67]–[69], the authors developed a voice disorder services such as telemedicine and disease diagnosis. The
detection and monitoring system. In [67], they collected cloud broker works in the middle to send and receive tasks
voice samples sent to the edge, which offers low latency and over the cloud.
reduces delays in data traffic flow. After processing data Authors in [76] proposed a tree-based deep model for
using edge computing, data were transferred to the cloud for efficient load distribution to edge devices. The input image
more processing and assessment. The medical information was divided into volume groups and a tree structure passed
was then sent to a specialist, who prescribed suitable through each volume. The tree structure had several branches
treatments for patients. The authors tested voice disorder and levels, each of which was defined by a convolutional
classification and detection and compared the results with layer.
VOLUME XX, 2017 9
In [77], Chung and Yoo increased the effectiveness of Ivanov et al. [81] introduced OpenICE-lite, a middleware
analyzing Big Data by proposing an edge-based health model for medical device interoperability designed to provide
using peer-to-peer DNNs. An edge-based health model and a security for IoMT devices. Several applications were
server model were established separately to tackle the issue investigated for this middleware, including a critical
of response time delay. The results showed that combining pulmonary shunt predictor and a remote pulmonary
DNN techniques and parallel processing models minimized monitoring system.
response time delay. Lu and Cheng [82] proposed a secure data-sharing scheme
Limaye and Adegbija [78] provided a comprehensive for IoMT devices. First, the system guarantees the protection
review of medical applications and algorithms in IoMT of and permitted access to mutual information. Second, the
architectures and their integration with edge computing. system conducts effective integrity tests until the customer
IoMT workloads were compared using MiBench, an existing opens mutual data to prevent an erroneous application or
open source embedded system benchmark suite. The calculation performance. Ultimately, the system provides a
comparison showed that the IoMT applications differed from lightweight procedure for both consumer and customer. The
MiBench, indicating the need for a new benchmark sufficient scheme removes the burden of generating encryption and
for the IoMT microarchitecture. A cloud-based healthcare decryption keys solely on end devices.
framework was proposed in [111]. In the framework, several Mohan [83] presented some cyber threats to IoMT devices
aspects of data transmission and latency were discussed. An and provided some solutions to these threats. As IoMT
edge-enabled DNN-based method was proposed in [111]. devices are limited by their battery life, they have only
Table 3 presents a summary of the papers described above limited encryption capability and are thus at risk in terms of
on edge- and cloud-based smart health care. integrity, confidentiality, and privacy. Sensitive patient data
can be leaked, and denial of service attacks can be made by
D. Security and privacy in IoMT-based health care draining the battery. As solutions, IoMT devices must be
installed during deployment and software details transferred
The security and privacy of medical data are very to the cloud-based system provider. IoMT devices encrypt all
important in smart health care frameworks. A patient’s data patient data using lightweight cryptographic methods and
should be handled privately. If privacy is breached, the store patient data on the cloud-based system. Only approved
patient may be harassed in public, which can lead patients to entities who send their verifiable attribute-based certificate to
become traumatized and depressed. If medical sports data are the cloud provider may access this data.
leaked, rival sports team members might use these data to Nkomo and Brown [84] proposed a cybersecurity
solicit illegal advantages. Therefore, medical data should be framework for IoMT devices in smart health care systems
dealt with privately and securely transmitted over that had five attributes: identify, protect, detect, respond, and
communication channels [123]. This important issue has recover. First, asset management and risk assessment should
been addressed in a great deal of prior research. be identified. Second, access control, data security, and
Alsubaei et al. [79] presented a taxonomy of security and protective technology should be developed. Third, anomalies
privacy in the IoMT. They categorized IoT layers and events should be detected. Fourth, response planning
(perception, network, middleware, application, business); should be designed through analysis and mitigation. Fifth, a
intruder types (individual, organized group, state-sponsored recovery path should be planned.
group); impact (life risk, brand value loss, data disclosure); Rathnayake et al. [85] realized a security mechanism for a
and attack method (social engineering, implementation layer, smart healthcare system using the IoMT. First, data from
software or hardware bugs, malware). The perception layer different IoMT devices were encrypted using asymmetric
includes wearable devices such as fitness trackers, BP cipher and advanced encryption standard (AES) keys. The
sensors, and respiratory sensors; implantable devices such as keys were protected using a ciphertext attribute-based
capsule cameras; ambient devices such as door sensors and encryption (ABE) protocol. The encrypted data were
daylight sensors; and stationary devices such as CT scanners transmitted through an insecure network. At the receiver end,
and X-rays. While there are many ways to fuse data from AES keys were decrypted using the ABE protocol. Data were
these devices, the authors did not discuss them in the paper. then decrypted using the ASE keys. This mechanism
In [80], the authors identified the potential security threats maintained the privacy and the security of patients’ data.
that can affect IoMT-based health care systems and Seliem and Elgazzar [86] proposed a blockchain for IoMT
recommended a series of security measures to tackle these (BIoMT) to preserve security and privacy in a smart health
threats. Some of the security issues mentioned in this paper care framework. The BIoMT had four layers. The first layer
include overlooking the aspects of built-in security, was a device layer, which contained IoMT and user interface
stakeholders’ unfamiliarity with security solutions and focus devices. The second layer was a facility layer, which had a
on marketing and financial gain, and a lack of consensus bolster to look after IoMT devices. The facility layer
between stakeholders for overlapping solutions. Based on provided the basic blockchain modules for attribute number
these threats, the authors proposed some ontology-inspired, selection, security generation, and identity issuance. The
stakeholder-centric, and scenario-based recommendations in third layer was a cloud layer that provided the computational
line with available guidelines.
VOLUME XX, 2017 9
power and storage, and the fourth layer was a cluster layer
which contained medical facilities and the service provider.
Table 3 Summary of papers regarding edge- and cloud-based smart health care.
Ref. Task IoMTs classifier Database Accuracy
[34] Seizure EEG's Kriging The University of Bonn dataset A training accuracy of
detection method consisting of 5 healthy subjects and 99.4%
5 epilepsy patients;
The Children’s Hospital Boston
(CHB) dataset having 22 patients
[64] Seizure A wearable EEG Swift In- EEG dataset [84] 98%
detection device network
Classification
(SIC)
[72] Voice Smart sensors CNN The Saarbrucken Voice Disorder 98.5%
disorder (SVD) database[85]
[77] Human ECG, RNN MHEALTH public dataset [86] 99.69%

activities magnetometer,
accelerometer
and gyroscope
sensors
[79] Stroke CT k-Nearest CT images dataset [87] 98.13% and 97.83%,

detection Neighbors 174 healthy + 142 hemorrhagic respectively
(KNN), SVM, stroke patients + 157 ischemic
Bayes, stroke patients
Multilayer
Perceptron
(MLP), and
Optimum
Path Forest
(OPF)
Wang et al. [87] designed a fog-based access control (AC) blockchain technologies to protect medical data. A unified
method for the IoMT. The authors developed a method that blockchain-based technique would solve many of the
installed an extra layer of control on fog servers to improve difficulties related to a centralized cloud solution. Authors in
protection for local mobile devices. A register in the AC [89] introduced an access management model that
server was important for compliance with devices. Data safeguarded patients’ medical data from internal information
access requests were submitted to the AC server, where the security attacks. It enabled only legally permitted people to
status of the application could be reviewed. The registry connect despite physical limitations. The suggested model
needed to ensure that the incoming function had been incorporated authorization consistent with permits and
recorded in the past. The comparison should be performed as responsibilities, rather than positions for medical personnel
the work form was recorded to ensure that the privacy setting only. It eliminated the contradictions of current AC models.
was changed. The architecture was situated in the fog layer, Omotosho et al. [90] identified and incorporated some of the
where functional-oriented servers could provide the required main characteristics of a patient’s health report that should be
AC service to each device. published and made accessible at all times as well as qualities
Dilawar et al. [88] introduced cryptography as a solution that should be disclosed only during emergency conditions or
for the safe exchange of patient safety records using pre-hospital treatment. Creating medical features from
VOLUME XX, 2017 9
patient health information that may be retrieved in critical cryptographic system was designed by analyzing fault
cases is a proactive step that allows technicians to obtain tolerance and differential power on a cloud platform.
access to required details in pre-hospital services while
protecting patients’ dignity and confidentiality. Survey papers on security and privacy in IoMT-based
Farahat et al. [91] introduced a data encryption scheme that health care
involved first encoding data, then encrypting those data with A survey on security and privacy in the IoMT was
a rotated key until they were sent across the network. Doctors presented in [97]. The authors identified four requirements
can recover the protected data using their login keys and for security and privacy: data integrity, data usability, data
credentials. The scheme was implemented using low-cost auditing, and patient information privacy. Existing solutions
equipment and reliable applications to ensure safety in the to these requirements were discussed and included data
delivery of medical information. encryption, access control, trusted third-party auditing, data
Guan et al. [92] proposed a differential private data search, and data anonymization. For example, some
clustering scheme to allow privacy-preserving IoMT using encryption methods for access control include attribute-based
the MapReduce system. For large-scale data sets, encryption and symmetric and asymmetric key encryption.
MapReduce is a parallel programming system that abstracts The paper ended by noting some future challenges, such as
parallel computing procedures into two functions: Map and how to deal with insecure networks, develop lightweight
Reduce. In this scheme, the authors refined the distribution of protocols for devices, and share patients’ private data.
privacy budgets and the collection of initial centroids to boost Hatzivasilis et al. [98] reviewed security and privacy in the
the performance of the k-means clustering algorithm. In IoMT. In an IoMT-based health care system, there are three
addition, an enhanced method for collection of the initial main application settings: hospitals, homes, and body
centroids was suggested to maximize the precision and sensors. Three security aspects—confidentiality, integrity,
reliability of the clustering algorithm. and availability—should be enforced in device, connectivity,
Hamidi [93] proposed a modern paradigm for the and cloud security. The survey analyzed different types of
application of biometric technologies to the advancement of security components. Various types of protection
smart health care using the IoMT, which, in addition to being mechanisms, identification and anonymity techniques, and
simple to use, requires broad-scope data access. While card data destruction for device reuse were also discussed.
IDs and passwords control entry, these systems can be Sun et al. [99] provided an outline of the latest problems,
quickly broken and are known to often be inefficient. A requirements, and possible risks to the protection and
biometric trait has four main features: universality, confidentiality of IoMT-based health care systems. To design
distinctiveness, permanence, and collectability. The author an IoMT networks, one must address postural body
anticipated four levels of security strategies: IoMT device, movements, rises in temperature, energy efficiency,
communication, analytical, and management. transmission range, quality of service, and heterogeneous
Alsubaei et al. [80] outlined a web-based IoMT security environments. The security and confidentiality requirements
assessment framework focused on an ontological scenario- have different attribute levels. At the data level, care must be
driven methodology to propose security steps in the IoMT taken regarding confidentiality, integrity, and availability. At
and to evaluate safety and deterrents in IoMT solutions. The the sensor level, the design must address tamper-proof
framework encouraged the development of a strategy that fits hardware, localization, self-healing, over-the-air
stakeholders’ protection goals and facilitates decision- programming, and forward and backward compatibility. At
making. the personal server level, device authentication and user
Elhoseny et al. [94] proposed a hybrid optimization of authentication should be considered, while at the medical
asymmetric encryption for IoMT security. An ideal private server level, important requirements include access control,
and transparent key-based authentication was used in IoT key management, trust management, and resistance to denial
therapeutic images. Various approaches were considered to of service attacks.
achieve optimal hybrid optimization, from which the Li et al. [100] provided a survey of secured IoMT with
researchers differentiated and analyzed the critical open- friendly-jamming schemes. The authors reviewed the
ended difficulties in enhancing IoT in healthcare. IoMT’s existing protection systems and defined key security
Shakeel et al. [95] introduced learning-based Deep Q- issues in the IoMT. They recommended friendly-jamming
Networks to reduce ransomware attacks when handling schemes to protect patients’ sensitive diagnostic data
health records using IoMT devices. The approach analyzed obtained from medical sensors. They concluded that, when
the medical knowledge in various layers per the Q-learning properly planned, friendly-jamming approaches could
principle, which allowed transitional attacks to be eliminated substantially reduce the probability of effectiveness of
with less difficulty. Efficiency was measured in terms of eavesdropping activity while having no substantial impact on
energy, lifetime, throughput, accuracy, and malware error legal transmission.
detection rate. Yi and Nie [96] proposed a multivariate Ahmed et al. [101] introduced a new medical image
quadratic equation–based cryptographic security system for forgery detection method to verify that health care images
IoMT devices. A physical analysis model of the had not been changed or altered. The method generates an
image noise map, realizes a multi-resolution regression filter
VOLUME XX, 2017 9
to the noise map, and feeds the output to SVM-based and detected; efficient computer scripting; and good data
ELM-based classifiers. Another copy-move image forgery packing, encryption, and encoding. An optimization
detection method was proposed in [112]; the method could algorithm is therefore required to maximize the usage of
be used in medical image forgery detection. sensors [105][109].
Ray et al. [102] reviewed the security and privacy issues, Wearable technologies have been introduced for different
challenges, and future directions in the IoMT field. There are forms of biomedical observation to enable people to enjoy
four major categories of medical sensors: disposable health long, stable lives. This is even more important for older
sensors, connected health sensors, IoT-supported sensors, people. These resources must also be easy and comfortable to
and IoT market cap sensors. The authors provided a use and fast to transport. These requirements are fulfilled by
systematic review of these sensors in terms of their security a portable, small, and well-structured device. The wearable
and privacy, followed by the challenges they present. Some interface is expected to be lightweight and thin and to
of these challenges included the integration of multiple function for a long period of time.
sensors with proper protocols, data bursts, and social Wearable sensors are equipped with batteries, Bluetooth,
acceptance. In a related survey, Yaacoub et al. [117] outlined and other materials and were designed to be attached to
some limitations and issues related to the security of IoMT human skin. For human safety, it is important to consider
devices and provided some recommendations. They listed toxicity, flammable materials, and other factors when
risks such as the disclosure of personal information, data designing wearable sensors. Wearable sensors that constrain
falsification, lack of training, and reasonable accuracy. Some body movement, such as a belt worn at the waist or ankle, are
recommended security layers included an accuracy layer with uncomfortable, especially for the elderly and children. One
a trust sub-layer; a prevention layer with authentication sub- challenge is to develop sensors that continuously monitor
layer, privacy sub-layer, and data confidentiality sub-layer; human vital signs using suitable materials and without
and a defensive layer consisting of two sub-layers (detection reducing user comfort.
and correction). As data is obtained by a microcontroller and transmitted to
a smart computer or cloud server, there is a risk of
IV. CHALLENGES AND FUTURE RESEARCH disconnection and consequent data loss. This must be
DIRECTIONS reduced as much as possible to ensure sound protection
The complexity of IoT-based healthcare rises in response monitoring. It may be essential to store temporary data
to three factors: number of attached sensors or IoMT devices, (buffering) in a large memory-providing microcontroller.
number of patients or end users, and timeline. It is also Because data from multiple sources are characterized by
affected by a broad variety of variables that have not been variable dimensions of characteristics, irregularity, and the
properly identified and measured. In certain health care retention of unnecessary or missing data, some appropriate
environments, the bulk of IoMT devices can be used to action should be taken. These solutions might involve AI-
identify and diagnose an illness, and the data collected from based feature collection, data transformation, data
heterogeneous sensors contains a variety of issues, such as calibration, and missing value generation.
hardware glitches, drained batteries, or connectivity Due to innovations in IoT-based smart health care, there is
problems [106]. There are certain basic problems that are no standard protocol for ensuring interoperability across
normal and unregulated. In particular, there are sometimes smart devices. In particular, these techniques must consider
unexplained errors in the usage of popular medical sensors, the issue of energy use, as treatments for some diseases
such as mobile phones and smartwatches. Irregular demand additional communication capability, which allows
complications, for example, may arise from breakdowns, people to control the different phases (early stage, middle
malfunctions, or the failure of a third-party device. There are stage, and final stage) of the disease.
also regular complexities, such as battery power, distinctions Automatic health care programs depend on self-healing,
between particular physical characteristics, and variations in self-optimization, self-configuration, and self-protection
the environment. [113][118]. As background such as sensor noise and
The above problems indicate that several difficulties exist recording environment, varies, multi-sensor fusion methods
in smart health care, though multimodal signals and several can deal with these modifications, since they can have a direct
IoMT devices are being used. A simplified and easier fusion impact on system properties such as precision. Information
solution should be discussed to facilitate the general adoption transfer methods for transfer learning should be used to allow
of such smart health care [115][119][120][121][123]. Below, the system to adapt to particular circumstances by collecting
we discuss some of these problems and potential solutions. and transferring knowledge from one context to another.
A strong IoT network will allow device connectivity and Explainable AI (XAI) and interpretable ML constructs can
system control functions at three levels of data collection, be used to combine multimodal signals for smart health care
portal data transfer, and continuous storage and monitoring at [1]. Because automatic diagnostic or disease detection tools
a medical facility. Such interventions ought to be are merely medical aids, the various levels of the tools should
safeguarded; thus, data security is necessary. In order to be interpretable in such a way as to reassure physicians.
retain basic device management and long-term regulation The practical usefulness IoMT activated healthcare
without interference, power loss is becoming increasingly systems is rarely addressed in literature. The main concern is
important. This is directly related to the amount of parameters
VOLUME XX, 2017 9
that the most relevant data is owned by companies and is not [7] T. V. Steenkiste, D. Deschrijver, and T. Dhaene, “Sensor Fusion
using Backward Shortcut Connections for Sleep Apnea Detection in
accessible to the public. The efficient deployment and
Multi-Modal Data,” p. 14.
utilization of data fusion in practice will allow for more
reliable measurement and evaluation of day-to-day physical [8] K. Lin, Y. Li, J. Sun, D. Zhou, and Q. Zhang, “Multi-sensor fusion
for body sensor network in medical human–robot interaction
activity utilizing low-cost monitors that can lead to easier and scenario,” Information Fusion, vol. 57, pp. 15–26, May 2020.
better preventive care for chronic diseases. We assume that
hosting medical data in a public archive with appropriate [9] F. Al-Shargie, “Fusion of fNIRS and EEG Signals: Mental Stress
Study,” Apr. 2019.
protection precautions and exploring current data fusion
strategies using such public data will be a crucial potential [10] A. A. Mergin and M. S. G. Premi, “Pixel level fusion of medical
signals using DCT, DWT and hybrid(DWT-DCT) transform based
direction for future research. on maximum selection rule-a comparison,” in 2017 International
5G-based edge computing involving light DL models can Conference on Computing Methodologies and Communication
be useful in ensuring a fast response. Edge processing can (ICCMC), Jul. 2017, pp. 898–903.
also be implemented so that privacy is preserved [11] J. Chen, L. Zhang, L. Lu, Q. Li, M. Hu, and X. Yang, “A novel
[2][104][107][108][114][116]. medical image fusion method based on Rolling Guidance Filtering,”
Internet of Things, p. 100172, Feb. 2020.
V. CONCLUSION [12] I. Cabria and I. Gondra, “MRI segmentation fusion for brain tumor
Smart healthcare is a well-researched area. In the smart detection,” Information Fusion, vol. 36, pp. 1–9, Jul. 2017, doi:
health care domain, there is a breadth of literature covering 10.1016/j.inffus.2016.10.003.
IoT, IoMT, medical signals, AI, edge and cloud computing at [13] V. Nathan and R. Jafari, “Particle Filtering and Sensor Fusion for
various rates and utilizing varied approaches. However, to the Robust Heart Rate Monitoring Using Wearable Sensors,” IEEE
Journal of Biomedical and Health Informatics, vol. 22, no. 6, pp.
best of our knowledge, there was a lack of a thorough and 1834–1846, Nov. 2018.
systematic analysis of state-of-the-art IoT, IoMT, medical
signals use and fusion, edge and cloud computing, privacy [14] M. Simjanoska, S. Kochev, J. Tanevski, A. M. Bogdanova, G. Papa,
and T. Eftimov, “Multi-level information fusion for learning a blood
and security in the smart health care domain. The purpose of pressure predictive model using sensor data,” Information Fusion,
this survey was thus to offer a formal classification and vol. 58, pp. 24–39, Jun. 2020.
specific comparative context for IoT, IoMT, edge and cloud
[15] D. Fabiano and S. Canavan, “Emotion Recognition Using Fused
computing, privacy and security in smart health care. The Physiological Signals,” in 2019 8th International Conference on
survey included the use of IoT, IoMT, and medical signals, Affective Computing and Intelligent Interaction (ACII), Sep. 2019,
the fusion of sensors, and the use of edge and cloud pp. 42–48.
computing in smart healthcare. It further provided a survey [16] C. Chen, R. Jafari, and N. Kehtarnavaz, “A Real-Time Human Action
of security and privacy issues involving IoMT devices. Recognition System Using Depth and Inertial Sensor Fusion,” IEEE
Finally, some research challenges and future research Sensors Journal, vol. 16, no. 3, pp. 773–781, Feb. 2016.
directions were discussed. [17] W. Zhang, J. Yang, H. Su, M. Kumar, and Y. Mao, “Medical data
fusion algorithm based on Internet of things,” Pers Ubiquit Comput,
REFERENCES vol. 22, no. 5, pp. 895–902, Oct. 2018.
[1] M. S. Hossain, G. Muhammad, and N. Guizani, “Explainable AI and [18] J. Du, W. Li, and H. Tan, “Intrinsic Image Decomposition-Based
Mass Surveillance System-based Healthcare Framework to Combat Grey and Pseudo-Color Medical Image Fusion,” IEEE Access, vol.
COVID-19 like Pandemics,” IEEE Network, vol. 34, no. 4, pp. 126- 7, pp. 56443–56456, 2019.
132, July/August 2020. [19] G. Qi, J. Wang, Q. Zhang, F. Zeng, and Z. Zhu, “An Integrated
[2] G. Muhammad, M. F. Alhamid, and X. Long, “Computing and Dictionary-Learning Entropy-Based Medical Image Fusion
Processing on the Edge: Smart Pathology Detection for Connected Framework,” Future Internet, vol. 9, no. 4, Art. no. 4, Dec. 2017.
Healthcare,” IEEE Network, vol. 33, no. 6, pp. 44–49, Nov. 2019. [20] Z. Baloch, F. K. Shaikh, and M. A. Unar, “A context-aware data
[3] F. Miao, Z.-D. Liu, J.-K. Liu, B. Wen, Q.-Y. He, and Y. Li, “Multi- fusion approach for health-IoT,” Int. j. inf. tecnol., vol. 10, no. 3, pp.
Sensor Fusion Approach for Cuff-Less Blood Pressure 241–245, Sep. 2018, doi: 10.1007/s41870-018-0116-1.
Measurement,” IEEE Journal of Biomedical and Health Informatics, [21] R. Dautov, S. Distefano, and R. Buyya, “Hierarchical data fusion for
vol. 24, no. 1, pp. 79–91, Jan. 2020. Smart Healthcare,” J Big Data, vol. 6, no. 1, p. 19, Feb. 2019.
[4] F. Yang, X. Zhao, W. Jiang, P. Gao, and G. Liu, “Multi-method [22] I. Herrera-Luna, E. J. Rechy-Ramirez, H. V. Rios-Figueroa, and A.
Fusion of Cross-Subject Emotion Recognition Based on High- Marin-Hernandez, “Sensor Fusion Used in Applications for Hand
Dimensional EEG Features,” Front Comput Neurosci, vol. 13, Aug. Rehabilitation: A Systematic Review,” IEEE Sensors Journal, vol.
2019, Accessed: Jul. 01, 2020. [Online]. Available: 19, no. 10, pp. 3581–3592, May 2019.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6714862/.
[23] M. Haghi, K. Thurow, and R. Stoll, “Wearable Devices in Medical
[5] Q. Gu, S. Jiang, M. Lian, and C. Lu, “Health and Safety Situation Internet of Things: Scientific Research and Commercially Available
Awareness Model and Emergency Management Based on Multi- Devices,” Healthcare Informatics Research, vol. 23, Jan. 2017.
Sensor Signal Fusion,” IEEE Access, vol. 7, pp. 958–968, 2019.
[24] M. Abdel-Basset, W. Ding, and L. Abdel-Fatah, “The fusion of
[6] M. Muzammal, R. Talat, A. H. Sodhro, and S. Pirbhulal, “A multi- Internet of Intelligent Things (IoIT) in remote diagnosis of
sensor data fusion enabled ensemble approach for medical data from obstructive Sleep Apnea: A survey and a new model,” Information
body sensor networks,” Information Fusion, vol. 53, pp. 155–164, Fusion, vol. 61, pp. 84–100, Sep. 2020, doi:
Jan. 2020. 10.1016/j.inffus.2020.03.010.
VOLUME XX, 2017 9
[25] R. Gravina, P. Alinia, H. Ghasemzadeh, and G. Fortino, “Multi- Mapping Visualization,” ACM Trans. Multimedia Comput.
sensor fusion in body sensor networks: State-of-the-art and research Commun. Appl., vol. 15, no. 1s, p. 10:1–10:17, Feb. 2019.
challenges,” Information Fusion, vol. 35, pp. 68–80, May 2017, doi:
10.1016/j.inffus.2016.09.005. [42] E. A. Bernal et al., “Deep Temporal Multimodal Fusion for Medical
Procedure Monitoring Using Wearable Sensors,” IEEE Transactions
[26] M. Sumithra and S. Malathi, “A Brief Survey on Multi Modalities on Multimedia, vol. 20, no. 1, pp. 107–118, Jan. 2018.
Fusion,” in Emerging Trends in Computing and Expert Technology,
Cham, 2020, pp. 1031–1041. [43] X. Xu, D. Shan, G. Wang, and X. Jiang, “Multimodal medical image
fusion using PCNN optimized by the QPSO algorithm,” Applied Soft
[27] C. Li and A. Zhu, “Application of Image Fusion in Diagnosis and Computing, vol. 46, pp. 588–595, Sep. 2016.
Treatment of Liver Cancer,” Applied Sciences, vol. 10, no. 3, Art.
no. 3, Jan. 2020. [44] “The Whole Brain Atlas.”
http://www.med.harvard.edu/aanlib/home.html (accessed Jul. 17,
[28] S. Swayamsiddha and C. Mohanty, “Application of cognitive 2020).
Internet of Medical Things for COVID-19 pandemic,” Diabetes &
Metabolic Syndrome: Clinical Research & Reviews, vol. 14, no. 5, [45] A. Moin, V. Bhateja, and A. Srivastava, “Weighted-PCA Based
pp. 911–915, Sep. 2020. Multimodal Medical Image Fusion in Contourlet Domain,” in
Proceedings of the International Congress on Information and
[29] Ting Yang, Mattia Gentile, Ching-Fen Shen, and Chao-Min Cheng, Communication Technology, Singapore, 2016, pp. 597–605.
“Combining Point-of-Care Diagnostics and Internet of Medical
Things (IoMT) to Combat the COVID-19 Pandemic,” Diagnostics, [46] C. Torres, S. D. Hammond, J. C. Fried, and B. S. Manjunath, “Sleep
vol. 10, no. 224, pp. 224–224, Apr. 2020. Pose Recognition in an ICU Using Multimodal Data and
Environmental Feedback,” in Computer Vision Systems, Cham,
[30] R. P. Singh, M. Javaid, A. Haleem, and R. Suman, “Internet of things 2015, pp. 56–66.
(IoT) applications to fight against COVID-19 pandemic,” Diabetes
& Metabolic Syndrome: Clinical Research & Reviews, vol. 14, no. [47] S. D. Ramlal, J. Sachdeva, C. K. Ahuja, and N. Khandelwal, “An
4, pp. 521–524, Jul. 2020. improved multimodal medical image fusion scheme based on hybrid
combination of nonsubsampled contourlet transform and stationary
[31] R. Pratap Singh, M. Javaid, A. Haleem, R. Vaishya, and S. Ali, wavelet transform,” International Journal of Imaging Systems &
“Internet of Medical Things (IoMT) for orthopaedic in COVID-19 Technology, vol. 29, no. 2, pp. 146–160, Jun. 2019.
pandemic: Roles, challenges, and applications,” Journal of Clinical
Orthopaedics and Trauma, May 2020. [48] M. Manchanda and R. Sharma, “An improved multimodal medical
image fusion algorithm based on fuzzy transform,” Journal of Visual
[32] K. Ullah, M. A. Shah, and S. Zhang, “Effective ways to use Internet Communication and Image Representation, vol. 51, pp. 76–94, Feb.
of Things in the field of medical and smart health care,” in 2016 2018.
International Conference on Intelligent Systems Engineering
(ICISE), Jan. 2016, pp. 372–379. [49] G. J. Joyia, R. M. Liaqat, A. Farooq, and S. Rehman, “Internet of
Medical Things (IOMT): Applications, Benefits and Future
[33] Y. Zhang, Y. Zhang, X. Zhao, Z. Zhang, and H. Chen, “Design and Challenges in Healthcare Domain,” JCM, 2017.
Data Analysis of Sports Information Acquisition System Based on
Internet of Medical Things,” IEEE Access, vol. 8, pp. 84792–84805, [50] M. Irfan and N. Ahmad, “Internet of medical things: Architectural
2020. model, motivational factors and impediments,” in 2018 15th
Learning and Technology Conference (L T), Feb. 2018, pp. 6–13.
[34] I. Chiuchisan, H.-N. Costin, and O. Geman, “Adopting the Internet
of Things technologies in health care systems,” in 2014 International [51] F. Al-Turjman, M. H. Nawaz, and U. D. Ulusar, “Intelligence in the
Conference and Exposition on Electrical and Power Engineering Internet of Medical Things era: A systematic review of current and
(EPE), Oct. 2014, pp. 532–535. future trends,” Computer Communications, vol. 150, pp. 644–660,
Jan. 2020.
[35] A. Sharipudin and W. Ismail, “Internet of Medical Things (IoMT) for
Patient Healthcare Monitoring System,” in 2019 IEEE 14th Malaysia [52] V. D. Calhoun and J. Sui, “Multimodal Fusion of Brain Imaging
International Conference on Communication (MICC), Dec. 2019, pp. Data: A Key to Finding the Missing Link(s) in Complex Mental
69–74. Illness,” Biological Psychiatry: Cognitive Neuroscience and
Neuroimaging, vol. 1, no. 3, pp. 230–244, May 2016.
[36] D. V. Dimitrov, “Medical Internet of Things and Big Data in
Healthcare,” Healthcare Informatics Research, vol. 22, no. 3, pp. [53] M. S. Hossain and G. Muhammad, “Emotion recognition using
156–163, Jul. 2016. secure edge and cloud computing,” Information Sciences, vol. 504,
pp. 589–601, Dec. 2019.
[37] A. Pazienza et al., “Adaptive Critical Care Intervention in the Internet
of Medical Things,” in 2020 IEEE Conference on Evolving and [54] I. V. Pustokhina, D. A. Pustokhin, D. Gupta, A. Khanna, K. Shankar,
Adaptive Intelligent Systems (EAIS), May 2020, pp. 1–8. and G. N. Nguyen, “An Effective Training Scheme for Deep Neural
Network in Edge Computing Enabled Internet of Medical Things
[38] S. Sanyal, D. Wu, and B. Nour, “A Federated Filtering Framework (IoMT) Systems,” IEEE Access, vol. 8, pp. 107112–107123, 2020.
for Internet of Medical Things,” in ICC 2019 - 2019 IEEE
International Conference on Communications (ICC), May 2019, pp. [55] T. Han, L. Zhang, S. Pirbhulal, W. Wu, and V. H. C. de Albuquerque,
1–6. “A novel cluster head selection technique for edge-computing based
IoMT systems,” Computer Networks, vol. 158, pp. 114–122, Jul.
[39] M. Luna-delRisco et al., “Adoption of Internet of Medical Things 2019.
(IoMT) as an opportunity for improving public health in Latin
America,” in 2018 13th Iberian Conference on Information Systems [56] M. Alhussein, G. Muhammad, M. S. Hossain, and S. U. Amin,
and Technologies (CISTI), Jun. 2018, pp. 1–5. “Cognitive IoT-Cloud Integration for Smart Healthcare: Case Study
for Epileptic Seizure Detection and Monitoring,” Mobile Netw Appl,
[40] T. Adali, Y. Levin-Schwartz, and V. D. Calhoun, “Multimodal Data vol. 23, no. 6, pp. 1624–1635, Dec. 2018.
Fusion Using Source Separation: Application to Medical Imaging,”
Proceedings of the IEEE, vol. 103, no. 9, pp. 1494–1506, Sep. 2015. [57] G. Muhammad, S. M. M. Rahman, A. Alelaiwi, and A. Alamri,
“Smart Health Solution Integrating IoT and Cloud: A Case Study of
[41] M. S. Hossain, S. U. Amin, M. Alsulaiman, and G. Muhammad, Voice Pathology Monitoring,” IEEE Communications Magazine,
“Applying Deep Learning for Epilepsy Seizure Detection and Brain vol. 55, no. 1, pp. 69–73, Jan. 2017.
VOLUME XX, 2017 9
[58] I. L. Olokodana, S. P. Mohanty, E. Kougianos, and O. O. Olokodana, edge-cloud architecture for brain CT image analysis,” Engineering
“Real-Time Automatic Seizure Detection Using Ordinary Kriging Applications of Artificial Intelligence, vol. 91, p. 103585, May 2020.
Method in an Edge-IoMT Computing Paradigm,” SN COMPUT.
SCI., vol. 1, no. 5, p. 258, Aug. 2020. [75] M. Elhoseny, A. Abdelaziz, A. S. Salama, A. M. Riad, K.
Muhammad, and A. K. Sangaiah, “A hybrid model of Internet of
[59] A. Awad Abdellatif, A. Emam, C.-F. Chiasserini, A. Mohamed, A. Things and cloud computing to manage big data in health services
Jaoua, and R. Ward, “Edge-based compression and classification for applications,” Future Generation Computer Systems, vol. 86, pp.
smart healthcare systems: Concept, implementation and evaluation,” 1383–1394, Sep. 2018.
Expert Systems with Applications, vol. 117, pp. 1–14, Mar. 2019.
[76] G. Muhammad, M. S. Hossain, and A. Yassine, “Tree-Based Deep
[60] Y. Zhang, X. Ma, J. Zhang, M. S. Hossain, G. Muhammad, and S. U. Networks for Edge Devices,” IEEE Transactions on Industrial
Amin, “Edge Intelligence in the Cognitive Internet of Things: Informatics, vol. 16, no. 3, pp. 2022–2028, Mar. 2020.
Improving Sensitivity and Interactivity,” IEEE Network, vol. 33, no.
3, pp. 58–64, May 2019. [77] K. Chung and H. Yoo, “Edge computing health model using P2P-
based deep neural networks,” Peer-to-Peer Netw. Appl., vol. 13, no.
[61] M. Chen, W. Li, Y. Hao, Y. Qian, and I. Humar, “Edge cognitive 2, pp. 694–703, Mar. 2020.
computing based smart healthcare system,” Future Generation
Computer Systems, vol. 86, pp. 403–411, Sep. 2018. [78] A. Limaye and T. Adegbija, “A Workload Characterization for the
Internet of Medical Things (IoMT),” in 2017 IEEE Computer Society
[62] Ç. Dilibal, “Development of Edge-IoMT Computing Architecture for Annual Symposium on VLSI (ISVLSI), Jul. 2017, pp. 302–307.
Smart Healthcare Monitoring Platform,” in 2020 4th International
Symposium on Multidisciplinary Studies and Innovative [79] F. Alsubaei, A. Abuhussein, and S. Shiva, “Security and Privacy in
Technologies (ISMSIT), Oct. 2020, pp. 1–4, doi: the Internet of Medical Things: Taxonomy and Risk Assessment,” in
10.1109/ISMSIT50672.2020.9254501. 2017 IEEE 42nd Conference on Local Computer Networks
Workshops (LCN Workshops), Oct. 2017, pp. 112–120.
[63] A. O. Akmandor and N. K. Jha, “Smart Health Care: An Edge-Side
Computing Perspective,” IEEE Consumer Electronics Magazine, [80] F. Alsubaei, A. Abuhussein, and S. Shiva, “Ontology-Based Security
vol. 7, no. 1, pp. 29–37, Jan. 2018. Recommendation for the Internet of Medical Things,” IEEE Access,
vol. 7, pp. 48948–48960, 2019.
[64] M. Alhussein and G. Muhammad, “Voice Pathology Detection Using
Deep Learning on Mobile Healthcare Framework,” IEEE Access, [81] R. Ivanov, H. Nguyen, J. Weimer, O. Sokolsky, and I. Lee,
vol. 6, pp. 41034–41041, 2018. “OpenICE-lite: Towards a Connectivity Platform for the Internet of
Medical Things,” in 2018 IEEE 21st International Symposium on
[65] G. Muhammad, M. S. Hossain, and N. Kumar, “EEG-Based Real-Time Distributed Computing (ISORC), May 2018, pp. 103–
Pathology Detection for Home Health Monitoring,” IEEE Journal on 106.
Selected Areas in Communications, pp. 1–1, 2020, doi:
10.1109/JSAC.2020.3020654. [82] X. Lu and X. Cheng, “A Secure and Lightweight Data Sharing
Scheme for Internet of Medical Things,” IEEE Access, vol. 8, pp.
[66] Z. Ali, G. Muhammad, and M. F. Alhamid, “An Automatic Health 5022–5030, 2020.
Monitoring System for Patients Suffering From Voice Complications
in Smart Cities,” IEEE Access, vol. 5, pp. 3900–3908, 2017. [83] A. Mohan, “Cyber Security for Personal Medical Devices Internet of
Things,” in 2014 IEEE International Conference on Distributed
[67] G. Muhammad, M. F. Alhamid, M. Alsulaiman, and B. Gupta, “Edge Computing in Sensor Systems, May 2014, pp. 372–374.
Computing with Cloud for Voice Disorder Assessment and
Treatment,” IEEE Communications Magazine, vol. 56, no. 4, pp. 60– [84] D. Nkomo and R. Brown, “Hybrid Cybersecurity Framework for the
65, Apr. 2018. Internet of Medical Things (IOMT),” in 2019 IEEE 12th
International Conference on Global Security, Safety and
[68] M. Alhussein and G. Muhammad, “Automatic Voice Pathology Sustainability (ICGS3), Jan. 2019, pp. 212–212.
Monitoring Using Parallel Deep Models for Smart Healthcare,” IEEE
Access, vol. 7, pp. 46474–46479, 2019. [85] R. M. P. H. K. Rathnayake, M. S. Karunarathne, N. S. Nafi, and M.
A. Gregory, “Cloud Enabled Solution for Privacy Concerns in
[69] A. Al-nasheri, G. Muhammad, M. Alsulaiman, and Z. Ali, Internet of Medical Things,” in 2018 28th International
“Investigation of Voice Pathology Detection and Classification on Telecommunication Networks and Applications Conference
Different Frequency Regions Using Correlation Functions,” Journal (ITNAC), Nov. 2018, pp. 1–4.
of Voice, vol. 31, no. 1, pp. 3–15, Jan. 2017.
[86] M. Seliem and K. Elgazzar, “BIoMT: Blockchain for the Internet of
[70] S. Oueida, Y. Kotb, M. Aloqaily, Y. Jararweh, and T. Baker, “An Medical Things,” in 2019 IEEE International Black Sea Conference
Edge Computing Based Smart Healthcare Framework for Resource on Communications and Networking (BlackSeaCom), Jun. 2019, pp.
Management,” Sensors, vol. 18, no. 12, Art. no. 12, Dec. 2018. 1–4.
[71] J. Kharel, H. T. Reda, and S. Y. Shin, “Fog Computing-Based Smart [87] X. Wang, L. Wang, Y. Li, and K. Gai, “Privacy-Aware Efficient
Health Monitoring System Deploying LoRa Wireless Fine-Grained Data Access Control in Internet of Medical Things
Communication,” IETE Technical Review, vol. 36, no. 1, pp. 69–82, Based Fog Computing,” IEEE Access, vol. 6, pp. 47657–47665,
Jan. 2019. 2018.
[72] Md. Z. Uddin, “A wearable sensor-based activity prediction system [88] N. Dilawar, M. Rizwan, F. Ahmad, and S. Akram, “Blockchain:
to facilitate edge computing in smart healthcare system,” Journal of Securing internet of medical things (IoMT),” International Journal of
Parallel and Distributed Computing, vol. 123, pp. 46–53, Jan. 2019. Advanced Computer Science and Applications, vol. 10, no. 1, pp. 82–
89, 2019.
[73] H. Wang, J. Gong, Y. Zhuang, H. Shen, and J. Lach, “HealthEdge:
Task scheduling for edge computing with health emergency and [89] M. A. Habib et al., “Privacy-based medical data protection against
human behavior consideration in smart homes,” in 2017 IEEE internal security threats in heterogeneous Internet of Medical
International Conference on Big Data (Big Data), Dec. 2017, pp. Things,” International Journal of Distributed Sensor Networks, vol.
1213–1222. 15, no. 9, p. 1550147719875653, Sep. 2019.
[74] F. F. X. Vasconcelos, R. M. Sarmento, P. P. Rebouças Filho, and V. [90] A. Omotosho, O. Adegbola, B. Adelakin, A. Adelakun, and J.
H. C. de Albuquerque, “Artificial intelligence techniques empowered Emuoyibofarhe, “Exploiting Multimodal Biometrics in E-Privacy
VOLUME XX, 2017 9
Scheme for Electronic Health Records,” Journal of Biology, [108] M. Chen et al., "Urban Healthcare Big Data System Based on
Agriculture and Healthcare, vol. 4, no. 18, pp. 22–33, Feb. 2015. Crowdsourced and Cloud-Based Air Quality Indicators," IEEE
Communications Magazine, vol. 56, no. 11, pp. 14-20, November
[91] I. S. Farahat, A. S. Tolba, M. Elhoseny, and W. Eladrosy, “A secure 2018.
real-time internet of medical smart things (IOMST),” Computers &
Electrical Engineering, vol. 72, pp. 455–467, Nov. 2018. [109] X. Yang, et al. "Deep Relative Attributes," IEEE Transactions on
Multimedia, vol. 18, no. 9, pp. 1832-1842, Sept. 2016.
[92] Z. Guan, Z. Lv, X. Du, L. Wu, and M. Guizani, “Achieving data
utility-privacy tradeoff in Internet of Medical Things: A machine [110] M. S. Hossain, M. Al-Hammadi and G. Muhammad, "Automatic
learning approach,” Future Generation Computer Systems, vol. 98, Fruit Classification Using Deep Learning for Industrial
pp. 60–68, Sep. 2019. Applications," in IEEE Transactions on Industrial Informatics, vol.
15, no. 2, pp. 1027-1034, Feb. 2019.
[93] H. Hamidi, “An approach to develop the smart health using Internet
of Things and authentication based on biometric technology,” Future [111] M. S. Hossain and G. Muhammad, “Cloud-based Collaborative
Generation Computer Systems, vol. 91, pp. 434–449, Feb. 2019. Media Service Framework for Health-Care,” International Journal of
Distributed Sensor Networks, vol. 2014, Article ID 858712, 11
[94] M. Elhoseny, K. Shankar, S. K. Lakshmanaprabu, A. Maseleno, and pages, February 2014.
N. Arunkumar, “Hybrid optimization with cryptography encryption
for medical image security in Internet of Things,” Neural Comput & [112] N. Muhammad, et al., "Copy-Move Forgery Detection Using Dyadic
Applic, Oct. 2018. Wavelet Transform," 2011 Eighth International Conference
Computer Graphics, Imaging and Visualization, Singapore, 2011, pp.
[95] P. Mohamed Shakeel, S. Baskar, V. R. Sarma Dhulipala, S. Mishra, 103-108.
and M. M. Jaber, “Maintaining Security and Privacy in Health Care
System Using Learning Based Deep-Q-Networks,” J Med Syst, vol. [113] G. Muhammad, et al., "Formant analysis in dysphonic patients and
42, no. 10, p. 186, Aug. 2018. automatic Arabic digit speech recognition," BioMedical Engineering
OnLine 2011, 10:41.
[96] H. Yi and Z. Nie, “On the security of MQ cryptographic systems for
constructing secure Internet of medical things,” Pers Ubiquit [114] M. A. Rahman, et al., “Secure and Provenance Enhanced Internet of
Comput, vol. 22, no. 5, pp. 1075–1081, Oct. 2018. Health Things Framework: A Blockchain Managed Federated
Learning Approach,” IEEE Access, vol. 8, pp. 205071-205087,
[97] W. Sun, Z. Cai, Y. Li, F. Liu, S. Fang, and G. Wang, “Security and November 2020.
Privacy in the Medical Internet of Things: A Review,” Security and
Communication Networks, Mar. 29, 2018. [115] S. U. Amin, et al., “Deep Learning for EEG motor imagery
classification based on multi-layer CNNs feature fusion,” Future
[98] G. Hatzivasilis, O. Soultatos, S. Ioannidis, C. Verikoukis, G. Generation Computer Systems, vol. 101, pp. 542-554, December
Demetriou, and C. Tsatsoulis, “Review of Security and Privacy for 2019.
the Internet of Medical Things (IoMT),” in 2019 15th International
Conference on Distributed Computing in Sensor Systems (DCOSS), [116] M. S. Hossain, M. A. Rahman, and G. Muhammad, “Cyber Physical
May 2019, pp. 457–464. Cloud-Oriented Multi-Sensory Smart Home Framework for Elderly
People: An Energy Efficiency Perspective,” Journal of Parallel and
[99] Y. Sun, F. P.-W. Lo, and B. Lo, “Security and Privacy for the Internet Distributed Computing, vol. 103, pp. 11-21, May 2017.
of Medical Things Enabled Healthcare Systems: A Survey,” IEEE
Access, vol. 7, pp. 183339–183355, 2019. [117] J.-P. A. Yaacoub, et al., “Securing internet of medical things
systems: Limitations, issues and recommendations,” Future
[100] X. Li, H.-N. Dai, Q. Wang, M. Imran, D. Li, and M. A. Imran, Generation Computer Systems, vol. 105, pp. 581–606, Apr. 2020.
“Securing Internet of Medical Things with Friendly-jamming
schemes,” Computer Communications, vol. 160, pp. 431–442, Jul. [118] M. F. Alhamid, et al., "Towards Context-Sensitive Collaborative
2020. Media Recommender System," Multimedia Tools and Applications,
Vol. 74, No. 24, pp. 11399-11428, Dec. 2015.
[101] A. Ghoneim, et al., “Medical Image Forgery Detection for Smart
Healthcare,” IEEE Communications Magazine, vol. 56, no. 4, pp. [119] M. S. Hossain, G. Muhammad, and A. Alamri, “Smart healthcare
33–37, Apr. 2018. monitoring: a voice pathology detection paradigm for smart cities,”
Multimedia Systems, vol. 25, no. 5, pp. 565-575, October 2019.
[102] P. P. Ray, D. Dash, and N. Kumar, “Sensors for internet of medical
things: State-of-the-art, security and privacy issues, challenges and [120] A. Vizitiu, C. I. Nä, A. Puiu, C. Suciu, and L. M. Itu, “Applying
future directions,” Computer Communications, vol. 160, pp. 111– deep neural networks over homomorphic encrypted medical data,''
131, Jul. 2020. Comput. Math. Methods Med., vol. 2020, pp. 1-26, Apr. 2020.
[103] M. S. Hossain, "Cloud-Supported Cyber–Physical Localization [121] N. Tsafack, et al., “A new chaotic map with dynamic analysis and
Framework for Patients Monitoring," in IEEE Systems Journal, vol. encryption application in Internet of health Things,” IEEE Access,
11, no. 1, pp. 118-127, March 2017. vol. 8, pp. 137731-137744, 2020,
[104] A. A. A. El-Latif, et al., "Secure Quantum Steganography Protocol [122] Y. Abdulsalam and M. S. Hossain, "COVID-19 Networking
for Fog Cloud Internet of Things," IEEE Access, vol. 6, pp. 10332- Demand: An Auction-based Mechanism for Automated Selection of
10340, 2018. Edge Computing Services," in IEEE Transactions on Network
Science and Engineering, doi: 10.1109/TNSE.2020.3026637.
[105] M. Masud, M. S. Hossain and A. Alamri, "Data Interoperability and
Multimedia Content Management in e-Health Systems," IEEE [123] E. M. Abou-Nassar, et al., “DITrust chain: Towards blockchain
Transactions on Information Technology in Biomedicine, vol. 16, no. based trust models for sustainable healthcare IoT systems,'' IEEE
6, pp. 1015-1023, Nov. 2012. Access, vol. 8, pp. 111223-111238, 2020,
[106] K. Lin, et al, "Green Video Transmission in the Mobile Cloud
Networks," in IEEE Transactions on Circuits and Systems for Video
Technology, vol. 27, no. 1, pp. 159-169, Jan. 2017.
[107] M. S. Hossain and G. Muhammad, “Emotion-Aware Connected
Healthcare Big Data Towards 5G,” IEEE Internet of Things Journal,
vol. 5, no. 4, pp. 2399-2406, August 2018.
VOLUME XX, 2017 9

A Comprehensive Survey of The Internet of Things I

Uploaded by

Copyright:

Available Formats

A Comprehensive Survey of The Internet of Things I

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Comprehensive Survey of The Internet of Things I

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

A Comprehensive Survey of the Internet of

Corresponding authors: Ghulam Muhammad (e-mail: ghulam@ksu.edu.sa).

I. INTRODUCTION enter health documents and link people, resources, and

VOLUME XX, 2017 1

VOLUME XX, 2017 9

The data categories that were collected from certain

III. RESEARCH AREAS

VOLUME XX, 2017 9

FIGURE 4. Taxonomy of the survey.

VOLUME XX, 2017 9

FIGURE 5. Overview of multi-sensor fusion framework

potential field clustering (PFC), and FOR. The proposed

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

Table 1 Summary of papers regarding IoT/IoMT and medical signals.

Ref. Task IoMTs Classifier Database Accuracy

VOLUME XX, 2017 9

cameras of eight subjects. When the performance of the

VOLUME XX, 2017 9

VOLUME XX, 2017 9

Table 2 Summary of papers regarding IoMT or medical signals fusion.

Ref. Task Modality Fusion / classifier Database Accuracy

[48] Human action Google CNN-LSTM Insulin 90% (clip)

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

Ref. Task IoMTs classifier Database Accuracy

[77] Human ECG, RNN MHEALTH public dataset [86] 99.69%

[79] Stroke CT k-Nearest CT images dataset [87] 98.13% and 97.83%,

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

You might also like