Theme Article: Special Section on Cognitive Science

and AI for Human Cognition and Communication

Human Arthritis Analysis

in Fog Computing
Environment Using
Bayesian Network
Classifier and Thread
Protocol
Sudeep Tanwar Neeraj Kumar
Nirma University Thapar Institute of Engineering and Technology
Jayneel Vora Vishal Sharma
Nirma University Soonchunhyang University
Shriya Kaneriya Ilsun You
Nirma University Soonchunhyang University
Sudhanshu Tyagi
Thapar Institute of Engineering and Technology

Abstract—Nowadays, many people are facing the problem of arthritis. Regular monitoring
and consultation of joint health from a specialist can help patients with this chronicle
disease. The ratio of orthopedic doctors to patients with arthritis is low, worldwide. Use of
smart devices can support the healthcare industry a lot. Motivated by these facts, here we
propose an architecture to track the hand movements of the patient. For regular

Digital Object Identifier 10.1109/MCE.2019.2941456

Date of current version 6 December 2019.

2162-2248 ß 2019 IEEE Published by the IEEE Consumer Electronics Society IEEE Consumer Electronics Magazine
88
Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 monitoring of patients with arthritis, fog and cloud gateways for real-time response
generation are used. Thread protocol and Bayesian network classifier have been included
in the proposed architecture to achieve reliable communication and anomaly detection,
respectively. A dataset of 431 patients with arthritis is taken in real time and simulated
on OMNet++ simulator. Observations show that the packet delivery ratio is improved by
15–20%, the response time is reduced by 20–30%, and the packet delivery rate is improved
by 25–35%, in comparison to not using the fog and thread protocol.

& HEALTH AWARENESS AMONG people is increas- Hence, it has been strengthened by adding a
ing day by day because diseases such as hyper- thread protocol that maintains an effective com-
tension, diabetes, arthritis, high blood pressure, munication between smart devices and gateway.
seizures, etc., have affected the communities The reliability of the gateway and sensing devi-
across the world for many decades. Rheumatoid ces has further prompted the need to take com-
and osteoarthritis are two major types of arthri- puting to the edge and make gateways more
tis, which are affecting a large number of people capable of handling the data and solving the
across the world. The current health situation of problem at hand. Observations have been made
females across the world for arthritis1 is critical, with the help of accelerometers and data has
as shown in Figures 1(a) and 1(b). Data gathered been collected from the wearable devices.
across the world shows that the U.S. is not the Sensed information has been stored and classi-
only country facing the problem of arthritis, but fied for anomaly detection, if any, using a Bayes-
India is also facing the same problem,2 as shown ian network classifier.
in Figures 1(c) and 1(d). Therefore, healthcare Following are the contributions of this article.
experts have a challenge to find the solution
 A WBAN-based architecture is proposed to
for the patients with arthritis. Literature shows
that the problem starts from a very young age; monitor the hand joints movement.
 The anomaly detection on the data received
hence, prior precautions and regular monitoring
can mitigate this problem. In order to collect reg- from smart devices is performed by using a
ular samples, various means of healthcare like Bayesian network classifier to reduce the
wireless body-area networks (WBAN), smart false detections.
wearable devices, etc., are used in hospitals.
To collect and manage the information sensed ROLE OF TECHNOLOGY IN HEALTHCARE
by the intelligent network, a huge amount of The application of WBAN is pivotal in the
storage is required. These types of devices healthcare sector like, ambience-assisted living
are helpful for doctors to keep track of their with monitoring the ambulatory motions of a
patients on the basis of their vital signs.3 Now, patient.4 WBAN also provides a point of care to
research community and infrastructure develop- patients in a hospital-based system. Since the
ers have the challenge to ensure the availability existing healthcare environment is shifting
of strong infrastructure to tackle this problem. from hospital-centric care to patient-centric care,
In order to manage the data generated from the related parameters like autonomous diagnosis,
healthcare sector and to avoid the network con- emergency services, real-time monitoring of
gestion, data segregation and classification is patients, managing of patient-related databases,
needed. etc., when combined in a group, have the poten-
Keeping in view of all issues, an integrated tial to give better services to all stakeholders. Sev-
architecture is required to handle various issues eral medical sensors were used to increase the
related to the healthcare domain. We have devel- capability of WBAN5; 7 such as electrocardiogram,
oped a WBAN-based architecture for the moni- electroencephalogram readers, blood oxygen lev-
toring of hand joints. A communication interface els, heart rates, pulse-oximeters, accelerometers,
is the backbone of the proposed architecture. and motion sensors, which are mounted on a

January/February 2020
89
Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 Special Section on Cognitive Science and AI for Human Cognition and Communication

Figure 1. Analysis of arthritis. (a) Levels of physical activity of women of the U.S. (b) Future projections of patients with
arthritis. (c) Prevalence of rheumatoid arthritis. (d) Osteoarthritis of the knee per 100,000 population in India.

microcontroller. These sensors are responsible aim is to keep the data nearer to edge9 of the net-
for handling the vital information of the patient work. The number of fog servers (FS) and com-
under observation. putations on individual server should be limited
Cloud Computing (CC) has been used in the on a fog layer to ensure the real-time access. Cer-
healthcare system with de-localization of data tain approaches have combined both FC9 and CC
and to provide a robust system with powerful together to ensure a robust system having faster
servers for storage and analysis. Fog computing communications and priority services.
(FC) was developed to align computations and Recent years have seen abundant responses
to reduce the latency during data transfers.8 Its from several technologists in healthcare

Table 1. Comparison of a thread over existing technologies.

Mesh Power IPv6 Open

Approaches Interoperability Failure Remarks
network consumption connectivity standards
Dispenses legacy
Thread10; 11 Yes No Yes Low Yes Yes
difficulties
Good for long
WiFi12 Yes Yes No High Yes Yes
range
Nonscalable
Zigbee IP SE 2.013 – Yes Yes High Yes Yes
routing

Zigbee Pro14 Yes Yes Yes Low No Low –

No Internet
Zwave15 Sometimes Yes Sometimes Low No No
connectivity

90 IEEE Consumer Electronics Magazine

Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 Figure 2. Illustration of the system architecture and analysis of human patients with arthritis based on the thread protocol.

automation. This automation involves a new net- With reference to the abovementioned dis-
working stack in the healthcare sector known as cussion, in proposed architecture, we have
the thread protocol (TP).10; 11 TP is targeted to connected the smart device with Arduino
enhance inter-device connectivity in a secure UNO. Here, WBANs serve as end devices, with
and reliable fashion. With no specified applica- one of them taking the role of the leader (usu-
tion layer in the TP-based architecture, the ally the border router). Since the number of
thread introduces a flexibility by rendering sup- devices in a range are limited, we require one
port to multiple application layer protocols. leader for every set of WBANs. Figure 2 shows
three classes of patients where we have incul-
PROPOSED ARCHITECTURE FOR ARTHRITIS cated the routers in the architecture to ensure
The WBAN architecture showing the monitor- better connectivity. These classes indicate the
ing of human arthritis with the help of TP and FC patients’ activities inside the room—standing,
is shown in Figure 2. Smart gloves are used for sitting, and sleeping. These measured activi-
the monitoring of arthritis of the patient under ties are used to monitor the level of disease,
observation. These are deployed to obtain the such as normal, moderate, and acute. This,
rotational as well as translational motion of the though uncalled for in normal patients, can be
wrist of the patient. The magnetic pickup unit adopted as an efficient mechanism for moni-
(MPU-6050) is mounted on the smart device to toring severe cases of arthritis. At the acute
capture the angular and transitional motions. level, patients with arthritis need to be con-
The packets received from the smart device con- tinuously observed, with any discrepancy
sist of the data sensed by the sensor placed in arousing immediate attention. In emergency
the gloves. The gateway collects data from the situations, a network failure cannot be tole-
smart devices using the microcontroller having rable at any cost. We have observed local
TP. The microcontroller has an MPU sensor and addressing in the thread network, which
a thread transmitter, which is connected to the allows the devices to address each other over
fog gateway. It also connects FS to the cloud a smaller header, while each retains their
using the HTTP protocol having cloud gateway respective IPV6 addresses to be recognized
or cellular networks. The analysis of an emer- from the outside.
gency situation has to be performed on FS and In the case of patients with arthritis who
accordingly appropriate action has to be taken. tend to move a lot, the portable battery may
A physician can visualize the received data and not be sturdy enough to supply the power for
generate the reports as and when required. A a long time. The readings of the sensors in this
person suffering from arthritis has a very wob- case are very sensitive, which are used to
bly-shaky movement of his hands, also known as avoid any untoward experience and, hence,
tremors. every single point of failure. Pre-existing

January/February 2020
91
Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 Special Section on Cognitive Science and AI for Human Cognition and Communication

Figure 3. Results for (a) packet delivery ratio with request, (b) packet delivery ratio with time, (c) packet delivery rate with
nodes, and (d) response time with concurrent users.

protocols for the microcontroller–gateway A high-dimensional model used for anomaly

communication have not proven to reduce the detection can be easily handled with Bayes
delay; hence, this study has taken TP as a lead- server.
ing protocol for communication. We have
incorporated the fog paradigm to take smart PERFORMANCE EVALUATION
decisions in emergency situations. However, To evaluate the performance of the pro-
the nonurgent data is not analyzed in this arti- posed scheme, a case study, using real-time
cle, and the data has been tested for urgency data of 431 patients with arthritis, on the
at the edge and sent to the cloud for storage OMNet++ tool was used. Packet delivery ratio
and analysis. Once the resources have been (PDR), response time (RT), and packet delivery
controlled and managed, then an anomaly- rate (PDA) were the performance metrics used
detection process is initiated through a to validate the results.
machine-learning classifier. In case of any As shown in Figure 3(a), with an increase in
emergencies, FS connects to the medical emer- the time frame, there is an increase in the PDR.
gency services using the HTTP protocol. The The obtained PDR is higher for the proposed
packets received via the cloud gateway from scheme as compared without the use of FC
the fog are stored on the cloud data storage, and TP, and if only FC techniques are used. As
which can be further retrieved for analysis and decisions related to the patients with arthritis
visualization of patient’s health data. in the proposed scheme have been performed
A lot of work have been done to identify at the edge of the network, it results in faster
various types of anomalies in the data sensed diagnosis and decision making by doctors,
by accelerometers and wearable devices making the life of the patients suffering from
through various classifiers such as supervised arthritis easier. Figure 3(b) shows the PDR of
neural network, support vector machine, K- patients with arthritis waiting for immediate
nearest neighbors, Bayesian networks, and diagnosis from the doctor. When the number
self-organizing maps. A Bayesian network7 was of data packets sent are increased, the pro-
selected because it can work with continuous posed scheme generates a higher PDR as com-
values as required in the proposed approach. pared to others. The PDR is higher because

92 IEEE Consumer Electronics Magazine

Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 Table 2. Percent improvements of proposed scheme.

Performance metrics Percent improvements

Packet deliver ratio 15%–20%

Response time (ms) 20%–30%

Packet deliver rate (Mb/s) 25%–35%

Figure 4. Screenshot of anomaly detection. provider via HTTP. The root mean square wave
of accelerometers is identified by the blue color
wave, and the red color wave depicts the angular
when the scheme without FC and TP is used, data shifts of the hand. Table 2 shows the percent
related to patients with arthritis is accessed far improvement range for the FC with TP approach
away from remote sites, which increases the with respect to the without FC and TP, and only
overload on the them. Hence, higher PDRs are FC approaches for various performance metrics.
achieved with the proposed scheme compared
to those without FC and TP, and those only with CONCLUSION
FC. Figure 3(c) shows that the proposed scheme We have presented an architecture to ana-
has a higher PDA with an increase in the number lyze real-time health monitoring of patients
of nodes with different data rates as compared with arthritis-related problems, using acceler-
with others. This higher PDA is achieved due to ometers and TP on fog servers. In this article,
computations of patients with arthritis data on we investigated whether a limb is prone to
the edge of the network. Any healthcare applica- the arthritis or not. Numerous health camps
tion running at the edge of network has a higher can be orchestrated without abundant physi-
data rate with the increase in number of nodes. cal or human resources. This article also used
Figure 3(d) shows the RT of patients with arthritis TP in observatory scenarios for the patients
waiting for immediate diagnosis from the doctor. with arthritis. It gives adaptive, complete con-
When the number of concurrent users are nectivity while fog computing provides de-
increased, the proposed scheme produces a localized, distributive structure, with both
lower RT as compared to others. The proposed maintaining a cost-effective solution. The per-
approach has a lower RT because when without formance evaluation results obtained clearly
FC and TP and only FC approaches are used, data indicate that FC with TP is a more effective
related to the patients with arthritis is accessed approach for the doctors to efficiently moni-
from remote sites located far way, which finally tor the patients with arthritis compared to
reduces the overload on them. However, in the the traditional core-based CC infrastructure.
proposed scheme, the data of patients with In future, the proposed architecture can be
arthritis is available to the doctor without delay extended in various domains to assist
for immediate diagnosis; therefore, FC with TP surveillance.
achieved a lower RT.
To find out the abnormality associated with
the data related to the patients with arthritis, we
& REFERENCES
have used the Bayesian network machine-learn- 1. National Statistics, “CDC, National Health Interview
ing classifier7 and TP.10; 11 It is a networking stack Survey (NHIS),” Apr. 2018. [Online]. Available: https://
introduced to facilitate automation at small www.cdc.gov/arthritis/data_statistics/national-
scales. This protocol is targeted to enhance statistics.html
the inter-device connectivity, in a secure and 2. Rheumatoid Arthritis Statistics, “Rheumatoid arthritis in
reliable fashion. Figure 4 shows an anomaly India statistics on overall impact and specific effect on
observed in the patients with arthritis data by demographic groups,” Feb. 2017. [Online]. Available:
the Bayesian network machine-learning classi- http://global-disease-burden.healthgrove.com/l/
fier, which intimated the emergency service 75957/Rheumatoid-Arthritis-in-India

January/February 2020
93
Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.
 Special Section on Cognitive Science and AI for Human Cognition and Communication

3. P. Sundaravadivel, E. Kougianos, S. P. Mohanty, and 14. W. Wang, G. He, and J. Wan, “Research on
M. K. Ganapathiraju, “Everything you wanted to know Zigbee wireless communication technology,” in Proc.
about smart health care: Evaluating the different Int. Conf. Elect. Control Eng., Yichang, 2011,
technologies and components of the Internet of Things pp. 1245–1249.
for better health,” IEEE Consum. Electron. Mag., vol. 7, 15. M. B. Yassein, W. Mardini, and A. Khalil, “Smart homes
no. 1, pp. 18–28, Jan. 2018. automation using Z-wave protocol,” in Proc. Int. Conf.
4. F. Schwiegelshohn, M. Hubner, P. Wehner, and Eng. MIS, Agadir, 2016, pp. 1–6.
D. Gohringer, “Tackling the new health-care paradigm
through service robotics: Unobtrusive, efficient,
Sudeep Tanwar is currently an Associate Profes-
reliable, and modular solutions for assisted-living
sor with the Department of Computer Science
environments,” IEEE Consum. Electron. Mag., vol. 6,
and Engineering, Institute of Technology, Nirma
no. 3, pp. 34–41, Jul. 2017.
University, Ahmedabad, India. Contact him at
5. A. Ahmadi et al., “3D human gait reconstruction and
[email protected].
monitoring using body-worn inertial sensors and
kinematic modeling,” IEEE Sensors J., vol. 16, no. 24,
Jayneel Vora is currently a graduate student with
pp. 8823–8831, Dec. 2016. Nirma University, Ahmedabad, India. Contact him at
6. A. L. S. Kawamoto and F. S. C. da Silva, “Depth-sensor [email protected].
applications for the elderly: A viable option to promote
a better quality of life,” IEEE Consumer Electron. Mag., Shriya Kaneriya is currently a graduate student
vol. 7, no. 1, pp. 47–56, Jan. 2018. with Nirma University, Ahmedabad, India. Contact
7. S.-C. Wang, R. Gao, and L.-M. Wang, “Bayesian her at [email protected].
network classifiers based on Gaussian kernel density,”
Expert Syst. Appl., vol. 51, pp. 207–217, 2016. Sudhanshu Tyagi is currently an Assistant
Professor, Department of Electronics and
8. A. Kumari, S. Tanwar, S. Tyagi, and N. Kumar, “Fog
Com-munication Engineering, Thapar Institute of
computing for Healthcare 4.0 environment:
Engineering and Technology, Deemed to be
Opportunities and challenges,” Comput. Elect. Eng.,
University, Patiala, Punjab, India. Contact him at
vol. 72, pp. 1–13, 2018. [email protected].
9. J. Vora, S. Kaneriya, S. Tanwar, S. Tyagi, N. Kumar,
and M. S. Obaidat, “TILAA: Tactile Internet-based Neeraj Kumar is currently working as a Professor,
ambient assistant living in fog environment,” Future Department of Computer Science and Engineering,
Gener. Comput. Syst., vol. 98, pp. 635–649, 2019. Thapar Institute of Engineering and Technology,
10. The THREAD Group, “Thread stack fundamentals, v2.0,” Deemed to be University, Patiala, Punjab, India. Con-
Jul. 2015. [Online]. Available: http://www.threadgroup.org/ tact him at [email protected].
11. The THREAD Group, “Thread usage of 6LoWPAN, v2.0,”
Jul. 2015. [Online]. Available: http://www.threadgroup.org/ Vishal Sharma is currently a Research
Assistant Professor with the Department of
12. V. Abinayaa and A. Jayan, “Case study on comparison
Information Security Engineering, Soonchunhyang
of wireless technologies in industrial applications,” Int.
University, Asan, South Korea. Contact him at
J. Sci. Res. Publ., vol. 4, no. 2, pp. 1–4, 2014.
[email protected].
13. A. J. Jara, L. Ladid, and A. Skarmeta, “The Internet of
everything through IPv6: An analysis of challenges, Ilsun You is currently an Associate Professor with
solutions and opportunities,” J. Wireless Mobile Netw., the Department of Information Security Engineering,
Ubiquitous Comput. Dependable Appl., vol. 4, Soonchunhyang University, Asan, South Korea.
pp. 97–118, 2013. Contact him at [email protected].

94 IEEE Consumer Electronics Magazine

Authorized licensed use limited to: VP's Kamalnayan Bajaj Institute of Eng. and Tech. Downloaded on July 17,2021 at 05:14:59 UTC from IEEE Xplore. Restrictions apply.