See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/365587896

Impact of IoT in Biomedical Applications Using Machine and Deep Learning

Chapter · November 2022

DOI: 10.1002/9781119861850.ch19

CITATIONS READS

4 307

6 authors, including:

Rehab A Rayan Imran Zafar

High Institute of Public Health Virtual University of Pakistan
64 PUBLICATIONS 254 CITATIONS 60 PUBLICATIONS 398 CITATIONS

SEE PROFILE SEE PROFILE

Husam Rajab Muhammad Asim Masoom Zubair

Budapest University of Technology and Economics 16 PUBLICATIONS 47 CITATIONS
15 PUBLICATIONS 309 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Rehab A Rayan on 22 November 2022.

The user has requested enhancement of the downloaded file.

 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 1
k

19

Impact of IoT in Biomedical Applications Using Machine and Deep Learning

Rehab A. Rayan 1 , Imran Zafar 2 , Husam Rajab 3 , Muhammad A. M. Zubair 4 , Mudasir Maqbool 5 , and
Samrina Hussain 6
AQ1 1
Department of Epidemiology, High Institute of Public Health, Alexandria University, Alexandria, Egypt
2
Department of Bioinformatics, Virtual University of Pakistan, Lahore, Punjab, Pakistan
AQ2 3
Department of Telecommunications and Media Informatics, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and
Economics, Hungary
AQ3 4
The Islamia University of Bahawalpur, Pakistan
5 Department of Pharmaceutical Sciences, University of Kashmir, Hazratbal, Srinagar, India
6
Department of Biotechnology, Lahore College for Women University, Pakistan

19.1 Introduction

k As we move into the 2020s, more devices than ever are connected to the internet, and this will continue. k
Accordingly, more than 21 billion devices will be connected to the internet around the world by 2020, which is
five times as many as there were four years ago [1]. The internet of things (IoT) is, at its most basic, a network
that connects items that can be used to identify them to the internet. This lets them send, store, and collect
information. IoT can be defined in terms of healthcare as any device that can collect health data from people. This
includes mobile phones, computers, wearables and smart bands, surgical devices, digital medications that are
implanted in the body, and other portable devices that can measure health data and connect to the internet [2].
As IoT technology has grown, it has garnered attention in a number of health practices that aim to improve the
health of the population as a whole [3]. In recent reviews, the many services, and uses of IoT in healthcare have
been discussed viz mobile health (mHealth), eHealth, semantic devices, ambient assisted living, smartphones and
wearable devices, and community-based healthcare [2, 4]. These solutions have been described in great detail
and can be used for a wide range of single-condition and cluster-condition management purposes, such as let-
ting health-care professionals monitor and track patient’s condition from a distance, making it easier for people
with chronic conditions to take care of themselves, helping to spot problems early, identifying symptoms and clin-
ical diagnoses faster, and so on. These apps can help make better use of healthcare resources while still giving
high-quality and low-cost care [5, 6].
IoT is a complicated network of “things” that each has a unique identifier and connects to a server that provides
the right services. They can talk to each other and people in the real world by sharing relevant information from
the real and virtual worlds. These things can react on their own to things that happen around them. Some of these
processes can be started by people or by machines talking to each other. They can also provide services. IoT oppor-
tunities will soon change the way healthcare is done. This technology will be a big part of tele-monitoring patients
in hospitals and, even more important, at home [6–8]. By capturing illnesses and hazardous situations early and
helping people avoid them, remote patient monitoring is a wonderful way to improve the quality of healthcare
and lower costs at the same time [9, 10].

Machine Learning Algorithms for Signal and Image Processing, First Edition.
Edited by Deepika Ghai, Suman Lata Tripathi, Sobhit Saxena, Manash Chanda, and Mamoun Alazab.
© 2023 by The Institute of Electrical and Electronics Engineers, Inc. Published 2023 by John Wiley & Sons, Inc.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 2
k

2 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

19.1.1 Artificial Intelligence and Machine Learning

Artificial intelligence (AI) is a field that includes machine learning (ML). The main goal of ML is to learn from
past situations and patterns. Instead of just making code, big data is fed into a generic algorithm, and analysis
is done using the available data. IoT and ML systems can quickly train a system to spot medical abnormalities
by using simple data and big data. The accuracy of predictions is related to how much big data has been taught
[11–13]. As a result, big data improves the predictive accuracy of ML algorithms used in healthcare prediction
platforms. Professionals now have access to a vast array of biological data, including diagnostic metrics and assess-
ments, socio-demographic factors, and diagnostic imaging technologies, because of advances in technology and
research. Biomedical data is unbalanced and nonstationary, with a prominent level of complexity, due to the abun-
dance of data and the veracity of certain circumstances [14]. In this situation, ML is still especially important to:
(i) To help doctors, naturalists, and health experts use and process substantial amounts of medical information
much better; (ii) To reduce the chances of medical mistakes; and (iii) make sure that predictive and therapeutic
rules and procedures work well together. Deep learning (DL) and artificial neural networks (ANNs) are the two
data-mining algorithms that are most often used for image processing and finding flaws. DL algorithms are used
at all levels of medicine in biomedical fields, from genomics tasks like figuring out how genes are expressed to
global health-management tasks like predicting population growth or the spread of a virus [15]. In recent years,
ML systems have been used in healthcare increasingly. ML techniques are used by several clinical decision sup-
port systems to create enhanced learning models that can be used to improve health-care service applications
[13, 16–18]. ANNs and support vector machines (SVMs) are two applications of ML in health. Such models are
used to accurately diagnose the type of cancer in a number of cancer classification applications. These algorithms
work by analyzing data from sensors and other sources to find a patient’s clinical problems and patterns of behav-
ior [19]. Like noticing changes in a patient’s behavior, in their daily routine, in how they sleep, eat, drink, and
k digest, and in how they move around. Then, health-care apps and clinical decision-support systems can use these k
algorithms to suggest changes to a patient’s lifestyle and routines, as well as to prescribe various kinds of special-
ized therapies and health-care programmers. This lets doctors make a caring plan to support patients make the
necessary changes to their lives [13, 16, 17, 20, 21]. ML is used to make patient load prediction techniques, which
makes it easy for hospitals to share information about how many patients they have. In a hospital, past data is
used to predict how many patients will come so that the right plans can be made. IoT devices with built-in ML
techniques are used to make a classifier that spots certain health incidents, like falls, in older people. The cluster-
ing algorithms could find forms of patient’s unruly behavior and let health-care workers know about them. In the
same way, IoT microchips are used to keep track of what a patient does every day through daily habit modeling.
The information is used to find things that are different about older people. Many countries have created new
skills and laws to make the most of the IoT in biomedical systems. Because of this, modern biomedical research
is now more interesting to investigate. This study aims to give a full review of innovative studies in biomedical
systems based on the IoT and a summary of how advanced studies in biomedical systems based on the IoT have
grown over time [22].
This chapter introduces the basic concepts of IoT, ML, and DL. Then it discusses the history of DL and ML
in Section 19.2. Next, it reviews ML and DL algorithms and classifications, emphasizing DL architectures in
Section 19.3. Further, it presents applied ML and DL techniques in the biomedical field in Section 19.4. Moreover,
it highlights some IoT-based ML and DL case studies in biomedical systems in Section 19.5. Finally, it concludes
with opportunities and challenges along with future insights in Section 19.6.

19.2 History of DL and ML

ANN was inspired by biological systems in the 1960s when it was discovered that different cells in the visual cortex
were active when cats looked at different objects. These tests showed that the eyes and cells in the visual cortex

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 3
k

19.3 Methods of ML and DL Algorithms and Classification 3

were connected and that the visual cortex processed data in layers. ANNs could copy the way we see things by
connecting artificial neurons in layers that could help us find out what they are.
After the 1960s, ANN development stopped because it could not do enough because its structures were too
shallow, and computers could not do enough computing. Thanks to improvements in computing and technology,
effective back propagation (BP) of RNA made it possible for pattern recognition studies [23]. First, the categoriza-
tions were done with an ANN model in a neural net with BP. Then, the parameters were changed by comparing
the predicted class labels to the real ones. Even though it helped to reduce mistakes by using gradient descent,
BP seemed to only work for some types of ANN. BP, adaptive learning rate, momentum, quasi-Newton methods,
least-square techniques, and conjugated gradients (CGs) methods were all suggested as ways to improve steeper
gradients through learning. Because ANN is so complicated, other fundamental ML approaches, like SVM support
machines, random forests, and k-nearest neighbor (k-NN) algorithms, have quickly caught up with it [24].
When an ANN has more hidden layers, it is much easier to figure out what its functions are. When RNA has deep
and complicated roles, it often sticks to what is best locally or moves through a gradient of spread. When a gradient
is passed backward through the layers, the slope quickly loses its steepness. This means that the weights of the
AQ4 layers closest to the entry do not change much. After that, a deep automatic encoder (AE) network for advanced
stock levels was suggested, bringing ANN into a new age. The layers of this network are trained by reducing the
difference between the original data and the data that has been recovered. Layered pre-training gets around the
gradient diffusion barrier and helps deep neural networks (DNNs) choose better weights. This keeps reconstructed
data from reaching a local optimum, which is often caused by choosing random weights at the start [25]. Graphics
processing units (GPUs) have given academics a new reason to be interested in DL. Deep understanding has
become more popular in recent years because people are paying more attention and putting in more effort. It is
used a lot in business right now. Deep belief networks (DBNs) and limited Boltzmann machine stacks (RBM) have
been used to recognize audio and video and to process natural language. Convolutional neural networks (CNNs)
k k
have been used a lot in image recognition, image segmentation, video recognition, and natural language process-
ing since they were first made to better mimic how animals see the world. In recurrent neural networks (RNNs),
which are a type of ANN, artificial neural cells are linked to time steps and show dynamic activity. RNN has been
used to handle sequential data in natural-language processing and document identification. In recent years, sparse
EAs, stacked EAs (SAEs), and denoising EAs have become more popular as part of in-depth online training [26].

19.3 Methods of ML and DL Algorithms and Classification

ML can frequently better explain data than model systems, giving technological answers as well as a helpful bench-
mark. It may also be utilized to assist people in their learning. A DNN is an ANN having multiple layers between
the input and output layers (DNN). Neurons, synapses, weights, biases, and functions are all core elements of ML,
which exist in various forms and dimensions [27].
ML allows a machine to autonomously analyze and understand a set of inputs as an experience without the
need for external help [13, 28]. The training and testing phases are crucial in the development of an effective ML
model. The training phase (which requires a lot of studies) entails giving the system labeled or unlabeled inputs.
The algorithm afterward keeps the training feeds in the feature space for future predictions to regard. Ultimately,
the system is inputted an unlabeled input and should expect the proper result in the testing phase. Simply defined,
ML predicts outcomes for unlabeled input by using known data in its feature space. As a result, a good ML model
can predict outputs by referring to prior encounters. The precision of this model is determined by the correctness
of the output and the training.
Advanced biological and medical technologies have provided us with explosive volumes of physical and physio-
logical data, medical imagery, functional magnetic resonance imaging, genomic, and protein complexes are just a
few examples of the ability to understand public health and illness is made more accessible by using this data [29].

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 4
k

4 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

CNNs-based DL approaches offer a lot of potential for extracting functions and predicting outcomes from massive
datasets. In biomedicine, ML and DL techniques enhance clinical care by utilizing the enormous volume of med-
ical information given by IoT technology [30]. Although these techniques have a lot of promise, they also have
several shortcomings. ML is measured in three realms: image acquisition, computational linguistics of hospital
data, and genetic data. These areas are concerned with diagnosing, discovering, and predicting outcomes [31].
A massive infrastructure for medical devices provides data because there is rarely any standard framework to use
such health information efficiently. Health records are available in multiple forms, which could also store future
formatting more challenging and raise distortion [32]. We study the history of ML and DL and the fundamentals
of techniques and technology in biomedical applications.
ML algorithms use patterns and experiences to improve the efficiency of activity. ML could be divided into three
types: supervised learning, unsupervised learning, and improved learning. Unsupervised learning, for example,
focuses on finding similarities in datasets while pooling samples [33]. Supervised learning is concerned with
identifying the best or most appropriate behavior to execute in a situation to maximize a reward, via putting it
differently. Semi-supervised learning is a technique of learning that falls in between supervised and unsupervised.
Here, the algorithm can operate with both labeled and unlabeled data. When the data given is scarce and learning
interpretations are necessary, this collection of methods is extraordinarily successful [34].
DL is a field of ML that is still relatively young and quickly expanding. It represents large-scale data abstraction
with deep multilayer neural networks (DNNs), which generate a sense of the data as pictures, audio, and text. DL
includes two attributes: multiple layers of irregular processing elements, each with a supervised or unsupervised
learning function. In the 1980s, ANNs were employed to establish the groundwork for DL, but it was not until
2006 that the true impact of DL was realized. DL has since been used in various fields, including automatic voice
recognition, picture recognition, natural language processing, drug development, and bioinformatics [35].

k k
19.3.1 Deep Learning Architectures
19.3.1.1 Auto Encoders
Auto encoders (AEs) take features from unlabeled data and set target values identical to the inputs, unlike regular
ANNs. Given the input vector {X 1 , X 2 , X 3 . . . . X n }, X (i) ∈ Rn , the AE tries to learn the model and is given by
Eq. (19.1):
hw,b (x) = g(Wx + b) ≈ x (19.1)
where W and b are the model’s parameters, g is the activation function (exact definition as in the following context),
and hw,b is the hidden units. The AE achieves a reduction of data dimensionality comparable to principal compo-
nent analysis when the number of hidden units, which reflects the dimension of features, is lower than the input
dimension. An AE with a classifier in the last layer may also do classification tasks and pattern recognition [36].

19.3.1.2 Deep Multilayer Perceptron

The well-known Rumelhart neural network, invented in 1986 and taught using the training algorithm approach,
was the forerunner to DL details mentioned in Figure 19.1. Simultaneously, the most concealed levels were at
least two or three, each containing only a few units [37]. Due to the advancement of various methods for training
big architectures, including GPU technologies, the NN scale can currently reach multiple concealed layers with a
little over 650 000 neurons and 630 million learned elements (e.g. AlexNet).

19.3.1.3 Deep Auto-encoders

A one hidden layer MLP is a particular instance of an auto-encoder. The goal of an AE is to reproduce the input
vector x^ = F(x), where x and x are the input and output vectors, respectively. In an AE, the input/output layers
have the same number of units as the hidden layer, but the hidden layer has fewer units. By stacking many AEs,
deep AEs may be created. DAEs are utilized in DL for feature extraction/reduction and pre-training elements for
compound networks [38].

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 5
k

19.3 Methods of ML and DL Algorithms and Classification 5

Optimizer part Constructive part

x41 x42 x43 x44 x45 x46 Tour

Single layer perceptron
competitive
Set Q Set R
layer
Feed forward
t4 Multi-layer perceptron
x 13 x 23 x 33 x 43 x 53 x 63 Link networks
competitive
layer Radial basis functions
t3

Neural
networks
Competetive networks

x12 x22 x32

Feedback Kohonen’s SOM
Link layer networks
x11 x21 x31 Hopfield network
Tour layer
Art models

Figure 19.1 Most common neural network structures applied in biomedical: (optimized and forward) feedforward neural
networks with various depths (one concealed layer, two concealed layers, and a deep structure with several concealed
layers); (X 1 , X 2 , and X n ) an auto-encoder and a deep auto-encoder, respectively; referrals to a confined Boltzmann machine
and a deep belief network, respectively; and an AlexNet, respectively.

19.3.1.4 Restricted Boltzmann Machine and Deep Belief Networks

A bounded Boltzmann machine is a dynamic probabilistic system that consists of a visible layer and a hidden
layer with no linkages (RBM). In other words, the observational and functional applicators are represented by
the visible and hidden layers, respectively. While the final model is a lower generic than a Boltzmann machine,
it may teach you how to extract critical functions [39]. In RBM, an energy-derived model, the energy throughout
the combined arrangement of transparent and concealed storage units is expressed as a Kohonen dynamical sys-
tem. This energy role allocates a probability to every pair of seen and concealed vectors in the modeled network.
k RBM typically replicated the probability of input data or the combination of data input and target classes as a k
combined distribution. In DL, RBM, like EA, may perform the characteristics of a complex network. A deep belief
network is a stack of RBMs, with the concealed conditions of every RBM serving as training data for a second RBM.
Consequently, each RBM recognizes pattern expressions at the lowest level and learns to code patterns indepen-
dent monitoring [40].

19.3.1.5 Convolutional Neural Networks

CNNs portray multidimensional networks that include input data, such as two-dimensional pictures with three
datasets. They are motivated via the visual cortex’s neurobiological model in which cells are liable to limited visual
field areas. The multilayer perceptron and the clustering layer are the three types of layers in CNN design, as
mentioned in Figure 19.2. Several neural maps, also known as function maps or filters, make up a precipitation
layer. In contrast to a highly centralized repository, every neuron in a function map is just linked to a small patch

Figure 19.2 The multilayer perceptron and the clustering layer of Hidden layers
the CNN design.
Input Output
layers layers

k
 k
Table 19.1 Review of the examined ML and DL algorithms.

Algorithm Scope Learning type Uses Applied technique Pros Cons

Trim Size: 8in x 10in Single Column

Support ML Supervised Binary classification, Decision boundary, SVM is very successful in Selecting the appropriate
vector nonlinear soft margin, kernel three-dimensional spaces. hyperplane and kernels
machine classification trick The fundamental strength of approach is difficult.
(SVM) [47] SVM rests in its application of
the kernel’s strategy.
Random ML Supervised Classification, Bagging In the context of random Random forest performs
forest [48] regression forest, the correlation badly when dealing with
coefficients among decision high-dimensional, data
trees (DTs) are weaker. sparsity.
Naïve Bayes ML Supervised Probabilistic Continuous Data scanning in which each A modest quantity of
(NB) [17] classification variables attribute is checked instructional analysis is
(maximum individually. Acquiring required. This function
likelihood) simple data from each simply computes the
characteristic helps to parameters’ eccentricities
improve the accuracy of the for each categorization.
premises.
Gradient ML Supervised Classification, Strong pre pruning GBDT is used to investigate In GBDT training takes
boosted regression and progressively improve an extended period and
decision trees forecasting accuracy. requires constant varied

k
k

[49] modification. It struggles

with sparse,
high-dimensional data.
Decision trees ML Supervised Prediction, Continuous target The design is straightforward. The skewed dataset and
(DTs) [50] classification variable (reduction It is applicable toward both interference influence
in variance) discrete and continuous the testing dataset,
Categorical target characteristics. Synthesis of making it more expensive
variable content is low to and using more RAM.
(Gini impurity) non-existent.
K-nearest ML Supervised Classification, Continuous K-NN decided to use a K-NN takes an
neighbor regression variables (Euclidean stochastic technique. Simple exceptionally long time to
(K-NN) [51] distance) to understand. It is simple to compute the comparison
Categorical implement. There is no need across databases. The
variables (Hamming for formal education. main disadvantage is that
Changes may be easily the numbers are skewed.
Tripathi861829

distance)
implemented by integrating Productivity is affected by
them into the current the classifier chosen
c19.tex

collection of labeled data. (K value). We must use

homogenous features
since knowledge may be
forgotten.
V1 - 09/04/2022 10:30pm Page 6

k
 k
Trim Size: 8in x 10in Single Column

Convolutional DL Spatial features Facial recognition Convolution Acquire the parameters CNN is unable to handle
neural systems operation dynamically without saying regionally fluctuating
network Analyzing and parsing anything explicitly. These Gaussian noise, which is
(CNN) [52] through documents conduct a search and users common in fuzzy
retrieve useful information pictures.
Smart cities (traffic
from data streams.
cameras, for example)
Recommendation
systems, among other
use cases
Recurrent DL RNN is created by Time Series data It captures the chronological All forms of neural
neural implementing a Text data data provided in the inputs, network models are
networks repeating Audio data such as the dependency of affected by the
(RNNs) [53] constraint on an Vanishing gradient terms in this case, while disappearing and
ANN’s hidden making assumptions. expanding potential
layers. Extension Distribute the parameters problems.
of recurrent neural over a number of time steps.
networks (RNNs) This reduces the number of
features to train as well as the
computational complexity.

k
k

Long DL In sequential Common applications Long-term Provide a diverse model Require many resources
short-term logistic regression, are cursive recognition contextual parameter, such as data for and time to be trained
AQ5 memory acquiring and production, connections may be training as well as input and and become ready for
(LSTMs) configuration knowledge represented without output preconceptions. real-world applications.
Networks dependence is representation and having to cope with backpropagation through They respond identically
[54]. crucial. interpretation, speech the minimization time (BPTT) and LSTMs to a feed-forward neural
acoustic modeling, concerns that beset lower the difficulty of system as a measure of
voice recognition, the main, resulting maintaining individual the initialization of
protein in a greater weighting to O. (1). several unrelated
conformational performance weights.
predictions, and
interactive multimedia
information
processing.
Tripathi861829

(continued)
c19.tex
V1 - 09/04/2022 10:30pm Page 7

k
 k
Table 19.1 (Continued)
Trim Size: 8in x 10in Single Column

Algorithm Scope Learning type Uses Applied technique Pros Cons

Stacked DL Unsupervised A study on P300 Random Reducing the dimensionality Processing time,
auto-encoders pre-training. Segment over-sampling of the data we are using, as hyperparameter tuning,
[55]. Supervised Identification and well as the learning time for and model validation
fine-tuning Tracking of 3D Spine your cases. The compactness before you even start
Modelling in and speed in coding using building the real model.
Adolescent Idiopathic backpropagation. Prone to overfitting,
Scoliosis was though this can be
published. mitigated via
regularization.
Deep DL Markov random Application in Backpropagation An appropriate selection of The high computational
Boltzmann field with multiple linguistics, robotics, method interactions between visible cost of inference, which
machine layers of hidden computer vision, and and hidden units can lead to is almost prohibitive
(DBM) [56] random variables artificial intelligence. more tractable versions of the when it comes to joint
model. Can capture many optimization in sizable
layers of complex datasets.
representations of input data,

k
k

and they are appropriate for

unsupervised learning since
they can be trained on
unlabeled data.
Deep belief DL Unsupervised, Aircraft engine fault Used in a Show a higher classification The approximate
networks probabilistic diagnosis feed-forward neural capability to multiple features inference procedure is
(DBN) [57] generative model network and of input patterns limited to a single
fine-tuned to hierarchically. With the bottom-up pass.
optimize pre-trained RBM. Training Ignoring top down
discrimination time is short on
influences on the
GPU-powered machines.
inference process are that
Exactly accurate compared to
the mode
a shallow net.
can fail to account for
uncertainty when
interpreting ambiguous
Tripathi861829

sensory inputs.
c19.tex
V1 - 09/04/2022 10:30pm Page 8

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 9
k

19.4 ML and DL Applications in Biomedicine 9

of neurons in the preceding layer, named receptive field. Following that, various precipitation filters are used to
corrupt the input data, progressively changing the convolution layers [41].
The precipitation filters employ the same settings in each tiny region of the picture, reducing hyperparameters
in the model. A clustering layer leverages the image’s stationing feature to decrease variance and capture essential
functions by taking the average, max, or other parameters for the functionalities at many sites on the function
maps. A CNN is made up of numerous unions and grouped layers that allow for the learning of increasingly
abstract data. In the final layers of a CNN, a totally linked classification is utilized to categories the information
retrieved by earlier aggregated and precipitated levels [42]. The most often used CNNs in ML applications are
AlexNet, Clarifai, VGG, and GoogleNet.

19.3.2 Findings of Applied ML and DL Techniques

ML is a connected field of AI. ML is a system based on AI algorithms that analyzes and interprets a collection of
inputs to determine knowledge without the need for human participation [28]. Classifiers and training models
are two essential components to develop an effective training model for learning approaches. During the training
phase, the machine may receive scored or unacknowledged inputs. After that, in the function area, the system
creates such instruction signals to represent forecasts. The logistics network appears to have used an unidentifi-
able input during the validation method to predict the correct results. The employment of diverse ML techniques
in transportation, for instance, proposes to suffice the challenges of rising travel requirements, safety concerns,
energy consumption, radiations, and environmental degradation [43]. As a result, an effective learning algorithm
may anticipate outcomes by highlighting its prior beliefs and viewpoints. A model’s efficiency is determined by
its output reliability and model design. Researchers have previously built and deployed several well-known ML
algorithms for categorization.
k k
The modern development in AI and large medical datasets has driven significant interest in developing DL
algorithms that would more quickly and accurately distinguish diagnostic tests than subjective evaluation and
other traditional methods. DL is a subgroup of ML, which performs excellent power and versatility by representing
the world as a nested scale of notions [44–46]. We examined ML and DL algorithms and summarized them in
Table 19.1 presenting the advantages and disadvantages along with techniques applied and deployment.

19.4 ML and DL Applications in Biomedicine

The field of medication gets advantages due to the structure of IoT that helps in the coordination of IoT technology
and cloud computing. The situation additionally spreads out conventions for spread of the patient’s information
from various sensors and clinical tools to a specified medical care organization. The geography of IoT is a plan of
various segments of an IoT medical care framework/network that are soundly associated with the medical care
climate [58]. An IoT system consists of three main components such as the publisher, broker, and subscriber.
An organization of associated sensors and other clinical gadgets that might work separately or in a group to
record a patient’s fundamental data. The data might be temperature, a saturation of oxygen blood pressure, heart
rate, EMG, electrocardiograph (ECG), electroencephalogram (EEG), etc. [2]. These pieces of information can be
sent consistently through an organization to a specialist. The broker is liable for the preparation and storage of
obtained information in the cloud. In the end, the patient’s data can be observed and taken through a phone, PC,
tablet, etc. the distributor can deal with this information and give criticism after getting the perception of any
humiliation or physiological peculiarity in the patient’s ailment. The IoT adopts distinct parts into an amalgam
grid, where a reason for the existence of every segment on the IoT organization and cloud in the medical services
organization is committed [59].

k
 k
Table 19.2 Review of the examined ML and DL applications.

Application Description Algorithm Solved problems Obstacles Future insights

Trim Size: 8in x 10in Single Column

ECG ECG stands for Deep learning based Used for the initial There is a problem of power The ECG and
monitoring electrocardiogram, and it is the on bio potential chip detection of heart consumption linked with an accelerometer data of
[62] movement of heart due to the to assemble virtuous aberrations with the help ECG monitoring system that is elderly patients will be
depolarization and eminence ECG data of ECG monitoring. wearable. real-time monitored using
repolarization of atria and this system.
ventricles.
Diagnosis of The interventions that need to Convolutional Diseases are diagnosed The sound legislation that To support the correct and
disease [63] be attempted are determined neural networks, using this that includes defines the use of ML in timely diagnosis of
by this. and back deep-learning systems and healthcare. diseases, the ML will be
It is applied to determine the propagation support vector machines. Obtaining well-annotated data assimilated into medical
hazard elements related to the networks Diagnosis of wrong for supervised learning is records
condition, also the signs and patients can become the challenging
symptoms, to enhance accuracy cause of incongruous
and detection proficiency mediations and
adversarial consequences.
Medical ML has several applications in ANNs (artificial In manual imaging, the Time-consuming and prone to To improve
imaging [64] the field of medical imaging. neural networks) use of ML resolves the errors. patient-centeredness by
Medical imaging is the process and CNNs issues of efficiency and Also, the outputs of deep improving the quality of
and method for the diagnosis (convolutional accuracy. training datasets.

k
k

learning techniques are

and treatment by creating neural networks). difficult to explain logically.
images of body parts. Great dependency on the
The health professional then volume of training sets and the
manually examines these images quality.
to detect the abnormalities.
Behavioral Encompasses helping a patient Several reasoning The incompetence for the The deficiency of a change in The consumption of
modification change undesirable behavior. and machine evidence that to be behavior intrusion awareness substantiation from ML
or treatment The amalgamation of ML into learning techniques, synthesized and delivered system consists of ontology, programs for the guidance
[65] behavioral alteration programs containing natural on behavioral changes process and assets for marking of change in behavior
can be helpful for the language needs of user reports, a computerized mediations.
determination of what works. administering involvements and annotator, machine learning
perspective to recover the and intellectual algorithms and
expediency of evidence. user interface.
Clinical trial There is an urge to advance ML Methods of DL The trouble of drawing The issue of using reflective The proceeded assortment
Tripathi861829

studies [5] algorithms accomplished for bits of knowledge from learning models to composite of preparing datasets for
ceaseless gaining from medical huge measures of clinical clinical datasets. advanced relevance of
c19.tex

information information utilizing Preparing datasets that are well deep learning in clinical
human abilities. categorized marks are required. exploration preliminaries.
V1 - 09/04/2022 10:30pm Page 10

k
 k
Trim Size: 8in x 10in Single Column

Smart The consideration of ML in DL, usual language The electronic well-being Planning information before Far-reaching selection of
electronic digital wellbeing records makes dispensation, and records stock clinical they are taken care of into a shrewd electronic
health shrewd frameworks with the administered ML information yet does not machine learning algorithm wellbeing records, so it can
records [66] ability to achieve a finding of help dynamic. remnant, a thought-provoking help the administration.
diseases, movement expectation, job. Furthermore, it is
and hazard assessment. troublesome to consolidate
patient-explicit elements in ML
models.
Epidemic Scrutiny of disease can be DNN (deep neural The trouble of getting Low precision of prescient To predict a variety of
outbreak beneficial from ML as it network), LSTM ready for and managing models. The trial indicating transferable diseases, the
prediction considers the forecast of (long short-term irresistible ailments constraints to exploit with ML usage of prognostic
[67] epidemics, henceforth, memory) learning, because of the deficiency models models
empowering the execution of and ARIMA of information.
proper protections. (autoregressive
integrated moving

k
k

average)
Diagnostic This study scrutinized the DL models There is a need to review The established simulations Assemble quality and
and forecast simulations for forecast simulations are challenging. high-volume datasets to
prognostic COVID-19 and concluded that train COVID-19 prediction
models for they are not well-designed. models.
COVID-19
[68]
Personalized ML algorithms run an Deep neural The incapability to offer It is essential for the sustained Making frameworks that
care [69] opportunity for contribution of networks (DNN), adapted care despite the accumulation of could be coordinated into
person-centered care. deep learning (DL) mounting amassing of elevated-quality training digital wellbeing records
data. datasets. to advance customized
medication.
Tripathi861829
c19.tex
V1 - 09/04/2022 10:30pm Page 11

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 12
k

12 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

Deep-learning application for IoT in healthcare

Medical Human Home based

Diseases
diagnosis and behavior and personal
prediction
differentiation recognition healthcare
applications
application application application

Automatic diagnosis Cardiovascular Address HAR Multi-morbidity

of heart diseases risk prediction problems chronic patients

Smartphones sensor
Automated EEG
Medicine data
disease diagnosis Seizure
prediction recognition system

Syndrome Smart dental health

differentiation loT system

Classifying skin Fall detection

lesions

Decode task-related Sport injury

information

Figure 19.3 Categorization of DL applications for IoT in the medical industry.

Since the connection of the devices for IoT relies upon the interest and application of the medical services,
it is difficult to recommend a comprehensive design for IoT. Various underlying changes have been embraced
in the past for an IoT framework [60]. It is pivotal to drill down completely related exercises identified with the
k k
ideal wellbeing application while planning another IoT-based medical services’ framework for continuous patient
observing. The accomplishment of the IoT framework relies upon how it is fulfilling the prerequisites of medical
care suppliers. Every disease needs devastating methods of medical services exercises; the topology should keep the
clinical guidelines and steps in the diagnosis system [61]. Distinguishing the right issue to be addressed utilizing
ML is the initial phase in building an ML product. Even though a model can be made to deliver understanding, it
can affect patient contemplation. Useful applications are listed below (Table 19.2).
Figure 19.3 shows a thorough categorization of DL applications for IoT in the medical industry, which includes
applications for medical forecasting, disparity, personal health-care programs, disease prediction technologies,
and disease human behavior tracking apps [70]. We looked at publications that dealt with DL applications
in the medical field in complicated subjects such as smartphone anomaly detection, chronic multi-morbidity,
learning the patient’s biological signal, perceptive dental health IoT systems, case identification models,
injuries in sports, and nutritional monitoring systems in the health and lifestyle and home automation systems
section [71].

19.5 Discussions of IoT-Based ML and DL Case Studies in Biomedical Systems

The internet of health things (IoHT) concept was investigated. By collecting and combining data on critical signs
required in hospitals, the exchange of objects leads to regulating patients’ physical conditions. IoT includes four
steps: collection, storage, processing, and presentation. IoT has two advantages: it can eliminate service interrup-
tions and efficiently distribute limited resources [72]. The number and intricacy of the data collected in this study
are the study’s principal flaws, making it a challenging chore for caretakers.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 13
k

19.5 Discussions of IoT-Based ML and DL Case Studies in Biomedical Systems 13

Tuli et al. anticipated HealthFog, a DL and IoT-based solution for automatic heart disease diagnosis. HealthFog
delivers a light mist and effective administration of data from numerous IoT devices relevant to heart disease
patients. The benefits of a new pattern for fog and edge equipment include energy-saving options and low-latency
data processing systems [73]. In terms of fog’s flaws in medical applications, it is crucial to understand the
response and latency and how difficult it is to make the best use of the Quality of Service (QoS) characteristics in
real-time fog.
Sarraf and Tofighi clarified that DL technologies have enabled neoteric technological improvements in the
EEC’s automated disease detection and diagnosis systems. The excellent efficiency of EEC deciphering due to
the autonomous extraction feature is one of the study’s positive findings. In addition, assessing your CEE can
reveal unique health concerns [74]. Moreover, finding EEC pathology datasets will be difficult because although
some are available online, most are small and inadequate for DL models.
Cerebral vascular accident (CVA), also known as cerebrovascular accident, is a condition in which specific brain
areas cease to function due to ischemia or blood clots. This issue can be resolved with prompt diagnosis. To diag-
nose a stroke, CT scans and MRIs are often employed [75]. IoT frameworks can also be used to identify strokes
based on CT images distributed by CNNs to determine whether the brain is robust, the stroke is ischemic, or a
hemorrhage causes the stroke. The advantages of implementing IoT in healthcare include regions that are less
reliant on humans, resulting in fewer human errors. The proposed architecture cannot be used in other medical
imaging. Hence, this study’s limitation is that it must expand.
Faust et al. established using pulse signals, the researcher created a DL model centered on long short-term
memory (LSTM) to detect atrial fibrillation (AF) episodes. With tagged heart rate signal data from Physio Net’s
Atrial Fibrillation Database, evaluated the DL LSTM-based system in 20 individuals. The enactment of ML meth-
ods is more limited than that of the DL model. Additionally, it can extrapolate the knowledge derived from a
restricted data set to a more extensive collection. The problem with this study is that it ignores the crucial concept
k of education [76]. k
Differentiation of the thymus is a fundamental aspect of Chinese medicine, used to treat infectious fevers.
In old classical Chinese medicine, it is difficult to distinguish between infectious febrile disorders and their
complications. Through convergence, DL is a promising approach for determining computer-assisted conditions
from contagious fever. The proposal is for a stacked automatic encoder with an adaptive abandon function [77].
The strengths of this study are considered to prevent over-assembly and increase classification accuracy, however,
deficiency in many clinical instances, infectious fever that it cannot distinguish.
Bray et al. investigated BP learning models for computer-assisted lung cancer diagnostic and therapeutic appli-
cations. Lung cancer is a severe concern to humanity today. Lung tumors can be benign or cancerous, and many
patients have both. Deep enhancement models can detect lung cancers and provide a reliable outcome. To cure
lung cancer utilizing deep reinforcement learning models, the most challenging part is developing an appropriate
function to update the Q value of each metric [78].
Ma and Tavares stated that melanoma is a dangerous kind of skin cancer that is more prone to spread. There
are three main types of melanocytic lesions: common nevi, atypical nevi, and melanoma. The skin lesions in this
study are classified using an IoT-based approach. To obtain images, the suggested method used CNN models on
the ImageNet dataset. The benefits of this strategy can be used in a variety of situations and are convenient [79].
Respecting internet access: A good connection must connect to the LINDA API (application programming
interface) and send photos. ConvNets learn to employ a wide range of power at high alpha, beta, and gamma
frequencies, according to Schirrmeister et al., who also gave unique methods for visualizing functions. Further-
more, the research depicts the ConvNets design process, which decodes the information and is linked to a task
derived from a smooth EEG without artisan functions [80]. One of the benefits is being able to provide edge learn-
ing and scalability for massive amounts of data. ConvNets have the flaw of displaying erroneous predictions and
need training data.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 14
k

14 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

19.6 Opportunities and Challenges

Biomedical systems have had substantial expansion in recent years, contributing significantly to income and
employment. Few years back, diseases and irregularities in the human body could be diagnosed only through
a physical examination in a hospital. During the duration of therapy, most patients were required to stay in the
hospital [81]. As a result, biomedical prices have risen and biomedical institutions in rural and isolated areas have
been overburdened. The technical advancements accumulated over the past have now enabled the diagnosis of
different ailments and the surveillance of health using tiny, embedded sensors in wristbands.
Furthermore, development has revolutionized the biomedical system from a hospital-centered to a
patient-centered approach. Various diagnostic tests, for example, can be conducted at home without the
assistance of a biomedical specialist (such as measuring pO2 , blood–glucose level, and blood pressure). Advanced
computing technologies also can be used to send clinical data to biomedical centers from remote locations.
The accessibility of biological resources has a qualitative part in the usage of various communication services in
conjunction with advancements in technology (e.g. big data analytics, cloud computing, mobile computers, IoT,
wireless sensing, and ML) [82].
In recent decades, advances in high-throughput technologies have resulted in a considerable rise in biomedical
data, including genetic sequences, protein structures, and medical imaging. Efficient and effective computation
techniques are necessary to store, analyze, and comprehend this flow of massive biological data. DL-based
algorithmic frameworks emphasize these challenging difficulties.
IoT has not only increased independence but has also broadened human capabilities to interact with the outside
world. IoT is a significant means of global communication by employing futuristic algorithms and protocols. It
connects many items to the internet, including wireless sensors, home appliances, and electrical devices [83].
k The IoT is gaining popularity due to its advantages of improved accuracy, reduced cost, and the capacity to k
better forecast events in the future [84]. Increased knowledge of computing technologies, software, apps, quick
access to wireless technology, the modernization of mobile, and a broader digital economy have all benefited
the IoT’s rapid development [83]. Sensors, actuators, and other IoT devices have been combined with some
further physical gadgets to exchange and monitor data via communication decorum, such as IEEE 802.11 Wi-Fi,
Zigbee and bluetooth. Sensors incorporated or transportable in the human body are often used in biomedical
applications to capture physiological data from the body of the patient, such as ECG, EEG, pressure frequency, and
temperature [85].
Furthermore, environmental conditions such as humidity, temperature, time, and date can be captured. Such
data assist in drawing relevant and accurate extrapolations about a patient’s health condition. Since much infor-
mation is acquired from a wide range of sources, storage capacity, and accessibility are especially vital in the IoT
system (sensors, mobile phones, email, software, and applications). Physicians and other authorized individuals
have access to the data collected by the above-mentioned sensors. Interacting with biomedical practitioners via
the cloud/server allows for faster patient diagnosis and, if necessary, medical intervention.
For a practical and secure transfer, coordination among users, patients, and the communication unit is preserved
[86]. Most IoT systems have a user experience that serves as a console for caregivers and allows them to regulate,
visualize, and be concerned about their data. Many studies on the advancement of the IoT system in biological
superintend, security, control, and integrity have been discovered in the literature [87]. These accomplishments
demonstrate the value of IoT in the biomedical field and its promising future. However, maintaining the quality
of the service matrix, which includes the integrity, security, cost, dependability, and availability of the information
exchange, is crucial in designing an IoT device [88].
Computer-based intelligence, and explicitly DL, upholds the formation of dynamic IoT frameworks both on the
plan of the correspondence foundation and on the investigation of information [89].
In any case, the huge measure of information gathered by IoT frameworks considers the neural networks’ prepa-
ration that can oversee the exhibition and security of the framework. Such points require added consideration from

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 15
k

19.6 Opportunities and Challenges 15

scholastic and modern specialists. It is true that QoS does not just affect the clients’ experience yet, additionally
assumes an urgent part in the administration of crisis of basic situations. The devices are required to have the
option to recognize between the high-need, crisis-related traffic, and ordinary traffic. Solid prioritization of the
network traffic and utilization of systems administration devices to satisfy certain QoS guidelines is still, an open
matter for a large part, that DL can explain with significant outcomes [90].
Besides, DL bids should be examined in the entire correspondence stack, from an actual level where sign aggre-
gation, encoding, and disaggregation of channels might be concluded because of AI models, conversely with
customary unequivocal improvement approach, at the level of application. Additionally, the organization layer
requires novel methods for choosing and adjusting steering calculations that might be absolutely or unconscious
of the organizations geography and state [91].
Without a doubt, the IoT networks development is frequently uncontrolled, and the circumstances of certain
regions are exceptionally unpredictable. Simulated intelligence will help in the plan or initiative-taking responsive
systems focused on guaranteeing the base QoS even in the event of irregularities. IoT security frameworks are
another area that will attract several examination endeavors before exceedingly long [92].
Scientific literature has received some interesting contributions due to the utilization of DL in IoT networks. Still,
abundant area for enhancements and additional directions needs to be investigated, comprising the conservation
of classification, and versatility to assaults that destabilize portions of the correspondence foundation (e.g. sticking
assaults, etc.).
The biggest region for upcoming placement of DL in IoT is the undeniable level applications, i.e. the exami-
nation of IoT gathered information [93]. The developed range of IoT devices considers a substantial number of
contextual analyses pointed toward making our social orders more comprehensive, secure, and, as a rule, vivid.
In the following several years, one critical difficulty that many societies must confront is the population aging and
the resulting distributed support and monitoring that vulnerable, many elders, and no autonomous citizens will
k k
require.
Dependable and secure IoT structures furnished with AI are the central progression to disseminate observing
of individuals needing consideration, identification of peculiarities (e.g. circulatory strain issues, falls strokes),
and giving data to a better life (e.g. calculation of person on foot courses without architectural obstruction). The
high transmission capacity needed by interactive media streams and the QoS severe prerequisites of these types of
uses present significant issues to the current IoT foundations, and advancement, both from the hypothetical and
practical are necessary [94].

19.6.1 Future Insights

In many cases, our health-care services are now more expensive than at any time before, and many patients are
compelled to be hospitalized for the length of their treatment. These difficulties can be overcome by using tech-
nologies that can remotely monitor patients. IoT technologies will cut the cost of health-care services by gathering
real-time medical data from patients and passing it to health-care providers. This will allow for the cure of health
problems before they become critical.
In the coming years, IoT technology will be extensively used in healthcare. The health-care industry is con-
tinually in the search for innovative methods to deliver services while reducing costs and quality improvement;
consequently, this sector’s dependency on IoT technology will continue to grow [95–97]. Patients are more able
to follow self-care principles using such technologies, resulting in improved cost-effectiveness of health services,
better self-management, and increased patient satisfaction.
IoT-based solutions can also be used for remote surveillance of physiological status in patients needing con-
stant supervision. Rapprochement of multiple IoT designs has also recently allowed the development of intelligent
health-care systems. IoT-driven solutions can be advantageous in developing a consistent system by the intercon-
nection of varied objects to gain a comprehensive picture of the patient’s health status [98–101].

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 16
k

16 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

19.6.2 Conclusions
The biomedical field is among the greatest tangled regarding the degree of responsibility and rigorous commands
that makes it a relevant and indispensable field for creativity. Biomedical systems have had substantial expansion
in recent years, contributing significantly to income and employment. The technical advancements accumulated
over the past have now enabled the diagnosis of different ailments and the surveillance of health using tiny, embed-
ded sensors in wristbands. This development has revolutionized the biomedical system from a hospital-centered
to a patient-centered approach. Moreover, advances in high-throughput technologies have resulted in a consider-
able rise in biomedical data, including genetic sequences, protein structures, and medical imaging. Efficient and
effective computation techniques are necessary to store, analyze, and comprehend this flow of massive biological
data. DL-based algorithmic frameworks emphasize these challenging difficulties.
The IoT has presented a realm of opportunities in the medical field and might be the answer to several
dilemmas. Introducing the IoT has not only increased independence but has also broadened human capabilities
to interact with the outside world. IoT is a significant means of global communication by employing futuristic
algorithms and protocols. Using the medical IoT would produce great possibilities for telemedicine, distant
supervision of patients’ status, among other applications. This might be with the sides of ML and DL frameworks.
Computer-based intelligence, and explicitly DL, upholds the formation of dynamic IoT frameworks both on the
plan of the correspondence foundation and on the investigation of information. However, the development of IoT
networks is frequently uncontrolled, and the conditions of certain regions are exceptionally erratic. The security
of IoT frameworks is another field that will draw many examination endeavors before exceedingly long.
The utilization of DL in IoT networks has led to some interesting contributions in the scientific literature. Still,
abundant space for enhancements and further directions still needs to be investigated, including the conserva-
tion of classification, and the versatility to assaults that destabilize portions of the correspondence foundation
k (e.g. sticking assaults, etc.). In this chapter, we reviewed the commonly dominant ML and DL algorithms, named k
some ML and DL applications in the biomedical domain, and examined IoT-based ML and DL applications in the
medical system.

References

1 Kelly, J.T., Campbell, K.L., Gong, E., and Scuffham, P. (2020). The internet of things: impact and implications
for health care delivery. Journal of Medical Internet Research 22: e20135. https://doi.org/10.2196/20135.
2 Dang, L.M., Piran, M.J., Han, D. et al. (2019). A survey on internet of things and cloud computing for
healthcare. Electronics 8: 768. https://doi.org/10.3390/electronics8070768.
3 Saarikko, T., Westergren, U.H., and Blomquist, T. (2017). The internet of things: are you ready for what’s
coming? Business Horizons 60: 667–676. https://doi.org/10.1016/j.bushor.2017.05.010.
4 Nazir, S., Ali, Y., Ullah, N., and García-Magariño, I. (2019). Internet of things for healthcare using effects
of mobile computing: a systematic literature review. Wireless Communications and Mobile Computing 2019:
e5931315. https://doi.org/10.1155/2019/5931315.
5 Yin, Y., Zeng, Y., Chen, X., and Fan, Y. (2016). The internet of things in healthcare: an overview. Journal of
Industrial Information Integration 1: 3–13. https://doi.org/10.1016/j.jii.2016.03.004.
6 Jara, A.J., Zamora-Izquierdo, M.A., and Skarmeta, A.F. (2013). Interconnection framework for mHealth
and remote monitoring based on the internet of things. IEEE Journal on Selected Areas in Communications
31: 47–65.
7 Tundis, A., Kaleem, H., and Mühlhäuser, M. (2020). Detecting and tracking criminals in the real world
through an IoT-based system. Sensors 20: 3795. https://doi.org/10.3390/s20133795.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 17
k

References 17

8 Trinugroho, D. and Baptista, Y. (2014). Information integration platform for patient-centric healthcare
Services: design, prototype and dependability aspects. Future Internet 6: 126–154. https://doi.org/10.3390/
fi6010126.
9 Ahmadi, H., Arji, G., Shahmoradi, L. et al. (2019). The application of internet of things in healthcare: a sys-
tematic literature review and classification. Universal Access in the Information Society 18: 837–869. https://doi
.org/10.1007/s10209-018-0618-4.
10 Taylor, R., Baron, D., and Schmidt, D. (2015). The world in 2025 – predictions for the next ten years. In: 2015
10th International Microsystems, Packaging, Assembly and Circuits Technology Conference IMPACT, 192–195.
AQ6 https://doi.org/10.1109/IMPACT.2015.7365193.
11 Hung, C.-Y., Chen, W.-C., Lai, P.-T. et al. (2017). Comparing deep neural networks and other machine learn-
ing algorithms for stroke prediction in a large-scale population-based electronic medical claims database.
In: 2017 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society
(EMBC), 3110–3113. https://doi.org/10.1109/EMBC.2017.8037515.
12 Ngiam, K.Y. and Khor, I.W. (2019). Big data and machine learning algorithms for health-care delivery.
The Lancet Oncology 20: e262–e273. https://doi.org/10.1016/S1470-2045(19)30149-4.
13 Al Dhahiri, A., Alrashed, B., and Hussain, W. (2021). Trends in using IoT with machine learning in health
prediction system. Forecast 3: 181–206. https://doi.org/10.3390/forecast3010012.
14 Yue, L., Tian, D., Chen, W. et al. (2020). Deep learning for heterogeneous medical data analysis. World Wide
Web 23: 2715–2737. https://doi.org/10.1007/s11280-019-00764-z.
15 Mahmud, M., Kaiser, M.S., McGinnity, T.M., and Hussain, A. (2021). Deep learning in mining biological data.
Cognitive Computation 13: 1–33. https://doi.org/10.1007/s12559-020-09773-x.
16 Alharbi, R. and Almawashi, H. (2019). The privacy requirements for wearable IoT devices in healthcare
domain. In: 2019 7th International Conference on Future Internet of Things and Cloud Workshops (FiCloudW),
k 18–25. https://doi.org/10.1109/FiCloudW.2019.00017. k
17 Kaur, G. and Oberoi, A. (2020). Novel approach for brain tumor detection based on Naïve Bayes classification.
In: Data Management, Analytics and Innovation (ed. N. Sharma, A. Chakrabarti and V.E. Balas), 451–462.
Singapore: Springer https://doi.org/10.1007/978-981-32-9949-8_31.
18 Ibrahim, N., Akhir, N.S.M., and Hassan, F.H. (2017). Predictive analysis effectiveness in determining the
epidemic disease infected area. AIP Conference Proceedings 1891: 020064. https://doi.org/10.1063/1.5005397.
19 Huang G, Hu J, He Y, Liu J, Ma M, Shen Z, et al. Machine learning for electronic design automation: a
AQ7 survey 2021.
20 Raeesi Vanani, I. and Amirhosseini, M. (2021). IoT-based diseases prediction and diagnosis system for health-
care. In: Internet of Things for Healthcare Technologies (ed. C. Chakraborty, A. Banerjee, M.H. Kolekar, et al.),
21–48. Singapore: Springer https://doi.org/10.1007/978-981-15-4112-4_2.
21 Deshields, T.L., Wells-Di Gregorio, S., Flowers, S.R. et al. (2021). Addressing distress management chal-
lenges: recommendations from the consensus panel of the American Psychosocial Oncology Society and the
Association of Oncology Social Work. CA: A Cancer Journal for Clinicians https://doi.org/10.3322/caac.21672.
22 Issac, A.C. and Baral, R. (2020). A trustworthy network or a technologically disguised scam: A biblio-
morphological analysis of bitcoin and blockchain literature. Global Knowledge, Memory and Communication
69: 443–460. https://doi.org/10.1108/GKMC-06-2019-0072.
23 Orhorhoro, E.K., Ebunilo, P., and Sadjere, G. (2017). Development of a predictive model for biogas yield using
artificial neural networks (ANNs) approach 2017. https://www.semanticscholar.org/paper/Development-of-a-
Predictive-Model-for-Biogas-Yield-Orhorhoro-Ebunilo/d65df521b9feceb147a01ca0949f318463bce7a9 (accessed
5 August 2021).
24 Khan, G.M. (2018). Evolution of Artificial Neural Development: In Search of Learning Genes. Springer Interna-
tional Publishing https://doi.org/10.1007/978-3-319-67466-7.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 18
k

18 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

25 Zhang, L., Tan, J., Han, D., and Zhu, H. (2017). From machine learning to deep learning: progress in machine
intelligence for rational drug discovery. Drug Discovery Today 22: 1680–1685. https://doi.org/10.1016/j.drudis
.2017.08.010.
26 Zhou, X., Li, Y., and Liang, W. (2021). CNN-RNN based intelligent recommendation for online medical
pre-diagnosis support. IEEE/ACM Transactions on Computational Biology and Bioinformatics 18: 912–921.
https://doi.org/10.1109/TCBB.2020.2994780.
27 Zendesk (2022). Deep learning vs. machine learning: what’s the difference? https://www.zendesk.com/blog/
machine-learning-and-deep-learning (accessed 5 August 2021).
28 Dike, H.U., Zhou, Y., Devarasetty, K.K., and Wu, Q. (2018). Unsupervised learning based on artificial neu-
ral network: a review. In: 2018 IEEE International Conference on Cyborg and Bionic Systems (CBS), 322–327.
https://doi.org/10.1109/CBS.2018.8612259.
29 Holzinger, A., Haibe-Kains, B., and Jurisica, I. (2019). Why imaging data alone is not enough: AI-based inte-
gration of imaging, omics, and clinical data. European Journal of Nuclear Medicine and Molecular Imaging
46: 2722–2730. https://doi.org/10.1007/s00259-019-04382-9.
30 Srivastava, A., Jain, S., Miranda, R. et al. (2021). Deep learning based respiratory sound analysis for detection
of chronic obstructive pulmonary disease. PeerJ Computer Science 7: e369. https://doi.org/10.7717/peerj-cs.369.
31 Myszczynska, M.A., Ojamies, P.N., Lacoste, A.M.B. et al. (2020). Applications of machine learning to diag-
nosis and treatment of neurodegenerative diseases. Nature Reviews Neurology 16: 440–456. https://doi.org/10
.1038/s41582-020-0377-8.
32 Bote-Curiel, L., Muñoz-Romero, S., Guerrero-Curieses, A., and Rojo-Álvarez, J.L. (2019). Deep learning and
big data in healthcare: a double review for critical beginners. Applied Sciences 9: 2331. https://doi.org/10.3390/
app9112331.
k 33 Bao, W., Lianju, N., and Yue, K. (2019). Integration of unsupervised and supervised machine learning algo- k
rithms for credit risk assessment. Expert Systems with Applications 128: 301–315. https://doi.org/10.1016/j.eswa
.2019.02.033.
34 Lai, W.-S., Huang, J.-B., and Yang, M.-H. (2017). Semi-supervised learning for optical flow with generative
adversarial networks. In: Advances in Neural Information Processing Systems, vol. 30. Curran Associates, Inc.
35 Lee, J.H., Shin, J., and Realff, M.J. (2018). Machine learning: overview of the recent progresses and implica-
tions for the process systems engineering field. Computers and Chemical Engineering 114: 111–121. https://doi
.org/10.1016/j.compchemeng.2017.10.008.
36 Won, Y.-S. and Bhasin, S. (2021). On use of deep learning for side channel evaluation of black box hard-
ware AES engine. In: Industrial Networks and Intelligent Systems. (ed. N.S. Vo, V.-P. Hoang and Q.-T. Vien),
185–194. Cham: Springer International Publishing https://doi.org/10.1007/978-3-030-77424-0_15.
37 Yao, X., Wang, X., Wang, S.-H., and Zhang, Y.-D. (2020). A comprehensive survey on convolutional neural
networks in medical image analysis. Multimedia Tools and Applications https://doi.org/10.1007/s11042-020-
09634-7.
38 Shao, H., Jiang, H., Lin, Y., and Li, X. (2018). A novel method for intelligent fault diagnosis of rolling bear-
ings using ensemble deep auto-encoders. Mechanical Systems and Signal Processing 102: 278–297. https://doi
.org/10.1016/j.ymssp.2017.09.026.
39 Liu, R., Rong, Z., Jiang, B. et al. (2018). Soft sensor of 4-CBA concentration using deep belief networks with
continuous restricted Boltzmann machine. In: 2018 5th IEEE International Conference on Cloud Computing
and Intelligence Systems (CCIS), 421–424. https://doi.org/10.1109/CCIS.2018.8691166.
40 Li, Z., Wang, Y., and Wang, K. (2020). A data-driven method based on deep belief networks for backlash error
prediction in machining centres. Journal of Intelligent Manufacturing 31: 1693–1705.
41 Yamashita, R., Nishio, M., Do, R.K.G., and Togashi, K. (2018). Convolutional neural networks: an overview
and application in radiology. Insights Into Imaging 9: 611–629. https://doi.org/10.1007/s13244-018-0639-9.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 19
k

References 19

42 Ulku, I. and Akagunduz, E. (2021). A survey on deep learning-based architectures for semantic segmentation
on 2D images. ArXiv191210230 Cs.
43 Abduljabbar, R., Dia, H., Liyanage, S., and Bagloee, S.A. (2019). Applications of artificial intelligence in trans-
port: an overview. Sustainability 11: 189. https://doi.org/10.3390/su11010189.
44 Asaoka, R., Murata, H., Hirasawa, K. et al. (2019). Using deep learning and transfer learning to accurately
diagnose rarly-onset glaucoma from macular optical coherence tomography images. American Journal of
Ophthalmology 198: 136–145. https://doi.org/10.1016/j.ajo.2018.10.007.
45 Shibata, N., Tanito, M., Mitsuhashi, K. et al. (2018). Development of a deep residual learning algorithm to
screen for glaucoma from fundus photography. Scientific Reports 8: 146–165. https://doi.org/10.1038/s41598-
018-33013-w.
46 Hood, D.C. and De Moraes, C.G. (2018). Efficacy of a deep learning system for detecting glaucomatous optic
neuropathy based on colour fundus photographs. Ophthalmology 125: 1207–1208. https://doi.org/10.1016/j
.ophtha.2018.04.020.
47 Zhou, X., Zhang, X., and Wang, B. (2016). Online support vector machine: a survey. In: Harmony Search
Algorithm (ed. J.H. Kim and Z.W. Geem), 269–278. Berlin, Heidelberg: Springer 10.1007/978-3-662-47926-1_26.
48 Khan, Z., Gul, A., Perperoglou, A. et al. (2020). Ensemble of optimal trees, random forest and random projec-
tion ensemble classification. Advances in Data Analysis and Classification 14: 97–116. https://doi.org/10.1007/
s11634-019-00364-9.
49 Shadkani, S., Abbaspour, A., Samadianfard, S. et al. (2021). Comparative study of multilayer perceptron-
stochastic gradient descent and gradient boosted trees for predicting daily suspended sediment load: The case
study of the Mississippi River, U.S. International Journal of Sediment Research 36: 512–523. https://doi.org/10
.1016/j.ijsrc.2020.10.001.
50 Sen, P.C., Hajra, M., and Ghosh, M. (2020). Supervised classification algorithms in machine learning: a sur-
k k
vey and review. In: Emerging Technology in Modelling and Graphics (ed. J.K. Mandal and D. Bhattacharya),
99–111. Singapore: Springer https://doi.org/10.1007/978-981-13-7403-6_11.
51 Hussain, W. and Sohaib, O. (2019). Analysing cloud QoS prediction approaches and its control parameters:
considering overall accuracy and freshness of a dataset. IEEE Access 7: 82649–82671. https://doi.org/10.1109/
ACCESS.2019.2923706.
52 Li, Z., Liu, F., Yang, W. et al. (2021). A survey of convolutional neural networks: analysis, applications, and
prospects. IEEE Transactions on Neural Networks and Learning Systems https://doi.org/10.1109/TNNLS.2021
.3084827.
53 Alkhodari, M. and Fraiwan, L. (2021). Convolutional and recurrent neural networks for the detection of
valvular heart diseases in phonocardiogram recordings. Computer Methods and Programs in Biomedicine 200:
105940. https://doi.org/10.1016/j.cmpb.2021.105940.
54 Lindemann, B., Müller, T., Vietz, H. et al. (2021). A survey on long short-term memory networks for time
series prediction. Procedia CIRP 99: 650–655. https://doi.org/10.1016/j.procir.2021.03.088.
55 Azarbik, M. and Sarlak, M. (2020). Real-time transient stability assessment using stacked auto-encoders.
COMPEL - The International Journal for Computation and Mathematics in Electrical and Electronic
Engineering 39: 971–990. https://doi.org/10.1108/COMPEL-12-2019-0477.
56 Gm, H., Gourisaria, M.K., Pandey, M., and Rautaray, S.S. (2020). A comprehensive survey and analysis of
generative models in machine learning. Computer Science Review 38: 100285. https://doi.org/10.1016/j.cosrev
.2020.100285.
57 Sohn, I. (2021). Deep belief network based intrusion detection techniques: a survey. Expert Systems with
Applications 167: 114170. https://doi.org/10.1016/j.eswa.2020.114170.
58 Oryema, B., Kim, H.-S., Li, W., and Park, J.T. (2017). Design and implementation of an interoperable messag-
ing system for IoT healthcare services. In: 2017 14th IEEE Annual Consumer Communications & Networking
Conference (CCNC), 45–52. https://doi.org/10.1109/CCNC.2017.7983080.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 20
k

20 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

59 Birje, M.N. and Hanji, S.S. (2020). Internet of things based distributed healthcare systems: a review. Journal of
Data, Information and Management 2: 149–165. https://doi.org/10.1007/s42488-020-00027-x.
60 Ahad, A., Tahir, M., and Yau, K.-L.A. (2019). 5G-based smart healthcare network: architecture, taxonomy,
challenges and future research directions. IEEE Access 7: 100747–100762. https://doi.org/10.1109/ACCESS
.2019.2930628.
61 Kadhim, K.T., Alsahlani, A.M., Wadi, S.M., and Kadhum, H.T. (2020). An overview of patient’s health sta-
tus monitoring system based on internet of things (IoT). Wireless Personal Communications 114: 2235–2262.
https://doi.org/10.1007/s11277-020-07474-0.
62 Tekeste, T., Saleh, H., Mohammad, B., and Ismail, M. (2019). Ultra-low power QRS detection and ECG com-
pression architecture for IoT healthcare devices. IEEE Transactions on Circuits and Systems I: Regular Papers
66: 669–679. https://doi.org/10.1109/TCSI.2018.2867746.
63 Xu, J., Xue, K., and Zhang, K. (2019). Status and future trends of clinical diagnoses via image-based deep
learning. Theranostics 9: 7556–7565. https://doi.org/10.7150/thno.38065.
64 Kim, M., Yun, J., Cho, Y. et al. (2019). Deep learning in medical imaging. Neurospine 16: 657–668. https://doi
.org/10.14245/ns.1938396.198.
65 Michie, S., Thomas, J., Johnston, M. et al. (2017). The Human Behaviour-Change Project: harnessing the
power of artificial intelligence and machine learning for evidence synthesis and interpretation. Implementa-
tion Science 12: 121. https://doi.org/10.1186/s13012-017-0641-5.
66 Shah, P., Kendall, F., Khozin, S. et al. (2019). Artificial intelligence and machine learning in clinical
development: a translational perspective. npj Digital Medicine 2: 1–5. https://doi.org/10.1038/s41746-019-
0148-3.
67 Lin, W.-C., Chen, J.S., Chiang, M.F., and Hribar, M.R. (2020). Applications of artificial intelligence to elec-
k tronic health record data in ophthalmology. Translational Vision Science & Technology 9: 13. https://doi.org/10 k
.1167/tvst.9.2.13.
68 Wynants, L., Calster, B.V., Collins, G.S. et al. (2020). Prediction models for diagnosis and prognosis of
covid-19: systematic review and critical appraisal. BMJ 369: m1328. https://doi.org/10.1136/bmj.m1328.
69 Ahmed, Z., Mohamed, K., Zeeshan, S., and Dong, X. (2020). Artificial intelligence with multi-functional
machine learning platform development for better healthcare and precision medicine. Database: The Journal
of Biological Databases and Curation 2020: baaa010. https://doi.org/10.1093/database/baaa010.
70 Atitallah, S.B., Driss, M., Boulila, W., and Ghézala, H.B. (2020). Leveraging deep learning and IoT big data
analytics to support the smart cities development: review and future directions. Computer Science Review
38: 100303. https://doi.org/10.1016/j.cosrev.2020.100303.
71 El-Sappagh, S., Ali, F., El-Masri, S. et al. (2019). Mobile health technologies for diabetes mellitus: current state
and future challenges. IEEE Access 7: 21917–21947. https://doi.org/10.1109/ACCESS.2018.2881001.
72 Da Costa, C.A., Pasluosta, C.F., Eskofier, B. et al. (2018). Internet of health things: toward intelligent vital
signs monitoring in hospital wards. Artificial Intelligence in Medicine 89: 61–69. https://doi.org/10.1016/j
.artmed.2018.05.005.
73 Tuli, S., Basumatary, N., Gill, S.S. et al. (2020). HealthFog: an ensemble deep learning based Smart Healthcare
System for Automatic Diagnosis of Heart Diseases in integrated IoT and fog computing environments. Future
Generation Computer Systems 104: 187–200. https://doi.org/10.1016/j.future.2019.10.043.
74 Sarraf, S. and Tofighi, G. (2016). Classification of Alzheimer’s disease using fMRI data and deep learning con-
AQ8 volutional neural networks. ArXiv160308631 Cs.
75 Kamnitsas, K., Ledig, C., Newcombe, V.F.J. et al. (2017). Efficient multi-scale 3D CNN with fully connected
CRF for accurate brain lesion segmentation. Medical Image Analysis 36: 61–78. https://doi.org/10.1016/j.media
.2016.10.004.
76 Faust, O., Shenfield, A., Kareem, M. et al. (2018). Automated detection of atrial fibrillation using a long
short-term memory network with RR interval signals. Computers in Biology and Medicine 102: 327–335.
https://doi.org/10.1016/j.compbiomed.2018.07.001.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 21
k

References 21

77 Ren, J.-L., Zhang, A.-H., and Wang, X.-J. (2020). Traditional Chinese medicine for COVID-19 treatment.
Pharmacological Research 155: 104743. https://doi.org/10.1016/j.phrs.2020.104743.
78 Bray, F., Ferlay, J., Soerjomataram, I. et al. (2018). Global cancer statistics 2018: GLOBOCAN estimates of
incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians 68:
394–424. https://doi.org/10.3322/caac.21492.
79 Ma, Z. and Tavares, J.M.R.S. (2017). Effective features to classify skin lesions in dermoscopic images. Expert
Systems with Applications 84: 92–101. https://doi.org/10.1016/j.eswa.2017.05.003.
80 Schirrmeister, R.T., Springenberg, J.T., Fiederer, L.D.J. et al. (2017). Deep learning with convolutional neural
networks for EEG decoding and visualisation. Human Brain Mapping 38: 5391–5420. https://doi.org/10.1002/
hbm.23730.
81 Mumpe-Mwanja, D., Barlow-Mosha, L., Williamson, D. et al. (2019). A hospital-based birth defects surveil-
lance system in Kampala, Uganda. BMC Pregnancy and Childbirth 19: 372. https://doi.org/10.1186/s12884-
019-2542-x.
82 Verma, P. and Fatima, S. (2020). Smart healthcare applications and real-time analytics through edge comput-
ing. In: Internet of Things Use Cases for the Healthcare Industry (ed. P. Raj, J.M. Chatterjee, A. Kumar and
B. Balamurugan), 241–270. Cham: Springer International Publishing https://doi.org/10.1007/978-3-030-37526-
3_11.
83 Pradhan, B., Bhattacharyya, S., and Pal, K. (2021). IoT-based applications in healthcare devices. Journal of
Healthcare Engineering 2021: e6632599. https://doi.org/10.1155/2021/6632599.
84 Perera, C., Zaslavsky, A., Christen, P., and Georgakopoulos, D. (2014). Context aware computing for the inter-
net of things: a survey. IEEE Communication Surveys and Tutorials 16: 414–454. https://doi.org/10.1109/SURV
.2013.042313.00197.
85 Jan, S.U., Ali, S., Abbasi, I.A. et al. (2021). Secure patient authentication framework in the healthcare system
k using wireless medical sensor networks. Journal of Healthcare Engineering 2021: e9954089. https://doi.org/10 k
.1155/2021/9954089.
86 Banerjee, A., Chakraborty, C., Kumar, A., and Biswas, D. (2020). Chapter 5 - Emerging trends in IoT and
big data analytics for biomedical and health care technologies. In: Handbook of Data Science Approaches
for Biomedical Engineering (ed. V.E. Balas, V.K. Solanki, R. Kumar and M. Khari), 121–152. Academic Press
https://doi.org/10.1016/B978-0-12-818318-2.00005-2.
87 Farahani, B., Firouzi, F., Chang, V. et al. (2018). Towards fog-driven IoT eHealth: promises and challenges of
IoT in medicine and healthcare. Future Generation Computer Systems 78: 659–676. https://doi.org/10.1016/j
.future.2017.04.036.
88 Aghdam, Z.N., Rahmani, A.M., and Hosseinzadeh, M. (2021). The role of the internet of things in healthcare:
future trends and challenges. Computer Methods and Programs in Biomedicine 199: 105903. https://doi.org/10
.1016/j.cmpb.2020.105903.
89 Haider, S.A., Adil, M.N., and Zhao, M. (2020). Optimization of secure wireless communications for IoT
networks in the presence of eavesdroppers. Computer Communications 154: 119–128. https://doi.org/10.1016/j
.comcom.2020.02.027.
90 Chen, Y., Song, B., Du, X., and Guizani, N. (2020). The enhancement of catenary image with low visibility
based on multi-feature fusion network in railway industry. Computer Communications 152: https://doi.org/10
.1016/j.comcom.2020.01.040.
91 Irshad, O., Khan, M.U.G., Iqbal, R. et al. (2020). Performance optimization of IoT based biological systems
using deep learning. Computer Communications 155: 24–31. https://doi.org/10.1016/j.comcom.2020.02.059.
92 Ali, R., Kim, B., Kim, S.W. et al. (2020). (ReLBT): a reinforcement learning-enabled listen before talk mecha-
nism for LTE-LAA and Wi-Fi coexistence in IoT. Computer Communications 150: 498–505. https://doi.org/10
.1016/j.comcom.2019.11.055.
93 Khan, S., Alvi, A.N., Javed, M.A. et al. (2021). An efficient medium access control protocol for RF energy
harvesting based IoT devices. Computer Communications 171: 28–38. https://doi.org/10.1016/j.comcom.2021.02
.011.

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm Page 22
k

22 19 Impact of IoT in Biomedical Applications Using Machine and Deep Learning

94 Grande, E. and Beltrán, M. (2020). Edge-centric delegation of authorization for constrained devices in the
internet of things. Computer Communications 160: 464–474. https://doi.org/10.1016/j.comcom.2020.06.029.
95 Anurag, M.S.R., Rahmani, A.-M., Westerlund, T. et al. (2014). Pervasive health monitoring based on
Internet of Things: Two case studies. In: 2014 4th International Conference on Wireless Mobile Communi-
cation and Healthcare-Transforming Healthcare Through Innovations in Mobile and Wireless Technologies
(MOBIHEALTH), 275–278. https://doi.org/10.1109/MOBIHEALTH.2014.7015964.
96 Thimbleby, H. (2013). Technology and the future of healthcare. Journal of Public Health Research 2: e28.
https://doi.org/10.4081/jphr.2013.e28.
97 Kulkarni, A. and Sathe, S. (2014). Healthcare applications of the internet of things: a review. International
Journal of Computer Science and Information Technologies 5 (5): 6229–6232.
98 Uckelmann, D., Harrison, M., and Michahelles, F. (2011). An architectural approach towards the future inter-
net of things. In: Architecting the Internet of Things (ed. D. Uckelmann, M. Harrison and F. Michahelles),
1–24. Berlin, Heidelberg: Springer https://doi.org/10.1007/978-3-642-19157-2_1.
99 Distefano, S., Bruneo, D., Longo, F. et al. (2017). Hospitalised patient monitoring and early treatment using
IoT and cloud. BioNanoScience 7: 382–385. https://doi.org/10.1007/s12668-016-0335-5.
100 Chan, W.M., Zhao, Y., and Tsui, K.L. (2017). Implementation of electronic health monitoring systems at the
community level in Hong Kong. In: Smart Health (ed. H. Chen, D.D. Zeng, E. Karahanna and I. Bardhan),
94–103. Cham: Springer International Publishing https://doi.org/10.1007/978-3-319-67964-8_9.
101 Segura, A.S. (2016). The internet of things: business applications, technology acceptance, and future prospects,
Doctoral Thesis. Universität Würzburg.

k k

k
 Trim Size: 8in x 10in Single Column Tripathi861829 c19.tex V1 - 09/04/2022 10:30pm
k

Author Queries
AQ1 Please note that ORCID number in affiliation is deleted.
AQ2 Please provide institute and city for affiliation 3.
AQ3 Please provide department, institute, and city for affiliation 4.
AQ4 Both automatic encoder and auto encoder are used for AE. Please check and specify the
which one should be used globally.
AQ5 Please check and confirm the representation of "O. (1)."
AQ6 Please provide publisher name and location for references [10, 11, 16, 28, 39, 58, 95].
AQ7 Please provide journal title, volume, and page range for reference [19].
AQ8 Please provide complete details for references [42, 74].

k k

k
View publication stats