Healthcare 4.0 Frontiers

Received: 20 November 2018 Revised: 20 November 2019 Accepted: 21 November 2019
DOI: 10.1002/widm.1350
OVERVIEW
Healthcare 4.0: A review of frontiers in digital health
Prem Prakash Jayaraman1 | Abdur Rahim Mohammad Forkan1 |

Ahsan Morshed2 | Pari Delir Haghighi3 | Yong-Bin Kang1
1
Department of Computer Science and
Software Engineering, Swinburne
Abstract
University of Technology, Melbourne, Healthcare 4.0 is a term that has emerged recently and derived from Industry 4.0.
Victoria, Australia Today, the health care sector is more digital than in past decades; for example,
2
Faculty of Engineering, Central
spreading from x-rays and magnetic resonance imaging to computed tomography
Queensland University, Melbourne,
Victoria, Australia and ultrasound scans to electric medical records. With the wide spectrum of digi-
3
Faculty of Information Technology, tal technologies underpinning Healthcare 4.0 to deliver more effective and effi-
Monash University, Melbourne, Victoria, cient health care services, in this article, we use the wisdom pyramid methodology
Australia
to conduct a systematic review of current digital frontiers in Healthcare 4.0.
Correspondence
Prem Prakash Jayaraman, Department of
This article is categorized under:
Computer Science and Software Technologies > Computer Architectures for Data Mining
Engineering, Swinburne University of Application Areas > Health Care
Technology, John Street, Hawthorn,
Melbourne, Victoria 3122, Australia. Application Areas > Data Mining Software Tools
Email: pjayaraman@swin.edu.au Fundamental Concepts of Data and Knowledge > Knowledge
Representation
KEYWORDS
artificial intelligence, digital health, Healthcare 4.0, Internet of things (IoT), ontology, social
network
1 | INTRODUCTION
Health care systems face major challenges with spending accounting to 10% of gross domestic product for the European
Union and Australia while 18% for the United States (Kaiser Family Foundation, 2018), increasing demand for provi-
sion of care in aging societies and better health and well-being management. Recent examples show that digital health
technologies have the potential to mitigate or even eliminate these challenges. Emerging digital health care services is
expected to reduce costs, improve user experience and increase the quality of care (Siemens, 2018).
Today, the health care sector is digital that is, from x-rays and magnetic resonance imaging (MRI) to computed
tomography (CT) and ultrasound scans to electronic medical records (EMRs; Mestres, 2017). For example, a recent sur-
vey (Frost and Sullivan, 2018) estimates that medical imagining data could soon reach the petabytes range forcing big
data tools and technologies for capturing, analyzing, managing, and viewing data. This data will be coupled with
emerging digital data sources such as the Internet of things (IoT) including sensors and wearables, smart mobile
devices, and social media. This trend enables the development of novel Healthcare 4.0 services that can help both of
health-care providers (e.g., health practitioners, hospitals, and clinics) and users to move toward demanding and
researching personalized, proactive, and predictive health-care models.
Prem Prakash Jayaraman, Abdur Rahim Mohammad Forkan, Ahsan Morshed, Pari Delir Haghighi, and Yong-Bin Kang contributed equally to this
study.
WIREs Data Mining Knowl Discov. 2019;e1350. wires.wiley.com/dmkd © 2019 Wiley Periodicals, Inc. 1 of 23
https://doi.org/10.1002/widm.1350
2 of 23 JAYARAMAN ET AL.
Healthcare 4.0 is a term that has emerged recently and derived from the Industry 4.0. Inspired by the fourth
paradigm—data-intensive scientific discovery (Hey, Tansley, Tolle, 2009), Healthcare 4.0 is a collective term for data-
driven digital health technologies such as smart health, mHealth (mobile health), wireless health, eHealth, online
health, medical IT, telehealth/telemedicine, digital medicine, health informatics, pervasive health, and health informa-
tion system. It describes the digital frontiers and disruptive innovation in the health care sector that is driving new busi-
ness models and value networks (Herrmann et al., 2018). Advancements and adoptions of Healthcare 4.0 are occurring
across many developed countries in the world with the digital health market expected to grow to $223.7 billion by 2023
(Prescient, & Strategic Intelligence, Sept, 2018).
In this article, we use the wisdom pyramid (Wikipedia, 2018) as presented in Figure 1 to conduct a systematic review
of frontiers in digital health. We break the digital underpinning of Healthcare 4.0 into a hierarchy of digital data foun-
dations, digital knowledge representation, and digital knowledge learning and prediction. The wisdom pyramid provides a
systematic approach to review data/information and digital knowledge representation and learning that provide the
ability to capture, store, and interpret existing knowledge as well as translate it into wisdom to empower decision-
making underpinned by machine learning (ML) and artificial intelligence (AI). The word wisdom is associated with
deep expertise in different health care related areas driving decision-making.
The data/information stem from multiple cutting-edge digital technologies such as IoT including wearables, mobile
smartphones, social networks (e.g., Facebook and Twitter), and the traditional EMRs that include medical imaging data.
The knowledge available within the health care areas is generally represented using ontologies such as geneontology,
pubmed, and semantic sensor networks (SSNs). The advance in AI techniques (e.g., ML, pattern recognition, natural lan-
guage processing (NLP), deep learning, semantic reasoning, image processing, and computer vision) provides the capabil-
ity to learn the deep expertise required to facilitate the decision-making process. The amalgamation of digital data
foundations, digital knowledge representation, and digital knowledge learning and prediction paves way to the development
of novel Healthcare 4.0 application such as remote health care, diseases monitoring, detection/prediction, captology,
assisted living, epidemic monitoring, automated clinical decision support, elderly care, mental health, health education,
and many more leading toward efficient personal/community care solutions and improved patient experience.
Figure 2 provides an illustration on how the wisdom pyramid methodology can be applied in developing a taxon-
omy of the digital technologies underpinning Healthcare 4.0. As illustrated, the digital data foundation layer comprises
multitude of digital data sources from IoT to social media that provide valuable data essential for digital health care.
The knowledge representation and learning layers fuse together existing domain knowledge from ontologies and
semantic data stores such as PubMed (PuMmed, 2018) and use AI techniques for learning and prediction and extract
wisdom, an essential requirement in aiding decision-making.
Based on the aforementioned wisdom pyramid methodology, in this article, we present a review of advancements in
the digital data foundation, knowledge representation, and learning that focus on the development of novel Healthcare
4.0 services and applications. We conclude the article with a comprehensive analysis of some open challenges that need
to be addressed for widespread adoption of Healthcare 4.0.
The article is organized as follows. Section 2 describes digital data sources relevant to Healthcare 4.0. Section 3 con-
tains an overview of how knowledge is represented digitally in Healthcare 4.0. Section 4 describes the contributions of
AI in Healthcare 4.0 and Section 5 discusses emerging application areas. Section 6 concludes the article with an over-
view of open issues and challenges.
FIGURE 1 Wisdom pyramid

JAYARAMAN ET AL. 3 of 23
F I G U R E 2 Healthcare 4.0: Taxonomy of digital

technologies, services, and applications
2 | H E A LT H C A R E 4 . 0 : D I G I T A L DA T A FO U N D A T I O N S
In this section, we describe the current emerging technologies, refering to them as digital data foundations, that under-
pin the development of Healthcare 4.0 applications and services. In particular, we provide a detailed review of technolo-
gies such as the IoT including mobile smartphones and wearables, social networks, and EMRs.
2.1 | Internet of things
IoT is the one of the fastest adopted technology pushing the frontiers of Healthcare 4.0 (Georgakopoulos & Jayaraman,
2016). It is estimated that the IoT health care market will be worth $158 billion by 2022 (MarketsandMarkets™, 2018). IoT
bring several key capabilities to Healthcare 4.0 that brings about effectiveness in applications such as remote monitoring of
patients with chronic conditions, elderly people and those requiring constant supervision (A. R. M. Forkan & Khalil, 2017).
IoT in digital health target highly personalized, affordable, accessible, and on-time Healthcare 4.0 services for everyone.
2.1.1 | Wearable IoT
Wearable IoT is defined as an electronic, mobile, context-sensitive device that can be incorporated into items worn by
the user (such as on the body or cloth) and can be operated and accessed without or with very little hindrance to user
activities (Perera, Liu, & Jayawardena, 2015). Wearable IoT devices are capable of delivering personalized, immediate,
and goal-oriented feedback based on specific tracking of health data obtained via various embedded sensors
(e.g., accelerometer, gyroscope, temperature sensor, moisture, location, heart rate sensor, and blood pressure monitor)
(Hiremath, Yang, & Mankodiya, 2014). Low cost wearable IoT devices can provide long lasting functionality without
requiring continual recharging. The flexible size makes them easier to wear and use continuously.
Millions of wearable IoT devices are being employed in various Healthcare 4.0 applications. Some examples are
shown in Figure 3. The popular categories include smart watches (e.g., Pebble, Samsung gear, Apple watch), wristband
sensors (e.g., fitbit, metawear, Garmin vivosmart), smart glasses (e.g., Microsoft HoloLens, Google glass), e-textile and
health monitors (e.g., heart rate, electrocardiogram [ECG], blood pressure monitors). Such wearable IoT devices allow
continual data acquisition of various physiological condition of human body such as ECG (Yang, Zhou, Lei, Zheng, &
Xiang, 2016), heart rate (Venkatraman & Yuen, 2016), blood pressure, reparation rate, blood oxygen saturation, body
temperature, blood glucose, electromyogram (EMG), photoplethysmogram (PPG), and electroencephalogram (EEG).
2.2 | Fabric and flexible sensors
Another new product category that is fast emerging as a common Healthcare 4.0 wearable technology is low-cost dis-
posable patches that are worn continuously for days at a time and then discarded (Lee et al., 2018). A few examples of
such patches are presented in Figure 4. One of the most exciting potential developments in wearable patches is Sano
Intelligence's continuous blood chemistry monitoring patches. It has been demonstrated to measure blood glucose and
potassium levels, and aims to measure a full metabolic panel, including kidney function and electrolyte balance (Sano
Intellignece, 2012). A promising concept pioneered by mc10 is stretchable digital tattoos named BioStamp for the con-
tinuous monitoring of vital signs with flexible electronics patches (USA Today, 2018). These stretchable electronics
track and wirelessly transmit information such as heart rate, brain activity, body temperature, and hydration level, and
may be available to athletes. The Zio Patch from iRhythm (2-week use) can be worn to monitor cardiac rhythm and
warn of arrhythmias (iRhythm, 2018). Another interesting example of new patch technology is a continuous blood pres-
sure monitoring patch from Sense A/S (SenseA/S, 2018). Instead of the cumbersome pressure cuff, there is a small arm
patch with electrodes that sense the changing impedance of tissue around a vessel and convert it into a blood pressure
reading via a waistband sensor. One of the classic use cases for wearable patches is the continuous glucose monitor
(CGM) worn by diabetics and other self-trackers. New developments mean that the current state-of-the-art technology
is available in several CGM solutions where an under-the-skin CGM uses a sensor and transmits glucose readings every
1–5 min to an external receiver or insulin pump.
Australian Centre for NanoMedicine (ACN) at UNSW has developed a new wearable sensor that informs exposure
to ultraviolet rays (Greg, 2018). This can change color in the sun and could provide an affordable tool to help prevent
deadly skin cancers. The stick-on patches can be tuned to suit individual skin types. L'Oréal-owned skincare brand La
Roche-Posay, ultra voilet sense (UV Sense) is a battery-free IoT wearable sensor that can measure your UV exposure
and store up to 3 months worth of data (L'Oréal, 2018).
2.2.1 | Ambient IoT
Nonwearable (i.e., ambient) IoT devices containing sensors such as motion, door, pressure, video, object contact, and
sound sensors are widely used in several Healthcare 4.0 applications (Plantevin, Bouzouane, Bouchard, & Gaboury,
2018) such as remote patient monitoring, fall detection, anomaly detection in daily behavior (A. R. M. Forkan, Branch,
Jayaraman, & Ferretto, 2019) and elderly care (A. Forkan, Khalil, & Tari, 2014). Such IoT sensors (Figure 5) have
proven to be very successful and useful in elderly care as elderly people generally dislike being obliged to wear a moni-
toring devices or if they are suffering from mild dementia and may not remember to wear them. Moreover, wearable
IoT sensors are difficult to install on the body and require professional adjustments. Although the data collection pro-
cess using wearable IoT is usually easier than using ambient sensors, restrictions regarding wearing the sensors on body
could discourage people from adopting them. Wearable devices also can generate an uncomfortable feeling during
FIGURE 3 Popular wearable Internet of things devices in Healthcare 4.0
FIGURE 4 Flexible disposable patch sensors that can be attached to skin

long-term skin attachment and so external or ambient sensors are highly accepted in application where users do not
prefer the use of wearable devices. Ambient IoT devices gather real-time data continuously and therefore can rapidly
detect any changes in pattern by analyzing historical data. A proper data analytics over the captured data from sensors
can provide a better view of daily behavior in real-world scenario.
2.3 | Mobile smartphones
Mobile smartphones play an important part in our daily activities and have revolutionized Healthcare 4.0. The mobile
smartphone-based health application market (mHealth apps) was valued at $4.19 billion in 2016 and is estimated to
increase by 44.2% by 2025 (Grand View Research, 2018). Mobile smartphones are no longer used only as a communica-
tion device but the plethora of sensors and the corresponding data they produce have opened immense opportunities
for personalized medical treatment, track behaviors, and movement (Aledavood et al., 2019) and, providing personal-
ized plan for a balanced lifestyle (Ernsting et al., 2017). A recent report (Statista, 2019) shows that a total of 44,384 iOS
health care apps were released in apple store and 41,377 in google play store between 2015 to 2019. According to
another survey, 58.23% of smart users have downloaded and used health-related applications on their mobile
smartphones attributed these devices as a major contributor of health data. (Krebs & Duncan, 2015).
Digital health technology has made it possible to manage self-health providing brief real-time snapshots which were
only seen in hospitals and clinics before. mHealth apps allow health-care providers to effectively streamline communi-
cations between patients, providers, and their caregivers. Also, these apps enable for 24/7 management of a patient's
condition along with the ability to personalize health care per patient. For example, Apple created the “Apps for
Healthcare Professionals” section within the medical category of the iTunes Appstore, a unique feature among mobile
app marketplaces. In 2013, this section was further divided into subcategories including: reference, medical education,
EMR and patient monitoring, nursing, imaging, patient education, and personal care. Google similarly launched a
“Google Play” shop that provides a wide variety of health care apps, including some for health professional, for mobile
devices that use the Android operating system (Cheng, Fang, Hong, & Yang, 2017).
2.4 | Social media
Social media refers to a group of Internet-based applications that build on the multifaceted and technological foundations
of Web 2.0 and allow the creation and exchange of user-generated contents (Kaplan & Haenlein, 2010). In 2017, there are
about 2.46 billion users on social networks and this number is rapidly growing (Statista, 2018).
There are many different social media sites available today such as Facebook, Twitter, Reddit, and Instagram
(Aichner & Jacob, 2015). These sites can be classified into 13 categories: social networks, video sharing, blogs, microblogs,
forums, business networks, collaborative projects, enterprise social networks, photo sharing, products and services review,
social bookmarking, social gaming, and virtual worlds (Statista, 2018).
Twitter is one of the most widely used social media providing microblogging services. Twitter has currently about
335 million monthly active users (Statista, 2018). In recent years, Twitter has been widely used as a platform to dissemi-
nate and share health-related information. Patients with chronic diseases are more actively using online social media to
express their feelings and symptoms, share their experiences, and provide mutual support (Chou, Hunt, Folkers, &
Augustson, 2011).
Twitter's Application Programming Interface (API) platform facilitates collecting recent and historical tweets using
free and paid access options. Aggregation and analysis of large volumes of tweets can provide valuable insights and
FIGURE 5 Ambient Internet of things sensors for unobtrusive monitoring

perception into the health and health care system of a given population/community. From the perspective of digital
health, analysis of social media data can lead to the following benefits: (a) monitoring and surveillance of pandemic dis-
eases (e.g., influenza (M. J. Paul & Dredze, 2011)), (b) gaining better knowledge about public health and health care needs
(Griffiths et al., 2012), (c) monitoring quality of health care delivery and detecting poor clinical care (Greaves, Ramirez-
Cano, Millett, Darzi, & Donaldson, 2013)), (d) sharing psychological support among certain patients such as in the cancer
community (Sugawara et al., 2012), (e) investigating risk factors of chronic diseases (Delir Haghighi, Kang, Buchbinder,
Burstein, & Whittle, 2017), (f) pharmacovigilance studies including understanding public perceptions toward drug thera-
pies (Sarker et al., 2015), and (g) understanding sentiments of clinical practitioners toward various policies, medications,
rural health, and so on.
2.5 | Electronic medical records
EMRs or electronic health record (EHR) is a digital version of a patient's medical history that includes the key adminis-
trative clinical data relevant to that person's care under a particular health-care provider. EMR data can include demo-
graphics, progress notes, description of health problems, medications, vital signs, past medical history, clinical data
generated during the diagnostic process, coded billing data, immunizations, admission and discharge summaries, pro-
gress notes, pathology or radiology test results, x-ray images, and many more technical reports (Figure 6). Moreover, cli-
nicians are now using voice recognition, captured images using mobile device camera, and documentation templates
that are directly recorded as information feeding into the EMR. Eighty percent of data in global EMR database are
unstructured text or images (Wong, Murray Horwitz, Zhou, & Toh, 2018). EMR are built to share information with
other health-care providers and organizations—such as pathological laboratories, general practitioners, specialists,
medical imaging facilities, pharmacies, emergency facilities, school, and workplace clinics—so they contain informa-
tion from all clinicians involved in a patient's care.
EMR has made it possible to easily access health data and make grouping of patients into the medical care, status,
and outcomes from a diverse patient population. EMR also has the ability to support care-related activities directly or
indirectly through evidence-based decision support, quality management, and outcomes reporting. Extracting and dis-
covering useful clinical information from these EMR data pose a challenge as it is difficult to define rules for such
unstructured data and manual review is not feasible on a large scale (Wong et al., 2018). EMR also provides the founda-
tions for several longitudinal studies.
3 | DIGITAL KNOWLEDGE REPRESENTATION
As presented in the Figure 1, data or information obtained from the technologies described in Section 2 are transcribed
and represented as digital knowledge. The digital knowledge underpins the rich domain knowledge that is pivotal to
several Healthcare 4.0 applications and services. In this section, we describe some of the technologies that provide the
ability to represent knowledge to support the development of Healthcare 4.0.
3.1 | Ontologies
Ontology is a formal specification of entities or objects and their relationships in a domain. Being machine interpret-
able, ontologies can be used for automated processing of tasks to support decision-making. Further, ontologies contain
domain information and hence can provide rich contextual knowledge for data analysis.
FIGURE 6 Examples of electronic medical records

Figure 7 shows the two types of ontologies that are widely employed in Healthcare 4.0. One is called medical
domain ontologies that describe specific domain concepts and terminologies used in health care. The other set of ontol-
ogies, named device ontologies, describe the data sensed from technologies described under digital data foundation in
particular the IoT. Some well-known and mostly cited medical domain ontologies are described as follows:
• Systematized Nomenclature of Medicine – Clinical Terms (SNOMED-CT) ontology, created over the years by health
community, is considered as the main ontology for presentation of clinical concepts, terms, and relationships. It
covers areas used in medical practice such as clinical findings, symptoms, diagnoses, pharmaceuticals, body struc-
tures, medical devices, and social contexts. Each concept is assigned a unique ConceptID and a Fully Specified Name
(FSD), which can be interpreted as being a unique human–readable description of that concept. SNOMED-CT pro-
vides a consistent way for indexing, storing, retrieving, and aggregating clinical data that can enhance the interopera-
bility between different health information systems (Schulz, Suntisrivaraporn, Baader, & Boeker, 2009).
• Logical Observation Identifiers Names and Codes (LOINCS) provide a universal code system for laboratory test results
and other clinical observations. For each observation, such as arterial blood gases, homo-gram or serum potassium,
the database includes a code, a short name, a long formal name, and synonyms. The aim of LOINCS is to provide
common codes and terminologies, so that when hospitals, pharmaceutical manufacturers, researchers, and public
health departments receive messages from multiple sources, they can easily or automatically file in the right slots of
their medical records, research, and public health systems (Forrey et al., 1996).
• MedDRA (Medical Dictionary for Regulatory Activities) ( MedDRA, 2004) is a rich and highly specific and standardized
terminology to facilitate sharing of regulatory information internationally for medical products used by humans
(MedRa, 2015). Products covered by the scope of MedDRA include pharmaceuticals, biological, vaccines, and drug-
device combination products. MedDRA is available to all for use in the registration, documentation, and safety moni-
toring of medical products both before and after a product has been authorized for sale. Products covered by the
scope of MedDRA include pharmaceuticals, biological, vaccines, and drug-device combination products.
• Common Terminology Criteria for Adverse Events (CTCAE) are a coding for adverse events that occur in the course of
cancer therapy. Adverse events are considered any unintended or unfavorable symptom, sign, or disease temporarily
associated with the use of a medical treatment or procedure that may or may not be considered related to the medical
treatment or procedure (Program, 2003).
• Foundation Model of Anatomy (Rosse & Mejino Jr, 2003) is an ontology developed and used to describe human anat-
omy. It contains approximately 75,000 classes and over 200,000 terms. Hundreds of relationships are used to semanti-
cally describe human anatomical structures, from microscopic parts such as macromolecules, cells and their parts,
and portions of tissues, to macroscopic anatomical structures such as body parts and the whole organism itself. The
FMA is available in Web Ontology Language (OWL) format and its subparts are used for specific anatomy or applica-
tion ontologies such as Semi-automated (SEMIA) ontology, Medical Ontology (Medico), and others.
• ICD-10 (International Classification of Diseases Tenth Revision) was issued by the World Health Organization (WHO)
as a classification system for diseases and other related aspects in medical practice, including symptoms, abnormal
findings, causes of diseases, death certificates, and health records (World Health Organization, 1993). The ICD is
used by member states of WHO for compilation of national mortality and morbidity statistics, for epidemiological
research and for assessment of trends in public health and illnesses. However, ICD-10 is not an ontology in the strict
sense, but is usually kept as thesaurus and used for classification. There are no relationships defined between the
F I G U R E 7 Overview of ontologies
employed in Healthcare 4.0
terms, as does not provide distinctions between terms on an ontological level. In recent years, an OWL version of
ICD-10 is developed to provide better and more formal way to describe the terms and their relationships at ontologi-
cal level. In addition, the next version of ICD is under construction and it is expected to overcome some existing
limitations.
• RadLex (Radiology Lexicon) (Mejino Jr, Rubin, & Brinkley, 2008) is a comprehensive lexicon that aims to provide a
unified language for standardized indexing and retrieval of radiological information resources. RadLex includes
many complex domains that are necessary for radiologists, such as image quality, treatment, anatomic location, and
uncertainty. RadLex although provides a unified lexicon for radiologist, it is not an ontological framework. It has the
potential to evolve to an ontology, but further steps are required.
• Gene ontology (GO) project provides ontologies to describe attributes of gene products in three nonoverlapping
domains of molecular biology. Within each ontology, terms have free text definitions and stable unique identifiers.
The vocabularies are structured in a classification that supports “is-a” and “part-of” relationships. GO molecular
function terms represent activities rather than the entities (molecules or complexes) that perform the actions, and do
not specify where, when, or in what context the action takes place. Examples of individual molecular function terms
are the broad concept of “kinase activity” and the more specifically “6-phosphofructokinase activity” (a subtype of
kinase activity).
• Disease ontology (DO) represents a comprehensive knowledge base of 8,043 inherited, developmental, and acquired
human diseases (Disease Ontology, 2018).
• DrOn is an ontology of drug products, their ingredients, and their biological activity. It allows semantic reasoning
and help to understand the meaning to drug properties. Classes derived from each source are serialized in separate
modules. For example, the classes in DrOn that are programmatically derived from RxNorm are stored in a separate
module and subsumed by classes in a manually curated, realist, upper-level module of DrOn with terms such as
“clinical drug role,” “tablet,” “capsule,” and so on (Hanna, Joseph, Brochhausen, & Hogan, 2013).
In addition, below we describe ontologies that have been used to describe digital data foundation in particular the
IoT sensors in order to provide a unified representation of such sensing data sources and enable them to be fused with
respective domain ontologies. p
FIGURE 8 Overview of semantic sensor network ontology

• Semantic Sensor Network Ontology (SSNO) is an upper ontology developed by W3C to provide a shared and agreed
set of common concepts to enable semantic interoperability across SSN domain. The main aim of SSNO was to begin
a formal process for building ontologies that represent semantically capabilities of sensors and sensor networks. The
full SSNO contains 41 classes and 39 object properties and extends DOLCE-Ultralite (DUL) upper ontology. Thus,
including DUL extension, it is composed of 117 concepts and 142 objects in total (Compton et al., 2012). Figure 8 pro-
vide an illustration of the SSNO. Recently, some health-related platforms such as kHealth (Anantharam, Barnaghi, &
Sheth, 2013) have been developed using SSNO. The SSNO specifically is used for knowledge modeling that help to
discover, access, search, integrate, and interpret data (and extract information). It is also used on modeling data (rep-
resentation of observations) from various sensors data sources such as mobile smartphone, temperature, humidity,
video cameras, and so on. These observation are provided a standardized representation by the SSNO.
• Federated Interoperable Semantic IoT/cloud Testbeds and Applications (FIESTA-IoT) ontology is build from the exis-
ting leverage a number of core concepts from various mainstream ontologies and taxonomies, such as SSN, M3-lite
(a lite version of M3 and also an outcome of this study), WGS84, IoT-lite, Time, and DUL (Agarwal et al., 2016).
• IoT-lite ontology is a lightweight ontology to represent IoT resources, entities, and services. IoT-lite is based on the
SSN ontology. The loT-lite can present IoT platform without consuming excessive processing time. However, it is also
a meta-ontology that can be extended in order to represent IoT concepts in a more detailed way in different domains.
In addition, it is used for knowledge discovery, finding the pattern and unified the information from the heteroge-
neous IoT platforms (Bermudez-Edo, Elsaleh, Barnaghi, & Taylor, 2015).
3.2 | Knowledge bases
A knowledge base is a machine readable technology which stores structured and unstructured information (Auer et al.,
2007). It provides real-world facts and associated information. A well-organized and rich knowledge base can provide
insightful information to the end users. More specifically, knowledge base is not a static collection of resources but
dynamic resources that may have self-capability to learn by using the AI. For example, PubMed (as depicted in Figure 9
is a knowledge base that stores bibliographic information of biomedical and life science. This bibliographic knowledge
base stores 5.1 million of articles including molecular biology databases. The skeleton of huge amount of bibliographic
FIGURE 9 An example of medical bibliographic knowledge base

information is based on the MESH terminologies, and GOs. Using these ontologies, PubMed search functionality pro-
vides right articles to the end users by filtering unnecessary articles from the database (Doms & Schroeder, 2005).
PubMed is maintained and updated by the National Library of Medicine on a weekly basis. It also provides links to
other knowledge base such as DbPedia (Auer et al., 2007) and PharmGKB (Hewett et al., 2002). Moreover, self-learned
devices can be learned from the knowledge base and provide the right answers when someone ask or search queries.
Such devices are Google home and Amazon Echo.
4 | TH E RI SE O F A I I N H EAL THC ARE 4 . 0: TOWA RD DI GI TA L

KNOWLEDGE LEARNING AND PREDICTION
A key aspect of the wisdom pyramid presented in Figures 1 and 2 is the ability to generate wisdom from data/informa-
tion and knowledge. From the perspective of Healthcare 4.0, wisdom provides applications the ability to make decision
and/or assist users to make informed decisions. AI is a branch of computer science that develops machines capable of
functioning as an intelligent human in reasoning, discovering meaning, and learning and generalizing past experience.
Powered by AI, these days, AI techniques are being used by many industries especially in digital health domains due to
their unprecedented improvement and learning/reasoning capabilities.
Nowadays, AI has rapidly advanced and been actively leveraged in a wider range of Healthcare 4.0 areas that
include disease detection (or diagnostics), personalized health care service provision, and drug discovery. AI facilitates
the development of several Healthcare 4.0 applications and services enabling the move from prescriptive to predictive
medicine. This advancement mainly comes from (a) the rapid increase in the computational power of computers,
(b) increased availability of large amounts of digital health care data described in Section 2 including the developments
in wearable IoT devices, intelligent sensors, needed to develop and test advanced and sophisticated AI techniques, and
(c) significant improvements in AI algorithms (e.g., deep learning based on neural networks).
One of the important features of current AI technologies is to intelligently handle increasing volume of structured,
unstructured, and heterogeneous (e.g., text vs. image vs. sensor data) health data. The goal of AI in Healthcare 4.0 is to
assist health-care professional in making accurate clinical decisions, provide users/patients more efficient personalized
services, and overall improve the efficiency of the health care system by reducing costs, making it accessible and reduce
medical errors. In this section, we provide a review of relevant branches of AI that have been repeatedly discussed in
recent years in the literature to facilitate the development of Healthcare 4.0 applications.
4.1 | Machine learning
ML plays a vital role in discovering valuable insights and patterns as well as deriving complicated rules and trends from
health care data. ML has been defined as a “program that learns to perform a task or make a decision automatically
from the data, rather than having the behaviour explicitly programmed” (AL & IS, 2018). In the health care sector, ML
has been effectively used to build data analytical models that are able to explore useful features from the enormous
amount of medical data as well as learn insightful patterns and knowledge from the data. Broadly, the data can be
divided into two parts (F. Jiang et al., 2017): (1) patients' traits and (2) their medical outcomes. The former commonly
include demographic data (e.g., demographic information, disease history, and disease-specific data (e.g., diagnostic
imaging, gene expressions, physical examination results, clinical symptoms). The latter often collected in clinical
research (e.g., disease indicators, patients' survival times and quantitative disease levels—e.g., tumor sizes).
Often used as synonym for AI, ML is a branch of AI that focuses on the development of algorithms to learn from data
and can be classified broadly into (1) supervised learning and (2) unsupervised learning. In the health care areas, supervised
learning focuses often on learning importance patterns or information from the data mentioned above as well as predicting
the best treatment processes with their probabilities closer the ideal treatment processes or other outputs under consider-
ations. For example, the outcome can be formed as the probability of getting epidemic diseases in a particular future time,
and the level of their critical factors, and so on. Today, supervised learning is considered more clinical relevant, thus health
care AI applications often use supervised learning. Examples of relevant techniques include classification algorithms such
as logistic regression algorithms, support vector machines, random forest, eXtreme Gradient Boosting (XGBoost) (Shi et al.,
2019), and neural network-based algorithms. Recently, XGBoost has been known as a dominating algorithm in Kaggle com-
petitions for structured or tabular data. XGBoost is an implementation of gradient boosted decision trees designed for speed
and performance (Hatton et al., 2019). Also, the logistic regression algorithms have been proven to be useful for predictive
analysis. Logistic regression is used to describe data and to explain the relationship between features of the data and the pre-
diction outcomes. Logistic regression can also be seen as a simpler deep learning algorithm.
Unsupervised learning (Y. Kwon, Kang, & Bae, 2014; Miotto, Li, & Dudley, 2016) is often used to understand distri-
bution of features from data. Clustering is a representative example in this learning. For example, clustering can group
patients with similar traits together into different clusters, maximizing and minimizing the similarity of the patients
within and between the clusters. K-mean, Gaussian mixture, and principal component analysis (PCA) have been widely
used for unsupervised learning. Especially, PCA is powerful for identifying the main components (or traits) and their
directions of subjects data without a loss of the original subject information.
Supervised learning requires well-annotated data that provides an accurate description of the reality (often referred
to as ground truth). Unsupervised learning algorithms use an exploratory approach to find patterns in data while
semisupervised uses a combination of unsupervised and supervised learning.
4.2 | Deep learning
Today millions of digital medical images such as MRIs and CT scans are stored in hospital database to capture thinner
slices of the body and three-dimensional (3D) and Four-dimensional (4D) medical images using advanced diagnostic
devices become the norm. Image processing is very important part of the AI. AI methods can tackle the huge amount
of data from images and signals (e.g., audio). Images and biosignals are analyzed by the deep learning algorithms such
as artificial neural networks (ANNs) (J. Jiang, Trundle, & Ren, 2010). ANNs are computing systems based on the net-
works that comprise animal brains.
Recent experiments have demonstrated cases where AI-driven algorithms using deep learning have outperformed
human doctors with regard to diagnosis of some medical conditions. Now deep learning-based approaches are used in
heart disease classification (Karthikeyan & Kanimozhi, 2017), diabetes detection (Tharani & Yamini, 2016), predictive
medicine based on patient history (Pham, Tran, Phung, & Venkatesh, 2016), and clinical management using EHRs
(Razavian, Marcus, & Sontag, 2016). Deep learning technology can also contribute significantly toward image
processing and interpretation of the medical images (e.g., X-days).
Recently, deep learning algorithms have been increasingly popular for analyzing and understanding diagnostic
image data and other prediction tasks. These algorithms have been quietly often used for supervised learning. These
algorithms can be seen as complex extensions of the logistic regression (see above) in general. The logistic regression
contains an input layer and an output layer, where the input layer has a node for each predictor variable (feature) and
all input nodes are connected to the output node where each connection has its own weight. The output node calculates
the sum of the weights of the input nodes and the logistic function finally measures the relevance degree of the sum as
the final output. Deep learning algorithms are commonly based on an idea of using neural networks. Neural networks
in a deep learning model expand on the structure of the logistic regression model by adding multiple hidden layers
between the input and output layers. The nodes in the hidden layers can make the model complex and nonlinear rela-
tionships between the predictor variables and the outcome. By adding multiple hidden layers, where each layer can
contain multiple nodes, neural networks allow deep learning models to model complex relationships and interactions
in between features as well as between the input data and the outcomes.
In the medical areas, convolutional neural network (CNN), recurrent neural network (RNN), and deep belief net-
work (DBN) have been often used. More descriptions of the use of these models in the medical areas can be found in
F. Jiang et al. (2017).
4.3 | Natural language processing
NLP is an integral part of AI that connects between human languages and technologies. Using advanced AI algorithms,
it is possible to crack down the semantic meaning from unstructured text. Specific tasks for NLP systems include sum-
marizing large blocks of narrative text such as a clinical note, mapping data elements present in unstructured text to
structured data in an EMR, converting data in the other machine-readable formats for reporting purposes, using of opti-
cal character recognition tool to read text from scanned documents, utilizing speech recognition to allow users to con-
vert clinical notes, or other information that can then be turned into text.
NLP uses automatic text classification that helps to automatically identify the text from the large collection and clas-
sify the contents according to the predefined classes. NLP is therefore very important for Healthcare 4.0, and it has
many AI-in-health care usages such as patient risk prediction by extracting free text data and vital sign data from clini-
cal texts in EMR, cohort building by capturing important information from medical documents, concept identification
for disease normalization, and extract information from biomedical literature to support clinicians.
NLP is built by a combination of rule-based and ML algorithms (e.g., support vector machines and conditional
random fields) (Heintzelman et al., 2012). Topic modeling is one of the most popular NLP techniques that can auto-
matically discover topics from a collection of clinical documents. It supports different algorithms such as latent
Dirichlet allocation (LDA) and Latent semantic indexing (LSI). LDA is a generative statistical model that is widely
used for topic identification based on Bayesian topic models (Blei, Ng, & Jordan, 2003). Maximum entropy
(MaxNet) modeling, another NLP approach, is used to recognize and categorize medications mentioned in the
unrestricted text of clinical documents generated in clinical practice. IBM Watson is the most famous example of
NLP usage in health care. Alchemy API in Watson service can quickly and easily extract and analyze metadata from
unstructured text.
4.4 | Social data analytics
Social data analytics is fast emerging as a connection between data, data science, and empowering communities that
harness the power of digitization and automation. Sentiment analysis and topic modeling are the most common
methods used for social media content analysis. Sentiment analysis mainly includes lexicon-based and learning-based
methods (Younis, 2015). The lexicon-based method uses a lexicon of opinion words including positive, negative, and
neutral words to assign a polarity score to each unigram that will then be added up to compute the polarity score of the
entire text. The learning-based methods train sentiment classifiers using features such as unigrams or bigrams. Com-
pared to lexicon-based methods, ML-based approach can produce higher accuracy but requires manual labeling and
training data (Zhang, Ghosh, Dekhil, Hsu, & Liu, 2011). On the other hand, Lexicon-based methods are more efficient,
but they can result in low recall.
Thematic analysis is a technique for identifying, detecting, and analyzing hidden thematic structure in large textual
datasets (Braun, Clarke, & Terry, 2014). Thematic analysis can be achieved by performing topic modeling or graph anal-
ysis and community detection. Topic modeling can be considered as an unsupervised text classification that is used to
cluster text into categories and detect the common themes within the documents of a corpus (Alghamdi & Alfalqi,
2015). Graph analysis is a different approach for thematic analysis that represents each term as a node in a network
and uses the edges that connect these nodes to show the relationships between the terms (Tighe, Goldsmith,
Gravenstein, Bernard, & Fillingim, 2015). Using graph analysis, community detection can be performed to identify
communities (clusters of related terms) in a large network.
4.5 | Semantic reasoning
Semantic web, linked data (i.e., resource description framework), and ontology are subsumed under AI. All of them
enable computer to understand the information by its own. One of the approaches is semantic modeling. This
approach is based on explicit, human interpreted concepts, relationship, and rules that contain the desired knowl-
edge of domain (Berners-Lee, 1998). This knowledge can be presented in a simple way, called knowledge base or
knowledge graph. The knowledge graph or knowledge base is also kind of linked data that combined the different
ontologies. To retrieve appropriate information from the knowledge base, reasoner can be used that can dynamically
combine the knowledge to answer of the given questions. A reasoner is a software program that infers logical effects
from a set of explicit asserted real-world facts or axiom that help for reasoning tasks such as classification,
debugging, and querying. In addition, reasoner rules can be integrated into the ML algorithms to predict informa-
tion relevant to a given context. Some of popular reasoners are developed in the recent years such Jena (McBride,
2001) inference system, Pellet, RACER, FACT++, Snorocket, Hermit, CEL, ELK, SWRL-IQ, FuzzyDL, Clipper, and
so on. These reasoners have different characteristics such as inference ML algorithms, supporting logic, degree of
completeness of reasoning, and others.
5 | H E A L T H C A R E 4 . 0 : E M E R G I N G A P P L IC A T I O N A R E A S
In Sections 2, 3, and 4, we provide a review of the current state of the art in Healthcare 4.0 using the proposed wisdom
pyramid methodology described in Figure 1. In this section, we provide a review of current emerging application areas
under the Healthcare 4.0 paradigm. These applications areas encompass one or more Healthcare 4.0 applications/
services.
5.1 | IoT wearable remote health care monitoring
IoT wearable monitoring is an emerging area in Healthcare 4.0 and it is changing the way we capture and process our bio-
logical data. The application domain includes remote patient monitoring services, smart home (Alaa, Zaidan, Zaidan,
Talal, & Kiah, 2017), hospital patient monitoring, rehabilitation (Bisio, Delfino, Lavagetto, & Sciarrone, 2017), self-track-
ing, performance sports (Camomilla, Bergamini, Fantozzi, & Vannozzi, 2018), and children, youths and elderly care
(Misra, Mukherjee, & Roy, 2018). IoT wearable devices supported by AI-driven intelligent data processing techniques have
great potential for early detection of physiological and behavioral changes and identify clinical episodes of several chronic
diseases (A. R. M. Forkan, Khalil, Tari, Foufou, & Bouras, 2015). A key goal is to identify abnormalities at the earliest pos-
sible stage. The data processing for such IoT applications can be made more cost-effective using Fog-based computational
model (Battula, Garg, Naha, Thulasiraman, & Thulasiram, 2019) (Liu, Chang, Guo, Mao, & Ristaniemi, 2018).
An example of IoT wearable devices is shown in Figure 10. IoT wearables are used in tracking of numerous chronic
diseases. Several commercially available wristband IoT wearables devices can track heart rate, steps, blood pressure, calo-
ries, and sleep quality (Bunn, Navalta, Fountaine, & Reece, 2018). These biological parameters can be used to detect health
abnormalities and disease progression. Various high-level user activities can be inferred by analyzing long-term captured
data from wearable devices that can be used to provide different health-related services (Mukhopadhyay, 2015). Body-
worn ECG sensors are used to detect different arrhythmias, myocardial infarction, and heart failure (Azariadi,
Tsoutsouras, Xydis, & Soudris, 2016; Lee et al., 2018; Yang et al., 2016). An AI-driven intelligent data processing algorithm
detects cardiac abnormalities based on real-time ECG data from ECG sensors. Similarly IoT wearable devices are used in
detecting epilepsy (Beniczky, Polster, Kjaer, & Hjalgrim, 2013) from convulsive seizures through electrodermal activity
and accelerometery. This is a useful improvement over lab-based EEG method in modern health as the device can be
worn continuously. Wearable solutions provide pain-free glucose monitoring to maintain daily habits for diabetes patients
FIGURE 10 Wearable Internet of things devices and their applications in Healthcare 4.0
FIGURE 11 An illustration of ambient assisted living
(Siddiqui, Zhang, Lloret, Song, & Obradovic, 2018). Levels of mental attention can be monitored with a small number of
nongelled EEG electrodes.
5.2 | Ambient-assisted living
Ambient IoT devices can be easily placed in different location of a home to monitor human behavior and their
health status. IoT sensors (e.g., motion sensors) installed in a home allow observation of daily behavior of individ-
uals in the home via a nonintrusive approach. Several IoT devices provide the effective infrastructure to continu-
ously monitor daily behavior of older adults in an unobtrusive way (Suryadevara, Mukhopadhyay, Wang, &
Rayudu, 2013).
We have developed such unobtrusive system for detecting anomalous situations in assisted living environment
named HalleyAssist (A. R. M. Forkan et al., 2019) using real-world data collected from different nonwearable IoT sen-
sors. Figure 11 presents an example deployment scenario where different motion and latch sensors along with speaker
and bulb are placed in different location in a house. A central hub gathers information from all IoT sensors. Magnetic
switch and latch sensor are easily installable and used to detect the opening and closing of doors and windows. This
data is then used to generate events and detect pattern such as number of toilet visits. Temperature sensor provides con-
tinuous data to detect the temperature of the ambient environment. Such sensors can detect the presence of users in
rooms by utilizing the changes in temperature. Pressure sensors detect the presence on chairs or in bed. They can be
used to detect sit-to-stand transfers and stand-to-sit transfers. Floor sensors detect situations such as falls and report
immediately to an emergency response team. Water flow sensor continuously measures the flow of water in taps or
showers and can be used to determine the hydration requirements of the user/person. Sensors such as microphones are
utilized to detect different events such as daily activities, for example, the sound that is generated while handling dis-
hes, when a person falls on the ground, or detecting Parkinson's disease (Kan, Kawamura, Hasegawa, Mochizuki, &
Nakamura, 2002). More advanced ambient IoT bed sensors can detect vital signs such as heart rate and respiration
(Kortelainen, Mendez, Bianchi, Matteucci, & Cerutti, 2010). All such ambient sensors do not raise privacy issues for in-
home monitoring and avoids the inconvenience of wearable devices.
Ambient IoT sensors are also shaping Healthcare 4.0 structure in hospitals. For hospitals, preventing infection is
imperative and given that thousands of sick and infected patients walk-in every day, hospital should strictly maintain
hygiene. Hand hygiene monitoring systems help in setting and detecting a degree of cleanliness among health care and
medical staff. The simplest function of hand hygiene IoT devices is to beep whenever medical staff comes in close prox-
imity of a patient bed without washing their hands (Mieronkoski et al., 2017).
Ambinet IoT device such as Sensordrone contains a selection of sensors, including gas, humidity, light, and pres-
sure and provide information about environment. It is now possible to use such environmental sensors to measure
a range of concerns including air quality, barometric pressure, carbon monoxide, capacitance, color, gas leaks,
humidity, hydrogen sulphide, temperature, and light. Other new home and environmental sensing solutions include
the air quality.
5.3 | Mobile smartphone applications
A growing number of health and wellness monitoring mobile smartphone applications (mHealth apps) have appeared in
recent past and are helping change people's perception toward digital technologies and how they manage and assess their
health. Various mHealth apps also now available in marketplace such as Apple and Google Play store to assist health pro-
fessionals with many important tasks, such as: information and time management, health record maintenance and access,
communications and consulting, reference and information gathering, patient management and monitoring, clinical
decision-making, and medical education and training. Some of these mHealth apps have been studied critically by medi-
cal professionals for their potential diagnostic capabilities, such as detecting acute coronary syndrome (Rathi, Kalantri,
Kalantri, & Rathi, 2016), arrhythmias, or blood pressure. The ability to download medical apps on smart mobile devices
has made a wealth of mobile clinical resources available to health practitioners. Different medical apps for many purposes
are available, including ones for electronic prescribing (Haffey, Brady, & Maxwell, 2014), diagnosis and treatment
(O'neill & Brady, 2012), practice management, coding and billing, CME or e-learning, and so on.
A broad choice of mHealth apps that assist with answering clinical practice and other questions at the point of care
exist, such as: drug reference guides, medical calculators, clinical guidelines and other decision support aids, textbooks,
and literature search portals (Mosa, Yoo, & Sheets, 2012). There are even mHealth apps that simulate surgical proce-
dures or that can conduct simple medical exams, such as hearing or vision tests. Many mHealth apps are not intended
to replace desktop applications but are meant to complement them in order to provide a resource that has the potential
to improve outcomes at the point of care. mHealth apps are also having impact on the home health care market,
enabling patients to leave the hospital for home sooner. A home health-care worker can use Health Insurance Portabil-
ity and Accountability Act (HIPAA)-compliant secure messaging to communicate with colleagues. They can even
retrieve a patient's EHRs securely using their smartphone or a tablet. The more a home health-care worker can do from
the field, the more time they get to spend with patients (Ventola, 2014).
Another emerging mHealth application area is Medication adherence. Medication adherence is an expensive and
damaging problem for patients and health-care providers. Patients adhere to only 50% of drugs prescribed for chronic
diseases in developed nations. Digital health has paved the way for innovative mHealth app solutions to tackle this
challenge. There are more than 97,000 mHealth applications/platforms available from different providers (Ahmed
et al., 2018) that assist patients to adhere to medication and monitor their medication intakes. It is evident from the
FIGURE 12 mHealth medication adherence application

literature that there has been a highly positive impact of using mHealth apps for medication adherence. One such
highly ranked app is medisafe (Santo et al., 2016). A total of 5,881 medication adherence apps are studied in Ahmed
et al. (2018) and concluded that the adoption of effective adherence mHealth apps can improve patient welfare in the
process. Figure 12 illustrates a mHealth medication adherence application.
5.4 | Clinical health care management
Continuous collection of digital health care data about patients can help identify potential anomalies early on and
provide proactive prescription to prevent potential risks (Watters, 2015). Given the enormous amounts of the data, AI-
driven Healthcare 4.0 can digest the data, detect patterns, analyze, and predict about future outcomes. These include
services for diagnosing potential diseases and helping health carers to make decisions about treatments. It is noted that
diagnostic errors happen to around 10% of patient deaths and between 6 and 17% of all hospital complications. Thus,
identifying the presence and absence of diseases or metastatic of diseases for pathologist is considered to be very impor-
tant, and ML can greatly help as an automated diagnostic tool. Electronic heath record (EHR) can also help identify
infection patterns and provide patients with risk information before they show symptoms. AI and ML algorithms can
help to enhance the prediction and accuracy of the emergence of infection faster as well as provide more accurate alert
services to health-care professionals. Several Healthcare 4.0 applications have been used to assist the diagnosis of cancer
and cardiology (F. Jiang et al., 2017). For example, it shows that IBM Watson for Oncology can be useful in assisting
cancer diagnosis through NLP techniques and ML from large amounts of unstructured data. Also, deep CNNs have the
potential in identifying different types of skin cancers. The CNNs were trained on more than 120 k clinical images con-
sisting of more than 2,000 different diseases, demonstrating highly achievable capabilities in classifying skin cancers
(Esteva et al., 2017).
According to S. Paul and Wetstone (2018), more than 17 million people died from cardiovascular diseases in 2015
corresponding to 31% of all global deaths. In this area, cardiac image data (e.g., ultrasound, CT) have been mainly used
that can assess heart anatomy and function as well as detect heart-related pathologies. The study in S. Paul and
Wetstone (2018) shows that deep learning techniques (i.e., CNNs) can be used to analyze these images automatically
and thus reducing manual work of radiologist (which is in shortage) and improving diagnostic efficiency.
For example, Wong et al. (2018) shows that ML can be used to model complex diagnostic decision-making process
from clinical inputs to identify individual with health outcomes. Also, it proposes that ML can be useful for extracting
and structuring diagnostic information, largely stored in free texts (e.g., clinical notes) and images. Modern ML
(e.g., deep learning) has the potential in automatically or semiautomatically extracting and leveraging useful clinical
information via the ability to understand and analyze human language models and powerfully learn the features of
images for image recognition.
5.5 | Virtual assistants and chatbots
Virtual personal assistance tools, such as Siri, Google Assistant, or Amazon Alexa, have limited capabilities in
terms of their intelligence (i.e., able to answer basic questions), but in the future they will have more robust
knowledge and intelligence of individuals' dietary behaviors, exercises, and medications from their daily routine
(Watters, 2015 Apr).
Personalized health assessment tools are currently being used for individual persons. For example, Ada Health
Companion has equipped with the AI chatbot app (Medical News Bulletin, 2017) that attempts to interactively commu-
nicate with an individual about his/her symptoms, provide an answer to what could be wrong using medical knowledge
base, and suggest what could be done next (e.g., visiting a doctor, pharmacist, or seeking emergency care). Another
examples is the IBM chatbot that provides information and advice to people living with arthritis (IBM Press Release,
2017) in which the machine is able to learn about patients through interactions (questions and answers) and provide
personalized information in a form that feels like a natural conversation.
Chatbots are also able to connect patients with clinicians for diagnosis or treatment (Figure 13). Conversation agents
such as Ada and IBM chatbot will emerge to be more prevalent. These tools will be able to provide day-to-day advice
through conversations on keeping more balanced and healthy lifestyle, offer guidance on whether an individual needs to
FIGURE 13 Health care chatbot
meet a health-care professional for his/her medical concern, and also assist with tasks such as medical reminders
(Watters, 2015).
Recently, mHealth chatbot apps have emerged providing personalized health care treatment (Watters, 2015) on
smartphones. For example, Your.Md is able to guide personal healthy life, find safe health information, and check med-
ical symptoms. Also, Sonde is a voice-based technology that has been designed to transform the way we monitor and
diagnose mental and physical health, working on mobile smartphones. It uses tones in voice to detect symptoms or
indicators for physical or mental illness. Safedrugbot offers assistant-like support to health professionals, doctors who
need appropriate data about the use of drugs during breastfeeding. Babylon Health uses patient's reported symptom
and checks them against a database of diseases using speech recognition, and then offers appropriate service
(e.g., connect with doctor in a live video consultation) (Babylon, 2018).
5.6 | Society 4.0: Toward empowering communities in Healthcare 4.0 using social data
The popularity of social media analysis in health and medical research has led to exploring and introducing more pow-
erful and sophisticated approaches. For example, by combining sentiment and thematic analysis, researchers could gain
an in-depth understanding of the public perception toward different health-related topics. (Christaki, 2016) presents
techniques combining different disease sources such as regional and global infectious disease surveillance, health-based
web search engines, and communication patterns in social media to predict and prevent future public health threats.
(Tighe et al., 2015) applied both sentiment analysis and community detection to study the patterns of communications
about pain-related medical events from social media data. They also analyzed the networks of retweets and compared
the connectivity patterns with other retweet networks.
The study presented in Delir Haghighi et al. (2017) investigated the subjective experience of patients with Fibromy-
algia and any association with coincident environmental factors. They first examined the association between negative
sentiment scores (indicative of more severe symptom density) and weather variables. The relationship between the
environmental variables and negative sentiment scores was tested to deduce that there was no significant correlation.
In this study, the date and time of posting of each tweet along with their location were collected, and this allowed them
to collect the climate data for each tweet from the public web APIs. In the second series of tests, the authors used graph
analysis to identify the most frequently occurring terms. Weather related terms were identified using community
detection.
In the study conducted by Bian et al. (2017) tweets related to Lynch syndrome were collected and analyzed to
understand the impact of promotional health information on laypeople's discussions. They first used a combination of
FIGURE 14 Example of using social media data (Twitter) in Healthcare 4.0
AI techniques including LDA model and CNN. They used the analysis results to examine the correlation between pro-
motional Lynch syndrome-related information and laypeople's discussions. An example scenario of sentiment analysis
using NLP tools is demonstrated in Figure 14.
Instagram, a photo and video-sharing social networking platform that has gained a great deal of popularity in recent
years for developing AI-driven Healthcare 4.0 application and services. Instagram provides users with creative ways to
filter and edit photos so they can express their personal styles. This aspect of Instagram has been utilized in a few stud-
ies for investigating and predicting personality traits (Ferwerda, Schedl, & Tkalcic, 2015). The study (Ferwerda et al.,
2015) found a significant correlation between personality traits and Instagram picture features applied by users and
based on these features they explored personality prediction. In a similar way, authors in Lay and Ferwerda (2018) used
“Liked” images on Instagram to predict user personality.
Online health forums as a different category of social media offer a safe and reliable platform where patients and
their families can seek information or share their experiences. Three online forums of Reddit, Mumsnet, and Patient
have been used for better understanding of health conditions (Cole, Watkins, & Kleine, 2016) including diabetes,
chickenpox, and human immunodeficiency virus (HIV), and chronic cough. While online health forums could provide
very relevant and useful data, however, accessing and extracting data programmatically from online forums can be diffi-
cult due to the lack of supporting APIs compared to Twitter. A recent study introduced a framework to model, rank,
and semantically analyze emerging topics across Twitter and Reddit that provides a wider view and better understand-
ing of emerging topics and trends (Pokharel, Delir Haghighi, Jayaraman, & Georgakopoulos, 2018).
Popularity of social media analysis under Healthcare 4.0 will keep increasing because it provides an ideal and low-
cost setting for the rapid and dynamic longitudinal collection of very large volume of data worldwide, compared to tra-
ditional data collection methods. User generated and first-person data are particularly suited to study chronic diseases
and conditions where there is little or poor understanding of the determinants of symptom expression in individuals
(Delir Haghighi et al., 2017).
5.7 | Privacy and security of health data
Healthcare 4.0 application has enormous potential to produce improved health outcomes and patient experience
(J. Kwon & Johnson, 2013). However, the resolution of privacy and security issues remains an ongoing and important
challenge. Health care data has become a popular target for hackers. The digital and connected (via Internet) nature of
Healthcare 4.0 make them a vulnerable target. Significant measures are required to ensure security and privacy of data
collected by IoT devices. Ensuing security using traditional privacy preserving techniques is a major challenge (Susilo &
Win, 2007). Health IoT devices can be connected to Internet through a wide range of wireless communication links
which made traditional wired security technique unusable and open the door to develop new security protocols. Low-
powered IoT devices are not capable of performing computationally expensive operation such as homomorphic encryp-
tion (Jayaraman, Yang, Yavari, Georgakopoulos, & Yi, 2017). Therefore, device-based authentication scheme such as
PMSec (Yanambaka, Mohanty, Kougianos, & Puthal, 2019) can be effectively used to solve security concern in low-
powered IoT medical devices. Many mHealth apps are not reliable enough and even well-known health care institu-
tions often cooperate with third-party developers for app development adding further complexity in the attempt to
adhere to relevant safeguards, regulations, and policies. Ensuring the security and privacy of electronic health data
stored in the cloud brings with it a new dimension of complication and challenges. Some recently proposed models in
information flow for smart health data to the cloud such as Lattice model (Puthal, 2018) can be explored to resolve such
challenges. The use of social media platforms is a great way to manage personal health but the implementation of its
security compliance is a big challenge.
6 | C ON C L U S I ON
A thorough understanding of current digital frontiers in Healthcare 4.0 is expected to be useful for various stakeholders
including health-care professionals, patients, and researchers interested in further research and development. This arti-
cle provided a systematic review of selected number of digital enablers underpinning developments in Healthcare 4.0
and how they are transforming the health care sector. Healthcare 4.0 is envisioned to deliver innovative and qualitative
services using advanced technologies (e.g., ML, IoT sensors and devices, social media) that are easily accessible, and
simple to personalize, bridge the gap between capturing health information and changing behavior, and drive the adop-
tion of healthy and active lifestyle and behaviors across the life which can result in positive health outcomes at every
stage of aging. We conclude this review by identifying some key issues and challenges that need to be overcome in order
to drive future development and adoption of Healthcare 4.0 application and services.
Healthcare 4.0 to be successful has to provide a multidisciplinary environment that brings together advances in the
different fields including computer science, engineering, economical, social science, public health, epidemiology, and
others. This will allow the digital underpinning of Healthcare 4.0 to go past technology to address the social and human
aspects in providing quality care.
Digital technologies underpinning Healthcare 4.0 lead to an entirely new data-driven health care applications and
systems. As discussed in this article, these data can be heterogeneous and hence required significant efforts toward
curation, verification, and validation. Moreover, expert knowledge is also a very important part of health care research
because of limited amount of medical data and their quality issues. For example, publicly available medical encyclope-
dia, Dbpedia, and PubMed are excellent sources of information that could be analyzed to extract valuable knowledge. A
critical challenge is to have a good metadata that describe the data to guide further analysis. Hence, due diligence and
policies need to be developed to ensure quality of the data being used in the decision-making process.
CONFLICT OF INTEREST
The authors have declared no conflicts of interest for this article.
A U T H O R C O N T R I B U T I O NS
Prem Prakash Jayaraman: Conceptualization; methodology; resources; writing-original draft, review, and editing.
Abdur Rahim Mohammad Forkan: Conceptualization; Investigation; Methodology; Writing-original draft, review,
and editing. Ahsan Morshed: Conceptualization; Methodology; Writing-original draft, review, and editing. Pari Delir
Haghighi: Conceptualization; Methodology; Writing-original draft, review, and editing. Yong-Bin Kang: Conceptuali-
zation; Methodology; Writing-original draft, review, and editing.
ORCID
Prem Prakash Jayaraman https://orcid.org/0000-0003-4500-3443
Abdur Rahim Mohammad Forkan https://orcid.org/0000-0003-0237-1705
Yong-Bin Kang https://orcid.org/0000-0002-0120-2582
R EL ATE D WIR Es AR TI CL E
Predictive data mining in clinical medicine: a focus on selected methods and applications
R EF E RE N C E S
Agarwal, R., Fernandez, D. G., Elsaleh, T., Gyrard, A., Lanza, J., Sanchez, L., Georgantas, N., & Issarny, V. (2016). Unified IoT ontology to
enable interoperability and federation of testbeds. In Internet of things (WF-IoT), 2016 IEEE 3rd world forum on (pp. 70–75). Reston, VA:
IEEE.
Ahmed, I., Ahmad, N. S., Ali, S., Ali, S., George, A., Danish, H. S., … King, D. (2018). Medication adherence apps: Review and content analy-
sis. JMIR mHealth and uHealth, 6, e62.
Aichner, T., & Jacob, F. (2015). Measuring the degree of corporate social media use. International Journal of Market Research, 57, 257–276.
AL, B., & IS, K. (2018). Big data and machine learning in health care. The Journal of the American Medical Association, 319, 1317–1318.
Alaa, M., Zaidan, A., Zaidan, B., Talal, M., & Kiah, M. L. M. (2017). A review of smart home applications based on internet of things. Journal
of Network and Computer Applications, 97, 48–65.
Aledavood, T., Torous, J., Hoyos, A. M. T., Naslund, J. A., Onnela, J.-P., & Keshavan, M. (2019). Smartphone-based tracking of sleep in
depression, anxiety, and psychotic disorders. Current Psychiatry Reports, 21, 49.
Alghamdi, R., & Alfalqi, K. (2015). A survey of topic modeling in text mining. International Journal of Advanced Computer Science and Applica-
tions, 6, 147–153.
Anantharam, P., Barnaghi, P., & Sheth, A. (2013). Data processing and semantics for advanced internet of things (iot) applications: modeling,
annotation, integration, and perception. In Proceedings of the 3rd international conference on web intelligence, mining and semantics, 5. ACM.
Auer, S., Bizer, C., Kobilarov, G., Lehmann, J., Cyganiak, R., & Ives, Z. (2007). Dbpedia: A nucleus for a web of open data. In The semantic
web (pp. 722–735). Busan, Korea: Springer.
Azariadi, D., Tsoutsouras, V., Xydis, S., & Soudris, D. (2016). Ecg signal analysis and arrhythmia detection on iot wearable medical devices.
In Modern circuits and systems technologies (MOCAST), 2016 5th international conference on (pp. 1–4). IEEE.
Babylon. (2018). Babylon health. Retrieved from https://www.babylonhealth.com/
Battula, S. K., Garg, S., Naha, R. K., Thulasiraman, P., & Thulasiram, R. (2019). A micro-level compensation-based cost model for resource
allocation in a fog environment. Sensors, 19, 2954.
Beniczky, S., Polster, T., Kjaer, T. W., & Hjalgrim, H. (2013). Detection of generalized tonic–clonic seizures by a wireless wrist accelerometer:
A prospective, multicenter study. Epilepsia, 54, e58–e61.
Bermudez-Edo, M., Elsaleh, T., Barnaghi, P., & Taylor, K. (2015). Iot-lite ontology. Retrieved from https://www.w3.org/Submission/2015/
SUBM-iot-lite-20151126/
Berners-Lee, T. (1998). Semantic web road map. Retrieved from https://www.w3.org/DesignIssues/Semantic.html
Bian, J., Zhao, Y., Salloum, R. G., Guo, Y., Wang, M., Prosperi, M., et al. (2017). Using social media data to understand the impact of promo-
tional information on laypeople's discussions: A case study of lynch syndrome. Journal of Medical Internet Research, 19, e414.
Bisio, I., Delfino, A., Lavagetto, F., & Sciarrone, A. (2017). Enabling iot for in-home rehabilitation: Accelerometer signals classification
methods for activity and movement recognition. IEEE Internet of Things Journal, 4, 135–146.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning Research, 3, 993–1022.
Braun, V., Clarke, V., & Terry, G. (2014). Thematic analysis. Qualitative Research in Clinical and Health Psychology, 24, 95–114.
Bunn, J. A., Navalta, J. W., Fountaine, C. J., & Reece, J. D. (2018). Current state of commercial wearable technology in physical activity mon-
itoring 2015–2017. International Journal of Exercise Science, 11, 503.
Camomilla, V., Bergamini, E., Fantozzi, S., & Vannozzi, G. (2018). Trends supporting the in-field use of wearable inertial sensors for sport
performance evaluation: A systematic review. Sensors, 18, 873.
Cheng, X., Fang, L., Hong, X., & Yang, L. (2017). Exploiting mobile big data: Sources, features, and applications. IEEE Network, 31, 72–79.
Chou, W.-Y. S., Hunt, Y., Folkers, A., & Augustson, E. (2011). Cancer survivorship in the age of youtube and social media: A narrative analy-
sis. Journal of Medical Internet Research, 13, e7.
Christaki, E. (2016). New technologies in predicting, preventing and controlling emerging infectious diseases. Virulence, 6, 558–565 Retrieved
from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4720248/
Cole, J., Watkins, C., & Kleine, D. (2016). Health advice from internet discussion forums: How bad is dangerous? Journal of Medical Internet
Research, 18, e4.
Compton, M., Barnaghi, P., Bermudez, L., GarcíA-Castro, R., Corcho, O., Cox, S., et al. (2012). The ssn ontology of the w3c semantic sensor
network incubator group. Web Semantics: Science, Services and Agents on the World Wide Web, 17, 25–32.
Delir Haghighi, P., Kang, Y.-B., Buchbinder, R., Burstein, F., & Whittle, S. (2017). Investigating subjective experience and the influence of
weather among individuals with fibromyalgia: A content analysis of Twitter. JMIR Public Health and Surveillance, 3, e4.
Disease Ontology. (2018). Disease ontology. Retrieved from http://disease-ontology.org
Doms, A., & Schroeder, M. (2005). Gopubmed: Exploring pubmed with the gene ontology. Nucleic Acids Research, 33, W783–W786.
Ernsting, C., Dombrowski, S. U., Oedekoven, M., LO, J., Kanzler, M., Kuhlmey, A., et al. (2017). Using smartphones and health apps to
change and manage health behaviors: A population-based survey. Journal of Medical Internet Research, 19, e101.
Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer
with deep neural networks. Nature, 542, 115–118.
Ferwerda, B., Schedl, M., & Tkalcic, M. (2015). Predicting personality traits with instagram pictures. In Proceedings of the 3rd workshop on
emotions and personality in personalized systems 2015 (pp. 7–10). ACM.
Forkan, A., Khalil, I., & Tari, Z. (2014). Cocamaal: A cloud-oriented context-aware middleware in ambient assisted living. Future Generation
Computer Systems, 35, 114–127.
Forkan, A. R. M., Branch, P., Jayaraman, P. P., & Ferretto, A. (2019). Halleyassist: A personalised internet of things technology to assist the
elderly in daily living. In Accepted in Hawaii international conference on system sciences (HICSS) 2019.
Forkan, A. R. M., & Khalil, I. (2017). A clinical decision-making mechanism for context-aware and patient-specific remote monitoring sys-
tems using the correlations of multiple vital signs. Computer Methods and Programs in Biomedicine, 139, 1–16.
Forkan, A. R. M., Khalil, I., Tari, Z., Foufou, S., & Bouras, A. (2015). A context-aware approach for long-term behavioural change detection
and abnormality prediction in ambient assisted living. Pattern Recognition, 48, 628–641.
Forrey, A. W., Mcdonald, C. J., DeMoor, G., Huff, S. M., Leavelle, D., Leland, D., et al. (1996). Logical observation identifier names and codes
(loinc) database: A public use set of codes and names for electronic reporting of clinical laboratory test results. Clinical Chemistry, 42, 81–90.
Frost and Sullivan. (2018). Big data opportunities in the US medical imaging market. Retrieved from https://store.frost.com/big-data-
opportunities-in-the-us-medical-imaging-market.html
Georgakopoulos, D., & Jayaraman, P. P. (2016). Internet of things: From internet scale sensing to smart services. Computing, 98, 1041–1058.
https://doi.org/10.1007/s00607-016-0510-0
Grand View Research. (2018). Retrieved from https://www.grandviewresearch.com/industry-analysis/mhealth-app-market
Greaves, F., Ramirez-Cano, D., Millett, C., Darzi, A., & Donaldson, L. (2013). Harnessing the cloud of patient experience: Using social media
to detect poor quality healthcare. BMJ Quality and Safety, 22, 251–255.
Greg, D. (2018). Australians develop new ‘wearable skin’ to monitor health conditions. Retrieved from https://www.sbs.com.au/news/
australians-develop-new-wearable-skin-to-monitor-health-conditions
Griffiths, F., Cave, J., Boardman, F., Ren, J., Pawlikowska, T., Ball, R., … Cohen, A. (2012). Social networks–the future for health care deliv-
ery. Social Science & Medicine, 75, 2233–2241.
Haffey, F., Brady, R. R., & Maxwell, S. (2014). Smartphone apps to support hospital prescribing and pharmacology education: A review of
current provision. British Journal of Clinical Pharmacology, 77, 31–38.
Hanna, J., Joseph, E., Brochhausen, M., & Hogan, W. R. (2013). Building a drug ontology based on rxnorm and other sources. Journal of Bio-
medical Semantics, 4, 44.
Hatton, C. M., Paton, L. W., McMillan, D., Cussens, J., Gilbody, S., & Tiffin, P. A. (2019). Predicting persistent depressive symptoms in older
adults: A machine learning approach to personalised mental healthcare. Journal of Affective Disorders, 246, 857–860.
Heintzelman, N. H., Taylor, R. J., Simonsen, L., Lustig, R., Anderko, D., Haythornthwaite, J. A., … Bova, G. S. (2012). Longitudinal analysis
of pain in patients with metastatic prostate cancer using natural language processing of medical record text. Journal of the American
Medical Informatics Association, 20, 898–905.
Herrmann, M., Boehme, P., Mondritzki, T., Ehlers, J. P., Kavadias, S., & Truebel, H. (2018). Digital transformation and disruption of the
health care sector: Internet-based observational study. Journal of Medical Internet Research, 20, e104.
Hewett, M., Oliver, D. E., Rubin, D. L., Easton, K. L., Stuart, J. M., Altman, R. B., & Klein, T. E. (2002). Pharmgkb: The pharmacogenetics
knowledge base. Nucleic Acids Research, 30, 163–165.
Hey, A. J., Tansley, S., Tolle, K. M. (2009). The fourth paradigm: Data-intensive scientific discovery (Vol. 1). Redmond, WA: Microsoft
Research.
Hiremath, S., Yang, G. and Mankodiya, K. (2014) Wearable internet of things: Concept, architectural components and promises for person-
centered healthcare. In Wireless mobile communication and healthcare (Mobihealth), 2014 EAI 4th international conference on
(pp. 304–307). IEEE.
IBM Press Release. (2017). Arthritis research UK introduces IBM Watson-powered ‘virtual assistant’ to provide information and advice to people
with arthritis. Retrieved from https://www-03.ibm.com/press/us/en/pressrelease/51826.wss.
iRhythm. (2018). Retrieved from https://www.irhythmtech.com/professionals/why-zio
Jayaraman, P. P., Yang, X., Yavari, A., Georgakopoulos, D., & Yi, X. (2017). Privacy preserving internet of things: From privacy techniques to
a blueprint architecture and efficient implementation. Future Generation Computer Systems, 76, 540–549 Retrieved from http://www.
sciencedirect.com/science/article/pii/S0167739X17303308
Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., … Wang, Y. (2017). Artificial intelligence in healthcare: Past, present and future. Stroke
and Vascular Neurology, 2, 230–243.
Jiang, J., Trundle, P., & Ren, J. (2010). Medical image analysis with artificial neural networks. Computerized Medical Imaging and Graphics,
34, 617–631.
Kaiser Family Foundation. (2018, February) How does health spending in the U.S. compare to other countries? Retrieved from https://www.
healthsystemtracker.org/chart-collection/health-spending-u-s-compare-countries/#item-start
Kan, Y., Kawamura, M., Hasegawa, Y., Mochizuki, S., & Nakamura, K. (2002). Recognition of emotion from facial, prosodic and written ver-
bal stimuli in parkinson's disease. Cortex, 38, 623–630.
Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of social media. Business Horizons, 53,
59–68.
Karthikeyan, T., & Kanimozhi, V. (2017). Deep learning approach for prediction of heart disease using data mining classification algorithm
deep belief network. International Journal of Advanced Research in Science, Engineering and Technology, 4, 3194–3201.
Kortelainen, J. M., Mendez, M. O., Bianchi, A. M., Matteucci, M., & Cerutti, S. (2010). Sleep staging based on signals acquired through bed
sensor. IEEE Transactions on Information Technology in Biomedicine, 14, 776–785.
Krebs, P., & Duncan, D. T. (2015). Health app use among us mobile phone owners: A national survey. JMIR mHealth and uHealth, 3, e101.
Kwon, J., & Johnson, M. E. (2013). Health-care security strategies for data protection and regulatory compliance. Journal of Management
Information Systems, 30, 41–66.
Kwon, Y., Kang, K., & Bae, C. (2014). Unsupervised learning for human activity recognition using smartphone sensors. Expert Systems with
Applications, 41, 6067–6074.
Lay, A., & Ferwerda, B. (2018). Predicting users' personality based on their ‘liked’ images on instagram. In The 23rd international on intelli-
gent user interfaces.
Lee, S. P., Ha, G., Wright, D. E., Ma, Y., Sen-Gupta, E., Haubrich, N. R., … Ghaffari, R. (2018). Highly flexible, wearable, and disposable car-
diac biosensors for remote and ambulatory monitoring. npj Digital Medicine, 1, 2.
Liu, L., Chang, Z., Guo, X., Mao, S., & Ristaniemi, T. (2018). Multiobjective optimization for computation offloading in fog computing. IEEE
Internet of Things Journal, 5, 283–294.
L'Oréal. (2018). La roche-posay. Retrieved from https://www.loreal.com/brand/active-cosmetics-division/la-roche-posay
MarketsandMarkets™. (2018). IoT healthcare market worth 158.07 Billion USD by 2022. Retrieved from https://www.marketsandmarkets.
com/PressReleases/iot-healthcare.asp
McBride, B. (2001). Jena: Implementing the rdf model and syntax specification. In Proceedings of the second international conference on
semantic web-volume 40 (pp. 23–28). CEUR-WS. org.
Medical News Bulletin. (2017). Artificial intelligence app Ada: your personal health companion. Retrieved from https://ada.com/
MedDRA. (2004). Medical dictionary for regulatory activities. Retrieved from http://www.meddra.org
MedRa. (2015). Medical dictionary for regulatory activities, MSSO presentations 2015. Retrieved from https://www.meddra.org/msso-
presentations-2015
Mejino Jr, J. L., Rubin, D. L., & Brinkley, J. F. (2008). Fma-radlex: An application ontology of radiological anatomy derived from the founda-
tional model of anatomy reference ontology. In AMIA annual symposium proceedings (vol. 2008, p. 465). American Medical Informatics
Association.
Mestres, J.-C. (2017). State of the art of health care: The cubism period (pp. 3–21). London, England: Palgrave Macmillan UK.
Mieronkoski, R., Azimi, I., Rahmani, A. M., Aantaa, R., Terävä, V., Liljeberg, P., & Salanterä, S. (2017). The internet of things for basic nurs-
ing care—A scoping review. International Journal of Nursing Studies, 69, 78–90.
Miotto, R., Li, L. and Dudley, J. T. (2016) Deep learning to predict patient future diseases from the electronic health records. In European
conference on information retrieval (pp. 768–774). Springer.
Misra, S., Mukherjee, A., & Roy, A. (2018). Knowledge discovery for enabling smart internet of things: A survey. WIREs: Data Mining and
Knowledge Discovery, 8, e1276.
Mosa, A. S. M., Yoo, I., & Sheets, L. (2012). A systematic review of healthcare applications for smartphones. BMC Medical Informatics and
Decision Making, 12, 67.
Mukhopadhyay, S. C. (2015). Wearable sensors for human activity monitoring: A review. IEEE Sensors Journal, 15, 1321–1330.
O'neill, S., & Brady, R. (2012). Colorectal smartphone apps: Opportunities and risks. Colorectal Disease, 14, e530–e534.
Paul, M. J., & Dredze, M. (2011). You are what you tweet: Analyzing Twitter for public health. Icwsm, 20, 265–272.
Paul, S., & Wetstone, J. (2018, February). Getting to the heart of it: How deep learning is transforming cardiac imaging. Retrieved from https://
medium.com/stanford-ai-for-healthcare/getting-to-the-heart-of-it-how-deep-learning-is-transforming-cardiac-imaging-22d34bf91a4e
Perera, C., Liu, C. H., & Jayawardena, S. (2015). The emerging internet of things marketplace from an industrial perspective: A survey. IEEE
Transactions on Emerging Topics in Computing, 3, 585–598.
Pham, T., Tran, T., Phung, D., & Venkatesh, S. (2016). Deepcare: A deep dynamic memory model for predictive medicine. In Pacific-Asia con-
ference on knowledge discovery and data mining (pp. 30–41). Springer.
Plantevin, V., Bouzouane, A., Bouchard, B., & Gaboury, S. (2018). Towards a more reliable and scalable architecture for smart home environ-
ments. Journal of Ambient Intelligence and Humanized Computing, 10, 2645–2656.
Pokharel, R., Delir Haghighi, P., Jayaraman, P. P., & Georgakopoulos, D. (2018). Analysing emerging topics across multiple social media plat-
forms. In Australasian computer science week 2019 (ACSW2019) (pp. 1–9). ACM.
Prescient & Strategic Intelligence. (2018, September). Digital health market overview. Retrieved from https://www.psmarketresearch.com/
market-analysis/digital-health-market
Program, N. C. I. U. C. T. E. (2003) Common terminology criteria for adverse events:(CTCAE). Cancer Therapy Evaluation Program. Retrieved
from https://ctep.cancer.gov/protocolDevelopment/electronic_applications/ctc.htm
PuMmed. (2018). PubMed. Retreived from https://www.ncbi.nlm.nih.gov/pubmed/
Puthal, D. (2018). Lattice-modeled information flow control of big sensing data streams for smart health application. IEEE Internet of Things
Journal, 6, 1312–1320.
Rathi, R. K., Kalantri, A., Kalantri, S., & Rathi, V. (2016). Symptom-based smartphone app for detecting acute coronary syndrome: A diag-
nostic accuracy study. Journal of the American College of Cardiology, 67, 632.
Razavian, N., Marcus, J., & Sontag, D. (2016). Multi-task prediction of disease onsets from longitudinal laboratory tests. In Machine learning
for healthcare conference (pp. 73–100).
Rosse, C., & Mejino, J. L., Jr. (2003). A reference ontology for biomedical informatics: The foundational model of anatomy. Journal of Bio-
medical Informatics, 36, 478–500.
Sano Intellignece. (2012). Retrieved from https://www.darkdaily.com/transdermal-patch-continuously-monitors-blood-chemistry-without-
needles-and-clinical-pathology-laboratory-testing-1031/
Santo, K., Richtering, S. S., Chalmers, J., Thiagalingam, A., Chow, C. K., & Redfern, J. (2016). Mobile phone apps to improve medication
adherence: A systematic stepwise process to identify high-quality apps. JMIR mHealth and uHealth, 4, e132.
Sarker, A., Ginn, R., Nikfarjam, A., O'Connor, K., Smith, K., Jayaraman, S., … Gonzalez, G. (2015). Utilizing social media data for
pharmacovigilance: a review. Journal of Biomedical Informatics, 54, 202–212.
Schulz, S., Suntisrivaraporn, B., Baader, F., & Boeker, M. (2009). Snomed reaching its adolescence: Ontologists' and logicians' health check.
International Journal of Medical Informatics, 78, S86–S94.
SenseA/S. (2018). New startup develops continuous blood pressure measurement without a counter-pressure cuff. Retrieved from http://
bloodpressure24by7.blogspot.com/2016/02/new-startup-develops-continuous-blood.html
Shi, H., Wang, H., Huang, Y., Zhao, L., Qin, C., & Liu, C. (2019). A hierarchical method based on weighted extreme gradient boosting in ecg
heartbeat classification. Computer Methods and Programs in Biomedicine, 171, 1–10.
Siddiqui, S. A., Zhang, Y., Lloret, J., Song, H., & Obradovic, Z. (2018). Pain-free blood glucose monitoring using wearable sensors: Recent
advancements and future prospects. IEEE Reviews in Biomedical Engineering, 11, 21–35.
Siemens. (2018, November). Retrieved from https://www.healthcare.siemens.com/news/mso-hbr-digitalization.html
Statista. (2018). Retrieved from https://www.statista.com/statistics/278414/number-of-worldwide-social-network-users/
Statista. (2019). Retrieved from https://www.statista.com/statistics/779910/health-apps-available-ios-worldwide/
Sugawara, Y., Narimatsu, H., Hozawa, A., Shao, L., Otani, K., & Fukao, A. (2012). Cancer patients on Twitter: A novel patient community
on social media. BMC Research Notes, 5, 699.
Suryadevara, N. K., Mukhopadhyay, S. C., Wang, R., & Rayudu, R. (2013). Forecasting the behavior of an elderly using wireless sensors data
in a smart home. Engineering Applications of Artificial Intelligence, 26, 2641–2652.
Susilo, W., & Win, K. T. (2007). Security and access of health research data. Journal of Medical Systems, 31, 103–107.
Tharani, S., & Yamini, C. (2016). Classification using convolutional neural network for heart and diabetics datasets. International Journal of
Advanced Research in Computer and Communication Engineering, 5, 417–422.
Tighe, P. J., Goldsmith, R. C., Gravenstein, M., Bernard, H. R., & Fillingim, R. B. (2015). The painful tweet: Text, sentiment, and community
structure analyses of tweets pertaining to pain. Journal of Medical Internet Research, 17, e84.
USA Today. (2018). Biostamp a big step forward in medical wearables. Retrieved from https://www.usatoday.com/story/tech/columnist/2016/
03/19/biostamp-big-step-forward-medical-wearables/82032978/
Venkatraman, S., & Yuen, S. G. J.. (2016). Wearable heart rate monitor. US Patent 9,456,787. Retrieved from https://patents.google.com/
patent/US8945017B2/en
Ventola, C. L. (2014). Mobile devices and apps for health care professionals: Uses and benefits. Pharmacy and Therapeutics, 39, 356.
Watters, S. (2015, April). Powered by AI: Technology, digital health, and a new era of care. Retrieved from https://medium.com/cossette-
health/powered-by-ai-technology-digital-health-and-a-new-era-of-care-a55e34cfd118.
Wikipedia. (2018). DIKW pyramid. Retrieved from https://en.wikipedia.org/wiki/DIKW_pyramid
Wong, J., Murray Horwitz, M., Zhou, L., & Toh, S. (2018). Using machine learning to identify health outcomes from electronic health record
data. Current Epidemiology Reports, 5, 331–342. https://doi.org/10.1007/s40471-018-0165-9
World Health Organization. (1993). The ICD-10 classification of mental and behavioural disorders: Diagnostic criteria for research (Vol. 2).
Geneva, Switzerland: World Health Organization.
Yanambaka, V. P., Mohanty, S. P., Kougianos, E., & Puthal, D. (2019). Pmsec: Physical unclonable function-based robust and lightweight
authentication in the internet of medical things. IEEE Transactions on Consumer Electronics, 65, 388–397.
Yang, Z., Zhou, Q., Lei, L., Zheng, K., & Xiang, W. (2016). An iot-cloud based wearable ecg monitoring system for smart healthcare. Journal
of Medical Systems, 40, 286.
Younis, E. M. (2015). Sentiment analysis and text mining for social media microblogs using open source tools: An empirical study. Interna-
tional Journal of Computer Applications, 112, 44–48.
Zhang, L., Ghosh, R., Dekhil, M., Hsu, M., & Liu, B. (2011). Combining lexiconbased and learning-based methods for Twitter sentiment analy-
sis. HP Laboratories, Technical Report HPL-2011, 89.
How to cite this article: Jayaraman PP, Forkan ARM, Morshed A, Haghighi PD, Kang Y-B. Healthcare 4.0: A
review of frontiers in digital health. WIREs Data Mining Knowl Discov. 2019;e1350. https://doi.org/10.1002/
widm.1350

Healthcare 4.0 Frontiers

Uploaded by

Copyright:

Available Formats

Healthcare 4.0 Frontiers

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Healthcare 4.0 Frontiers

Uploaded by

Copyright:

Available Formats

Received: 20 November 2018 Revised: 20 November 2019 Accepted: 21 November 2019

Healthcare 4.0: A review of frontiers in digital health

Prem Prakash Jayaraman1 | Abdur Rahim Mohammad Forkan1 |

FIGURE 1 Wisdom pyramid

F I G U R E 2 Healthcare 4.0: Taxonomy of digital

2.1 | Internet of things

2.1.1 | Wearable IoT

2.2 | Fabric and flexible sensors

2.2.1 | Ambient IoT

FIGURE 3 Popular wearable Internet of things devices in Healthcare 4.0

FIGURE 4 Flexible disposable patch sensors that can be attached to skin

2.3 | Mobile smartphones

2.4 | Social media

FIGURE 5 Ambient Internet of things sensors for unobtrusive monitoring

2.5 | Electronic medical records

3 | DIGITAL KNOWLEDGE REPRESENTATION

FIGURE 6 Examples of electronic medical records

FIGURE 8 Overview of semantic sensor network ontology

3.2 | Knowledge bases

FIGURE 9 An example of medical bibliographic knowledge base

4 | TH E RI SE O F A I I N H EAL THC ARE 4 . 0: TOWA RD DI GI TA L

4.1 | Machine learning

4.2 | Deep learning

4.3 | Natural language processing

4.4 | Social data analytics

4.5 | Semantic reasoning

5.1 | IoT wearable remote health care monitoring

FIGURE 11 An illustration of ambient assisted living

5.2 | Ambient-assisted living

5.3 | Mobile smartphone applications

FIGURE 12 mHealth medication adherence application

5.4 | Clinical health care management

5.5 | Virtual assistants and chatbots

FIGURE 13 Health care chatbot

FIGURE 14 Example of using social media data (Twitter) in Healthcare 4.0

5.7 | Privacy and security of health data

You might also like