2017 7th International Conference on Communication Systems and Network Technologies

IoT-based Decision Support System for Intelligent Healthcare – Applied to

Cardiovascular Diseases

Parag Chatterjee Leandro J. Cymberknop Ricardo L. Armentano

Universidad Tecnológica Nacional Universidad Tecnológica Nacional Universidad Tecnológica Nacional
Buenos Aires, Argentina Buenos Aires, Argentina Buenos Aires, Argentina
paragc@ieee.org lcymber@ieee.org armen@ieee.org

Abstract— Internet of Things (IoT) has been one of the

revolutionary technologies in recent times, which has redefine the healthcare system by connecting all the
catalyzed the paradigm shift in traditional healthcare stakeholders and the state-of-art technologies making the most
methods. Thanks to the concept of seamless data exchange of the information shared across the closely communicating
between connected devices followed byefficient data devices using the IoT platform. Typically, IoT is expected to
analytics, IoT-based eHealth strives to cater intelligent and offer advanced connectivity of devices, systems, and services
personalized healthcare services. This paper is focused on that goes beyond machine-to-machine communications
the aspects of IoT and decision support systems to proffer
preventive and intelligent healthcare. The objective of this
work is two-fold – to design the components ofan IoT- (M2M) and covers a variety of protocols, domains, and
healthcare platform and decision support system, applications[1].The field of IoT is spanned vastly among
satisfying the goals of an efficient technology-led sensors, networks, data processing and analytics, and security.
healthcare system, and to illustrate the same by applying Considering its aspects in healthcare, each of these parts of
to cardiovascular diseases and ascertain the risk groups
IoT has a specific set of applications. In identifying the key
from a sample set of people. Especially, to determine the
goals of technology-led healthcare, the following goalsin
risk groups from a sample set of people in terms of
Table I are of high impact.
cardiovascular disease, this system has been used.The risk
groups were identified based on their possibility of having
TABLE I
a cardiovascular disease.Such a system offers a common FOCAL OBJECTIVES OF IOT-HEALTHCARE
platform for all the stakeholders of healthcare andcan be
extended in other domains too as it works efficiently in Goals Description
predicting diseases, using the decision support system. Improved Looking at healthcare from a holistic view, the
health of the objective is to have an overall upliftment in health
Index Terms— Internet of Things, decision support mass standards of the mass.
system, eHealth, intelligent, healthcare, aging
Lowered cost Considering eHealth, the objective of using IoT is to
of high quality make the healthcare services affordable to the mass.
healthcare

Improved IoT strives to provide a connected platform of all the

I. INTRODUCTION experience for medical entities for an efficient exchange of

I NTERNET of Things (IoT) has been a path-breaking

technology backed by many handshaking research areas to
establish a high-end connectivity and communication between
patient and
medical
personnel
information backed by intelligent analysis. The
objective is to improve the patient experience like
minimizing the length of treatment or personalized
medication for fast recovery on one hand, and make
several mutually related devices to share information and the system more efficient and simpler for the
interact, toward a better user experience. In the field of healthcare providers on the other hand.
healthcare, the task of IoT is not only to proffer a truly Best use of Connected medical devices (also sensor fusion)
efficient and personalized healthcare to the users but also to resources makes the task of monitoring more efficient. Also,
IoT provides the common platform for the devices to
share and communicate the reports between those
entities.

Ambient One of the important objectives of IoT-enabled

assisted living healthcare is to assist the elderly people.

978-1-5386-1860-8/17/$31.00 ©2017 IEEE 362

DOI 10.1109/CSNT.2017.69
 knowledge to the health records of the patients as well as
deeper insights to the data. This work discusses anIoT-based
model that serves as the one-stop platform for all the inter and
, regarding the application of IoT, the following areas are of intra entity communications in healthcare, as well as the
focal importance in healthcare. assistive tool for the patients as well as to medical
personnel.Thereby, the decision support system is illustrated
A. Health Records
by performing a risk group identificationon 600 patients using
One of the fundamental aspects of IoT is data management. k-means clustering and considering several physiological
Medical and healthcare records constitute a huge share in this parameters impactful for cardiovascular diseases. It illustrates
respect. However, there is a constant shift in paradigm to the foundation of decision support system in this scenario, for
electronic health records. Less than a decade ago, nine out of in-system prediction of the risk of a person for cardiovascular
ten doctors in the U.S. updated their patients’ records by hand diseases. Also, it ascertains the category of risk group a new
and stored them in color-coded files. By the end of 2017, person belongs to depending on the analysis in the decision
approximately 90% of office-based physicians nationwide will support system.
be using electronic health records (EHR) [2]. IoT contributes
actively in this area by making the entire method of data
collection, storing and processing more efficient. The II. DESIGN OF THE UBIQUITOUS IOT PLATFORM FOR
foundation of eHealth being data, it is crucial to manage health HEALTHCARE
records well.
The objective of the IoT-based system discussed here is to
B. Monitoring serve as a one-stop common platform to healthcare services as
Considering recent paradigms, monitoring the health of a well as a decision support system for the medical
person is no more restricted within the limits of a hospital or personnel.Fig. 1 gives a holistic view to the IoT-based
medical center. Thanks to the IoT devices, monitoring is healthcare system.
widespread, even from one’s home. Another interesting
aspect lies in the field of lifestyle monitoring, since lifestyle
data counts significant in predicting several chronic diseases
beforehand. Therefore, apart from connecting the monitoring
devices, IoT provides the common platform to collect, store
and share all the data related to healthcare.
C. Personalized Treatment
A large pool of all historical data of a person is a huge plus for
any medical personnel before starting a thorough treatment.
IoT provides that possibility of collecting and storing the
health data in a single place, ready for analysis. This opens the
pathway to personalized and tailored treatment plans, where
before prescribing a diagnoses and treatment, the pool of
patient’s health-related data is considered.
D. Remote Care
To extend medical care beyond the limits of clinical centers,
IoT acts as a facilitator. Providing the platform for seamless
communication between the medical and healthcare entities
remotely, IoT strives towards ‘virtual hospital’, a
revolutionary trend of remote care.
Fig. 1. Holistic View to IoT-based Healthcare System
E. Ambient Assisted Living
One of the important aspects where IoT specifically functions The design of the system is illustrated as several modules
is to facilitate elderly lifestyle, by assisted living. By making based on its functionalities and domain of users.
immediate needs accessible and simplified making use of its
A. Electronic Health Records (EHR) Management
connected things and intelligence, IoT stands quite significant
in aging.
In an IoT environment, one of the fundamental properties
F. Decision Support System the EHRs should possess is ubiquitous addressability and
Decision support system is a quintessential part of recent interoperability. This raises the need of a unique addressing
healthcare which analyzes data and supports healthcare scheme for all the entities. However, looking from the user-
providers to take clinical decisions. But thanks to IoT, end, a unique user id is also mandatory to keep the pool of
decision support systems can do more, by spanning its user data linked to the user.TheEHRs are broadly classified
363
from the perspectives of its origin and type, as shown in Table parts. In the current healthcare scenario, patients need to visit
II. a clinical institution to obtain an assessment of their health.
TABLE II This involves costs on patient transportation, the clinician’s
KEY COMPOSITION OF ELECTRONIC HEALTH RECORDS
Source of Description Constituents
consultation and the hospital infrastructure. An automatic and
Data at-home procedure for assessment would reduce these costs
Patient Personal data (Name, Age, Sex, Text, Number significantly. Furthermore, the subjective nature of the test
Unique Identification, Family makes the results dependent on the clinician’s expertise and
details etc.) opinion. This introduced bias may be also minimized with an
E-health Data obtained from wearable Multivariant
devices eHealth devices (often used to (text, number,
automatic and objective instrumented test setting (IoT eHealth
measure lifestyle data or to monitor signals) devices), or other sensor-based information [3].
in post-hospitalization phase). This The interconnectivity among the different wearable sensors
data consists of physiological
parameters (heart rate, weight,
backed with intelligent analytics could provide an amazing
body temperature, blood pressure way in monitoring patients remotely. Analysis of the
etc.) physiological data measured through the IoT devices leads to
Investigation Data obtained from medical Multivariant medical decisions which when endorsed by a medical
devices for analysis and (text, number,
investigation signals, images) professional, can be reverted to the people. The IoT-based
Physician Prescription with diagnoses and Text, Number healthcare system could also work for passive medication or
treatment plan, along with lifestyle [4].
medication
Intermediate Intermediate data generated during Multivariant
Unclassified treatment (text, number,
Treatment signals, images) III. DECISION SUPPORT SYSTEM
Data
One of the focal goals of IoT-based healthcare is
empowering the physicians with deeper level of insights into
B. Data Storage and Linkage patients’ data, therefore enabling them to devise a tailored and
The data having its sources from multiple entities, needs to efficient treatment. The system discussed here consists of
be stored for ubiquitous access. This poses the need of a three parts – a knowledge base, an inference engine, and a
distributed data storage system. But the crucial task to communication mechanism. Though the focal goal is to use
maintain such a distributed database is the aspect of linkage of the decision support system as an assistive tool for the
the modules. The system discussed here consists of the healthcare provider for prescribing a treatment, it has other
following modules (Table III). aspects too, like havingdeeper insights to the patients’ data.
The system is usually trained with a dataset. Also, externally
TABLE III input logic is inserted into the system. Therefore, the
DATABASE MODULES IOT-BASED E-HEALTH SYSTEM
knowledge of the system is two-fold – based on the patients’
No. Module Description Query
Access data and on the input logic.
1 Patient Uniquely identified and linked to 2,5 The knowledge base primarily contains the comprehensive
a virtual storage space for information about the patient’s historic health-related data. It
ubiquitous access of all health- includes the comprehensive data including the lifestyle data as
related data
well, to make the decision support system make use of its
2 Doctor Uniquely identified and secured 1,3,4,5
access to all health-related data analytics fully.
of a patient The decision support system provides the following
3 Device-sourced Data sourced from IoT devices, 2,6 functionalities to the system
Data (Automatic stored automatically in the
recording) database A. Historical data
4 Device-sourced Data manually entered by 2,6
Data (Non- medical personnel, stored in the The IoT-based system maintains a pool of historical data
Automatic) database about healthcare of a patient, before starting a new diagnosis.
5 Prescription Decision Support System 1,2,6 This narrows down the diagnosis and eases in pin-pointing
assisted prescription, validated
and finalized by Doctor several issues. The first level of decision assistance starts with
6 Virtual Patient Comprehensive storage of 1,2,3,4, the comprehensive data of all instances of previous illness and
Storage health-data of patient, enabling 5 consultations with medical personnel.Next, the health-related
strict access control and
conditional query processing data (of specific physiological parameters) is also stored in the
service same platform, for ubiquitous access. A major part of this data
often happens to be the lifestyle data, which counts crucial
too.The effectiveness of this IoT-based decision support
C. Remote and Virtual Care system is that the patient need not to exclusively keep track of
all prior medical history. Instead all the historical data is
Despite widespread availability of healthcare services, it is
stored in the knowledge base for ubiquitous access.
still a challenge to ensure the reach of the same to the obscure

364
B. Predictive analytics The data has been classified into two clusters (Fig. 3). But
One of the most important aspects of the decision support one of the intended tasks of the decision support system is to
system is that, it provides a deeper level of insights to patient’s identify the risk groups, which becomes easier due to mapping
health-related data. The purpose of predictive analytics is to of the centroids. Clearly, in Fig. 3, the clusters are identified
perform statistical analysis and pattern recognition on the into two risk groups – high (red) and low (blue).It eases the
existing pool of data to predict the future trend of the same. task of a medical practitioner to determine easily the risk of a
Chronic diseases take a huge toll on human health worldwide. person since the decision support system calculates the risk
However, predictive analytics supports this early detection of and locates that in a specific risk group. Also, the
cardiometabolic and chronic diseases. meanprovides a clear visualization of the average risk of the
sample.

IV. DETERMINATION OF RISK GROUPS

(APPLIED TO CARDIOVASCULAR DISEASES)
To look at the treatment procedures from a holistic point of
view, identifying the severity of a disease in a group of people
and addressing the risk groups. This involves the introduction
of several risk scoring schemes into the decision support
system. Therefore, using the scoring scheme as the decisional
attribute, risks groups are being identified from the pool of
data. On principle, both supervised and unsupervised learning
methods count significant based on the data and classification
resources available.

In this scenario, the objective was to design a decision

support system to determine the amount risk a person has for
cardiovascular disease considering some physiological
parameters and to identify the specific risk group that person Fig. 3. Cluster assignments and centroid mapping for identification of high
belongs to, depending on the severity of the risk. and low risk groups in cardiovascular diseases using Framingham score
In this scenario, 600 people were considered, along with
their physiological parameters (age, gender, systolic and Considering from the point of view of a decision support
diastolic blood pressure, cholesterol, diabetes and smoking system, this scheme works intuitively with the user of the
habits). Then the Framingham Scores for each one ofthe entire decision support system. The logic of Framingham score being
sample set were calculated using Framingham’s Equation for embedded within the inference engine, an input of a person
cardiovascular diseases. Simple k-means clustering was with all the parameters considered would return the risk score.
performed on this data and Fig. 2 shows an instance of But the important aspect is the visualization of the same with
visualization of the clusters. respect to a sample set, and to ascertain the nature of the input
data (high risk, low risk, outlier etc.). This gives the possibility
to the medical personnel to have a deeper insight into the data
and to compare a new data to the historic ones.

Apart from the scenario of risk assessment, such a decision

support system can visualize the impact of physiological
parameters over the risk for a specific disease.

Fig. 2. Visualization of Clusters after k-means clustering on Framingham

Scores of the sample set
Fig. 4. Plotting of physiological parameters against risk score

365
 Applying this logic into the inference engine of the
decision support system (Fig. 4), it helps to visualize the
relative impact of a physiological parameter on the risk
determination. Especially visualizing outliers turns simple and
intuitive using this approach.

V. CONCLUSION AND FUTURE SCOPE

One of the focal goals of IoT-based healthcare is empowering
the physicians with deeper level of insights into patients’ data,
therefore enabling them to devise a tailored and efficient
treatment. Intelligent machines change the efficiency of entire
healthcare system dramatically as the machine-generated
reports could be analyzed easily by such machines with a
minimum human intervention [5]. The focal objectives of a
technology-led healthcare system as identified in the
beginning of the paper, is facilitated by the IoT-based decision
support system. The decision support system enables
preventive healthcare by data analysis on one hand, and
supports personalized treatment on the other hand, thanks to
the connected knowledge base, ubiquitously accessible by a
healthcare provider. Especially in the field of cardiovascular
diseases where the system is applied and discussed in this
paper, IoT-based intelligent decision support system counts
truly significant for predictive analytics and prediction of
cardiovascular diseases.
However, the background idea of the system discussed here is
to serve as a ubiquitous platform for all the healthcare
services, especially shifting the treatments to remote and
virtual paradigms, thanks to IoT, decision support systems and
predictive analytics. The future aspect of this work extends to
develop an automated analysis system of running data
(automated machine-generated data for vitals) and perform a
running analysis for better visualization and insights.

REFERENCES
[1] J. Höller, V. Tsiatsis, C. Mulligan, S. Karnouskos, S. Avesand, D.
Boyle, “From Machine-to-Machine to the Internet of Things:
Introduction to a New Age of Intelligence,” Elsevier, 2014, ISBN 978-0-
12-407684-6.
[2] R. Shay (2017, March). EHR adoption rates – 20 must-see stats.
Practice Fusion Blog. [Online] Available:
http://www.practicefusion.com/blog/ehr-adoption-rates/
[3] L. Romero, P. Chatterjee, and R. L. Armentano, “An IoT approach for
integration of computational intelligence and wearable sensors for
Parkinson’s disease diagnosis and monitoring”, Health and Technology,
6, (3), 167-172, Springer, 2016. doi: 10.1007/s12553-016-0148-0
[4] P. Chatterjee and R. L. Armentano, "Internet of Things for a Smart and
Ubiquitous eHealth System," 2015 International Conference on
Computational Intelligence and Communication Networks (CICN),
Jabalpur, 2015, pp. 903-907. doi: 10.1109/CICN.2015.178
[5] L. J. Cymberknop, P. Chatterjee, D. Dujovne, L. Romero, and R. L.
Armentano, “Internet of Things toward Efficient Analysis of Aging,
Cardiometabolic, and Neurodegenerative Diseases—An E-Health
Perspective”, The Internet of Things: Foundation for Smart Cities,
eHealth and Ubiquitous Computing, CRC Press (Taylor & Francis),
2017, ISBN: 9781498789028

366