Toward a Digitalized Holistic and Integrated Healthcare Vision

Driven by Deep Learning and IoT

Hayat Zaydi and Zohra Bakkoury

École Mohammadia d’Ingénieurs, Mohammed V University in Rabat, Ibn Sina Street, RABAT, Morocco

Keywords: Survey, Deep Learning, Applications, IoT, Healthcare, Holistic Approach.

Abstract: Internet of Things, deep learning, and smart Healthcare are terms that have been very popular over the last

decade, with hundreds of searches being conducted around the world on one or more aspects related to all

three words. Some of them are surveys on the use of Deep Learning in Healthcare, and others are surveys on

the use of Deep Learning in sensor networks; some have focused on the emergence of IoT in Industry 4.0 for

the Healthcare sector or deployment of deep neural network architecture in sensor networks. The present work

is a shortcut to several studies on different aspects, all dealing with three elements mentioned above, giving a

critical analysis of missed aspects or which would be bringing more value to these works.

INTRODUCTION

Deep Learning, IoT, and smart Healthcare have been

increasingly co-existing and cooperating in the last

few years.

Indeed, the fast progress in technology that the

world of data has experienced, specifically the

emergence of management, storage, and processing

of voluminous data platforms, contributed to this.

This volume, which was a significant challenge for

infrastructure that existed not very long ago, is itself

a valuable asset that has given the height of artificial

intelligence technology with all its sub-domains,

particularly Machine Learning and Deep Learning.

This huge volume of data is also due, among other

sources, to the connected objects explosion with

arrival of IoT concept, which is one of the areas that

have experienced an exponential emergence and

development of fastest in history of information

technology, with more than 50 billion devices at the

end of the previous year 2020 (Ray et al., 2016). This

field [(Abawajy & Hassan, 2017)] has a wide range of

use cases, from industry, telecommunications,

entertainment, smart cities, smart homes to one of the

most sensitive areas for humanity; I refer to the field

of health and medicine, continues to generate interest

and is driving studies and research, especially in the

age of industry 4.0.

Our interest focuses on the intersection of these

three technologies, in particular deep learning

techniques applied to data collected from connected

medical objects, which is an Artificial Intelligence

aspect for the internet of medical things IoMT.

The present work offers a shortcut to studies and

research related to one or two elements of the triplet

of our paper (DL, IoT, and Healthcare), with the

purpose to summarize, analyze and discuss research

addressed in this work; moreover, we will bring out

points and axis that these researches could have

strengthened, that can open new researches

perspectives.

This paper is structured as follows, after

introduction, section 2 presents methodology

followed, Section 3 discusses the core of the work,

namely the presentation of related work, discussion

and criticism with the opening of new horizons and

new research axis and, then a conclusion and

references used to construct this paper.

METHODOLOGY

The survey in this article is conducted as follows:

After identifying an extensive collection of articles

from 2016 to 2020. An in-depth reading has been

done with a view to identifying the essence of

concerned article, article type and, key points

addressed.

After, a cross-analysis was carried out on all of

these summaries, which allowed us to identify the

Zaydi, H. and Bakkoury, Z.

Toward a Digitalized Holistic and Integrated Healthcare Vision Driven by Deep Learning and IoT.

DOI: 10.5220/0010727900003101

In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning (BML 2021), pages 33-37

ISBN: 978-989-758-559-3

shortcomings in the insights discussed above and to

conclude the need for a shortened point to these

research panoplies presented in our paper; figure 1

summarizes said methodology.

Figure 1: Followed methodology

SURVEY: LITERATURE AND

CRITICAL ANALYSIS REVIEW

In (Gavrilović & Mishra, 2020) paper’s, the software

architecture part dedicated to the IoT for Healthcare

was addressed. They first proposed a global

architecture for all sectors potentially affected by IoT,

as shown in Figure 2.

Figure 2: Global IoT architecture for a variety of industries

(Sethi & Sarangi, 2017)

They then proposed dedicated architectures for

three sectors: Smart Cities, Healthcare, and

Agriculture.

They also emphasized the importance of using a

software architecture dedicated to the Healthcare

sector, as shown in Figure 2. To meet

IoT system

requirements

, any IoT architecture must be functional,

scalable, available, and resilient. In the Healthcare

field, an IoT architecture allows the interconnection

of several functionalities and collects data from

several sensors types (portable, inertial, location, and

physiological sensors) using a dedicated sensor layer

(Kamienski et al., 2017).

(Yao et al., 2018) have attempted to provide

answers to 4 key questions to propose and design a

Deep Learning framework applied to IoT data from

sensor networks; these questions are: What neural

network structures can efficiently process and merge

data from connected objects? How to take into

account the low consumption of resources by the

connected objects in the design of neural network

architectures, given that these architectures have a

very high resources consumption? What are the

metrics to be measured to determine the relevance of

Deep Learning models for the IoT? And finally, they

highlighted the significant need for labelled data in

the training process in these cases and the need to

minimize it for performance gains. They also

proposed using a neural network compression

algorithm called Deep IoT, and for measures of

accuracy of obtained results, they proposed the use of

the RdeepSense algorithm (Yao et al., 2017), which

allows making precise and well-calibrated estimates

by changing objective function.

In the same spirit (Fadlullah et al., 2018)

have proposed a solution based on deep neural

networks, specifically the CNN convolutional neural

network on the data retrieved from the IoT devices of

individual users to translate the analysis and

processing of these data from the Cloud to the edge

of the IoT network to overcome the problem of non-

tolerance to delay of certain types of data that are

sensitive in the Healthcare process.

The proposed solution involves three phases,

namely data collection, chosen neural network

training , and prediction from new data via the model

generated in the second phase.

They evaluated the relevance of the proposed

solution through Python programming of a simple

neural network using the Keras library, the Theano

library and Tensorflow.

This solution could have been extended by testing

its operation and measuring its relevance at the three

possible sites in IoT network architectures, namely

Cloud, Edge, and Fog Computing.

On the other hand, the application of Deep

Learning and Machine Learning algorithms to

Healthcare data, in general, has been the subject of

many articles and research works.

Among these researches, we have (Pandey &

Janghel, n.d.); according to this work, the most

relevant deep neural network architectures for these

use cases are Autoencoder, Restricted Boltzman

Machine, Deep Belief Network, Recurrent Neural

Literature review on the interval

[2016, 2020]

Summaries cross-analysis and

deficiencies identification.

Result Need for a shortcut to

these studies.

BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)

Network, Convolutional Neural Network, and

Generative Adversarial Networks.

All these Neural Network architectures apply to

the Healthcare field, with both advantages and

disadvantages.

The application of these deep neuronal

architectures on data obtained from medical IoTs has

been the subject of another research.

Indeed, (Dourado et al., 2021) have proposed an

online Framework in the context of computational

intelligence to be used with IoT devices.

This Framework has the particularity of not

requiring any advanced skills or knowledge in AI or

image processing: any user can load his images, and

perform training of algorithm and then generate the

appropriate model.

This Framework is validated using data from three

medical databases and is based on the CNN

(Convolutional Neural Network), which is one of the

most important neural network architectures; CNNs

are mainly used to classify images, group them by

similarities, and then perform object recognition.

Many algorithms using NDCs can identify faces,

traffic signs, and animals.

(Granados et al., 2018) have proposed an IoT

platform for ECG analysis (electrocardiogram, which

is a graphical representation of heart's electrical

activity using connected electrodes placed on

patient's skin in the area of heart); this intelligent

system has consisted of connected ECG sensors, a

web gateway from the smartphone and a server in the

Cloud integrating a deep neural network as shown in

the following figure 3:

Figure 3: Cloud-based ECG analysis platform based on Deep Neural Network and the IoT (Granados et al., 2018)

(Suneetha et al., 2020) proposed a review of some

works that have proposed platforms for analyzing

Healthcare data collected from IoT devices to

diagnose and predict diseases to anticipate and

improve people's health global condition.

The authors of this work reviewed 12 papers.

{ (Kaur et al., 2018); (Lakshmanaprabu et al., 2019);

(F. Ali et al., 2018) ; (S. A. Ali et al., 2020) ; (Chen

et al., 2017) ;(Din et al., 2019) ;(Satapathy et al.,

2015); (Gupta et al., 2017) ; (Liu et al., 2019) ;(Kaur

et al., 2018) ; (Samuel et al., 2017) ; (Zhang et al.,

2018) }

And then, they listed the types of Deep Neural

Architectures used, the number of diseases involved

in each work, the constraints faced as well as the

metrics and methods used to measure the relevance of

each model generated.

Most of the articles approached by this review

deal with heart disease cases; thus, it is concluded that

the study is interesting. It would be more interesting

if it covered a wide spectrum of diseases; moreover,

the datasets used are not all generated by the IoT

devices.

As concerns (Zikria et al., 2020), they discussed

the

application of Deep Learning algorithms on smart

IoT;

they also highlighted the need to orient the

conception and design of next-generation wireless

networks towards high autonomy and robustness in

order to overcome the limitations of IoT devices,

especially in terms of computing power, energy

autonomy, and memory. They also highlighted the

significant challenge of merging the techniques of

Deep Learning and Machine Learning with the

functionalities provided by IoT devices to improve

IoT applications. The optimization of these

applications depends on the software and hard

architectures as well as the site of implementation;

they can be deployed in the Cloud, Edge, or Fog

computing.

Toward a Digitalized Holistic and Integrated Healthcare Vision Driven by Deep Learning and IoT

(Irshad et al., 2020) have gone beyond the building

of intelligent systems using Deep Learning, IoT for

health care data analysis; they have focused on

optimizing these systems, especially biological

systems. In this perspective, they proposed a smart

system based on RNN (Recurrent Neural Network)

and LSTM (Long short-term memory) combined with

state-of-the-art probabilistic methods to predict

performance needs and anticipate situations. This

system would have more impact and usability if

expanded to cover all intelligent Healthcare systems,

specifically those based on AI and IoT. Other research

has focused on the security aspect of Deep learning

and IoT and the privacy respect in the Healthcare

field. In this context (Thakkar & Lohiya, 2020) have

focused on using Deep Learning and Machine

Learning algorithms to predict potential intrusions in

IoT devices in all fields, particularly the detection of

attacks in IoT networks. These intrusions are

increasingly numerous, and they are very harmful, in

particular, in the healthcare domain because there is a

need to preserve patients' privacy while handling data

collected from IoMTs;

In this research, there is neither a segmentation of

security risks nor a focus on the criticality of this issue

at Healthcare level.

DISCUSSION

Nowadays, research is focused on challenges posed

by data specificities collected from IoMT, as well as

on nature of devices that generate these data, in

particular, sensors with memory, energy autonomy,

and computing power constraints.

Others focused on the technical aspect by

presenting the mathematical and algorithmic details

of the various deep neuronal architectures. While

others have been focused on the aspect of using these

algorithms for a specific purpose, which may be the

intrusions detection in IoT networks, the automation

of Healthcare processes through the application of DL

on medical sensor data, or purely medical purpose

including the prediction of various diseases through

images analysis, patient’s data analysis and patient's

vital signal This is done for several purposes,

including early intervention by the medical

profession, the prescription of more appropriate

treatments, prediction of the evolution of a disease

and more.

In this paper, we have discussed the gaps, the

omitted facets, the aspects that could have been

included in each research overview as it is discussed.

The main thing we have concluded is the fact that

while there has been an enormous amount of research

on these topics, mainly on Deep Learning, Machine

Learning, IoT, Healthcare, security, software

architecture, we lack a holistic approach to bring all

these elements together in the same Framework, the

same study, the same survey, thus providing a single,

shortcut access to this large and vast world structured

around three dimensions, notably Deep Learning, IoT

and Healthcare, with all the hidden aspects behind

each of these three dimensions.

We propose, after this study, to design, develop

and deploy a Framework that fulfills this holistic

approach.

CONCLUSIONS

In this paper, we have tried to summarize all these

aspects describing synergies and interactions between

Deep Learning, IoT, and the health industry, to create,

first, a shortcut to these different aspects from a single

point which is this article, second; we have

highlighted the lack of holistic study and research that

integrates all these elements into a single survey.

The purpose of this article is to bring all these

elements together in one place with, as a scientific

contribution, a layer of analysis and discussion.

This work also allows us to focus, for the next

work, on application aspects related to these same

topics to propose design, development, and

deployment of a global solution, and this, by

combining Deep Learning, Machine Learning, and

IoT techniques for Healthcare.

Indeed, based on this study, we propose to design,

develop and deploy a Framework that meets this

global approach. It is planned to design the

Framework to be customizable according to a specific

configuration (various types of Deep Neural

Architectures) and depending on the use case and the

location of its deployment (Cloud, Edge or Fog or

other depending on the evolution of science) along

with a security dimension considering its crucial

aspect.

REFERENCES

Abawajy, J. H., & Hassan, M. M. (2017). Federated

Internet of Things and Cloud Computing Pervasive

Patient Health Monitoring System. IEEE

Communications Magazine, 55(1), 48–53.

Dourado, C. M. J. M., Da Silva, S. P. P., Da Nobrega, R.

V. M., Reboucas Filho, P. P., Muhammad, K., & De

Albuquerque, V. H. C. (2021). An Open IoHT-Based

BML 2021 - INTERNATIONAL CONFERENCE ON BIG DATA, MODELLING AND MACHINE LEARNING (BML’21)

Deep Learning Framework for Online Medical Image

Recognition. IEEE Journal on Selected Areas in

Communications, 39(2), 541–548.

Fadlullah, Z. M., Pathan, A. S. K., & Gacanin, H. (2018).

On Delay-Sensitive Healthcare Data Analytics at the

Network Edge Based on Deep Learning. 2018 14th

International Wireless Communications and Mobile

Computing Conference, IWCMC 2018, 388–393.

Gavrilović, N., & Mishra, A. (2020). Software architecture

of the internet of things (IoT) for smart city, Healthcare

and agriculture: analysis and improvement directions.

Journal of Ambient Intelligence and Humanized

Computing, 0123456789.

Granados, J., Westerlund, T., Zheng, L., & Zou, Z. (2018).

IoT platform for real-time multichannel ECG

monitoring and classification with neural networks. In

Lecture Notes in Business Information Processing (Vol.

310). Springer International Publishing.

Irshad, O., Khan, M. U. G., Iqbal, R., Basheer, S., & Bashir,

A. K. (2020). Performance optimization of IoT based

biological systems using deep learning. Computer

Communications, 155(September 2019), 24–31.

Kamienski, C., Jentsch, M., Eisenhauer, M., Kiljander, J.,

Ferrera, E., Rosengren, P., Thestrup, J., Souto, E.,

Andrade, W. S., & Sadok, D. (2017). Application

development for the Internet of Things: A context-

aware mixed criticality systems development platform.

Computer Communications, 104(2016), 1–16.

Pandey, S. K., & Janghel, R. R. (n.d.). Recent Deep

Learning Techniques, Challenges and Its Applications

for Medical Healthcare System: A Review. 50(2), 1907–

1935.

Ray, S., Jin, Y., & Raychowdhury, A. (2016). The Changing

Computing Paradigm with Internet of Things: A

Tutorial Introduction. IEEE Design and Test, 33(2),

76–96.

Sethi, P., & Sarangi, S. R. (2017). Internet of Things:

Architectures, Protocols, and Applications. Journal of

Electrical and Computer Engineering, 2017.

Suneetha, K. C., Shanmuka Shalini, R., Kumar Vadladi, V.,

& Mounica, M. (2020). Disease prediction and

diagnosis system in cloud based IoT: A review on deep

learning techniques.

Thakkar, A., & Lohiya, R. (2020). A Review on Machine

Learning and Deep Learning Perspectives of IDS for

IoT: Recent Updates, Security Issues, and Challenges.

In Archives of Computational Methods in Engineering

(Issue 0123456789). Springer Netherlands.

Yao, S., Zhao, Y., Shao, H., Zhang, A., Zhang, C., Li, S., &

Abdelzaher, T. (2017). RDeepSense: Reliable Deep

Mobile Computing Models with Uncertainty

Estimations. ArXiv, 1(4).

Yao, S., Zhao, Y., Zhang, A., Hu, S., Shao, H., Zhang, C., &

Su, L. (2018). Cover Feature Embedded Deep

Learning.

Zikria, Y. Bin, Afzal, M. K., Kim, S. W., Marin, A., &

Guizani, M. (2020). Deep learning for intelligent IoT:

Opportunities, challenges and solutions. Computer

Communications, 164, 50–53.

Ali, F., Islam, S. M. R., Kwak, D., Khan, P., Ullah, N., Yoo,

S. jo, & Kwak, K. S. (2018). Type-2 fuzzy ontology aided

recommendation systems for IoT–based

Healthcare. Computer Communications, 119, 138–155.

Ali, S. A., Raza, B., Malik, A. K., Shahid, A. R., Faheem,

M., Alquhayz, H., & Kumar, Y. J. (2020). An Optimally

Configured and Improved Deep Belief Network (OCI-

DBN) Approach for Heart Disease Prediction Based on

Ruzzo-Tompa and Stacked Genetic Algorithm. IEEE,

Access, 8, 65947–65958.

Chen, M., Hao, Y., Hwang, K., Wang, L., & Wang, L.

(2017). Disease Prediction by Machine Learning over

Big Data from Healthcare Communities. IEEE Access,

5(c), 8869–8879.

Din, I. U., Guizani, M., Rodrigues, J. J. P. C., Hassan, S.,

& Korotaev, V. V. (2019). Machine learning in the

Internet of Things: Designed techniques for smart

cities. Future Generation Computer Systems, 100, 826–

843.

Gupta, N., Ahuja, N., Malhotra, S., Bala, A., & Kaur, G.

(2017). Intelligent heart disease prediction in cloud

environment through ensembling. Expert Systems,

34(3), 1–14.

Kaur, P., Sharma, M., & Mittal, M. (2018). Big Data and

Machine Learning Based Secure Healthcare

Framework. Procedia Computer Science, 132, 1049–

1059.

Lakshmanaprabu, S. K., Mohanty, S. N., S., S. R.,

Krishnamoorthy, S., Uthayakumar, J., & Shankar, K.

(2019). Online clinical decision support system using

optimal deep neural networks. Applied Soft Computing

Journal, 81, 105487.

Liu, Y., Zhang, L., Yang, Y., Zhou, L., Ren, L., Wang, F.,

Liu, R., Pang, Z., & Deen, M. J. (2019). A Novel Cloud-

Based Framework for the Elderly Healthcare Services

Using Digital Twin. IEEE Access, 7, 49088–49101.

Samuel, O. W., Asogbon, G. M., Sangaiah, A. K., Fang, P.,

& Li, G. (2017). An integrated decision support system

based on ANN and Fuzzy_AHP for heart failure risk

prediction. Expert Systems with Applications, 68, 163–

172.

Satapathy, S. C., Govardhan, A., Srujan Raju, K., &

Mandal, J. K. (2015). Emerging ICT for Bridging the

Future - Proceedings of the 49th Annual Convention of

the Computer Society of India (CSI) Volume 1.

Advances in Intelligent Systems and Computing,

337(V), 371–372.

Zhang, C., Zhu, L., Xu, C., & Lu, R. (2018). PPDP: An

efficient and privacy-preserving disease prediction

scheme in cloud-based e-Healthcare system. Future

Generation Computer Systems, 79, 16–25.

Toward a Digitalized Holistic and Integrated Healthcare Vision Driven by Deep Learning and IoT