A System Architecture to Implement Deep Learning Techniques for

Patients Monitoring with Heart Disease: Case of Telerehabilitation

Khaoula Slime

, Abderrahim Maizate

, Larbi Hassouni

, and Najat Mouine

ENSEM, UH2, Casablanca, Morocco

Mohamed V Military Hospital, UM5, Rabat, Morocco

Keywords: Heart disease, telemedicine, deep learning, telerehabilitation.

Abstract: The development of information technologies and the introduction of artificial intelligence in the healthcare

sector is becoming a great wealth. It allows timely remote diagnosis by specialists and then reduces the cost

of unnecessary transfers. Nowadays, older people and the increase of chronic diseases pose new challenges

to health systems. We can overcome these challenges by the use of new technologies and artificial

intelligence tools. This work aims to present a new approach to use machine learning to predict cardiac

problems. As known, heart disease is the deadliest reason for morbidity. However, early detection of cardiac

anomalies can prevent heart attack and solve patient life. That is why we think to design and build an

e-health platform that will enable telemedicine acts such as remote monitoring and remote assistance. We

will first focus on the telerehabilitation of patients suffering from heart disease.

https://orcid.org/0000-0002-4261-1198

1 INTRODUCTION

Even if life expectancy has increased since 2000 and

become five years more than before, the health

system in Morocco still faces many challenges:

 Material and especially human resources still

largely insufficient

 Significant disparities in access to care

(including primary care) between urban and

rural areas

 Epidemiological transition situation still

significant share of mortality linked to 6

diseases/health priorities

 ~ 270 rural communes are in critical health

isolation (distance > one hour from a hospital

structure), including 160 priority communes

representing ~ 2 Million inhabitants.

All these challenges are due to the lack of

infrastructure, technical and human resources. To

overcome these problems, we think to benefit from

telemedicine which is very promising and very

useful. Indeed, telemedicine combines scientific

knowledge and the development of IT. Also, it has

shown its benefits all over the world, and it can have

a significant contribution to the resolution of the

above problems. Thanks to telemedicine, people can

benefit from permanent medical monitoring from

their houses; this, can reduce patient visits to the

hospital and save them travel costs and time.

Figure 1 is a graph that describes six health

priorities addressable by telemedicine in Morocco.

We had this graph from a public health congress. As

shown, in morocco, as in the whole world, heart

disease is the primary cause of death. This is due to

the emergency of heart attacks, and as known, only a

few minutes between detection of the heart anomaly

and the intervention of doctors/specialists can make

a huge difference and can save many lives.

Furthermore, real-time diagnosis and preventives

alerts help patients feel more secure (Shishvan et al.,

2020). Thanks to the Internet of Things (IoT) and

the development of Artificial Intelligence

technologies, health services have been omnipresent

in patient life (Benjemmaa et al., 2020).

The deep learning approach is also used to

automate cardiac diagnosis (Bernard et al., 2018)

and help in the development of smarter healthcare

(Simsek et al., 2020)

ith better performance than

handcrafted features (Habibzadeh and Soyata, 2020).

Our work will treat the use of telemedicine and

especially the advances in artificial intelligence (AI)

in cardiology. We will discuss at first previous

Slime, K., Maizate, A., Hassouni, L. and Mouine, N.

A System Architecture to Implement Deep Learning Techniques for Patients Monitoring with Heart Disease: Case of Telerehabilitation.

DOI: 10.5220/0010737100003101

In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning (BML 2021), pages 483-487

ISBN: 978-989-758-559-3

483

works on the prevention and monitoring of heart

diseases. In the second section, we will propose a

new architecture and talks about methods and tools

and we will conclude with a discussion and

perspectives.

Figure 1: Health priorities addressable by telemedicine

(Chaachou, 2019)

2 RELATED WORKS

This section provides existing literature on cardiac

prediction, reviews the different tools and

technologies used, and gives challenges and

limitations of existing approaches.

We can avoid cardiovascular disease if cardiac

anomalies/problems had been detected early and in

real-time happening. This issue is the leading cause

behind the implementation of many studies in this

field. Using sensors and wearable devices connected

to mobile applications make it possible to pursue

patients, analyze ECG diagram, detect heart attacks

and eventual cardiac arrest and then transfers results

to doctors and emergency unit (Leijdekkers and

Gay, 2008).

Cardiologists or practitioners need to check

multiple parameters such as blood pressure, oxygen

saturation, heart rate, and analyze electrocardiogram,

before making any decision (Darwaish et al., 2019).

We can get all these parameters are given from

biomedical devices (ECG, Holter, pulse oximeter

…), which collecting and transmitting health data

remotely using wireless communication. MyHeart

(Luprano et al., 2006) is an example of a project

which uses wearable sensors to extract data,

analyses ECG variability, and then give the

classification of body activities.

Some literature reviews highlight the great

interest of having the possibility to diagnosis cardiac

activities via wireless communication technologies.

This method will reduce the mortality rate for

patients with heart diseases (Ghosh et al., 2021).

Besides, as cardiac patients are threatened to have a

heart attack anytime, a new concept of remote

applications is developed using wearable sensors,

mobile applications, and web applications (Ltifi et

al., 2016).

Data are collected and transferred via Bluetooth

or Wi-Fi to the smartphone, which transmits it to the

web application (Kakria et al., 2015). Data mining

and multi-agent systems provide real-time analysis

for complex and various data (Jemmaa et al., 2016).

One of the significant challenges of all these

applications and studies is to reduce the delay time

between the onset of a heart attack and the signal

alert sent to the emergency services (Raihan, M., et

al, 2021).

Table 1 summarises the major applications of

deep learning in health sectors and their associated

techniques.

Table 1: A summary of deep learning & telemedicine

applications in the health sector.

Theme /

lication

Description

Medical

imaging:

Organ

Identification

Development and Validation of a

Deep Learning Algorithm for

Detection of Diabetic Retinopathy in

Retinal Fundus Photographs

(Stacked Autoencoder) (Gulshan et

al., 2016

)

Stacked Autoencoders for

Unsupervised Feature Learning and

Multiple Organ Detection in a Pilot

Study Using 4D Patient Data

(Stacked Autoencoder) (Shin et al.,

2012)

Medical

imaging:

Tumor

detection

DeepMitosis: Mitosis Detection via

Deep Detection, Verification and

Segmentation Networks (CNN)(Li et

al., 2018)

Medical

imaging:

Echocardiography

Fast and accurate view classification

of echocardiograms using deep

learning (CNN) (Madani et al., 2018)

Clinically Feasible and Accurate

View Classification of

Echocardiographic Images Using

Deep Learning (CNN)(Kusunose et

al., 2020

)

Bioinformatic:

Protein

Structure

prediction

Predicting Backbone Cα Angles and

Dihedrals from Protein Sequences

(Sparse autoencoder)(Lyons et al.,

2014

)

Boosted Categorical Restricted

Boltzmann Machine fo

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

484

Computational Prediction of Splice

Junctions

(

RBM

)(

Lee et al., 2015

)

Bioinformatic:

Cancer

detection and

identification

Prostate cancer diagnosis using deep

learning with 3D multiparametric

MRI

(

CNN

)(

Liu et al., 2017

)

Deep Learning Model Based Breast

Cancer Histopathological Image

Classification (Wei et al., 2017)

Bioinformatic:

Gene

expression

Deep learning of the tissue-regulated

splicing code (DNN)(Leung et al.,

2014

)

A deep learning framework for

modeling structural features of RNA-

binding protein targets (DNN)(Zhang

et al., 2016

)

Predictive

analytics:

Patients health

prediction

Deep Patient: An Unsupervised

Representation to Predict the Future

of Patients from the Electronic Health

Records (Denoising

autoencode

)(

Miotto et al., 2016

)

DeepCare: A Deep Dynamic Memory

Model for Predictive Medicine

(Denoising autoencoder)(Pham et al.,

2016)

Telemedicine:

Monitoring

the Health

Status

Building a Telemedicine System for

Monitoring the Health Status and

supporting the social adaptation of

children with Autism spectrum

disorders(Artificial neural

network)(Lebedev et al., 2019)

MediumTerm Effectiveness of a

Comprehensive Internet-Based and

Patient-Specific Telerehabilitation

Program With Text Messaging

Support for Cardiac Patients:

Randomized Controlled

Trial(Artificial neural

network

)

(

Frederix et al., 2015

)

Application of Telemedicine for the

Control of Patients with Acute and

Chronic Heart Diseases (Artificial

neural network)(Escobar-Curbelo et

al., 2019)

3 PROPOSED APPROACH

In this part, we will discuss our new architecture for

remote monitoring heart disease systems. We will

explain the components of our system and compare

our methods with previous ones.

3.1 System Architecture

Figure 2 presents an overview of our proposed

architecture for monitoring heart systems.

Through telerehabilitation, heart patients can be

more independent and will be under continuous

surveillance.

Figure 2: System architecture

 Data acquisition: Data acquisition mainly

consists of sensors equipped with a wireless

data transmission device allowing the

recovery of the patient's health indices such as

temperature, pulse, blood pressure, ECG ...

These sensors can be placed on different parts

of the patient's body or integrated into

intelligent clothing. The big challenge in this

part is to reduce power (Krachunov et al.,

2017) consumption and offer a very high level

of communication.

 Gateway: The gateway part will consist of a

smartphone which will allow acquiring the

data emitted by the wearable sensors, and then

it will use the collected data in two ways:

 The first operation is ensured by a local deep

learning application installed on the

smartphone. It analyses data and

communicates alerts or recommendations to

the patient.

 The second operation is to send this data and

the patient geolocation coordinates to the

central system.

 Central system: The last part of our

architecture is the central system. It consists

of a Big Data infrastructure that will store the

data received from the gateway and host deep

A System Architecture to Implement Deep Learning Techniques for Patients Monitoring with Heart Disease: Case of Telerehabilitation

485

learning applications. As the first step, the

deep learning applications will allow

performing an in-depth and complete analysis

of acquired data which allows sending more

precise alerts and recommendations in real-

time to the patient and at the same time gives

accurate information for the services

concerned and the practitioner who monitors

the patient. In the second step, deep learning

applications will perform analysis on the data

stored over a long period to display statistical

data in a graphical form, allowing visualizing

the evolution of the patient's state of health.

3.2 Discussion

Our proposed architecture, as we have seen, is a

combination of three primary layers. We choose this

structure to ensure a real-time response.

This is the main problem of all previous

approaches: they are not feasible in real-time.

Besides, this architecture can analyze many

heterogeneous data thanks to the big data platforms

and AI tools. Alarms and alerts are the significant

challenges of our work. We aim that patients feel

safe every second; that is why we think about the

integration of a mobile deep learning application

which no one of previous studies has deployed.

Streaming data will give our application a wealth of

information that can be used to produce real-time

analysis and also to develop machine learning

algorithms.

4 CONCLUSIONS

In this study, we have discussed the utility of

telemedicine to pursue cardiovascular disease. In

fact, because of the high level of heart problems, one

minute can make a significant change in the cardiac

patient life. Hence, to have permanent diagnostic,

we propose in this work a global architecture of a

remote monitoring and rehabilitation system.

We have presented several studies which have

been realized to minimize the risk of heart attack and

cardiac problems. Following our analysis, we

deduced that the development of IoT and AI

technologies provide a revolution in telemedicine,

and based on previous works, we have improved a

new telerehabilitation conception of cardiac disease.

Our system is a part of an e-health platform that

aims to prevent and reduce heart attack

consequences. Thanks to alert notifications, the

patient and doctor can be connected and warned of

any abnormal changes.

In global architecture, we divide our system into

three main parts: Data acquisition, gateway, and the

central system. We will equip each patient with

sensors and deep learning mobile application which

gives him measures and alerts when something goes

wrong. Thanks to the communication with the other

parts of the system, doctors, and healthcare centers

are also alerted.

In our future work, we plan to develop a mobile

deep learning application, which will catch sensors

values and give first analysis and alert patients and

health centers.

REFERENCES

Benjemmaa A., Ltifi H., Ayed M.B. (2020) Design of

Remote Heart Monitoring System for Cardiac Patients.

In:Barolli L., Takizawa M., Xhafa F., Enokido T. (eds)

Advanced Information Networking and Applications.

AINA 2019.Advances in Intelligent Systems and

Computing, vol 926. SPRINGER, CHAM.

Bernard, O., Lalande, A., Zotti, C., Cervenansky, F.,

Yang, X., Heng, P. A., ... &Sanroma, G. (2018). Deep

learning techniques for automatic MRI cardiac multi-

structures segmentation and diagnosis: is the problem

solved?. IEEE transactions on medical imaging,

37(11), 2514-2525.

Simsek, M., Obinikpo, A. A., &Kantarci, B. (2020). Deep

Learning in Smart Health: Methodologies,

Applications, Challenges. In Connected Health in

Smart Cities (pp. 23-46). SPRINGER, CHAM.

Shishvan, O. R., Zois, D. S., &Soyata, T. (2020).

Incorporating Artificial Intelligence into Medical

Cyber Physical Systems: A Survey. In Connected

Health in Smart Cities (pp. 153-178). Springer, Cham.

Habibzadeh, H., &Soyata, T. (2020). Toward uniform

smart healthcare ecosystems: A survey on prospects,

security, and privacy considerations. In Connected

Health in Smart Cities (pp. 75-112). Springer, Cham.

Chaachou, S. (2019, December). Deployment of a

telemedicine initiative in Morocco. 18th Study days,

Information systems at CHU towards anchoring

hospital performance.

Raihan, M., et al. "Smartphone-Based Heart Attack

Prediction Using Artificial Neural Network."

Proceedings of International Joint Conference on

Advances in Computational Intelligence. Springer,

Singapore, 2021.

Darwaish, A., Naït-Abdesselam, F., & Khokhar, A.

(2019). Detection and Prediction of Cardiac

Anomalies Using Wireless Body Sensors and Bayesian

Belief Networks. arXiv preprint arXiv:1904.07976.

Luprano, J., Solà, J., Dasen, S., Koller, J. M., &Chételat,

O. (2006, April). Combination of body sensor

networks and on-body signal processing algorithms:

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

486

the practical case of MyHeart project. In International

Workshop on Wearable and Implantable Body Sensor

Networks (BSN'06) (pp. 4-pp). IEEE

Ghosh, Pronab, et al. "Efficient Prediction of

Cardiovascular Disease Using Machine Learning

Algorithms With Relief and LASSO Feature Selection

Techniques." IEEE Access 9 (2021): 19304-19326.

Ltifi, H., Ben Mohamed, E., & ben Ayed, M. (2016).

Interactive visual knowledge discovery from data-

based temporal decision support system. Information

Visualization, 15(1), 31-50.

Kakria, P., Tripathi, N. K., &Kitipawang, P. (2015). A

real-time health monitoring system for remote cardiac

patients using smartphone and wearable sensors.

International journal of telemedicine and applications,

2015, 8.

Jemmaa, A. B., Ltifi, H., &Ayed, M. B. (2016,

November). Multi-agent architecture for visual

intelligent remote healthcare monitoringsystem. In

International Conference on Hybrid Intelligent

Systems (pp. 211-221). SPRINGER, CHAM.

Gulshan, V., Peng, L., Coram, M., Stumpe, M. C., Wu, D.,

Narayanaswamy, A., ... & Webster, D. R. (2016).

Development and validation of a deep learning

algorithm for detection of diabetic retinopathy in

retinal fundus photographs. Jama, 316(22), 2402-

2410.

Shin, H. C., Orton, M. R., Collins, D. J., Doran, S. J., &

Leach, M. O. (2012). Stacked autoencoders for

unsupervised feature learning and multiple organ

detection in a pilot study using 4D patient data. IEEE

transactions on pattern analysis and machine

intelligence, 35(8), 1930-1943.

Li, C., Wang, X., Liu, W., &Latecki, L. J. (2018).

DeepMitosis: Mitosis detection via deep detection,

verification and segmentation networks. Medical

image analysis, 45, 121-133.

Madani, A., Arnaout, R., Mofrad, M., & Arnaout, R.

(2018). Fast and accurate view classification of

echocardiograms using deep learning. NPJ digital

medicine, 1(1), 1-8.

Kusunose, K., Haga, A., Inoue, M., Fukuda, D., Yamada,

H., &Sata, M. (2020). Clinically Feasible and

Accurate View Classification of Echocardiographic

Images Using Deep Learning. Biomolecules, 10(5),

665.

Lyons, J., Dehzangi, A., Heffernan, R., Sharma, A.,

Paliwal, K., Sattar, A., ... &Yang, Y. (2014).

Predicting backbone Cα angles and dihedrals from

protein sequences by stacked sparse auto‐encoder deep

neural network. Journal of computational chemistry,

35(28), 2040-2046.

Lee, T., & Yoon, S. (2015, June). Boosted categorical

restricted Boltzmann machine for computational

prediction of splice junctions. In International

conference on machine learning (pp. 2483-2492).

PMLR.

Liu, S., Zheng, H., Feng, Y., & Li, W. (2017, March).

Prostate cancer diagnosis using deep learning with 3D

multiparametric MRI. In Medical imaging 2017:

computer-aided diagnosis (Vol. 10134, p. 1013428).

International Society for Optics and Photonics.

Wei, B., Han, Z., He, X., & Yin, Y. (2017, April). Deep

learning model based breast cancer histopathological

image classification. In 2017 IEEE 2nd international

conference on cloud computing and big data analysis

(ICCCBDA) (pp. 348-353).IEEE

Leung, M. K., Xiong, H. Y., Lee, L. J., & Frey, B. J.

(2014). Deep learning of the tissue-regulated splicing

code. Bioinformatics, 30(12), i121-i129.

Zhang, S., Zhou, J., Hu, H., Gong, H., Chen, L., Cheng,

C., & Zeng, J. (2016). A deep learning framework for

modeling structural features of RNA-binding protein

targets. Nucleic acids research, 44(4), e32-e32.

Miotto, R., Li, L., Kidd, B. A., & Dudley, J. T. (2016).

Deep patient: an unsupervised representation to predict

the future of patients from the electronic health

records. Scientific reports, 6(1), 1-10.

Pham, T., Tran, T., Phung, D., & Venkatesh, S. (2016,

April). Deepcare: A deep dynamic memory model for

predictive medicine. In Pacific-Asia conference on

knowledge discovery and data mining (pp. 30-41).

SPRINGER, CHAM.

Lebedev, G., Klimenko, H., Fartushniy, E., Shaderkin, I.,

Kozhin, P., &Galitskaya, D. (2019). Building a

Telemedicine System for Monitoring the Health Status

and Supporting the Social Adaptation of Children with

Autism Spectrum Disorders. In Intelligent Decision

Technologies 2019 (pp. 287-294). SPRINGER,

SINGAPORE.

Frederix, I., Hansen, D., Coninx, K., Vandervoort, P.,

Vandijck, D., Hens, N., ... &Dendale, P. (2015).

Medium-term effectiveness of a comprehensive

internet-based and patient-specific telerehabilitation

program with text messaging support for cardiac

patients: randomized controlled trial. Journal of

medical Internet research, 17(7), e185.

Escobar-Curbelo, L., & Franco-Moreno, A. I. (2019).

Application of telemedicine for the control of patients

with acute and chronic heart diseases. Telemedicine

and e-Health, 25(11), 1033-1039.

Krachunov, S., Beach, C., Casson, A. J., Pope, J., Fafoutis,

X., Piechocki, R. J., & Craddock, I. (2017, June).

Energy efficient heart rate sensing using a painted

electrode ECG wearable. In 2017 Global Internet of

Things Summit (GIoTS) (pp. 1-6). IEEE.

A System Architecture to Implement Deep Learning Techniques for Patients Monitoring with Heart Disease: Case of Telerehabilitation

487