ITU Kaleidoscope 2019 - ICT For Health
ITU Kaleidoscope 2019 - ICT For Health
ITU Kaleidoscope 2019 - ICT For Health
ITUKALEIDOSCOPE
ATLANTA 2019
ICT for Health:
Networks, standards and innovation
4-6 December
International
Telecommunication
Union
Telecommunication
Standardization Bureau
Place des Nations
CH-1211 Geneva 20
Switzerland
ISBN 978-92-61-24301-2
ISBN: 978-92-61-28401-5
Disclaimer
The opinions expressed in these Proceedings are those of the paper authors and do not necessarily reflect the views
of the International telecommunication Union or of its membership.
ITU 2019
All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior
written permission of ITU.
Foreword
Chaesub Lee
Director
ITU Telecommunication Standardization Bureau
The ITU Kaleidoscope academic conference has gained a reputation for providing an in-depth
discussion on matters relevant to the ITU membership. This year, in collaboration with the World
Health Organization (WHO), Kaleidoscope 2019: ICT for Health: Networks, standards and
innovation provided a forward-looking perspective on the future developments for better healthcare
delivery.
Kaleidoscope is ITU’s flagship academic event. Now in its eleventh edition, the conference supports
productive dialogue between academics and standardization experts. I wish to thank the Georgia
Institute of Technology for stimulating this dialogue and providing the space for such engagement in
the hosting of Kaleidoscope 2019 in Atlanta, Georgia, USA.
The research presented at this conference focused on how information and communication
technologies (ICTs) are set to further revolutionize the heath sector, looking into the technical aspects
such as digital health strategies, smart technologies and access networks for healthcare, as well as
issues of safety, security and data protection. The various sessions, including the special panel
designed by the WHO Department of Evidence and Intelligence for Action in Health, highlighted
how we can use ICT developments to ensure that the goal towards universal, quality health coverage
is achieved. These discussions also aided in the understanding of how ITU’s work on standardization
can advance the digitization of the health sector.
I would like to express my great appreciation to the Kaleidoscope community and the larger ITU
Academia membership for their enduring support to this series of conferences. With over
160 academic and research institutes now members of ITU, the Kaleidoscope series is certain to
continue growing in strength.
My sincerest thanks go to WHO for collaborating with us on Kaleidoscope 2019, to our host, the
Georgia Institute of Technology, Atlanta, Georgia; our technical co-sponsors, the Institute of
Electrical and Electronics Engineers (IEEE), the IEEE Communications Society and The Lancet
Digital Health. I would also like to thank our academic partners and longstanding ITU members,
Waseda University, the Institute of Image Electronics Engineers of Japan (I.I.E.E.J.), the Institute of
Electronics, Information and Communication Engineers (IEICE) of Japan, the Chair of
Communication and Distributed Systems at RWTH Aachen University, the European Academy for
Standardization (EURAS), and the University of the Basque Country.
–i–
I would especially like to thank the members of the Kaleidoscope 2019 Technical Programme
Committee (TPC) and the members of our Steering Committee: Michael Best, Georgia Tech;
Christoph Dosch, IRT GmbH; Kai Jacobs, RWTH Aachen University; Mistuji Matsumoto, Professor
Emeritus Waseda University; Sameer Pujari, WHO; Rupa Sarkar, The Lancet Digital Health; Mostafa
Hashem Sherif, USA (also TPC Chair) and Daidi Zhong, Chongqing University. I would also like to
thank the distinguished General Chairman of Kaleidoscope 2019 and Executive Vice Director of
Research at Georgia Tech, Chaouki Abdallah.
Chaesub Lee
Director
ITU Telecommunication Standardization Bureau
– ii –
Chairman’s message
Chaouki Abdallah
General Chairman
The use of innovative applications and advanced information
and communication technologies (ICTs) are set to continue to
affect the health sector globally, providing significant
developments and ensuring that communities around the world
are capable of providing necessary and efficient healthcare.
Georgia Institute of Technology is proud to provide a space for the presentation and discussion of
essential research towards this year’s ITU Kaleidoscope academic conference on ICT for Health:
Networks, standards and innovation, at our campus in Georgia, Atlanta, USA, 4-6 December 2019.
The establishment of the ITU Academia membership category in 2011 brought greater significance to
Kaleidoscope’s role in fostering academic engagement in the work of ITU. As a member within this
category, Georgia Tech is committed to continuing its support to the Union, and particularly in the
pursuit of research and academic engagement.
The Technical Programme Committee chaired by Mostafa Hashem Sherif selected 20 papers through a
double-blind peer-review process supported by 75 international experts. I would like to thank the
Committee and the reviewers for selecting high-caliber papers for presentation at the conference and
identifying papers eligible for awards.
Among the various keynotes presented in this year’s programme, the first by Valerie Montgomery Rice,
President and Dean of the Morehouse College of Medicine, explored the possibilities of leveraging
digital health technology to advance health equity. Ian F. Akyildiz, the Kenneth G. Byers Professor in
the School of Electrical and Computer Engineering here at Georgia Tech, offered insightful research
into the technical aspects of health applications in the context of an Internet of Bio-Nanothings. Both
keynotes emphasized the importance of investigating the convergence of engineering and medical
research in the pursuit of the global good. John Vertefeuille, of the US Centers for Disease Control and
Prevention, delivered his keynote speech titled, “Polio eradication and how technology is reaching the
last mile,” discussing how digital health plays a key role in combatting disease.
The first Kaleidoscope 2019 invited paper, “Towards international standards for the evaluation of
Artificial Intelligence for health,” co-authored by Markus A. Wenzel and Thomas Wiegand, from
Fraunhofer Heinrich Hertz Institute, explored how international standards are necessary for thoroughly
validating AI solutions for health, and how such standards could create trust among stakeholders. This
presentation also highlighted the achievements of the ITU/WHO focus group on “AI for Health.”
Kaleidoscope 2019 was developed as a joint collaboration between ITU and the World Health
Organization (WHO). In light of this partnership, Marcelo D’Agostino, WHO’s Senior Advisor on
Information Systems and Digital Health, delivered a keynote speech as part of the opening plenary on
“Digital Health in the Information Society: Working together to leave no one behind.” Mr. D’Agostino
also moderated the WHO special panel session titled, “Digital transformation of the health sector: The
power of Artificial Intelligence and the potential of unstructured and Big Data for public health.” Yuri
Quintana from Harvard Medical School discussed the potential power of Artificial Intelligence to
support patients, families and healthcare providers. Ian Brooks of NCSA University of Illinois explored
what potential there might be for public health, given the uses of unstructured data and Big Data today,
and Jennifer Nelson from the Interamerican Development Bank in the United States focused on the
challenges and opportunities surrounding digital transformation in Latin America.
– iii –
Brian Scarpelli, presented part of the second special panel on “Essential considerations for
policymakers addressing the role of Artificial Intelligence in healthcare,” from Connected Health
Initiative, USA. Ilise Feitshans, Fellow in Law at the European Scientific Institute in France, presented
her research on “Global health impacts of personal data protections under European laws and beyond.”
This presentation focused on understanding the role of privacy in society as well as its influence on
personal health, including whether health concerns affect the application, use and disclosure of personal
data in light of the GDPR provisions.
Selected papers from each year’s Kaleidoscope conference are considered for publication in a special-
feature section of IEEE Communications Standards Magazine. In addition, special issues of the
International Journal of Technology Marketing (IJTMKT), the International Journal of IT Standards
and Standardization Research (IJITSR) and the Journal of ICT Standardization may publish extended
versions of selected Kaleidoscope papers. Authors of outstanding Kaleidoscope 2019 papers have also
been invited to contribute to the work of the ITU/WHO Focus Group on ‘AI for Health.’
All papers accepted and presented at the conference will be submitted for inclusion in the IEEE Xplore
Digital Library. The Conference Proceedings from 2009 onwards can be downloaded free of charge
from http://itu-kaleidoscope.org.
I would like to thank our technical co-sponsors, supportive partners and Alessia Magliarditi and her
team at the ITU for their role in ensuring the continued success of the Kaleidoscope series of academic
conferences.
Chaouki Abdallah
General Chairman
– iv –
TABLE OF CONTENTS
Page
Foreword ............................................................................................................................................. i
Chairman's Message ............................................................................................................................. iii
S1.1 5G-enabled health systems: Solutions, challenges and future research trends
Di Zhang; Teng Zhang; Yunkai Zhai; Joel J. P. C. Rodrigues; Dalong Zhang;
Zheng Wen; Keping Yu; Takuro Sato ....................................................................... 1
S1.2 Community healthcare mesh network engineering in white space frequencies
Hope Mauwa; Antoine Bagula; Emmanuel Tuyishimire; Tembisa Ngqondi ........... 9
S1.3 Exploration of the non-intrusive optical intervention therapy based on the indoor
smart lighting facility
Jian Song; Xiaofei Wang; Hongming Zhang; Changyong Pan ................................ 17
S1.4 Access technologies for medical IoT systems
Junaid Ahmed Siddiquee........................................................................................... 23
Session 2: Medical ICT
S3.1 Facilitating healthcare IoT standardization with open source: A case study on
OCF and IoTivity
Hongki Cha; Younghwan Choi; Kangchan Lee........................................................ 49
S3.2 Empirical study of medical IoT for patients with intractable diseases at home
Kentaro Yoshikawa; Masaomi Takizawa; Akinori Nakamura; Masahiro Kuroda .. 59
Session 4: Digital health strategies
S4.1 Invited paper - Towards international standards for the evaluation of artificial
intelligence for health
Markus A. Wenzel; Thomas Wiegand ....................................................................... 67
S4.2 Redesigning a basic laboratory information system for the global south
Jung Wook Park; Aditi Shah; Rosa I. Arriaga; Santosh Vempala ........................... 77
–v–
Page
S4.3 #RingingTheAlarm: Chronic "Pilotitis" stunts digital health in Nepal
Ichhya Pant; Anubhuti Poudyal................................................................................ 85
S4.4 Designing national health stack for public health: Role of ICT-based knowledge
management system
Charru Malhotra; Vinod Kotwal; Aniket Basu......................................................... 95
Session 5: Smart technologies for caregivers
S5.1 Elderly health monitoring system with fall detection using multi-feature based
person tracking
Dhananjay Kumar; Aswin Kumar Ravikumar; Vivekanandan Dharmalingham;
Ved P. Kafle .............................................................................................................. 105
S5.2 A healthcare cost calculator for older patients over the first year after renal
transplantation
Rui Fu; Nicholas Mitsakakis; Peter C. Coyte........................................................... 115
S5.3 Automatic plan generating system for geriatric care based on mapping similarity
and global optimization
Fei Ma; Chengliang Wang; Zhuo Zeng .................................................................... 125
Session 6: Data and artificial intelligence era
S6.1 Invited paper - Preparing for the AI era under the digital health framework
Shan Xu; Chunxia Hu; Dong Min............................................................................. 135
S6.2 Operationalizing data justice in health informatics
Mamello Thinyane .................................................................................................... 145
Session 7: Safety and security in healthcare
S8.1 Technical and legal challenges for healthcare blockchains and smart contracts
Steven A. Wright ....................................................................................................... 173
S8.2 Design of a credible blockchain-based e-health records (CB-EHRs) platform
Lingyu Xu; Antoine Bagula; Omowunmi Isafiade; Kun Ma; Tapiwa Chiwewe ....... 183
S8.3 The GDPR transfer regime and modern technologies
Melania Tudorica; Trix Mulder ................................................................................ 191
Abstracts ............................................................................................................................................. 199
Index of Authors ................................................................................................................................... 209
– vi –
COMMITTEES
Steering Committee
Host Committee
Secretariat
– viii –
Technical Programme Committee
Marco G. Ajmone Marsan (Polytechnic University of Turin and Institute IMDEA Networks, Italy)
Ahmad Alaiad (Jordan University of Science and Technology, Jordan)
Rafael Asorey-Cacheda (Technical University of Cartagena, Spain)
Chaodit Aswakul (Chulanlongkorn University, Thailand)
Luigi Atzori (University of Cagliari, Italy)
Antoine Bagula (University of the Western Cape, South Africa)
Paolo Bellavista (University of Bologna, Italy)
Michael Bove (Massachusetts Institute of Technology, USA)
Marcelo Carvalho (University of Brasilia, Brazil)
Shelly Chadha (WHO, Switzerland)
Periklis Chatzimisios (Alexander TEI of Thessaloniki and Bournemouth University, Greece)
Kejia Chen (Nanjing University of Posts and Telecommunications, China)
Luca Chiaraviglio (University of Rome Tor Vergata, Italy)
Mahmoud Daneshmand (Stevens Institute of Technology, USA)
Alessio Diamanti (Orange & Cnam, France)
Christoph Dosch (ITU-R Study Group 6 Chairman; IRT GmbH, Germany)
Tineke Mirjam Egyedi (Delft University of Technology, The Netherlands)
Marcos Fagundes Caetano (University of Brasilia, Brazil)
Erwin Folmer (University of Twente, The Netherlands)
Luca Foschini (University of Bologna, Italy)
Ivan Ganchev (University of Limerick, Ireland / University of Plovdiv "Paisii Hilendarski", Bulgaria)
Joan Garcia-Haro (Universidad Politécnica de Cartagena, Spain)
Antonio Javier Garcia-Sanchez (Technical University of Cartagena, Spain)
Katja Gilly (Miguel Hernandez University, Spain)
William J. Gordon (Brigham and Women's Hospital, USA)
Smrati Gupta (Microsoft Corporation, USA)
Dijiang Huang (Arizona State University, USA)
Eva Ibarrola (University of the Basque Country, Spain)
Kai Jakobs (RWTH Aachen University, Germany)
Ved P. Kafle (National Institute of Information and Communications Technology, Japan)
Tim Kelly (World Bank, USA)
Katarzyna Kosek-Szott (AGH University of Science and Technology, Poland)
Ken Krechmer (IEEE, USA)
Dhananjay Kumar (Anna University, India)
Andreas Kunz (Lenovo, Germany)
Tsung-Ting Kuo (University of California San Diego, USA)
Mark Leeson (University of Warwick, UK)
– ix –
Jie Li (Shanghai Jiaotong University, China)
Fidel Liberal (University of the Basque Country, Spain)
Luigi Logrippo (Université du Québec en Outaouais, Canada)
Rafael Marin-Perez (OdinS, Spain)
Mitsuji Matsumoto (Waseda University, Japan)
Arturas Medeisis (ITU Arab Office, Riyadh station)
Ahmed Mohammed Mikaeil (Shanghai Jiao Tong University, China)
Alejandro Molina Zarca (University of Murcia, Spain)
Yoshitoshi Murata (Iwate Prefectural University, Japan)
Kazuhide Nakajima (NTT Corporation, Japan)
David Palma (Norwegian University of Technology and Science, Norway)
Vitaly Petrov (Tampere University of Technology, Finland)
RangRao Venkatesha Prasad (Delft University of Technology, The Netherlands)
Alexander Raake (Technische Universität Ilmenau, Germany)
Julia Rauscher (University of Augsburg, Germany)
Anna Riccioni (Università degli Studi di Bologna, Italy)
Domenico Rotondi (FINCONS SpA, Italy)
Mihoko Sakurai (Keio University, Japan)
Andreas Sciarrone (University of Genoa, Italy)
Stefano Secci (Cnam, France)
Cristina Serban (AT&T Security Research Center, USA)
Mostafa Hashem Sherif (Consultant, USA)
Minrui Shi (Shanghai Telecom, China)
Antonio Skarmeta (University of Murcia, Spain)
Michele Solimando (University of Bologna, Italy)
Jian Song (Tsinghua University, China)
Duncan Sparrell (sFractal Consulting LLC, USA)
Christian Timmerer (Information Technology (ITEC) Alpen-Adria-Universität Klagenfurt, Austria)
Marco Torello (University of Bologna, Italy)
Valerio Torti (European University of Rome, Italy)
Taavi Valdlo (Estonian IT Standardization Technical Committee, Estonia)
Riccardo Venanzi (University of Bologna & University of Ferrara, Italy)
Honggong Wang (University of Massachusetts, Dartmouth & College of Engineering, USA)
Jinsong Wu (University of Chile, Chile)
Keping Yu (Waseda University, Japan)
Richard Yu (Carleton University, Canada)
Daidi Zhong (Chongqing University, China)
–x–
KEYNOTE SUMMARY
PANACEA: AN INTERNET OF BIO-NANOTHINGS APPLICATION FOR
EARLY DETECTION AND MITIGATION OF INFECTIOUS DISEASES
Ian F. Akyildiz
The state-of-the-art diagnostics, monitoring, and therapy are limited by the imprecise nature of
current methods and use of devices that are either external, or when implanted, suffer from large size.
A breakthrough is eminent since we are at a critical crossroad in biomedical research in which our
ability to miniaturize sensors and electronics is unprecedented, and our understanding of biological
systems enables fine-grained manipulation and control of behavior of cells down to the molecular
level. These technologies will be leveraged to create Internet of Bio-NanoThings (IoBNT), which
is envisioned to be a heterogeneous network of nanoscale bio-electronic components and engineered
biological cells, so called Bio-NanoThings (BNT), communicating via electromagnetic waves, and
via molecular communication. The objective of this concept is to directly interact with the cells
enabling more accurate sensing and eventually control complicated biological dynamics of the human
body in real time.
As the enabler of IoBNT, Molecular Communication (MC) arises from the observation of chemical
communications in and among the basic units of life, i.e. biological cells, where the information is
represented, exchanged and stored in the form of molecules. The key processes of chemical reactions
and molecular transport are at the basis of encoding, propagation, and processing of information
bearing molecular signals. The focus of this discipline is on the modeling, characterization, and
engineering of information transmission through molecule exchange, with immediate applications in
biotechnology, medicine, ecology, and defense, among others. In the past decade of MC research, the
first studies focused on the physical layer characteristics of communication channels where MC
techniques are defined based on the transport mechanism such as diffusion-based and flow-based MC,
chemotaxis, and molecular motors. However, there is still limited investigation on the definition of
technologies for practical applications of MC. Here, we present a novel perspective on the theory of
MC by expanding on existing and future studies for its application to healthcare.
To illustrate how MC brings together biological and cyber worlds for healthcare applications, we
introduce the concept of a new cyber-physical system called, PANACEA (a solution or remedy for
all difficulties or diseases in Latin), which is a closed-loop solution to the problem of monitoring
infections. PANACEA leverages cutting-edge technologies in the cyber (i.e. machine learning, big
data analytics, cloud computing, security) and physical (i.e. bio-nanosensors, magnetic and wireless
communications) domains to continuously monitor the tissues at risk of serious infection for early
detection and mitigation of infections. By tapping into cell-to-cell communication mechanisms of
bacteria infecting human body, it is possible to estimate the increase in the population of the bacteria
indicating an infection even before the patient shows symptoms. Bio-nanosensors sense
communication molecules, so-called quorum sensing molecules, exchanged among the infectious
bacteria. Quorum sensing is the major cell-to-cell communication mechanism where bacteria produce
and release chemical signal molecules whose external concentration increases as a function of
increasing cell-population density. Therefore, by sensing the concentration of its quorum sensing
molecules, it is possible to estimate the density of the infectious bacteria population. This can be used
to detect infection, which is the invasion of various healthy human tissues by pathogenic bacteria that
are multiplying and disrupting tissues’ operation, causing diseases.
– xiii –
The physical domain of PANACEA will comprise all the bio-nanosensors and actuators (e.g. drug
delivery devices, pacemakers, etc.) embodied by the RIMOR (explorer in Latin) subsystem, which
consists of 3 parts: bio-nanosensor, sensor interface chip, and a coil/inductor for wireless
communication to wearable hub outside of the body. The bio-nanosensor can be diversified by
sensing quorum sensing signals directly or via a reporter bacteria. Moreover, the signals generated
by bacteria can be sensed by utilizing electro-chemical or fluorescence methods. The bio-nanosensor
of RIMOR, has two parts, namely, the bacterial sensor and the physical sensor. The bacterial sensor
senses molecular communication signals generated by the bacteria in the body, and produces light
detected by the physical sensor which converts light to electrical current. This way, MC signals are
transduced to electrical signals to be further relayed to the wearable hub. Interactions between
physical and cyber domains are established by heterogeneous wireless communication modules that
utilize radio-frequency (RF), ultrasonic and molecular communications through RIMOR and
wearable devices.
The cyber part of the PANACEA is in charge of collecting sensing data and performing complex data
processing and learning procedures for the early detection of diseases and infections. The access to
PANACEA is made possible by the Human-Machine Interface (HMI), which provides an easy and
intuitive Data Visualization Interface (DVI) enabling the visualization of relevant information of each
patient and provides alert message management to notify both caregivers and patients when an
infection occurs. The DVI allows human-in-the-loop control thus making it possible for caregivers to
dynamically and actively interact with the system and to regulate drug delivery through ad-hoc
control primitives and APIs exposed by actuator devices. PANACEA not only facilitates interactions
with humans, but it also enables advanced automated drug delivery systems that rely on supervised
machine learning. The learning block is fed with both data collected by the physical system and
supervised input data generated by caregivers. Such an approach makes it possible to train
PANACEA with patient-dependent data so that individual medical treatments can be achieved for
each patient.
Even though applications such as PANACEA are very promising since they are based on the better
defined and more studied MC technique of bacterial communication, a plethora of biomedical
applications can be enabled by the rest of the MC techniques such as calcium signaling, nervous
networks, endocrine network, and molecular motors. The standardization efforts in molecular
communication started in 2014 with the IEEE P1906.1.1 - Standard Data Model for Nanoscale
Communication Systems and they have released IEEE 1906.1-2015 - IEEE Recommended Practice
for Nanoscale and Molecular Communication Framework. Although this standard is a step towards
developing MC as an implementable technology, it only covers the basic diffusion-based molecular
communication and it also includes THz band communication under the nano-communication
umbrella which overlooks underlying challenges arising from the biological nature of MC. Despite
the prior work in the field on the channel characterization, estimation, and capacity calculations of
these aforementioned techniques, a unifying information-theoretic framework that captures the
peculiarities of an MC channel over classical communication systems for all the various MC
techniques, is currently missing.
We aim at filling the aforementioned research gap by providing a mathematical framework rooted in
chemical kinetics and statistical mechanics to define the main functional blocks of MC, to abstract
any MC system and determine or estimate the information capacity of their communication channels.
By using the general formulation of the Langevin equation of a moving nanoscale particle subject to
unavoidable thermally driven Brownian forces as a unifying modeling tool for molecule propagation,
we build a general mathematical abstraction of an MC system. Then, we derive a methodology to
determine (or estimate, whenever closed-form analytical solutions are intractable) the MC channel
capacity based on the decomposition of the Langevin equation into two contributions, namely,
propagation according to the Fokker–Planck equation followed by a Poisson process.
We classify diverse implementations of MC based on their underlying physical and chemical
processes and their representation in terms of the Langevin equation. MC systems based on random
– xiv –
walk, such as calcium signaling in cell tissues, neuron communication by means of neurotransmitters,
and bacterial quorum sensing, include only the contribution of the Brownian stochastic force f. MC
systems based on drifted random walk, such as MC in the cardiovascular system, microfluidic
systems, and pheromone communication between plants, include both f and a drift velocity vn(t) as
function of the time t for each molecule n, which is independent of the Brownian motion. MC systems
based on active transport, such as those based on molecular motors and bacteria chemotaxis, include
instead a deterministic force Fn(t) added to f. For each of these categories of MC systems, and based
on the aforementioned Langevin equation decomposition, we provide a general information capacity
expression under simplifying assumptions and subsequently discuss these results in light of the
functional blocks of more specific MC system models, including cases where a closed-form capacity
expression cannot be analytically derived. This statistical-mechanics-based framework provides a
common ground that not only allows existing researchers in this field to formalize their direction
taken in the last decade in this high-level framework but also provides future researchers with a well-
defined methodology to evaluate the performance of the existing and to-be-discovered MC systems.
We believe this contribution will be foundational for this discipline on the way to standardization,
and an important milestone for the engineering of future MC systems.
MC promises to better understand communications in biological systems, and reciprocally develop
biologically-inspired approaches for communication systems. Since it provides a disruptive
technology based on interfacing directly with living cells and organisms which enables an
unprecedented way of reaching health information in the living body, which we believe will be at the
core of next-generation ICT technologies for human health.
_______________________
*
This talk is based on the following three papers:
1. Akyildiz, I. F., Guler, U., Ozkaya-Ahmedov, T., Sarioglu, A. F., Unluturk, B. D., “PANACEA: An Internet of
Bio-NanoThings Application for Early Detection and Mitigation of Infectious Diseases,” submitted to IEEE
Access, 2019.
2. Akyildiz, I. F., Pierobon, M., Balasubramaniam, S., “An Information Theoretic Framework to Analyze
Molecular Communication Systems Based on Statistical Mechanics,” Proceedings of the IEEE, vol. 107, no. 7,
pp. 1230-1255, 2019.
3. Akyildiz, I. F., Pierobon, M., Balasubramaniam, S., and Koucheryavy, Y., "Internet of BioNanoThings,"
IEEE Communications Magazine, vol. 53, no. 3, pp. 32-40, March 2015.
– xv –
SESSION 1
S1.1 5G-enabled health systems: Solutions, challenges and future research trends
S1.2 Community healthcare mesh network engineering in white space frequencies
S1.3 Exploration of the non-intrusive optical intervention therapy based on the indoor smart
lighting facility
S1.4 Access technologies for medical IoT systems
5G-ENABLED HEALTH SYSTEMS: SOLUTIONS, CHALLENGES AND
FUTURE RESEARCH TRENDS
Di Zhang 1 ; Teng Zhang 2 ; Yunkai Zhai 3,4 ; Joel J. P. C. Rodrigues 5,6 ; Dalong Zhang 1 ; Zheng Wen 7 ; Keping Yu 7 ;
Takuro Sato 7
1
School of Information Engineering, Zhengzhou University
2
Interventional Operating Theater, the First Affiliated Hosital of Zhengzhou University
3
School of Management Engineering, Zhengzhou University
4
National Engineering Laboratory for Internet Medical Systems and Applications, China
5
Federal University of Piauí (UFPI), Brazil
6
Instituto de Telecomunicações, Portugal
7
Waseda University, Japan
[1]. All these characteristics match the requirements of 5G-assisted health systems is a more realistic choice. In
ICT-assisted health systems perfectly. In 5G, massive literature, the ICT-assisted health systems have been
multi-input multi-output (MIMO) [6], non-orthogonal discussed a lot. However, due to the current restrictions
multiple access (NOMA) [7], and full duplex (FD) are of latency, transmission speed and number of connected
emerging technologies for these claimed targets. devices, the telemedicine services are inefficient.
The 5G and beyond wireless technologies provide The 5G-enabled health systems, on the contrary, can
perfect solutions to ICT-assisted health systems, which solve these problems. The most promising technologies
inspires 5G-enabled health systems. For example, with of 5G being applied to the health systems is the URLLC.
the help of ultra-reliable low latency communications Additionally, the massive machine type communications
(URLLC), a 5G-enabled ambulance can provide remote (mMTC) and enhanced mobile broadband (eMBB)
diagnosis and operation. 5G-enabled ambulance can characteristics can further improve the experience of
also automatically respond to an emergency call and ICT-assisted health systems [7, 8]. As is known
plan an optimal route in advance to save precious rescue to all, compared to the fourth generation wireless
time. Besides, URLLC is also a critical issue for remote communications (4G), 5G aims to offer less than 1 ms
surgery. Otherwise, the wound area will be big and latency, more than 20 billion connected devices and up
might cause some risks to the patient’s life. 5G’s large to 1 Gb/s experienced transmission speed. All these
volume real-time medical image transmission also makes features make remote diagnosis and treatment a reality.
remote expert consultation a reality. Moreover, in
general and specialist hospitals, patients or their escorts
2.1 5G-enabled remote diagnosis and
need to press the widely used emergency call-buttons to
treatment
call the medical staff whenever an emergency happens.
It is time-consuming and might even waste precious Telemedicine
rescue time. The 5G-enabled monitoring systems, on vehicle
Remote diagnose
the other hand, can reduce the consumed time and
save the patient’s life especially in emergency conditions. Remote treatment
–2–
ICT for Health: Networks, standards and innovation
–3–
2019 ITU Kaleidoscope Academic Conference
and the remote clinic (hospital) scenario. The control massive Internet of things (IoT) devices connected to
center is in FAHZZU, whereas the remote surgery robot the Internet in order to continuously monitor the vital
workbench can be installed at either the ambulance side signs.
(Figure 3) or the remote clinic (hospital, control center, As depicted in Figure 5, in the 5G-enabled smart
etc.) side. In order to offer reliable connections between ward, we connect ward equipment such as the body
the control center and the surgery robot workbench, we vital sign monitoring sensors, intravenous injector, to
employ both the 5G wireless and extranet connections. a 5G wireless network. The patient’s vital signs and
The 5G wireless connections are used for the ambulance venous transfusion can be remotely monitored and
scenario because of the frequently moving demands. On controlled. Whenever the transfusion is about to finish,
the other hand, the extranet connections are used for the the transfusion monitoring system can inform the nurse
remote clinic (hospital) scenario. Different from the 5G station via 5G networks to withdraw the needles. In
wireless networks, the extranet network can offer more contrast, patients and their escorts need to watch closely
reliable connections. the transfusion speed, and call the nurse by pressing the
Asides from the remote diagnosis and treatment, emergency button if needed.
specialized doctors can remotely operate the surgery if Secondly, doctors and nurses in ward rounds can share
needed. Severely injured patients do not need to travel the patient’s information with each other, and also
to the big hospitals in case of long distance traveling share with the doctor’s office. Real-time information
may cause serious damage. Additionally, it can save can also be transferred to the security office in case
precious rescue and operation time. The remote clinic of encountering a medical dispute. Security offices can
scenario, on the other hand, can provide remote surgery immediately respond to these disputes and record the
operations with less risk compared to the ambulance videos as evidence if needed.
scenario as it does not need any travel. However, due
to the operational risk, we so far have not demonstrated 3. CHALLENGING ISSUES
or tested the 5G-enabled remote surgery yet.
We have introduced the potential solutions and our
implementations of the 5G-enabled health systems in
2.3 5G-enabled smart monitoring the previous section. In the sequel, we will discuss
challenging issues that are faced on implementing
5G-enabled health systems. Currently, eMBB and
mMTC can be easily accomplished by emerging 5G
NR technologies. However, it is still difficult to realize
the URLLC requirements, especially the less than 1 ms
latency. In this regard, we might need some trade-off
strategies between the latency and reliability. Moral
ethics is another challenging issue for implementing
5G-enabled health systems.
Pt
C = log(1 + SINR) = log(1 + ), (1)
Pi + σ 2
–4–
ICT for Health: Networks, standards and innovation
However, story will be different when talking about precision in surgical procedures. Thanks to 5G, we
URLLC. In prior wireless evolution, most of our can reconstruct the three-dimensional view and transmit
attention has been on the transmission speed and these high-definition streaming media [13]. Remote
network capacity enhancement, less attention has been body vital sign monitoring sensors can be used for the
paid to the latency and reliability (successive packet in-home health monitoring, diagnosis and treatment and
rate delivery (SPRD) [10]). This is because that the 10 rescue ambulance. It can save rescue time, travel costs,
ms magnitude latency and 1 − 10−2 reliability are not and reduce the number of outpatients.
challenging technical issues (e.g., the channel coding and However, ICT-assisted health systems propose some
re-transmission can achieve 1 − 10−2 SPRD). However, new challenging issues to the legal and ethical fields. In
when it comes to less than 1 ms latency and 1 − literature, the relationship between robots and humans
[10−9 , 10−5 ] reliability, things become difficult. Albeit has been long argued even after Isaac Asimov’s “Three
the less than 10 ms and 1 − 10−2 are almost enough Laws of Robotics”. One of the widely discussed problems
for the majority of wireless communications, but not is the conflict of interactions between human verdict
in remote surgery. In addition, less latency and higher and robot command. For example, when faced with a
reliability always the mean the better performance and potential traffic accident, should the automatically drive
less operational risk. ambulance hit the pedestrian or avoid it even through
In reality, it is a dilemma to achieve the ultra-reliable it might cause some serious injury to its passengers?
connections and low latency communications. For ICT-assisted health systems also raise some pitfalls
example, hybrid automatic repeat request (HARQ) to the ethics, e.g., patient-doctor relationship erosion,
re-transmissions is a good choice for higher reliability, threat to patient information privacy. If this and
but it will cause higher latency performance [10]. similar ethic dilemmas cannot be perfectly solved, we
Most of the existing works on low latency divide the may not be able to largely and widely implement the
information into short packets, which will generate ICT-assisted health systems. The rapidly and even
network jams because of the large volume of short accelerating technical advances on AI and bio-robot
packets. In addition, the short packet strategy is raise new moral dilemmas, e.g., when the robot is
incapable of VR/AR streaming media information intelligent enough, shall we treat it (or him/she) as a
transmission. This is because AR/VR streaming media human or just a robot?
information transmission requires intensive computing
and a large packet transmission strategy. Recently 4. FUTURE RESEARCH TRENDS OF THE
transmission without guarantee emerges as a hot topic 5G-ENABLED HEALTH SYSTEMS
in terms of low latency [11]. However, since there is no
We focus on the future research trends of 5G-enabled
transmission guarantee, transmission reliability will be
health systems in this section. Due to the author’s
reduced. In this case, for the URLLC requirements,
background, we mainly talk about the 5G NR
some trade-off strategies might be more reasonable
technologies and network architecture redesigning topics
choices [12].
here.
Besides, URLLC solely for the wireless access part might
be easy, yet it will be difficult from the whole network 4.1 5G new radio technology
perspective. Redesigning the network architecture is
required in this regard. For the upper layer technologies, As discussed before, eMBB, mMTC and URLLC
mobile edge computing can offload the network load to are inevitable elements of the 5G-enabled health
edge server, reduce the distance from the subscriber systems to transmit high-definition streaming media
to the vendor and provide edge computing ability for data, to connect more devices to monitor the
signal processing at the edge side. It thus can greatly vital signs, and to reduce the response time. In
enhance latency performance [12]. Network slicing literature, massive multi-input multi-output (MIMO),
is emerging as a promising technology for the new non-orthogonal multiple access (NOMA) and full duplex
network architecture design and URLLC requirements. (FD) are some emerging technologies to accomplish the
It can create some delegate network slicing services eMBB, mMTC and URLLC requirements. For instance,
for specific applications. In network slicing studies, in the work of [6], it is proved that with antenna numbers
substantial works are still needed, for instance, the increasing, uncorrelated noise and fast fading effects
routing algorithm, labeling method, file division and have vanished. Increasing the antenna numbers also
cache strategy, orchestration of various network slicing leads to less required transmitted power per bit, which
pieces. yields better capacity and faster transmission rates per
user. However, the merits of massive MIMO are greatly
3.2 Ethics of 5G-enabled health systems hampered by the pilot contamination. Novel precoding
and beam-forming algorithms are good topics for future
The ethics of 5G-enabled health systems is another research on massive MIMO.
challenging issue on its implementation. As we know, Asides from massive MIMO, NOMA is another
a remote surgery robot has contributed to the greater emerging 5G technology. It utilizes the superposition
–5–
2019 ITU Kaleidoscope Academic Conference
18
ergodic capacity of FD-NOMA can be given as [7].
16
M X
N n+1 t+1
X XX ( γ̄ 1 )
14
C ≈ π log2 e e i,j ×
i=1 j=1 k=1 s=1 (2)
System capacity (Bit/s/Hz)
12 p (−bk bs γ̄ 1 )
ak bk as e i,j ,
10
where M, N denote the transmitter number and
8 receiver number, respectively. We denote αi,k
as the FD self-interference from transmitter i to
6
receiver j. Moreover, γ̄i,j can be given by γ̄i,j =
ραi,j
4 ρ( N
P
α +ηα )+1
, with αi,j , ρ, η, αi,k the NOMA
l=i+1 i,l i,k
–6–
ICT for Health: Networks, standards and innovation
while the TCP-based network throughput is growing some strategies to balance the volume of engaged label
exponentially until reaching the system’s limitation. parameters and the processing accuracy. To this end,
This is because in an edge computing-assisted ICN a joint force from the off-the-shelf cooperative edge
scenario, we may always obtain the required content computing chips, dedicated image processing chips and
from the nearest caches without routing back to the state-of-the-art algorithms, are comprehensively needed.
remote server through the BS.
Edge computing-assisted ICN indeed has some 5. CONCLUSION
drawbacks, one of which is sub-network congestion.
In this work, the 5G-enabled health systems are
It happens especially when a bunch of sub-network
introduced. By leveraging the 5G NR and AI-based
subscribers simultaneously request the same content.
technologies, we can greatly improve the medical service
For future studies, effective routing and cache
quality for the remote areas, and upgrade in-hospital
distribution strategies can be good topics. On the
medical services. The solutions and demonstrations
other hand, optimal content division and labeling
of the 5G-enabled health systems are introduced.
strategy are some other topics for future studies.
For future studies, some new 5G NR technologies,
1000
network architecture redesigned from being data-centric
900 to information and user-centric, the image processing
800 algorithms and specialized devices are needed for better
implementation of the 5G-enabled health systems.
System throughput (MBit/s)
700
600 ACKNOWLEDGMENT
500
This work is supported by the Zhengzhou University
400
Research Startup Foundation under grant:124-32210907
300 and 124-32211247; the Natural Science and
200
Technology Major Projects of China under Grants:
2017ZX03001001-004; the JSPS KAKENHI of Japan
100
under Grant JP18K18044, the National Funding from
0 the FCT-Fundação para a Ciência e a Tecnologia
100 101 102 103
Client number
through the UID/EEA/500008/2019 Project; and
by the Brazilian National Council for Research and
Figure 7 – System throughput comparison between Development (CNPq) via Grant No. 309335/2017-5.
ICN-assisted edge computing and conventional scheme.
REFERENCES
4.3 VR/AR, medical image processing and the [1] D. Zhang, Z. Zhou, S. Mumtaz, J. Rodriguez, and
AI-based technologies T. Sato, “One integrated energy efficiency proposal
for 5G IoT communications,” IEEE Internet of
Currently, image label and detection are needed for Things Journal, vol. 3, no. 6, pp. 1346–1354, Dec.
the implementation of remote surgery and rescue. For 2016.
example, in machine learning based image detection,
we need to label the targets beforehand. We can [2] Y. Fan, L. Yang, D. Zhang, G. Han, and D. Zhang,
unmistakenly detect the target after training with “An angle rotate-qam aided differential spatial
existing objectives. However, it is still hard for the AI modulation for 5G ubiquitous mobile networks
to detect the unexperienced targets. The omitted target (accept),” Mobile Networks and Applications, 2019.
poses risks to the automatic ambulance and remote
surgery. Similarly, for the medical image applications, [3] J. J. P. C. Rodrigues, I. de la Torre, G. FernÃąndez,
we also need to label these disease symptoms, once and M. LÃşpez-Coronado, “Analysis of the security
some labels are omitted or mistakenly studied by the and privacy requirements of cloud-based electronic
algorithm, it will result in misdiagnosis. health records systems,” Journal of Medical
Internet Research, vol. 15, no. 8, p. e186, Aug. 2013.
Apart from the labeling and detection method, ML
algorithm’s processing speed is slowed down by the large [4] Y. Xue and H. Liang, “Analysis of telemedicine
scale and even growing AR/VR-based three-dimensional diffusion: The case of china,” IEEE Transactions
data volume. This is mainly due to the limited ability on Information Technology in Biomedicine, vol. 11,
of existing processing devices. In order to accelerate no. 2, pp. 231–233, Mar. 2007.
the processing speed, large scale deep learning (DL)
high-rank matrix factorization (MF) algorithms and [5] S. Thelen, M. Czaplik, P. Meisen, D. Schilberg,
dedicated processors are needed in the future. Due to and S. Jeschke, “Using off-the-shelf medical devices
the large scale and even growing data, we also need for biomedical signal monitoring in a telemedicine
–7–
2019 ITU Kaleidoscope Academic Conference
system for emergency medical services,” IEEE [16] H. Harai, K. Fujikawa, V. P. Kafle, T. Miyazawa,
Journal of Biomedical and Health Informatics, M. Murata, M. Ohnishi, M. Ohta, and
vol. 19, no. 1, pp. 117–123, Jan. 2015. T. Umezawa, “Design guidelines for new generation
network architecture,” IEICE transactions on
[6] T. L. Marzetta, “Noncooperative cellular wireless communications, vol. 93, no. 3, pp. 462–465, Mar.
with unlimited numbers of base station antennas,” 2010.
IEEE Transactions on Wireless Communications,
vol. 9, no. 11, pp. 3590–3600, Nov. 2010. [17] V. P. Kafle, Y. Fukushima, and H. Harai, “Internet
of things standardization in ITU and prospective
[7] D. Zhang, Y. Liu, L. Dai, A. K. Bashir, networking technologies,” IEEE Communications
A. Nallanathan, and B. Shim, “Performance Magazine, vol. 54, no. 9, pp. 43–49, Sep. 2016.
analysis of FD-NOMA-based decentralized V2X
systems,” IEEE Transactions on Communications, [18] A. Laghrissi, T. Taleb, and M. Bagaa, “Conformal
vol. 67, no. 7, pp. 5024–5036, July 2019. mapping for optimal network slice planning based
on canonical domains,” IEEE Journal on Selected
[8] D. Zhang, Z. Zhou, S. Mumtaz, J. Rodriguez, and Areas in Communications, vol. 36, no. 3, pp.
T. Sato, “One integrated energy efficiency proposal 519–528, Mar. 2018.
for 5G IoT communications,” IEEE Internet of
Things Journal, vol. 3, no. 6, pp. 1346–1354, Dec. [19] Z. Wen, D. Zhang, K. Yu, and T. Sato,
2016. “Information centric networking for disaster
information sharing services,” IEICE Transactions
[9] M. Mikhail, K. Mithani, and G. M.Ibrahim, on Fundamentals of Electronics, Communications
“Presurgical and intraoperative augmented reality and Computer Sciences, vol. 98, no. 8, pp.
in neuro-oncologic surgery: Clinical experiences 1610–1617, Aug. 2015.
and limitations,” vol. 128, pp. 268–276, Aug. 2019.
–8–
COMMUNITY HEALTHCARE MESH NETWORK ENGINEERING IN WHITE SPACE
FREQUENCIES
and the results show the designs can guide network engineers The main goal of the backbone-based techniques is to find
to select the most relevant performance metrics during a a connected subset of nodes in a network that guarantee
network feasibility study aimed at guiding the implementation connectivity by allowing every other node in the network
process. to reach at least one node on the backbone in a direct way
The rest of the paper is structured as follows: section 2 [11]. A communication backbone can be created by selecting
introduces topology reduction and discusses the approaches nodes that form a connected dominating set (CDS). From
used to achieve it; section 3 discusses the proposed graph theory, a CDS of a graph is a connected subset in which
network optimization function that is used to introduce all other nodes that do not belong to that subset have at least
hierarchical backbone network topology from sparse network one adjacent neighbor inside the subset. Advantages of this
topology; section 4 discusses the proposed link-based CDS-based topology control are collisions control, protocol
topology reduction algorithm for reducing dense mesh overhead control and energy consumption reduction, efficient
network topology to sparse mesh network topology; section network organization and scalability improvement [10].
5 discusses the backbone network topology algorithm used
to introduce hierarchical backbone network topology from 3. NETWORK OPTIMIZATION FUNCTION
sparse network topology; section 6 is a performance
evaluation of the proposed designs; and section 7 concludes The network design consists of finding a network
the paper. configuration expressed by the graph G = (N, L), where N
is the set of nodes while L is the set of links connecting the
nodes with the objective of optimizing an objective function
2. TOPOLOGY REDUCTION AND APPROACHES
representing a penalty to be minimized or a profit/reward
to be gained. In this paper, the network engineering profit
While algorithms discussed in this section are designed for
function P(G) is considered. It combines reliability and
application in physical networks, the designs proposed in
quality of service (QoS) features, which are based on three
this paper are for predesigning a network topology offline
metric measures; node degree, link margin and Euclidean
before it is replicated in reality. In general, topology control
distance.
can be achieved through three main mechanisms: power
control technique, power mode mechanism and hierarchical
3.1 Network engineering design
formation technique.
In power control technique the communication range of the The profit function P(G) is expressed as follows:
wireless nodes is controlled by modifying the transmission Õ
power parameter of the nodes in the network. This way the P(G) = P(i) (1)
network nodes are able to better manage their neighborhood i ∈N
size, interference level, power consumption and connectivity P(i) = α ∗ ndi + β ∗ lmi + γ ∗ spi (2)
[9]. In power mode mechanism, the node activity is controlled
by switching between active and sleep operation modes to where, α, β and γ are coefficients of proportionality used to
dispense with redundant nodes and still achieve the desired express the preference for a given metric measure. A high
connectivity [10]. The main idea of the algorithms using value of one of the coefficients reveals a preference for the
these first two mechanisms is to produce a connected topology corresponding metric measure. The profit P(i) expresses the
by connecting each node with the smallest necessary set of resultant preference of node iN to be part of the backbone.
neighbors and with the minimum transmission power possible The metric measures are explained below.
[11]. These first two techniques are the main options for
1. Node degree: Nodes with a higher node degree lead
flat networks, where all nodes have essentially the same role
to reduced network topology for the backbone network,
[7, 13], i.e., in an homogeneous infrastructure.
which is preferred to nodes with a lower node degree.
Controlling the transmission power of the nodes or their Therefore, preference is given to nodes with a higher
activities only reduces the network topology to help save node degree than nodes with a lower node degree. The
energy but the approach does not prevent the transmission of node degree nd(i) of node i in a network graph with N
redundant information when several nodes are close to each number of nodes is calculated as:
other and may not simplify the network topology enough
for scalability [11]. The hierarchical formation technique N
Õ
addresses the scalability problem. In hierarchical formation nd(i) = xi j (3)
technique, a reduced subset of the nodes in the network j=1
is selected and given more responsibilities on behalf of a where xi j = 1 if there is a link between node i and node
simplified and reduced functionality for the majority of the j and xi j = 0 otherwise.
nodes [11]. This approach greatly simplifies the network
topology and saves additional energy by assigning useful 2. Link margin: Links with higher link margins are
functions, such as information aggregation and filtering and better for communication than links with lower link
routing and message forwarding to the reduced subset of margins. Furthermore, nodes whose corresponding
nodes [11]. A hierarchical topology can be constructed by links have smaller differences in link margins are better
using either a backbone network or a cluster-based network. for communication than nodes whose corresponding
– 10 –
ICT for Health: Networks, standards and innovation
– 11 –
2019 ITU Kaleidoscope Academic Conference
into a sparse mesh network topology. The objective of the 5.1 Backbone network design algorithm
Algorithm 1: LTR algorithm The algorithm for creating a hierarchical backbone network
1 mark all links in dense mesh network as non-visited; topology is provided by Algorithm 2. It uses a graph coloring
2 for each non-visited link of the network do approach, where the nodes of the network are initially
3 select worst non-visited link of the network; // i.e., assigned a white color and thereafter, they are colored black or
link with lowest link margin. gray, depending on whether they have qualified for backbone
4 artificially delete the link; or edge status. This algorithm returns a network configuration
5 run the K-shortest path to detect if the network is still
k-connected; // it is k-connected if you Algorithm 2: Backbone formation
can find k-disjoint shortest paths for
1. Initialisation.
each source-destination pair of the
Assign a white colour and zero height to all nodes of the
reduced network.
network,
6 if it is k-connected then
Select a node n from W hite whose profit/reward is highest,
7 remove the link permanently;
Backbone ← {n},
8 else
Grey ← all neighbours of n,
9 leave the link and mark it as visited;
W hite ← N \ ({n} ∪ Grey).
10 end
2. Select a node k from Grey whose profit/reward is highest and
height is lower.
algorithm is to improve i) quality of the links by retaining
Include k into the Backbone,
the links of high margin and pruning those of low margin Assign a black colour to k and update its height,
and ii) maintain the reliability of the network at a predefined Remove k and its neighbours from W hite,
level. In order to design fault-tolerant networks, the algorithm Include the neighbours of k in Grey.
uses the K-Shortest Path (K-SP) algorithm in [14] to compute
3. Repeat Step 2 whenever W hite , ∅.
K-shortest paths between source-destination pairs where K >
1. Links that provide K-disjoint shortest paths from each
node to the network sink are considered and included in the where the backbone nodes are colored into black and the edge
sparse network. nodes are colored into gray.
Õ
τ̂opt (Copt ) = max P(k) (7)
Cn ∈ G
k ∈N[Cn ]
subject to
((7).1) lm (x, y) > τlm ∀ x, y ∈ Copt
((7).2) k sp (x, y) > τsp ∀ x, y ∈ Copt
((7).3) ∀n ∈ Copt : n ∈ N̂ ∨ ∃m ∈ N̂ : (n, m) ∈ L
((7).4) N̂ ∪ Ň = ∅ ∧ N̂ ∩ Ň = N
where N(X) is the set of nodes in the configuration X. Note
that constraints (7).1 and (7).2 express the QoS in terms of link
margin and reliability respectively, while constraints (7).3
and (7).4 represent the topology control model in terms
of backbone connectivity based on the K-dominated set
model [17, 18, 19]. Figure 2 – Network engineering process
– 12 –
ICT for Health: Networks, standards and innovation
– 13 –
2019 ITU Kaleidoscope Academic Conference
– 14 –
ICT for Health: Networks, standards and innovation
REFERENCES
– 15 –
2019 ITU Kaleidoscope Academic Conference
– 16 –
EXPLORATION OF NON-INTRUSIVE OPTICAL INTERVENTION THERAPY
BASED ON THE INDOOR SMART LIGHTING FACILITY
1
Tsinghua University, Beijing, P. R. China
Keywords – Alzheimer's disease, human physiological In 2018, Edward S. Boyden and Li-Huei Tsai showed that
rhythm, Internet of light, LED, non-intrusive optical optogenetically driving fast-spiking parvalbumin-positive
intervention therapy (FS-PV)-interneurons at gamma (40 Hz) can reduce levels of
amyloid-β (Aβ)1–40 and Aβ 1–42 isoforms [2].They
1. INTRODUCTION designed a non-invasive 40 Hz light-flickering regime that
successfully reduced levels of Aβ1–40 and Aβ1–42 in the
The artificial light source is perhaps one of the most visual cortex of pre-depositing mice and mitigated plaque
important inventions for human beings. Since it is not always load in aged, depositing mice to attenuate Alzheimer’s-
possible to enjoy the natural light day and night, human disease-associated pathology [3][4].
beings started to explore artificial light sources, and the
electric light source is an important outcome from this effort. On the other hand, LED lamps based on semiconductor
Thomas Edison first invented the incandescent lamp, lighting are becoming more and more popular in the world
marketed on a large scale in 1879, which is taken as the first due to their low cost, high luminous efficiency and long-life
leap in the development of electric light source. In 1938, the expectancy. Unlike the incandescent lamps or other light
birth of the fluorescent lamp (low pressure gas discharge sources in the past, one can easily adjust the intensity and
lamp) made the electric light source achieve its second giant color temperature of LED lights to accommodate people’s
leap. Later, in 1993, the famous blue-light LED technology, needs. Preliminary research results show that it could
invented by Dr. Suiji Nakamura, and successfully promoted potentially provide a new type of non-intrusive treatment by
to commercialization of the LED lighting source is changing the intensity and color temperature for the indoor
considered as the third great leap in the history of the lighting environment. At present, the problem of an aging
society in most countries around the world becomes much example, Internet of Radio and Light in [8][9], are discussed
more severe than ever before. Depression is a high-risk in section 5.
disease for the senior people, which seriously affects the
physical and mental health, and eventually jeopardizes the 2. SYSTEM DESCRIPTION
quality of life [5]. Numerous studies have shown that light
stimulation can effectively alleviate depression or other The strategic roadmap from 2015 to 2025 of the European
psychological disorders. Some literature has attempted to Lighting Association is shown in Figure 1, indicating that the
systematically analyze the research trends for the optical semiconductor lighting started with the environment
intervention therapy for senile depression under the indoor protection purpose is currently in the stage of intelligent
lighting environment, and to combine the current light lighting, which is about to transit to human-oriented lighting,
environment status of retirement buildings with the visual, that is, smart lighting [10]. The goal is to achieve smart
psychological and physiological characteristics of the senior lighting that supports healthy lighting by 2025 and to provide
people. Following this, key considerations of the healthy people with a healthy and comfortable indoor living
light environment of the nursing space against the depression environment. However, during the current stage of
of the seniors [6] [7], are given. As an effective treatment, intelligent lighting, the main goal is still to save energy.
the optimal dose of phototherapy treatment time, light Considering that the intensity and color temperature of LED
intensity and duration of illumination need to be intelligently light sources can be easily controlled, ICT technology
adjusted depending on the type of illness, the severity of the combined with sensors and the intelligent driver in the
condition, and individual characteristics. Phototherapy has luminaire can be used to monitor and track environmental
advantages that are easy to control and implement with changes in real time.
negligible side effect when compared with traditional drug
treatments. More importantly, phototherapy provides a
compatible aid with the regular drugs for the treatment of
mental illness, which can accelerate improvement and
alleviate the symptoms.
– 18 –
ICT for Health: Networks, standards and innovation
senior people. The research focus on IoL can be summarized To deal with the challenges presented by the design and
as the three following issues: optimization on IoL systems, the research should focus on (1)
joint sensing and key data extraction to improve both the
First is theoretical analysis on the multidimensional joint accuracy and coverage from a sensing perspective; (2)
sensing and intrinsic data extraction under the dense, in- resource coordination and control mechanism of the
depth illumination coverage to support the construction of heterogeneous network (wired and wireless) for reliable
ubiquitous multifunctional light-borne-based sensor access under the burst mobile environment; and (3)
networks. intelligent light control for user-centric applications which
potentially provide the possibility for non-intrusive optical
Second is investigation of a heterogeneous network structure intervention therapy.
and channel characteristics consist of wireless and wired
communication (powerline communication, PLC) systems
and corresponding resource coordination mechanisms for the
reliable, burst mobile access under massive interconnection.
In this system, each LED can serve as the sensing node after
being integrated with sensors such as the hazard gas, Figure 3 - Schematic diagram of IoL with the focus on
occupancy and moving objects; the sensing data collected by main technical challenges and research objectives
the node will be sent to the operating center any by the wired
system (i.e., PLC) or the wireless (using radio frequency, RF, Other than the advantages of energy saving, as well as the
signals) system or the combination of both for robust data low operation and maintenance costs of lighting equipment
delivery. Then the operating center will analyze those inside the building, defining and producing the specific work
received data, make decisions on the operations to be taken, environment or atmospheres and supporting LED lamp
and send the control commands to each node eventually. For interconnectivity for the value-added services based on
example, if the hazard gas is detected by one node or several intelligent lighting systems, IoL is expected to be handily
neighboring nodes, the alarm signal will be sent immediately adjusted for the circadian rhythm control of the human body
to the operating center, and then to the corresponding people and with the function of serving healthy lighting. Here, we
or the agency while the ventilation system will start working name it as non-intrusive optical intervention therapy, which
automatically. With this arrangement, IoL can be established is different from the well-known photo dynamic treatment.
by combining the naturally combined illumination and It covers quite different applications, not only the visual
power supply networks to provide information services in a health needs such as suitable brightness, no glare and no
very cost-effective way. stroboscopic illumination, but also psychological and
physical health for working place safety and working
efficiency improvement, circadian rhythm regulation and
disease rehabilitation. The natural light changes during one
day showing that the high color temperature environment
under moderate brightness can inhibit melatonin secretion,
induce alertness and improve work efficiency while the low
color temperature environment stimulates melatonin,
promotes relaxation and sleep. It is believed for an
environment with high color temperature but insufficient
brightness, people feel gloomy and depressed. Through an
intelligent lighting control system which mimics the daily
changes of the natural light for the indoor environment, LED
physiological lighting that meets people's health needs can
be used to improve lighting comfort, adjust physiological
rhythm, improve psychological mood and improve work
efficiency as well, assist and treat diseases, etc.
– 19 –
2019 ITU Kaleidoscope Academic Conference
– 20 –
ICT for Health: Networks, standards and innovation
– 21 –
2019 ITU Kaleidoscope Academic Conference
Study on the neurological mechanism of different factors [4] Chinnakkaruppan Adaikkan, Steven J.Middleton,
of visible light radiation (color temperature, frequency, Asaf Marco, Ping-Chieh Pao, Hansruedi Mathys,
modulation mode, radiation intensity). The significance David Nam-Woo Kim, FanGao, Jennie Z.Young,
of light in biological evolution is self-evident, yet its Ho-Jun Suk,Edward S.Boyden,Thomas J.McHugh,
molecular mechanism is still quite unclear, especially the Li-HueiTsai, “Gamma Entrainment Binds Higher-
mechanism of the direct effect of the light on the nervous Order Brain Regions and Offers Neuroprotection”,
system. One can continue to carry out mouse experiments Neuron 102, 929–943 June 5, 2019
on this platform to accumulate more knowledge in this
area. It is anticipated that successful implementation of [5] World Health Organization: Group Interpersonal
this work into the human medical applications will open Therapy (IPT) for Depression, 2016:
a new field for human recognition. https://www.who.int/mental_health/mhgap/interpers
onal_therapy/en/
There is no doubt ICT technology could play a much more
important role with the combination of the non-intrusive [6] Cui Zhe, Hao Luoxi, Xu Junli, “A Study on the
optical intervention therapy by introducing an adaptive Emotional and Visual Influence of the CICU
feedback mechanism through artificial intelligence, machine Luminous Environment on Patients and Nurses”,
learning and other methods. The therapeutic effect can be Journal of Asian Architecture and Building
tracked and the treatment process and intensity can be Engineering, 2017
flexibly adjusted to further enhance the effectiveness. Other
than that, there might be another advantage by conducting [7] World Health Organization, Mental health of older
this research. As visible light communication is considered adults. December 2017:
as a promising technology for indoor applications, it is also https://www.who.int/en/news-room/fact-
necessary to evaluate the potential impact from the low- sheets/detail/mental-health-of-older-adults
frequency operation of LED in VLC application on human
wellness. Therefore, it is quite important to consider and [8] ITU-R REPORTS
coordinate standardization efforts from the perspectives of Visible light for broadband communications:
communication, Internet of light application, and the non- https://www.itu.int/pub/R-REP-SM.2422
intrusive optical intervention therapy.
[9] J. Cosmas, Y. Zhang and X. Zhang, "Internet of
6. ACKNOWLEDGEMENT Radio-Light: 5G Broadband in Buildings," European
Wireless 2017; 23th European Wireless Conference,
The authors would like to thank Dr. Wei Shi for her great Dresden, Germany, 2017, pp. 1-6.[10]
help in arranging mouse experiments. This work was Strategic Roadmap 2025 of the European
supported by the National Key Research and Development Lighting Industry:
Program of China (2017YFB0403402) and also by the https://www.lightingeurope.org/images/160404-
Natural Science Foundation of Guangdong Province (Grant LightingEurope_Roadmap---final-version.pdf
No. 2015A030312006).
[11] openbci website: https://openbci.com/
REFERENCES
[12] EEGLAB website:
[1] Nobel prize website 2017: https://sccn.ucsd.edu/eeglab/index.php
https://www.nobelprize.org/prizes/medicine/2017/pr
ess-release/ [13] neuroexplorer website:
https://www.neuroexplorer.com/
[2] HF Iaccarino, AC Singer, AJ Martorell, A Rudenko,
F Gao, TZ Gillingham, H Mathys, J Seo, O Kritskiy,
F Abdurrob, C Adaikkan, RG Canter, R Rueda, EN
Brown, ES Boyden, LH Tsai, “Gamma frequency
entrainment attenuates amyloid load and modifies
microglia” , Nature 540:230–235, December 8, 2016
– 22 –
ACCESS TECHNOLOGIES FOR MEDICAL IOT SYSTEMS
Ericsson, India
sensors. Device types in healthcare can be mobile handsets, Table 1 – Access technologies for IoT (Compilation)
laptops/computers, screens, cameras, diagnostic tools,
monitors and other advanced tools and equipment. 3GPP/ 2G- 3G- 4G- NB- 5G
3GPP2 GSM, WCDMA, LTE/ IOT
CDMA HSPA LTE-M
Applications: The users of the m-health applications can be:
[3] a) health professionals (physicians, nurses, midwives, Zigbee
etc.); b) public including patients and healthy individuals; c) IEEE 802.11 802.15.6
health institutions (hospitals, insurance companies, drug (Wi- (WBAN)
stores, etc.). These users would be interested in the various Fi)
lines of preventive and general treatment. Information
availed from the end points will be analyzed resulting in the LoRa
future course of action or to bring about new insights. This SIGFOX
will enable medical expertise at a central location to quickly
Bluetooth BLE BR/EDR
diagnose and send expert advice.
Weightless
5. ARCHITECTURE AND TECHNOLOGIES
6. ACCESS REQUIREMENTS OF MEDICAL IOT
There are several ways to visualize the layers making up the SYSTEMS
IoT architecture. Here we show three layers; the lowermost
layer has the IoT end points: the devices, sensors and other e-health and m-health services can be availed remotely. For
equipment that will communicate through intervening layers these services to be effective, the communication access
to talk to the application (s). The middle layer is the one that systems that talk to the end devices and the upper layers,
provides the connectivity between the devices to the including the applications, need to fulfill certain criteria. We
different modules and functions residing in the upper layer. can categorize the deployment requirements for such access
This is the access layer or the connectivity layer. The upper systems into two categories: current requirements and
layer is a conglomeration of many sublayers: the upcoming requirements. By current we define the access
connectivity management, device management and systems that are presently serving the IoT needs.
functions as the operations, billing and revenue management.
Data resides here. 6.1 Current Access Fulfillment
– 24 –
ICT for Health: Networks, standards and innovation
– 25 –
2019 ITU Kaleidoscope Academic Conference
3. Quality of end devices and overall cost of 10. Standardization: Standardization organizations
ownership: exist in the telecommunication and ICT domains, where
these bodies work on enhancements in existing features and
a) Resiliency: from cyberattacks, equipment and network functionalities. There exist medical associations and trade
architecture to enable availability. Node architectures like organizations that define the ethics and ways of working for
CU-DU split in radio access nodes, CP-UP in core networks, the medical fraternity. Research is ongoing in the medical
network slices, container-based cloud applications will help field in newer and better ways of treatment and drugs. The
in application recovery and resilience. Implementation of challenge lies in ensuring that the various research and
such changes is not expected to take place rapidly and will standard organizations and associations operating for the
take time for full-scale deployment. b) End-to-end various layers are working in tandem. This is a tall order in
ownership: spanning from the devices, access systems, today’s world.
gateways, internet, applications and databases spread in the
cloud. Service Level Agreements (SLAs) of the network, 8. MODEL FOR IOT DEPLOYMENT
SLAs of the medical equipment will become relevant and
needed. Currently, e2e SLAs for IoT systems spanning The paper puts forward a model for IoT keeping in view the
across multiple layers with different ownerships are non- requirements of the healthcare sector. Multiple factors
existent or evolving. contribute to a healthy ecosystem for enabling IoT for
healthcare. The importance of each factor cannot be
5. Security a) Network security: algorithms used for discounted as the diffusion of IoT hinges on each of the
encryption and ciphering at the access layer, application enabling factors.
layer and database layer with adequate protection for the
control and traffic layers. b) User security: the end-to-end
encryption from user to application. SIM-based
– 26 –
ICT for Health: Networks, standards and innovation
– 27 –
2019 ITU Kaleidoscope Academic Conference
– 28 –
ICT for Health: Networks, standards and innovation
11. WAY FORWARD applications and takes it forward to capture the probable
This paper posits some recommendations to leverage the needs for future systems. The challenges for the deployment
potential of IoT in the ever-increasing domain of medical of IoT systems to fulfill the needs of medical systems are
healthcare. However, to effectively harness the potential of analyzed and two technological standards, 5G and Wi-Fi 6
technology, certain steps need to be taken in the overall that hold promise for meeting the future needs of medical
ecosystem. IoT systems, are analyzed. Deployment models for the
access technologies are presented. However, significant
1. Build reference frameworks changes are needed in the ecosystem and the paper suggests
This paper suggests that a reference model be formulated some actions to bring out the latent potential for the use and
keeping in consideration the unique needs of the healthcare diffusion of IoT systems in the healthcare sector.
sector. A model is proposed in this paper. This will help each
actor in the ecosystem to understand the requirements and REFERENCES
work accordingly to bring in synergies. This aspect is
significant in a fragmented environment where multiple [1] "eHealth," WHO, [Online]. Available:
OEMs and service providers are present. https://www.who.int/ehealth/en/. [Accessed 1 July
2. Regulatory guidance 2019].
In the face of fast-moving developments in the use of
technology in the healthcare industry, the regulation in any [2] WHO, "mHealth New horizons for health through
dispensation will need to be lightweight. This will help mobile technologies," Global Observatory for eHealth
innovation. Regulation needs to come up with series - Volume 3, 2011.
recommendations on individual privacy and security needs
and protecting the end consumer of medical services. Areas [3] ,. G. M. Mobin YASINIa, "Toward a use case based
like the development of technology, devices, the architecture classification of mobile health applications," Digital
of deploying the IoT systems, etc. are best left to the market. Healthcare Empowering Europeans R. Cornet et al.
Besides institutional regulation, the role of self-regulation in (Eds.) © 2015 European Federation for Medical
this emerging field calls for high ethical standards from Informatics (EFMI). , pp. 175-179, 2015.
professional bodies, manufacturers, service providers and
users.
[4] S. &. S. S. K. &. W. X. Bhandari, "Latency
3. Certification Minimization in Wireless IoT Using Prioritized
A plethora of standard bodies, 3GPP, IEEE, Open stack, Channel Access and Data Aggregation,"
OpenFOG, IETF and OMA exist today. The future is 0.1109/GLOCOM.2017.8255038, 2017.
increasingly heterogeneous and interoperability and
interworking between different standards will be the need. [5] R. Pegoraro, "You're buying a 4K TV. How much
Quality has to be the essence on which medical IoT systems Internet bandwidth do you need?," 2017. [Online].
would hinge on. An independent certification process that Available:
ensures the quality of the products and meet minimum https://www.usatoday.com/story/tech/columnist/201
standards like power consumption, interference to other 7/12/10/youre-buying-4-k-tv-how-much-internet-
systems, security aspects need to be followed. Regulation bandwidth-do-you-need/933989001/.
and certification are normally independent functions. The
certification process may need to conform to the guidelines [6] A. Kishore, "Worried About Bandwidth for 4K? Here
emanating from the regulation aide. Learnings from the Comes 8K!," 2017. [Online]. Available:
certification process may, in turn, lead to changes in https://www.lightreading.com/video/4k-8k-
regulatory guidelines. video/worried-about-bandwidth-for-4k-here-comes-
4. Demonstrate POC in real conditions 8k!/d/d-id/737330.
Healthcare conditions are unique in different countries. In
the western world, facilities are remarkably different from [7] A. P. L. D. N. &. F. F. Noemi Scarpato, "E-health-
the developing world. Proof of concepts and trial systems IoT Universe: A Review," International Journal on
help enhance the knowledge of the application of Advanced Science Engineering Information
technologies like IoT to real-world problems. Healthcare is Technology, vol. Vol.7 (2017) , no. No 6, 2017.
an area that directly touches the user and success and
confidence of IoT-based technologies will help in the [8] D. V. Dimitrov, "Medical Internet of Things and Big
acceptance of such solutions. Data in Healthcare," Healthc Inform Res, doi:
10.4258/hir.2016.22.3.156, 2016.
12. CONCLUSION
Affordable healthcare through m-health and e-health [9] U. &. D. R. Kar, "Application of Artificial
applications riding on ICT will play an important role in Intelligence in Healthcare: Past, Present and Future,"
providing healthcare to remote and rural areas. This paper Arch Biomed Eng & Biotechnol.1(1):
brings out the needs of IoT-based systems for medical 2018.ABEB.MS.ID.000503, 2018.
– 29 –
2019 ITU Kaleidoscope Academic Conference
[13] "Therapeautic Goods Administration,Department of [24] Recommendation ITU-T Y.2060 (2012), Overview
Health, Australian Givernment," TGA, 19 July 2019. of the Internet of things.
[Online]. Available:
https://www.tga.gov.au/publication/medical-device-
[25] "Infrastructure, Internet Protocol Aspects And Next-
cyber-security-guidance-industry.
Generation Networks- Overview Of The Internet Of
Things".
[14] A. Mulero, "France Drafts Medical Device
Cybersecurity Recommendations," 23 July 2019.
[26] D. Darwish, "Improved Layered Architecture for
[Online]. Available: https://www.raps.org/news-and-
Internet of Things".- International Journal of
articles/news-articles/2019/7/france-drafts-medical-
Computing Academic Research (IJCAR) ISSN 2305-
device-cybersecurity-recomme.
9184 Volume 4, Number 4 (August 2015), pp.214-223
© MEACSE Publications
[15] K. e. a. Boeckl, "Considerations for Managing http://www.meacse.org/ijcar.
Internet of Things (IoT) Cybersecurity and Privacy
Risks," National Institute of Standards and
[27] ETSI, "Why do we need 5G?," [Online]. Available:
Technology, June 2018.
https://www.etsi.org/technologies/5g.
– 30 –
SESSION 2
MEDICAL ICT
1
Faculty of Software and Information Science, Iwate Prefectural University
2
Solution Strategic Department, DOCOMO Technology, Inc.
ABSTRACT One of the main reasons why the cost of introducing existing
powered prosthetics is too expensive may be that the
Several million people around the world live with limb loss. prosthetic market is not open. A number of manufacturers
Prosthetics are useful to improve their quality of life, and provide them as an integrated device, and the components
some powered prosthetics enable them to walk naturally. are not compatible between different manufacturers.
However, most are too expensive for most amputees to afford.
We propose a module structure for a foot prosthetic and Introducing a module structure to prosthetics and
standardized interfaces between modules to lower the price standardizing the interface between modules will enable the
of powered ones. The prosthetic is battery-powered and price of powered prosthetics to be lower. In this paper, we
controlled by data from sensors built into the heel of a shoe propose a module structure for foot prosthetics with
for a healthy foot. Some modules can be applied to people standardized components.
with walking disabilities. Such standardization can lower the
price of such modules, and many amputees and people with Our paper is outlined as follows. Existing power-assist
walking disabilities, such as hemiplegia, can easily afford prosthetic leg designs are introduced in section 2. The results
them, which can help improve their quality of life. of our research related to the gait of stroke patients are
introduced in section 3. We developed a walking assist shoe
Keywords – Amputee, foot prosthetic, gait assist, walking that has a coil and leaf spring to easily raise the heel. Its
disability structure and effect of raising the heel are introduced in
section 4. Our proposed module structure for a foot
1. INTRODUCTION prosthetic is introduced in section 5. The heel-up spring,
which is one of the modules comprising the foot prosthetic,
As the percentage of elderly people in the world’s population is derived from the results of the walking assist shoe
is increasing [1], the number of functionally impaired people, introduced in section 4. We conclude in section 6.
such as those with cerebrovascular diseases, will also
increase. People with such diseases often have walking 2. EXISTING POWER-ASSIST FOOT
disabilities, which increases their risk of falling and PROSTHETIC
consequently injuring themselves [2]. One main cause of this
is due to their inability to raise their heel and swing their toes In this section, we introduce existing powered foot
up because of muscle weakness [3]. prosthetics. Ottobock in Germany and Ösuur in Iceland
provide such prosthetics to consumers and the
There are nearly 2 million people living with limb loss in the Biomechatronics Group, a research group within MIT Media
United States [4]. Maurice LeBlanc estimated the number of Lab., has also developed some models.
amputees was approximately 10 million in the world, with
30% comprising arm amputees [5]. Therefore, the number of Ottobock provides a power-assist foot prosthetic called “1B1
leg amputees was 7 million. Leg amputees use foot Meridium [6].” Its mechanism is shown in Figure 1. It adopts
prosthetics to improve their quality of life. However, low- a hydraulic pressure mechanism, in which a hydraulic
priced foot prosthetics have rigid ankle parts and no power pressure cylinder pushes and pulls a lever on the toe plate,
drive mechanism, which makes it difficult to raise the heel causing the instep to rise and fall, respectively.
and swing the toes up. Therefore, most users need more
power to move their foot. Powered foot prosthetics enable Össur provides a power-assist foot prosthetic called
users to move their foot easily and walk more naturally. “PROPRIO FOOT® [7]” shown in Figure 2. Due to a lack of
However, such prosthetics are too expensive for most relevant material on the product's operation, we assume from
amputees. One example is that in Japan it costs more than 2 observations that an air cylinder positioned in the area of the
million yen ($18,000). Achilles' tendon raises and lowers the foot part.
This prosthetic, however, is not a commercial product. Prices Figure 4 – Measuring device and WD mounting method
of the PROPRIO FOOT® and 1B1 Meridium are not
available to the public, but are assumed to be more than 2
million yen ($18,000) in Japan, which is too expensive for
most amputees.
– 34 –
ICT for Health: Networks, standards and innovation
As described in section 3, people with walking disabilities, Figure 8 – Kicking power when heel is raised with normal
such as those who suffer from hemiplegia, clearly have a and proposed assist shoes for a stroke patient
weaker kicking power when raising their heel and swing
power when swinging their toe up. We have developed a
shoe, shown in Figure 7, that assists people with walking
disabilities. This shoe has a coil spring and leaf spring to
enable a user to easily raise their heel. The spring force of
the coil spring is 15 kg. The shoe has a roller to avoid the toe
accidentally tripping.
– 35 –
2019 ITU Kaleidoscope Academic Conference
– 36 –
ICT for Health: Networks, standards and innovation
There are no such differences as the foot prosthetic moves reliability, etc. must be defined in standardization. On the
synchronously with the healthy foot. However, a sensor that other hand, since a computer system controls several
monitors the healthy foot is needed. From our existing modules in a powered prosthetic, not only physical
research, we determined that monitoring the heel position of information but also data-level information must be
the healthy foot is best. This is why a sensor is built into the standardized.
heel of the shoe.
The following eight interfaces shown in Figure 12 do not
The central terminal is required to control the instep have exchange information between modules, so their
push/pull module in collaboration with the control board. We interface level is physical:
believe the upper position of the foot module is ideal for - INT-1: size, connection method, and reliability
mounting the battery and control board to the foot prosthetic. between socket and ankle joint;
- INT-2: size, connection method, and reliability
We plan to use a single motor cylinder for the instep between ankle joint and instep push/pull;
push/pull module. The module is attached to the front of the - INT-3: size, connection method, and reliability
shin as shown in Figure 12. However, we plan to determine between ankle joint and foot;
whether the module should be placed on the front of shin or - INT-4: size, connection method, and reliability
on part of the Achilles' tendon on the basis of experimental between instep push/pull and foot;
results. - INT-5: size, connection method, and reliability
between foot and toe;
For the single motor cylinder, we used an Oriental Motor DR - INT-6: size, connection method, and reliability
series with a 30-mm stroke, 2-kg carrying force, and a 100- between heel sensor and shoe;
mm/sec maximum stroke speed [11]. The heel-up spring has - INT-7: DC/AC, voltage, and connector type between
a motor-driven spring-release mechanism. However, we battery and instep push/pull;
estimate that the release timing must be controlled by a - INT-8: DC/AC, voltage, and connector type between
sensor built into the heel-up spring, not one built into the heel battery and toe.
of a shoe for a healthy foot.
On the other hand, the following four interfaces include data-
We estimate the instep push/pull module and heel-up spring level information in addition to the physical level
can be applied to people with walking disabilities. In information;
particular, the heel-up spring is useful as it compensates for - INT-9:
muscle weakness, as described in section 4. This means that Physical level: connector type;
the price of the heel-up spring can be lowered. Data level: pulses from the control board to the
instep. The control board changes the direction,
speed and number of pulses to control the cylinder
speed and stroke.
- INT-10:
Physical level: connector type;
Data level: pulses from the control board to a
cylinder of the toe module. The control board
changes the direction, speed and number of pulses
to control the cylinder speed and stroke.
- INT-11:
Physical to session level: wireless connection
(Bluetooth);
Application level: controls direction, speed and
maximum angle of foot rotation.
- INT-12:
Physical to session level: wireless connection
(Bluetooth);.
Figure 12 – Module structure foot prosthetic Application level: controls direction, speed and
maximum angle of foot rotation.
5.2 Interface and standardization items
6. CONCLUSION
In the case of unpowered prosthetics, interfaces between
modules are at the physical level, since there is no control There are several million people living with limb loss in the
information transferred between them. Physical level world. Powered leg and/or foot prosthetics enable amputees
information, such as size, weight, connecting method,
– 37 –
2019 ITU Kaleidoscope Academic Conference
to walk naturally. However, most of them are too expensive [4] K. Ziegler‐Graham, E. J. MacKenzie, P. L.
for most leg amputees to afford. Ephraim, T. G. Travison, R. Brookmeyer,
"Estimating the Prevalence of Limb Loss in the
We determined through experimentation that people with United States: 2005 to 2050," Archives of Physical
walking disabilities, such as hemiplegia, clearly have a Medicine and Rehabilitation 2008, vol. 89 IS. 3,
weaker kicking power when raising their heel and a weaker pp.422‐429, 2008.
swing power when swinging their toes up than unimpaired
people. Our proposed shoe design has springs in the heel that [5] Estimated of Amputee Population 11/09/2008,
compensate for muscle weakness. https://web.stanford.edu/class/engr110/2011/LeBlan
c-03a.pdf , [retrieved: September 2019].
We proposed a module structure for a foot prosthetic derived
from our research results and observations of existing [6] Ottobock, 1B1 Meridium,
powered foot prosthetics. We also proposed standardizing https://professionals.ottobock.com.au/Products/Prost
interfaces between modules, enabling third-party hetics/Prosthetics-Lower-Limb/Feet/1B1-
manufacturers to produce prosthetic components at lower Meridium/p/1B1 , [retrieved: September 2019].
costs.
[7] Ossur, PROPRIO FOOT®,
The introduction of such modules and standardization can https://www.ossur.com/prosthetic-
lower overall prices for prosthetics, enabling them to be solutions/products/dynamic-solutions/proprio-foot ,
more affordable for foot amputees and people with walking [retrieved: September 2019].
disabilities, which, as a result, will improve their quality of
life. [8] M. F. Eilenberg, H. Geyer, and H. Herr, “Control of
a Powered Ankle–Foot Prosthesis Based on a
7. ACKNOWLEDGEMENTS Neuromuscular Model,” IEEE, Transaction on
Neural System and Rehabilitation Engineering,
Thanks to Mr. Takashi Ushizaki, Dr. Koya Sato for helping VOL. 18, NO. 2, pp. 164-173, 2010.
with this research. This work was supported by JSPS
KAKENHI Grant Number 19K11326. [9] Wireless EMG logger, Logical Product Corporation,
http://www.lp-d.co.jp/EMGSensor.html , [in
REFERENCES Japanese, retrieved: September 2019].
[1] World Population Ageing: 1950-2050, Population [10] A. Staros, “The SACH (Solid-Ankle Cushion-
Division, Department of Economic and Social Heel),” Orthopedic & Prosthetic Appliance Journal,
Affairs, United Nations, pp. 23-31, June-August 1957.
http://www.un.org/esa/population/publications/worl
dageing 19502050/ [retrieved: September, 2019]. [11] Single motor cylinder, DR series, Oriental Motor
Ltd.,
[2] W. P. Berg, H. M. Alessio, E. M. Mills, and C. https://www.orientalmotor.co.jp/products/new/1806l
Tong, "Circumstances and consequences of falls in 134l135dr/ [in Japanese, retrieved: September
independent community-dwelling older adults", Age 2019].
Ageing, Vol. 26, pp. 261–268, 1997.
– 38 –
DEVELOPMENT OF HEARING TECHNOLOGY WITH PERSONALIZED SAFE
LISTENING FEATURES
Shayan Gupta1,2; Xuan Xu2; Hongfu Liu2; Jacqueline Zhang2; Joshua N Bas2; Shawn K. Kelly2
1
Audition Technology, LLC., Pittsburgh, PA, USA
2
Carnegie Mellon University, USA
Standards have been proposed to promote safe listening. In The WHO defines mobile health (m-health), a subset of e-
the US, there are safe listening standards from National health, as ‘the use of mobile wireless technologies for health’
Institute for Deafness and Communication Disorders and recognizes the value of digital technologies to contribute
(NIDCD) [5], National Institute for Occupational Safety and to advancing health aims of the Sustainable Development
Health (NIOSH) [6], and Occupational Safety and Health Goals [15]. Our selection of development of an app
Administration (OSHA) [7] to limit occupational noise acknowledges the WHO m-health directive and is guided by
• daily sound exposures, as a summation of A-weighted Of the various standards, only the OSHA PEL and the WHO-
sound pressure levels (SPL in dBA) over time, based on ITU provide methodology for computing A-weighted sound
daily activities related to occupation, lifestyle and pressure level exposures. The resultant exposure
recreational choices; assessments can be used to address the noise limits
• estimation of the user’s cumulative daily exposure vs recommended by the other agencies listed in Table 1. Also,
sound dosage from recommended US (occupational, OSHA recommends that when daily noise exposure is
military) and UN (WHO-ITU) safe listening standards; composed of at least two periods of different sound pressure
• risk notifications about unsafe noise exposures; levels, the combined effect should be considered, rather than
• incorporation of user voice on engagement strategies; the individual effect of each. The exposure calculations take
• option for connection to a personal hearing device, e.g., this recommendation into account.
PSAP, via Bluetooth.
The Hearing Health app calculates the user’s occupational
App functionalities and user voice implementation were environmental exposure compared to OSHA’s PEL exposure
based on user feedback from a wide demographic. using the following formula:
– 40 –
ICT for Health: Networks, standards and innovation
the different exposure volume for the 8-hour period, the The app samples environmental noise exposure using an
percent dosage calculation is multiplied by a factor f, where external microphone or internal phone microphone and can
f is the ratio of OSHA’s recommended 8-hour noise level (85 sample audio sound exposure from the mobile phone’s
dBA) over the other standard’s recommended 8-hour noise system audio player. The app leverages the mobile phone’s
level. short-range wireless connection to a PSAP consisting of
digital signal processor (DSP) packaged with a Bluetooth
2.4 Mitigation of discrepancies in measuring sound Low Energy (LE) module to more accurately measure the
exposure environmental and streamed sound exposure from a closer
distance to the user’s eardrum (Figure 1). Following the
There are a few causes of discrepancy when measuring a Bluetooth LE protocol will allow the app to receive data
user’s sound exposure. One such cause is the difference in points from the PSAP, subject to the particular DSP. For
distance between the ear and microphone sampling example, volume settings, battery level and ambient noise
environmental noise. Because sound attenuation is inversely levels may be available data points the app can query. When
proportional to the distance from the source, squared, the possible, hardware tests of the PSAP device can be
perceived sound at a point closer to the source of the sound performed by ensuring that input and output voltages do not
can be substantially louder than the perceived sound at a exceed those as listed in the DSP specification, as well as
point farther away. Another such cause is the type of accessing test points from the Bluetooth LE module to
microphone that is sampling. Different microphones have program and debug the module. Note that the PSAP is
different sensitivities, meaning that different microphones indicated for users without hearing impairment.
can possibly register different sound pressure levels from the
same sound signal. To mitigate these two causes of Privacy and security measures to safeguard personal
discrepancy, we propose using a microphone from a specific information include: (i) limiting data collection to that
PSAP product because the distance between the PSAP required specifically for app execution, (ii) implementing
microphone and the user’s eardrum is decreased and because Health Insurance Portability and Accountability Act
the software can be standardized to that specific PSAP (HIPAA) and General Data Protection Regulation (GPDR),
microphone’s sensitivity. and (iii) using Amazon Web Services (AWS) for cloud
security. To ensure data security, the app only collects
3. RESULTS amplitudes of the sound in decibels and stores them in
dynamoDB using s3 provided by AWS, which highly
3.1 App overview emphasizes security and strictly meets US and international
compliance requirements.
The Hearing Health app (Figure 1) is designed as a software
tool to serve as a companion for personal hearing health that 3.2 User voice assessment for app personalization
prompts a user to make informed decisions about personal
listening behaviors based on personal listening trends. An initial assessment of user awareness of NIHL and
Throughout the day, the app monitors sound levels to preferences for app personalization features was conducted
estimate the user’s sound exposure, while also presenting in several user segments to represent a wide demographic
alerts and notifications to the user to indicate how the and is summarized below. The user segments included
personal listening behavior compares to sound doses military and civilian, young adults (age 18-25 years) and
prescribed by safe listening standards. The app also provides adults (age >25 years), with or without perceived hearing
personalized recommendations to limit or counteract unsafe impairment.
noise exposure that is relatable to daily lifestyles.
CIVILIAN, young adult, not aware of personal impairment:
– 41 –
2019 ITU Kaleidoscope Academic Conference
– 42 –
ICT for Health: Networks, standards and innovation
app), considerations for selecting hearing protection based As depicted in Figure 3, the app presents the highlights of
on their needs and preferences, and other recommended the user’s noise exposure in the Hearing History and the
practices aimed at preventing hearing loss. Listening Profile modules. The Hearing History module
displays graphs of the user’s measured noise exposure over
The main goal of the app is to encourage and facilitate the time period to which the user toggles (i.e. past hour, past
healthy listening behaviors by creating a personalized sound day, past week). The listening profile keeps track of the
exposure profile and with engagement tools personalized to qualitative aspects of the user’s listening experience, such as
the user. Based on the functionalities described above, there what genre of music to which the user frequently listens and
are three main approaches for user interface personalization: periods of the user’s day where the noise exposure is
(i) active and customizable monitoring, (ii) awareness of heightened (i.e. an exercise class or a noisy commute).
personal exposure and (iii) feedback.
– 43 –
2019 ITU Kaleidoscope Academic Conference
– 44 –
ICT for Health: Networks, standards and innovation
hearing aids, preservation of residual hearing and [7] OSHA standard 1910.95, Occupational noise
improvement of quality of life. exposure, accessed 2019:
https://www.osha.gov/laws-
The app development has been influenced by the global need, regs/regulations/standardnumber/1910/1910.95
priorities for systems and services for assistive products as
announced by the UN Global Partnership for Assistive Appendix A Noise exposure computation
Technology [19]. For hearing aids, one of the five identified https://www.osha.gov/laws-
‘priority’ products, the proposed app could serve as a tool to regs/regulations/standardnumber/1910/1910.95App
advance overall hearing healthcare while addressing needs A
for global awareness, access and user-centric innovation.
[8] NIOSH Sound Level Meter App, 2019:
In the US, the National Academies of Science, Engineering https://www.cdc.gov/niosh/topics/noise/app.html
and Medicine (NASEM) report on improving access and
affordability for hearing healthcare identifies the need for [9] Department of Defense, Design criteria noise limits.
innovative solutions for patient-centered care with a 2015:
recommendation for a new category of over-the-counter https://www.arl.army.mil/www/pages/343/MIL-
hearing devices [20]. The report also recommends engaging STD-1474E-Final-15Apr2015.pdf
a wider community by awareness, education and support.
We envision the Hearing Health app to be a component of [10] U.S. Army Public Health Command. Hearing loss
an OTC hearing aid system that would engage the user in and noise. 2019:
better management of personal hearing health needs https://phc.amedd.army.mil/PHC%20Resource%20
supporting effective use of the hearing aid. Library/HEARING_LOSS_AND_NOISE.pdf
Cumulatively, the experience from users with and without [11] Department of Defense, Hearing Center of
hearing impairment could be of significance for potential Excellence, accessed 2019:
extensions of Recommendation ITU-T H.870 to other use https://hearing.health.mil/Prevention/Preventing-
cases. Noise-Induced-Hearing-Loss/How-Loud-is-Too-
Loud
REFERENCES
[12] ITU-T. H.870. Guidelines for safe listening
[1] J. Eichwald, F. Scinicariello, J.L. Teffer, and Y.I. devices/systems, 2018. https://www.itu.int/rec/T-
Carroll. “Use of personal hearing protection devices REC-H.870-201808-I/en
at loud athletic or entertainment events among adults
– United States, 2018. Morbidity and Mortality [13] WHO-ITU Standard. Safe Listening Devices and
Weekly Report, vol. 67 no. 41, pp 1151-1155, 2018. Systems, 2019.
https://apps.who.int/iris/bitstream/handle/10665/280
[2] National Institute on Deafness and other 085/9789241515276-eng.pdf?ua=1
Communication Disorders, 2017:
https://www.nidcd.nih.gov/news/2017/us-adults- [14] World Health Organization, hearWHO, 2019.
aged-20-69-years-show-signs-noise-induced- https://www.who.int/deafness/hearWHO/en/
hearing-loss
[15] WHO guideline, Recommendations on digital
[3] World Health Organization, 2019: interventions for health system strengthening, 2019:
https://www.who.int/deafness/en/ https://apps.who.int/iris/bitstream/handle/10665/311
941/9789241550505-eng.pdf?ua=1
[4] Centers for Disease Control and Prevention, 2017:
https://www.cdc.gov/vitalsigns/HearingLoss/ [16] Principles of digital development, accessed 2019:
https://digitalprinciples.org/principles/
[5] National Institute on Deafness and other
Communication Disorders, 2019: [17] WHO-ITU toolkit for safe listening devices and
https://www.noisyplanet.nidcd.nih.gov/kids- systems, 2019:
preteens/keep-listening-to-the-beat https://apps.who.int/iris/bitstream/handle/10665/280
086/9789241515283-eng.pdf?ua=1
[6] NIOSH Occupational noise exposure, revised
criteria, 1998: https://www.cdc.gov/niosh/docs/98- [18] FDA Draft Guidance. Clinical and Patient Decision
126/pdfs/98- Support Software, 2017:
126.pdf?id=10.26616/NIOSHPUB98126 https://www.fda.gov/media/109618/download
– 45 –
2019 ITU Kaleidoscope Academic Conference
[19] ATScale, Global Partnership for Assistive [20] Hearing health care for adults. Priorities for
Technology, 2019: improving access and affordability, NASEM, 2016.
https://static1.squarespace.com/static/5b3f6ff171069
9a7ebb64495/t/5ca3cfd3fa0d60051a9a7703/155423
9448526/ATscale_Strategy_Overview_February_20
19.pdf
– 46 –
SESSION 3
MEDICAL IOT
S3.1 Facilitating healthcare IoT standardization with open source: A case study on OCF and
IoTivity
S3.2 Empirical study of medical IoT for patients with intractable diseases at home
FACILITATING HEALTHCARE IOT STANDARDIZATION WITH OPEN SOURCE:
A CASE STUDY ON OCF AND IOTIVITY
Keywords – healthcare, IoT, IoTivity, OCF, open source, However, the proliferation of different international
standardization healthcare standards has not brought about a wide
deployment of healthcare IoT services and business
1. INTRODUCTION opportunities. The fragmentation of standards hindered the
access and usage of developed standards because
Digital health, an integral part of the "Fourth Industrial implementing such standards did not always guarantee
Revolution", revolutionizes the medical and healthcare interconnectivity and compatibility between services and
domain, and in turn, helps us overcome the technical and devices. In addition, it was difficult to assure that such
socio-economic challenges we face around health. Leaping healthcare devices can interoperate with other devices from
to the next level of connected healthcare services is only different silos. Such concerns eventually isolated healthcare
possible when taking advantage of the Internet of things devices to communicate with other devices from other
(IoT), 5G and other emerging technologies. Such a globally domains.
connected healthcare ecosystem facilitates remote medical
services which act as a bridge between medical practitioners Nevertheless, there have been some efforts in taking
and people who are especially in need. advantage of open-source projects as an enabler to better
deploy healthcare IoT standards. Most recent and remarkable
Healthcare IoT opens up seamless opportunities by efforts include the standardization activities in the Open
unleashing possibilities to better implement healthcare Connectivity Foundation (OCF). The OCF sponsors an
services to patients, physicians and hospitals. Patients with open-source project to provide reference implementations of
wirelessly connected devices can seamlessly track their the OCF specifications. IoTivity, an open-source project of
health conditions and request emergency assistance from the Linux Foundation, is sponsored by the OCF and aims to
physicians and hospitals. Physicians can track or monitor bring the open-source community together to accelerate the
deployment of OCF specifications. The reference [6], loss of productivity and motivation at work [7], and
implementation of healthcare devices such as blood pressure violation of license policy [8].
monitors helps developers to install pre-written code very
quickly and allows them to test and enhance the software. 2.2 Governance, organization and the process of
innovation in open-source software projects
In this paper, the authors develop a case study of their efforts
to standardize healthcare IoT with IoTivity, with the OCF. Competitive dynamics enforced by open-source innovation
First, the authors conduct a literature review on open-source in organizations, from the governance perspective, has led to
innovation to provide a theoretical background to the case more interest and resolutions to better adoption. It was found
study. Second, the authors introduce the OCF, its policy that the share of corporate contributions is much larger in
called “3-pillar alignment” and IoTivity to illustrate what large and growing projects [9], which implies that the
open-source implementation means to the OCF. Third, the contributions of companies are growing in volume.
authors discuss how they proposed new healthcare devices
and data models and how they eventually published the first Meanwhile, the operational efficiency potential and business
healthcare devices in the OCF. Finally, the authors discuss agility of open-source adoption are expected to mitigate the
the benefits of taking advantage of IoTivity and how it led to difficulty of accepting new models for designing, developing,
the enhancement of standardization efficiency and testing and deploying network solutions to the
acceleration in healthcare IoT. telecommunications industry [10]. In this sense, the 2016
World Telecommunication Standardization Assembly
2. RELATED WORKS (WTSA-16) resolved that the Telecommunication
Standardization Advisory Group (TSAG) pursue their work
In this section, the authors review relevant literature and on the benefits and disadvantages of the implementation of
studies on open-source innovation to provide a theoretical open-source projects in relation to the work of the ITU
background to this paper. A large volume of research on Telecommunication Standardization Sector (ITU-T), as
open-source innovation has found that such organizational appropriate [11]. More general research on this subject was
principles and operational culture has attracted competent a review that authors provided a framework which classified
individuals and organizations. The authors mainly focus on six distinctly different ways in which organizations adopt
existing findings, especially to understand why open-source open-source software [12].
innovation and projects attract people and organizations. The
review helps us understand why the OCF has put a particular Based on the common understanding of open-source
emphasis on taking advantage of open-source projects to innovation and why they attract individuals, the authors
provide reference implementations of OCF specifications proceeded to discuss a specific case of the OCF on how they
since its establishment. took advantage of open-source projects to develop healthcare
standards.
2.1 Motivations of open-source software
contributors 3. INTRODUCTION TO OCF
Motivations of open-source software contributors have In this section, the OCF, its unique policy called "3-pillar
always been the leading interest for research on open-source alignment" and its open-source project “IoTivity” are
innovation. These studies on open-source innovation introduced to provide a basic understanding of the OCF.
claimed that the popularity growth of open-source products
and platforms from various perspectives are: career concern 3.1 OCF
[1], ego gratification incentive [2] that motivate
programmers to participate in open-source projects, and Founded in 2014, the OCF comprises of over 400 member
organization and process of innovation [3]. These authors companies including Samsung Electronics, Intel, LG
commonly stressed that the success of an open-source Electronics, and Qualcomm to provide an IoT framework
project is attributed to its modularity, fun challenges to that works in various vertical domains including smart home,
pursue and credible leadership. healthcare, etc. [13] The OCF aims to provide
interoperability among IoT devices not only in the OCF
There also have been studies which summarized five ecosystem but also outside OCF boundaries to support
characteristics that have led to the proliferation of multi- multiple verticals. Legacy vertical services are traditionally
disciplinary research with regard to open-source innovation. designed as silos where there are no universal ways to
They include impact, transparency, theoretical tension, interwork between them. However, the OCF sets its goals to
communal reflexivity and proximity [4]. It is noted that specify and provide a foundational middleware platform for
open-source licenses are designed to ensure the rights of heterogeneous vertical applications. They develop technical
future users against appropriation [5]. Nevertheless, there are specifications to provide a foundational architecture, security,
significant advantages to open source whereby resources and bridging and other requirements, which in turn allow devices,
contributors involved in innovation are widely distributed regardless of manufacturers, to communicate between
throughout the globe. However, other studies have stressed themselves. Table 1 describes the list of OCF specifications
the drawbacks of open-source innovation such as free-riding and OCF healthcare devices that the authors proposed and
– 50 –
ICT for Health: Networks, standards and innovation
Table 2 – OCF specification development process and validation to OCF specifications. The OCF provides the
conformance test tool (CTT) to authorized test laboratories
Process Description (ATLs) and specification developers to validate devices.
1 Identify use cases, scope, and requirements Specification developers must define new test cases to the
2 Write draft specification
CTT to test whether devices comply with the proposed
specification or not. The OCF gives logos to the devices that
3 Develop open source code and test cases
pass the certification test. The final pillar is the open-source
4 Distribute draft specification for IPR review reference implementation of the OCF specifications.
5 Publish final specification Specification developers have to provide a properly working
defined as part of the OCF Device Specification. The open-source code to guarantee whether the proposed
development of OCF healthcare devices are further specification is valid or not. The CTT validates the reference
elaborated in section 4. implementation. A new specification is allowed to be
published if all three pillars are satisfied.
The OCF offers RAND-Z as its intellectual property rights
(IPR) to facilitate the deployment of OCF technology among 3.3 IoTivity
members. Offering manufacturer-friendly IPR policies
enables the growth of the market by attracting not only large IoTivity is an OCF-sponsored open-source project, which
enterprises but also start-ups. After a continuous merger with implements all mandatory features of the OCF specification
UPnP and AllSeen Alliance, the OCF has become one of the along with some optional features [15]. It provides a
most significant industrial consortia for IoT standardization. reference implementation of OCF specifications to ensure
Most recently in November 2018, the OCF announced that interoperability between OCF devices and certification of
OCF 1.0 specifications have been ratified as international OCF products. OCF members take advantage of IoTivity
standards by ISO/IEC JTC 1 and approved as ISO/IEC because publication of OCF specifications requires open-
30118 (Parts 1-6) [14]. source implementation and certification. In this sense,
IoTivity provides the fastest and the easiest way to develop
3.2 3-pillar alignment not only standards but also products. IoTivity can be
installed in many IoT devices, even in class 2 constrained
devices [16] to minimize CPU load, traffic and bandwidth.
Unlike any other industry groups or SDOs, the OCF has a
IoTivity uses an Apache 2.0 license with its accompanying
unique policy when developing its specifications: the policy
patent grant.
is titled the "3-pillar alignment", which mandates
specification developers to provide not only specifications,
but also a proof of passing the certification program provided 4. HEALTHCARE SPECIFICATION
by the OCF, and finally, an open-source reference DEVELOPMENT IN OCF USING IOTIVITY
implementation of the proposed specification. The ultimate
goal of 3-pillar alignment is to ensure full interoperability In this section, the authors carry out an in-depth analysis of
between devices and compatibility to specifications. their standardization efforts for the development of
healthcare IoT devices for the OCF specification and how
The first pillar is the specifications which define baseline they engaged in open-source reference implementations. The
functionalities and vertical profiles for OCF devices as authors primarily focus on the interaction with IoTivity and
described above. The second pillar is the certification why using the open-source solution was crucial during the
program to carry out conformance testing for interoperability specification development. The authors explain how IoTivity
offered a solution to meet the challenge of developing
– 51 –
2019 ITU Kaleidoscope Academic Conference
Table 3 – OpenAPI code snippet of blood pressure The authors also emphasized that healthcare IoT devices
could keep track of the people at risk of developing chronic
{
"swagger": "2.0", disease and manage their ailments. These devices can weave
"info": { into everyday lives, seamlessly collect activity metrics, and
"title": "Blood Pressure", encourage maintaining healthy behaviors. Furthermore, the
... authors stressed that standardizing the healthcare ecosystem
}, was the way forward to ensuring peer to peer independence,
... user’s confidence and data, which eventually lead to
"definitions": { actionable business insights. Since the healthcare industry is
"BloodPressure": { affected by many regulatory bodies, the new group intended
"properties": {
to account for privacy, security and global regulatory bodies.
...
"systolic": { The OCF approved the establishment of the Healthcare
"description": "Systolic blood pressure", Project and the specification development was initiated.
"minimum": 0.0,
"readOnly": true, 4.2 Write draft specification
"type": "number"
}, Defining a new OCF device is twofold: Specify the behavior
"diastolic": { and requirements of a new device, and design any necessary
"description": "Diastolic blood pressure",
data models which need to support the new device. The
"minimum": 0.0,
"readOnly": true,
authors leveraged the existing core functionalities including
"type": "number" the protocol used for transmission [17], data modeling
}, practices [18], and security [19] to ensure the OCF healthcare
... devices are interoperable with other OCF devices from
"type": "object", different silos.
"required": [
"systolic", The authors designed each healthcare device as follows. The
"diastolic" authors specified a minimal set of resources that shall be
]
implemented by the device and an additional optional set of
}
} resources that may be exposed by the device. A blood
} pressure monitor, for example, must expose blood pressure
information (systolic blood pressure, diastolic blood pressure,
healthcare applications. Table 2 summarizes the etc.) but may expose mean arterial pressure (MAP), pulse
specification development process in the OCF. The table rate, units (mmHg or kPa), associated timestamp [20] and
provides an overall view of how the authors developed OCF user identification. Defining mandatory and optional
specifications. resources separately allowed minimizing payload
transmitting between OCF devices.
4.1 Identify use cases, scope and requirements
Next, the authors defined healthcare data models based on
New specification development in the OCF starts with the OpenAPI specification [21]. The OpenAPI specification
identifying use cases, scope and technical requirements. In is an open-source project which defines programming
this sense, the authors proposed the Healthcare Project to language-agnostic RESTful APIs. The specification is
initiate standardization on healthcare IoT. The objective of expressed by Swagger [22]. Table 3 is a snippet of authors’
the new group was to evaluate use cases, derive proposed data model for blood pressure. The data model is
interoperability requirements and develop technical uploaded to OCF Github [23] and oneIoTa [24], which is the
specifications for the healthcare vertical within the OCF's official web-based data model repository. The data
framework of the OCF. model specifies how blood pressure data shall be exposed on
the wire. For example, all systolic blood pressure
The goal of the specification was to define OCF healthcare communicated within the OCF must be a read-only floating
devices in healthcare, fitness and medical domains of the number titled “systolic” whose minimum value is 0. The
OCF ecosystem. Regardless of devices, the authors proposed proposed data model was approved by OCF reviewers as
a simplified operational scenario which involves OCF final prior to open-source development because the open-
servers (e.g. body scale, glucose meter, etc.) and OCF clients source reference implementation had to comply with the
(monitoring devices such as smartphones). Eventually, the predefined approved data model.
collected data could be used for tracking users’ fitness
conditions or transferring them to medical institutes to In addition, the authors defined an additional functionality
receive remote services. The technical requirements for and requirement to the core specification [17] to ensure
healthcare specifications were to provide additional better interoperability among OCF devices. In certain use
healthcare device types and data models and, if deemed cases, healthcare devices require that the information of
necessary, additional functional requirements to be added to multiple resources be only accessible as a group and
the core specification [17]. individual access to this information of each resource by an
– 52 –
ICT for Health: Networks, standards and innovation
Table 4 – C++ Code snippet of blood pressure monitor First of all, the authors wrote a C/C++ code which
encapsulates the requirements and functionalities of the
//-----------------------------------------
// Title: [IoTivity][Blood Pressure Monitor] Linked Resource proposed additional functional requirement (atomic
Type: Blood Pressure measurements), which had the potential to affect the core
// Description: Defines "oic.r.blood.pressure" and its behaviors specification [17]. There existed two ways to develop: first
//----------------------------------------- was to modify the core code of IoTivity by adding new APIs
... to the IoTivity library [25] with regard to the new functional
OCRepPayload* getBP1Payload(const char* uri) requirement, and second was to build an application which
{ runs on top of the existing core IoTivity code but not adding
OCRepPayload* payload = OCRepPayloadCreate(); new APIs to the IoTivity library. The authors chose the latter
if(!payload) because building a separate application would save much
{
more time.
OIC_LOG(ERROR, TAG, PCF("Failed to allocate
Payload"));
return nullptr; In this sense, the authors developed a proof of concept (PoC)
} of blood pressure monitor reflecting the atomic
size_t dimensions[MAX_REP_ARRAY_DEPTH] = { 0 }; measurements requirements and uploaded to GitHub for peer
review [26]. Table 4 is a C/C++ code snippet which describes
dimensions[0] = 1; the payload transferred from the blood pressure monitor. The
char * rtStr[] = {"oic.r.blood.pressure"}; payload had to comply with not only the schema in Table 3
OCRepPayloadSetStringArray(payload, "rt", (const char but also the atomic measurement requirements that the
**)rtStr, dimensions);
authors proposed. Thus, several APIs defined in IoTivity C
OCRepPayloadSetPropString(payload, "id",
"user_example_id"); SDK [27] and IoTivity C++ SDK [28] were used to describe
OCRepPayloadSetPropInt(payload, "systolic", 0); the payload. For example, OCRepPayloadCreate() function
OCRepPayloadSetPropInt(payload, "diastolic", 0); was used to create a new payload and
OCRepPayloadSetPropString(payload, "units", "mmHg"); OCRepPayloadSetPropInt() was used to set an integer
property to that payload. Meanwhile, the way to encapsulate
return payload; the overall payload and send them on the wire required
} certain routines of combining functions and triggering error
OCF client is prohibited. For example, users need to be able messages. Taking advantage of the existing SDK allowed the
to retrieve their blood sugar level and the time of authors to easily duplicate, test and eventually improve the
measurement simultaneously from his or her glucose meter, deliverables in the future.
in order to properly keep track of the daily glucose level
fluctuation. In this sense, the authors named this additional Finally, the authors proposed new healthcare device
feature as "atomic measurement" and defined its common definitions and asked for a pull request by adding lines of
properties, normative behavior, security considerations and device and resource definitions in JSON format to the OCF’s
other requirements. The proposal was also drafted but had to GitHub repository where the OCF manages all device,
be discussed at a different group (Architecture Task Group) resource and enumeration definitions [29]. In the repository,
where the Core Specification [17] is developed. there is a folder where the information of the device
specification in a machine-readable format is stored. This
4.3 Develop open-source code and test cases repository intends to provide the information in a machine-
readable format for CTT, which in turn, ensures the latest
While the authors were developing the specification for new information of OCF devices for certification program. The
healthcare devices and additional functional requirements, pull request was eventually merged as final.
they had to simultaneously develop the open-source
reference implementation of the proposed specification and Concurrently, the authors had to develop test cases while
test cases for the certification and CTT. developing the code. Test cases in the OCF aim to verify if
the written code well complies with the specification. All
In the OCF, developers are free to choose any open-source mandatory requirements of OCF specifications must have
code for reference implementations. Similarly, the authors corresponding test cases which ensure compatibility and
had the freedom to choose from existing open-source interoperability of OCF devices. The test cases should be
projects which implement OCF specifications or could start able to run in all OCF devices regardless of manufacturers
from scratch. As introduced above, IoTivity has been an and platforms. The approach here for authors was also
OCF-sponsored open-source project since the establishment twofold: the test cases for mandatory resources of each
of the OCF. The project was initiated and developed by healthcare device and the test cases for atomic measurement.
architects of Intel and Samsung Electronics, and periodically The authors developed all necessary test cases, and the test
published stable releases. Any developers can download, cases were incorporated into the CTT. The CTT was
fork, commit and contribute to the existing project to file eventually ready for Plugfest, which is an official OCF
bugs and improve the code. The authors thus decided to take interoperability and compatibility certification event for
advantage of the existing architecture and APIs of IoTivity OCF members to test their device against the CTT and the
using C/C++ [25]. OCF devices of other companies.
– 53 –
2019 ITU Kaleidoscope Academic Conference
Figure 1 – Healthcare PoC at CES 2018 motivations, such reference implementations lowered the
boundaries of OCF specifications, and it was clear that
IoTivity led to better organizational innovation [3] because
the authors fully benefited from the existing code provided
by IoTivity and its contributors. The authors were able to
spend dramatically less resource compared to what the
authors might have spent when developing from scratch. The
authors found that it is especially useful in the healthcare
domain because developers can put more efforts on privacy,
security and other regulatory aspects, which is considered
much stricter than those in the smart home domain.
4.4 Distribute draft specification for IPR review The authors stress that reference implementations using open
source enhances standardization efficiency. There is a
When all three pillars, including the specification, the test considerable population of companies, start-ups and
cases for certification, and the open-source reference hobbyists who are low on resources but interested in
implementation, were considered stable enough by OCF developing applications and products based on the published
members, the draft specifications including the new standards. An open-source community discloses extensive
healthcare devices and data models were put through IPR opportunities for those who need help from external experts
review. In general, the OCF posts new draft specifications around the world. Requesting a public review by other
for 60-day member IPR review when the test cases and open- experts can decrease debugging time. Also, it accelerates
source code are also ready for distribution. The comments better exposure and deployment of emerging technology
are subsequently submitted to the OCF and are reviewed by standards. It must also be noted that free-riding [6] by other
the relevant members if deemed necessary. If any OCF stakeholders could discourage development. However, it is
members intend to exclude necessary claims from the OCF’s highly recommended to use appropriate open-source licenses
IPR policy, they must provide a complete, appropriate and to protect ownership and track history of code change [5].
timely written notice of such intent to the OCF, no later than
the IPR review deadline. Since sufficient discussion was 5.2 Acceleration in healthcare IoT
made during the specification development and data models,
the authors received no additional comments with regard to Developing reference implementation using IoTivity opened
the new healthcare devices and data models. the potential and accelerated the healthcare IoT best
practices of IoT application developers. Reference
4.5 Publish final specification implementation eventually allowed manufacturers of other
domains to connect healthcare devices to their devices.
After the IPR review, the OCF published the new healthcare When the authors developed their PoC, the IoTivity code
devices and data models as part of OCF 2.0.0 Specification was also used and developed by other OCF members who
in June 2018 [30]. Publication of new OCF specifications were smart home device vendors. Thanks to the same code
implies full support of certification and open-source base, which acted as the same foundational software
reference implementation. In addition to the first four architecture, the IoT devices from different domains were
published healthcare devices, the authors defined five able to be tested with less challenge.
additional healthcare devices, which in turn, were published
as part of OCF 2.0.4 Specification in July 2019. The active corporate participation and contribution to
IoTivity lured more member companies into joining IoTivity.
5. DISCUSSION While contributions were growing in volume [9], the OCF
was able to exhibit the devices of its members at CES 2018.
In this section, the authors discuss the benefits of taking The devices were developed with IoTivity. At CES 2018,
advantage of IoTivity and how it has led to the enhancement The authors presented their healthcare PoC, including blood
of standardization efficiency and acceleration in healthcare pressure monitors, body thermometers, body scales and other
IoT. healthcare devices. The devices were able to connect and
interoperate with other smart home devices. Figure 1
5.1 Enhancement of standardization efficiency displays a user's health information, which was measured
using healthcare devices developed by the authors, to another
The main benefits of using IoTivity as an enabler was to OCF member's refrigerator. The exhibition at CES 2018
improve the development efficiency of the reference provided the viewers with insight on how the future of
implementation of the OCF healthcare specification. healthcare IoT would look like upon its realization.
Typically, sharing the same open-source project allows
developers to share best practices among peers while using 6. CONCLUSION
the code. These peer IoTivity developers may contribute to
the IoTivity project hoping for better promotion [1], or In this paper, the authors discussed the standardization of
perhaps pursuing self-satisfaction [2]. Regardless of their healthcare IoT at the OCF while focusing on IoTivity as its
– 54 –
ICT for Health: Networks, standards and innovation
enabler for the reference implementation. First, a literature incentives when standards developers provide reference
review on open-source innovation was carried out to provide implementations of Recommendations. For example, inside
a theoretical background to this paper. Second, the OCF, "3- the ITU-T’s template for A.1 justification for a new work
pillar alignment" and IoTivity were introduced to offer a item (NWI) proposal [35], a new line asking for plans for
contextual background to this paper. Third, a case study on reference implementations of Recommendations could be
the authors' efforts to specification development for added. By the time of completion of the Recommendation,
healthcare devices in the OCF using IoTivity was illustrated. the draft Recommendation could be evaluated and consented
Finally, the benefits of taking advantage of IoTivity and the based on the reference implementation, which complies with
implications for better enabling healthcare IoT the proposed Recommendation.
standardization were discussed.
From the technological perspective, developing reference
6.1 Prospective future works implementations would require Recommendation
developers to take advantage of existing open-source
IoTivity was discussed as a best practice for a reference projects to save time and resource. During code development,
implementation of OCF specifications, as well as an enabler developers would seek guidance for external experts, which
of facilitating the healthcare IoT. To harmonize with other will in turn, produce more mature deliverables. From the
SDOs outside the OCF, however, ensuring interoperability marketing perspective, active participation in open-source
between different healthcare standards groups using a projects would result in better promotion and exposure of
universal healthcare bridge could be a next step forward. The ITU-T Recommendations.
OCF has already defined an OCF bridging specification to
specify a high-level framework to translate between OCF Taking into account some best practices in the industry
devices and other ecosystems [31]. including IoTivity, the authors strongly recommend that
ITU-T, within their core mandates, continue their efforts to
However, work on developing a specific healthcare interface seek the roles of open-source implementation for faster and
mapping standard is not yet initiated. For example, HL7 equal adoption of not only healthcare IoT but also their
FHIR specifies over 100 data formats and elements, APIs to overall Recommendations. Such efforts could lead to the
exchange electronic health records (EHR) in the medical active participation of developers who can provide codes
domain [32]. Similar to IoTivity, there exists an open-source immediately applicable to production and deployment,
project called HL7 Application Programming Interface which helps companies who are especially in need of
(HAPI) to provide reference implementations of HL7 FHIR standardization development resources. Such practice could
[33]. In this sense, connecting the OCF and HL7, by eventually facilitate wider adoption of ITU-T
developing an interface mapping standard could surely bring Recommendations.
synergy not only to the OCF and HL7 but also to future
healthcare IoT ecosystems. Such prospective work could be 7. ACKNOWLEDGEMENT
accompanied by IoTivity and other possible open-source
activities for developers to understand the collaborating This work was supported by the Institute for Information &
standard more rapidly. communications Technology Promotion (IITP) grant funded
by the Korea government (MSIT) (No. 2019-0-00137,
6.2 Proposed efforts to ITU-T Standards Development of Platform and Networking
Interworking for IoT Interoperability).
International standards organizations are recognizing the
importance of open-source innovation. The WTSA-16 REFERENCES
instructed all applicable and relevant Study Groups of the
ITU-T, to consider output from TSAG on open source [1] J. Lerner and J. Tirole, “Some Simple Economics
mentioned in section 2, within available financial resources, of Open Source,” J. Ind. Econ., vol. 50, no. 2, pp.
to study the benefits of using open source to develop 197–234, Mar. 2003.
reference implementations of ITU-T Recommendations [11].
Likewise, JTC 1 recently reconstituted the Advisory Group [2] J. Lerner and J. Tirole, “The Economics of
on Open Source Software (ISO/IEC JTC 1/AG 3) to assess Technology Sharing: Open Source and Beyond,” J.
JTC 1’s current and potential requirements and opportunities Econ. Perspect., vol. 19, no. 2, pp. 99–120, 2005.
with regard to open-source software [34]. The ISO/IEC JTC
1/AG 3 also seeks to identify areas where open-source [3] G. Von Krogh and E. Von Hippel, “The Promise of
software can be used to initiate or accelerate standardization Research on Open Source Software,” Manage. Sci.,
projects and work programs within JTC 1 scope. vol. 52, no. 7, pp. 975–983, 2006.
ITU-T requires industry’s adoption and exposure of their [4] G. Von Krogh and S. Spaeth, “The open source
Recommendations. One possible way to facilitate the software phenomenon: Characteristics that promote
adoption and exposure of their Recommendations is granting research,” J. Strateg. Inf. Syst., vol. 16, pp. 236–
253, 2007.
– 55 –
2019 ITU Kaleidoscope Academic Conference
[5] G. Von Krogh and E. Von Hippel, “Special issue on [18] ISO/IEC 30118-4:2018 - Information technology --
open source software development,” Res. Policy, Open Connectivity Foundation (OCF) Specification
vol. 32, pp. 1149–1157, 2003. -- Part 4: Resource type specification:
https://ituk2019.page.link/ey3j
[6] C. Y. Baldwin and K. B. Clark, “The Architecture
of Participation: Does Code Architecture Mitigate [19] ISO/IEC 30118-2:2018 - Information technology --
Free Riding in the Open Source Development Open Connectivity Foundation (OCF) Specification
Model?,” Manage. Sci., vol. 52, no. 7, pp. 1116– -- Part 2: Security specification:
1127, 2006. https://ituk2019.page.link/B8Ro
[7] C. Melian and M. Mähring, “Lost and Gained in [20] RFC 3339 - Date and Time on the Internet:
Translation: Adoption of Open Source Software Timestamps:
Development at Hewlett-Packard,” in International https://ituk2019.page.link/uUB8
Federation for Information Processing 2008, pp.
93–104, 2008. [21] OpenAPI Specification:
https://ituk2019.page.link/2m6g
[8] M. Henley and R. Kemp, “Open Source Software:
An introduction,” Comput. law Secur. Rep., vol. 24, [22] Swagger:
pp. 77–85, 2008. https://ituk2019.page.link/8h2g
[9] J. Lerner, P. A. Pathak, and J. Tirole, “The [23] Blood pressure definition in OpenAPI (GitHub):
Dynamics of Open-Source Contributors,” Am. https://ituk2019.page.link/g8n6
Econ. Rev., vol. 96, no. 2, pp. 114–118, 2006.
[24] Blood pressure definition in OpenAPI (oneIoTa):
[10] S. A. Wright and D. Druta, “Open source and https://ituk2019.page.link/DLNP
standards: The role of open source in the dialogue
between research and standardization,” in 2014 [25] Official repository of IoTivity:
IEEE Globecom Workshops (GC Wkshps), pp. 650– https://ituk2019.page.link/3cXL
655. 2014.
[26] Reference implementation of blood pressure
[11] Proceedings of WTSA-16: monitor:
https://ituk2019.page.link/nM9x https://ituk2019.page.link/tega
[13] Open Connectivity Foundation (OCF): [29] Information of the OCF device specification in
https://ituk2019.page.link/sReo machine readable format:
https://ituk2019.page.link/cx9A
[14] OCF official announcement of ratification of
International Standards: [30] OCF - The OCF Healthcare Project: Improving
https://ituk2019.page.link/4xSE health outcomes through IoT data:
https://ituk2019.page.link/9i2F
[15] IoTivity:
https://ituk2019.page.link/KPM8 [31] ISO/IEC 30118-3:2018 - Information technology --
Open Connectivity Foundation (OCF) Specification
[16] RFC 7228 - Terminology for Constrained-Node -- Part 3: Bridging specification:
Networks: https://ituk2019.page.link/U3E6
https://ituk2019.page.link/EBMe
[32] HL7 FHIR:
[17] ISO/IEC 30118-1:2018 - Information technology -- https://ituk2019.page.link/1nmD
Open Connectivity Foundation (OCF) Specification
-- Part 1: Core specification: [33] HAPI FHIR:
https://ituk2019.page.link/oySD https://ituk2019.page.link/bjy4
– 56 –
ICT for Health: Networks, standards and innovation
– 57 –
EMPIRICAL STUDY OF MEDICAL IOT FOR PATIENTS WITH
INTRACTABLE DISEASES AT HOME
1
Shinshu University School of Medicine, Japan
2
Nagano Prefectural Kiso Hospital, Japan
3
Central Corridor Communications 21, Japan4Goleta Networks Co. Ltd
yoshikaw@shinshu-u.ac.jp; takizawa@ccc21.co.jp; anakamu@shinshu-u.ac.jp; marshkamakura@yc5.so-net.ne.jp
ABSTRACT assistance [1], even though there are still legal issues
associated with remote patient care. In cases of monitoring
Telemedicine for chronic disease management is extending those patients remotely, critical events are false positive
to the home through the use of medical devices and ICT alarms and floods of similar alarms generated by highly-
technologies. Patients with intractable diseases, such as controlled medical devices. These events cause alarm fatigue
amyotrophic lateral sclerosis (ALS) and lethal not only in professional caregivers but also in families. There
neurodegenerative diseases, have been returning to their have been trials reducing the number of these events in
homes rather than remaining hospitalized. Reliable alarms hospitals intensive care units (ICUs), but the need for home
for condition changes of patients and burden reduction of care means that we also need to deal with intractable-disease
their families are taking root as foundations of telemedicine patients at home. There are investigations and discussions on
for patients with intractable diseases. This paper discusses alarms at ICU in hospitals [2-3]. Along with the trend to
reliable alarm delivery and expected medical IoT features extend the use of highly-controlled medical devices, such as
for those patients. A patient’s family has difficulty in setting an artificial ventilator, to the patient’s home, we need to
optimal parameters of life-support medical devices following tackle technical issues on alarms and also the management
patient condition changes. Also, caregivers and patients’ of the devices.
families expect reliable alarms and false alarm reduction
from tele-alarm systems used at home. We need to provide Medical devices have been developed for hospital use and
both anxiety relief for patients’ families and patient safety by have proprietary functions to output alarm sounds and to
reliably monitoring the patients. We designed and provide nurse calls via a wired interface [4]. When these
implemented an alarm delivery system for patients with devices are used at home, users need to carefully integrate
intractable diseases, and here we propose a prototype false- and operate them in their home network. A user needs to
alarm reduction mechanism for highly-controlled medical know the detailed meanings of functions in sophisticated
device systems including an artificial ventilator. We devices provided by manufacturers and to understand the
investigated alarms of a patient for one year, cooperating meaning of the data. In hospitals, medical doctors and
with the patient’s family. We need both hardware standard engineering staff take care of these issues, whereas at home
interfaces and consistent alarm functions between artificial it is difficult for them to be taken care of by the patient’s
ventilators. We conclude with our further work for patients family. There are two ways to develop standard interfaces
with different types of intractable diseases and for and procedures reducing the burden that artificial ventilators
standardization of medical IoT networks integrating false- impose upon medical doctors, visiting nurses and patient
alarm reduction systems. families. One is to integrate any artificial ventilator
following a standard operating interface and the other is to
Keywords – Artificial ventilator, crying wolf, false alarm, provide consistent alarm functions between ventilators.
intractable disease, medical IoT, telemedicine
In this paper, we first introduce a tele-alarm system and
1. INTRODUCTION explain real alarm-related issues. We explain issues that must
be dealt with when managing a highly controlled medical
The number of patients with intractable diseases is device, such as an artificial ventilator, at home. We then
increasing worldwide. The increasing number of ventilator- describe and discuss an approach to integrate various
dependent patients is exceeding the capabilities of hospitals ventilators manufactured by different companies expecting
and health organizations, shifting the burden of chronic care standard operating interfaces. We propose a software
to patients’ families. The need to reduce the burden on framework to integrate with a home IoT network and
caregivers and to increase safety has accelerated the describe our implementation of an alarm delivery system for
development of remote monitoring for home ventilator patients with intractable diseases and of a prototype false-
alarm reduction system having an interface to connect Most artificial ventilators have a nurse-call switch which is
various highly-controlled medical devices including an implemented as a means to inform caregivers of an urgent
artificial ventilator. When consistent alarm functions are situation caused by delayed notice of an alarm sound. Most
defined, they can be designed in various false-alarm ventilators are equipped with an emergency call connector,
reduction systems. We investigated alarms of a patient for and each model is connected to a gateway server by a
one year in cooperation with the patient’s family and connector.
discussed false alarm reductions that are required especially
for home use. We raise issues and conclude with our further There are two types of connecters to plug in a gateway, one
work for patients with different types of intractable diseases using a cable with 3 core wires for detection of cable
and for standardization of alarms and their treatment in disconnection and the other using 2 core wires. The left
medical IoT networks integrating false-alarm reduction figure and photo in Figure 2 show the normally closed (NC)
systems. type connection of an HT70Plus ventilator using a 3-core
wire cable, whereas the right figure and photos show the
2. BACKGROUND AND RELATED WORKS normally open (NO) type connection of a Vivo50 ventilator
using a 2-core wire cable. Each ventilator uses either type of
2.1 Tele-alarm experimental system connections.
Along with the number of uses, severe sequelae or death Figure 3 - Experimental scheme of vital signs and
cases are increasing because medical staff do not always alarm delivery
notice alarm events immediately.
– 60 –
ICT for Health: Networks, standards and innovation
– 61 –
2019 ITU Kaleidoscope Academic Conference
the results are delivered to a person close to the patient or to frequent, as shown in Table 1. The alarm is mainly caused
a remote person in the home IoT platform. by an intentional and temporal removal of the respiration
circuit by a caregiver for suction of sputum or change of the
patient’s body position. An accidental removal may also be
a possible reason. The second most frequent alarm is
peak_press_hi (peak air way pressure is higher than the
setting). This alarm is mainly caused by obstruction of the
air way by sputum or secretion. These alarms are clinically
important to prevent artificial ventilator associated accidents.
– 62 –
ICT for Health: Networks, standards and innovation
Figure 7 shows the relationship between three alarm Table 3 - Expected alarm reduction effect
priorities and the medical treatments. When an escalation
occurs, alarm priority changes from a lower level to a higher
level, as specified in Table 2.
– 63 –
2019 ITU Kaleidoscope Academic Conference
[3] Xiwei Xu, Liming Zhu et al., “Crying Wolf and [13] https://www.iso.org/standard/63718.html .
Meaning It: Reducing False Alarms in Monitoring of
Sporadic Operations through POD-Monitor”, DOI:
10.1109/COUFLESS.2015.18, 2015.
– 64 –
SESSION 4
S4.1 Invited paper - Towards international standards for the evaluation of artificial intelligence
for health
S4.2 Redesigning a basic laboratory information system for the global south
S4.3 #RingingTheAlarm: Chronic "Pilotitis" stunts digital health in Nepal
S4.4 Designing national health stack for public health: Role of ICT-based knowledge
management system
TOWARDS INTERNATIONAL STANDARDS FOR THE EVALUATION OF
ARTIFICIAL INTELLIGENCE FOR HEALTH
1
Fraunhofer Heinrich Hertz Institute, Berlin, Germany
2
Technische Universität Berlin, Berlin, Germany
In contrast to the growing interest and impressive exchanged liaison statements, in view of common use cases
advancements, the healthcare sector has only hesitantly addressed.
adopted these powerful innovations in practice so far,
because any technical fault can affect people’s health, The Institute of Electrical and Electronics Engineers (IEEE)
privacy, and lives [25]. Providing conclusive evidence about has established an “Artificial Intelligence Medical Device
the performance, reliability and limits of the ML/AI models Working Group” that has started working on two projects for
is required for harnessing the benefits of trustworthy new IEEE standards in 2018. “P2802” is a “Standard for the
solutions, while avoiding the risks of inadequate Performance and Safety Evaluation of Artificial Intelligence
implementations. Due to the high complexity of the ML/AI Based Medical Device: Terminology” and “P2801” is about
models and the addressed health tasks, it is not trivial to the “Recommended Practice for the Quality Management of
demonstrate conclusively whether a particular Datasets for Medical Artificial Intelligence” [30].
implementation solves a task adequately and reliably under
realistic conditions. For safe usage, it is of paramount The U. S. Consumer Technology Association (CTA) started
importance that future international standards can give clear a working group on “Artificial Intelligence in Health Care
recommendations about how to validate the models. These (R13 WG1)” in April 2019, with the participation of AT&T,
standards are expected to promote interoperability and Google, IBM, Philips, Samsung, and other companies [32].
dismantle trade barriers too. Moreover, the development of This initiative has “launched a new standards effort
these standards is in line with the Sustainable Development addressing The Use of Artificial Intelligence in Health Care:
Goals (SDG) of the United Nations (UN), in particular with Trustworthiness”. Moreover, CTA has released a “White
“SDG 3: Ensure healthy lives and promote wellbeing for all Paper on Use Cases in Artificial Intelligence” in December
at all ages” [26]. 2018, which includes use cases in healthcare [33].
– 68 –
ICT for Health: Networks, standards and innovation
Deutsches Institut für Normung (DIN) began drafting an “AI Typically, AI serves as a multivariable prediction model that
roadmap” in May 2019 “to create a framework for action for maps multidimensional input variables to one or
standardization” [42]. DIN has also founded an multidimensional output variables, e.g. pictures to disease
interdisciplinary AI Working Committee [43] and is working classification codes. Accordingly, the TRIPOD statement for
on two DIN SPECs related to AI [44, 45]. the “transparent reporting of a multivariable prediction
model for individual prognosis or diagnosis” can serve as a
Large companies lead the field in the area of AI and have landmark for AI methods too. These guidelines have been
started joint activities on safe AI, which potentially can published by the EQUATOR Network, an organization
establish de-facto standards fast. The “Partnership on aiming to enhance the quality and transparency of health
Artificial Intelligence to Benefit People and Society” is led research [49, 50, 51]. Cf. [52] for a discussion about how the
by representatives from large technology firms and several TRIPOD statement relates to AI.
other member organizations, also from academia and civil
society. The first goal of this initiative is “to develop and ML/AI models are implemented as pieces of software and
share best-practice methods and approaches in the research, hence belong to digital technologies in almost all cases (in
development, testing, and fielding of AI technologies”. This principle, they can be analogue hardware, too [53]). The
includes addressing “the trustworthiness, reliability, International Medical Device Regulators Forum has outlined
containment, safety, and robustness of the technology”. They principles for the clinical evaluation of software as a medical
are particularly interested in “safety-critical application areas” device in a draft from 2017 [54]. Three main topics structure
and mention healthcare as an example [46]. this clinical evaluation process: (a) Assuring that there is a
“valid clinical association” between the software output and
The “OpenAI” research center, which is well known in the the “targeted clinical condition”. (b) Correct processing of
ML/AI research community and backed by large investors, the “input data to generate accurate, reliable, and precise
has recently published a policy paper on “the role of output data”. (c) Achieving the “intended purpose in your
cooperation in responsible AI development”, “across target population in the context of clinical care” using the
organizational and national borders”, discussing “joint software output data. The English National Institute for
research into the formal verification of AI systems’ Health and Care Excellence (NICE) has published an
capabilities and other aspects of AI safety”. In particular, “evidence standards framework for digital health
they mention “various applied ‘AI for good’ projects whose technologies” in March 2019 [55]. This document
results might have wide ranging and largely positive “describes standards for the evidence (…) of effectiveness
applications (e.g. in domains like [...] health); coordinating relevant to the intended use(s) of the technology”. Moreover,
on the use of particular benchmarks; joint creation and the document states that the framework is applicable to
sharing of datasets that aid in safety research”. Moreover, digital health technologies “that incorporate artificial
they raise the question of the role of “standardization bodies intelligence using fixed algorithms”, excluding adaptive AI
in resolving collective action problems between companies”, algorithms.
in particular internationally [47]. OpenAI claims, “AI
companies can work to develop industry norms and 4. ML/AI PERFORMANCE EVALUATION
standards that ensure systems are developed and released
only if they are safe, and can agree to invest resources in The ML/AI models are expected to return meaningful results
safety during development and meet appropriate standards that are accurate, plausible and reliable, when processing
prior to release”. They “anticipate that identifying similar completely novel data points that the model has never seen
mechanisms to improve cooperation on AI safety between before, during the actual usage in the “real world”. Out-of-
states and with other non-industry actors will be of sample tests make it possible to assess this capability to some
increasing importance in the years to come” [48]. degree, if the tests are conducted appropriately. These tests
can be largely conducted in silico, at least as a first step,
3. VALIDATING DIGITAL HEALTH without posing the potential hazards of clinical trials, by
TECHNOLOGIES confronting the model with previously recorded test samples,
and by comparing the model output with the “ground truth”
Previous work can provide orientation for future for the respective task. This characteristic allows conducting
international standards for the validation of novel ML/AI- systematic tests at large scale (e.g. using databases with
based health technologies. Physicians, regulators, scientists thousands of MRT images), replicable and fast (e.g. in the
and engineers have long-ranging experience in dealing with case of software updates, or adaptive algorithms).
complex safety-critical health interventions and technologies
that require careful validation checks prior to usage. These The machine learning community evaluates the performance
technologies include, for instance, clinical interventions, of ML/AI models usually as follows: First, the model is
surgical procedures, pharmaceutics, medical devices and tested out-of-sample, but in-house, by splitting the available
software. Randomized controlled clinical trials, peer-review data in a training and a test set, often in a cross-validation
of scientific literature and standard tests in accredited testing scheme. The trained model computes labels or other output
laboratories are examples of well-established methods for variables from the input data of the test set, which are
assessing these interventions, substances or devices. statistically compared with the “true” labels or annotations
(the comparison is summarized in a score). Then, method
– 69 –
2019 ITU Kaleidoscope Academic Conference
and in-house test results are reported in a scientific paper that ML/AI evaluation are still missing. Standardization bodies
is reviewed by peers for publication in a journal or have merely started to address health ML/AI technologies (cf.
conference proceeding. Occasional open-source releases of section 2). Principles for prediction models, software and
the software code can allow the reviewers and other peers to digital health technologies can provide some overall
reproduce the results, in principle. Yet, the model orientation (section 3), but can only serve as a starting point,
performance is evaluated in-house only in many cases, e.g. and need to be transferred to the characteristics of the novel
because the code/model is not published or because of legal technologies. State-of-the-art procedures for ML/AI
or other barriers to share the test data. Therefore, it remains performance evaluation are a sound foundation, but the
unclear if the evaluation was conducted properly, if common limits discussed in section 4 need to be addressed.
pitfalls were avoided [cf. 56] such as leakage between test
and training data, or if the test data set was (un)intentionally The mission of the ITU/WHO focus group on “AI for Health”
curated, which can all result in overestimating the model is to undertake crucial steps towards evaluation standards
performance and in spurious results. Performance reports of that are applicable on a global scale, an approach that offers
different models can often not be compared, because of substantial potential for synergies. A large number of
individual data preprocessing and filtering. This problem is national regulatory institutions, public health institutes,
even more severe for commercial AI developers that physicians, patients, developers, health insurances, licensees,
typically refrain from publishing details of their methods or hospitals and other decision-makers around the globe can
the code [57]. profit from a common, standardized benchmarking
framework for health ML/AI. Standards live on being
For a range of tasks, human experts are required to label or sustained by a broad community. Therefore, the focus group
annotate the test data. In fact, experts can disagree, which is creating an ecosystem of diverse stakeholders from
leads to questions related to the so called “ground truth” or industry, academia, regulation, and policy with a common,
“gold standard”. How many experts of which level of substantial interest in health ML/AI benchmarking. ITU and
expertise [57] need to be asked? Crucially, in-house test data WHO officials monitor and document the overall process.
are often very similar to the training data, e.g. when Since its foundation in July 2018, the focus group has been
originating from the same measurement device, due to organizing a series of free workshops with subsequent multi-
practical reasons (cost, time, access and legal hurdles). day meetings in Europe, North America, Asia, Africa and
Therefore, the capacity of the AI to generalize to potentially India (and South America in January 2020) every two or
different, previously unseen data is often unclear, e.g. to data three months for engaging the regional communities.
from other laboratories, hospitals, regions or countries [cf. Participation in the focus group is encouraged by attending
58]. the events on site or via the Internet remotely. In addition,
further virtual collaboration allows for carrying forward
Researchers from the medical and machine learning work in between meetings. These online participation
communities are aware of these open questions and problems. possibilities and the generous support from a charitable
The medical journal “The Lancet - Digital Health” sets a foundation, with travel grants for priority regions, foster the
good example and requires “independent validation for all global participation, considering time and resource
AI studies that screen, treat, or diagnose disease” [59]. constraints.
Machine learning scientists urge towards reproducibility and
replicability by organizing challenges (also known as The structure of FG-AI4H is shown below, Figure 1. Two
competitions), where an independent, neutral arbiter types of sub-groups are generating the main deliverables:
evaluates the AI on a separate test data set [e.g. 60]. These working groups (WGs) and topic groups (TGs).
challenges are conducted at scientific conferences (e.g.
NeurIPS, MICCAI, CVPR, SPIE, etc.) and on Internet
platforms (e.g. Kaggle, AIcrowd, EvalAI, DREAM
Challenges, Grand Challenge etc.). Challenge design is not
trivial and research shows that many design decisions can
have a large impact on the benchmarking outcome [61].
Aspects beyond mere performance have not been addressed
sufficiently so far, including the benchmarking of robustness
[62], and of uncertainty [63], which is important for the
practical application in healthcare. Moreover, further in-
depth discussions with domain experts, e.g. physicians, are
required, in order to find out if the used evaluation metrics
are actually relevant with meaningful (clinical) endpoints
Figure 1 – Structure of FG-AI4H
[64].
WGs consider matters such as data and AI solution handling,
5. ITU/WHO FOCUS GROUP ON AI FOR HEALTH
assessment methods, health requirements, operations, and
regulatory considerations. Many of these matters are cross-
While there is considerable experience and previous work to
cutting subjects that affect a specific aspect of an AI for
build upon, generally accepted, impartial standards for health
– 70 –
ICT for Health: Networks, standards and innovation
health application. The deliverables of the WGs are planned topics and the specific problems involved with a number of
to be a number of documents that cover topics including: AI for health tasks and data modalities. At present, the topic
• AI ethical considerations, groups address AI-based cardiovascular disease risk
• AI legal consideration, prediction, dermatology, histopathology, outbreak detection,
• AI software life-cycle, ophthalmology, radiotherapy, symptom assessment,
• reference data annotation specification, tuberculosis prognostics/diagnostics and several further
• training and test data specification, domains. In each topic group, different stakeholders,
• AI training process specification, including competing companies, with a common interest in
the topic are working together. “Calls for topic group
• AI test process specification,
participation” are published on the website
• AI test metric specification, and
(https://www.itu.int/go/fgai4h), introduce the respective
• AI post-market adaptation and surveillance
topic group and invite participation. The creation of many
specification.
other topic groups in response to the open “call for proposals:
use cases, benchmarking, and data” is expected. Selection
An overview of the technical output of the WGs is given in criteria include the prospect for a widespread and, ideally,
Figure 2. global impact, a clear concept described in sufficient detail,
and preliminary evidence for feasibility.
Every topic group defines its scope, the specific ML/AI tasks
and the evaluation procedures with corresponding test data
and metrics in a topic description document in full detail.
Statistical metrics for assessing the model performance are,
e.g. precision, specificity, F1 score and area under curve, but
can be multiple or combined metrics too [61]. In particular,
it should be assured that the (e.g. clinical) endpoints are
Figure 2 − Overview of the technical output of the WGs meaningful in practice. Further criteria should be considered,
e.g. robustness to noise and to other variations in the input
The WG data and AI solution assessment methods reviews data [62], or to manipulations [65]. Humans prefer
the topic description documents (see below), in collaboration transparent decision-making: Can the model adequately
with independent experts with substantial records of quantify the uncertainty [63] and plausibly explain the
accomplishment in the respective health topic, with decision [66, 67]? These criteria beyond mere performance
proficient knowledge in ML/AI, and with transversal should also be considered.
competences from areas such as ethics and statistics. During
a repeated review cycle, the working group and the experts The topic description document must capture a range of
check that the topic description documents are accurate, aspects related to the test data, because they determine
complete, sound, understandable and objective, and give largely if the evaluation procedure is appropriate and
according feedback for improvement to the respective topic meaningful. The procedure can return conclusive results if,
group and the entire focus group. The WG is in charge of and only if, the test data are realistic, i.e. close to the actual
providing a number of technical deliverables, given above. application, of representative coverage, and of traceable
provenance from different sources. Data acquisition must be
The working group data and AI solution handling takes transparently documented in full detail [cf. 68], including
charge for a range of tasks related to conducting the tests, annotation guidelines, for reproducibility, replicability, and
which requires bringing the test data and the to-be-tested AI scalability. All ethical and legal questions related to the
solutions together. Relevant aspects include, e.g. transfer acquisition, storage and processing of health data must be
agreements, secure data and solution transfer, data checks, taken into careful consideration. Bias must be controlled and
IT infrastructure, access rights, traceability, IT security, test documented clearly. The document shall specify quality and
implementation and report generation. quantity criteria for the test data, including corresponding
references. The annotation needs to be conducted by experts
The working group for regulatory considerations is involved with a defined level of expertise, with potentially several
in the entire process, with representatives of FDA (USA), independent annotations per sample (if applicable).
CMDE/ NMPA (China), CDSCO (India), EMA (Europe) Technical matters, e.g. data formats [cf. 69, 70] and data
and BfArM (Germany) so far. In close collaboration with the management [71], need to be specified. A reference model
WHO, the working group facilitates subsequent steps (e.g. can potentially be defined (e.g. “average human performance
AI testing process specification, clinical evaluation, for this task”, “best in class”). Limiting factors for data
certification etc.) towards deployment of the health AI availability should be referred to, such as finances or time.
solution in practice. The plan detailed in the topic description document must be
implemented in practice. The test data must be provided or
The topic groups, TGs, take charge of specific health acquired, and measures for quality assurance taken. The
domains with corresponding ML/AI tasks. They are evaluation routine must be implemented, and the code
providing the connection of the WGs with actual health published together with at least a few example data with
– 71 –
2019 ITU Kaleidoscope Academic Conference
references (e.g. annotated images) to enable the developers [29] and a white paper on the website, where also the full
carrying out a trial run of their code. documentation of all previous meetings is published.
For a clean and fair evaluation, a trusted third party should 6. OUTLOOK
receive the trained model, as independent arbiter, and
conduct the tests on data that have never been published In summary, the ITU/WHO focus group on “AI for Health”
before. This cautious procedure prevents unfair conduct, e.g. has taken the first exploratory steps towards international
tuning the model for optimal performance on this particular health ML/AI evaluation standards. For the future, we expect
test set (“overfitting”), without actually being able to that a wide spectrum of health ML/AI topics will be
generalize well to real-world data, which can be expected in addressed and that insights from the evaluation will be
practice. Therefore, widely available, public data sets cannot brought back to research and development. The evaluation
be used for the evaluation and the entire test data set must procedure will be continuously refined in a repeated cycle,
remain secret, i.e. neither labeled nor unlabeled test data considering further quality criteria beyond mere
should be made available. The model performance should be performance, and including high quality test data with
evaluated in a closed computing environment without increasing geographic coverage. For the years to come, we
Internet access. Otherwise, test data could be leaked, against also anticipate further deepening of cooperation on ML/AI
the rules, and the model be tweaked on the test data. Besides, between standard setting organizations. While the
leaderboard probing and other potential pitfalls known from standardization activities on ML/AI differ in their thematic
ML challenges must be kept in mind [72, 73]. The trusted scope and particular objective (see section 2), they can profit
third party is responsible to protect both test data and ML/AI from collaboration, because different application areas of
model. The test data have to remain secret for subsequent ML/AI often share problems and data modalities. For
meaningful testing and the AI models may contain business- instance, assuring robust automatic image interpretation can
relevant trade secrets of the developer. be relevant for a range of safety-critical application domains,
and is not limited to healthcare. At the same time, a generic
In this spirit, focus group members have conducted a first approach is often not possible, because the cross-sectional
proof-of-concept benchmark for digital pathology, where an ML/AI technologies require cooperation with the respective
ML/AI model can provide diagnostic support by quantifying domain experts. A good example for this multidisciplinary
tumor infiltrating lymphocytes in breast cancer, from whole cooperation is the joint focus group of ITU and WHO, which
slide histopathology images, which is relevant for prognosis brings together expertise from information technology and
and therapy selection [cf. 74, 75]. The topic group had health standardization bodies. In particular, this initiative
defined the evaluation task and procedure, and had acquired shows that global collaboration can leverage synergy effects,
and annotated test data. The developer had trained a model since many relevant issues are common across the world.
on own training data to predict the annotations that a
pathologist would give from the images. A focus group REFERENCES
member as arbiter provided the computing infrastructure
according to specifications of the developer (here a desktop [1] World Health Organization (2019) Global Strategy
computer with a certain graphics processing unit, operating on Digital Health 2020-2024. Retrieved from
system, package manager, and ML framework installed) and https://www.who.int/DHStrategy
granted the developer access via the Internet to install the
prediction routine. Few annotated example data enabled the [2] U.S. Food and Drug Administration (2018). FDA
developer to test the prediction routine. After disconnecting News Release. Retrieved from
the computer from the Internet, the arbiter uploaded https://www.fda.gov/news-events/press-
undisclosed test data, directly received from the topic group, announcements/fda-permits-marketing-artificial-
on the machine, and executed the prediction routine, which intelligence-based-device-detect-certain-diabetes-
processed the data and predicted the annotations. Finally, related-eye
scores (true positive rate and true negative rate) were
computed by comparison with the reference annotations, and [3] Mesko, B. (2019). FDA Approvals For Smart
reported back to the topic group and the developer. Naturally, Algorithms In Medicine In One Giant Infographic.
this manual procedure can be automatized and scaled, e.g. The Medical Futurist. Retrieved from
with one of the ML challenge frameworks mentioned in https://medicalfuturist.com/fda-approvals-for-
section 4, potentially installed on a server on ITU or UN algorithms-in-medicine
premises.
[4] U. S. Food and Drug Administration (2019).
Interaction with further health institutions will be Proposed Regulatory Framework for Modifications
strengthened, potentially, e.g., with the International to Artificial Intelligence/Machine Learning
Association of National Public Health Institutes, the (AI/ML)-Based Software as a Medical Device
InterAcademy Partnership and the World Health Summit. (SaMD) - Discussion Paper and Request for
Further information about the scope and general process of Feedback. Retrieved from
the focus group can be found in a commentary in The Lancet https://www.regulations.gov/document?D=FDA-
2019-N-1185-0001
– 72 –
ICT for Health: Networks, standards and innovation
[5] Editorial (2019). Medicine in the digital age. Nature [15] Park, H. S., Rinehart, M. T., Walzer, K. A., Chi, J.
Medicine, 25, 1. https://doi.org/10.1038/s41591-018- T. A., & Wax, A. (2016). Automated detection of P.
0322-1 falciparum using machine learning algorithms with
quantitative phase images of unstained cells. PloS
[6] Editorial (2019) Machine Learning in Medicine. one, 11(9), e0163045.
Nature Materials 18, 407. https://doi.org/10.1371/journal.pone.0163045
https://doi.org/10.1038/s41563-019-0360-1
[16] Esteva, A., Kuprel, B., Novoa, R. A., Ko, J.,
[7] Editorial (2019). A digital (r)evolution: introducing Swetter, S. M., Blau, H. M., & Thrun, S. (2017).
The Lancet Digital Health. The Lancet Digital Dermatologist-level classification of skin cancer
Health, 1(1). https://doi.org/10.1016/S2589- with deep neural networks. Nature, 542(7639), 115.
7500(19)30010-X https://doi.org/10.1038/nature21056
[8] Trendall, S. (2019). NHSX to create policy guide for [17] Ching, T., Himmelstein, D. S., Beaulieu-Jones, B.
use of AI in healthcare. Public Technology. K., Kalinin, A. A., Do, B. T., Way, G. P., ... & Xie,
Retrieved from W. (2018). Opportunities and obstacles for deep
https://www.publictechnology.net/articles/news/nhs learning in biology and medicine. Journal of The
x-create-policy-guide-use-ai-healthcare Royal Society Interface, 15(141), 20170387.
https://doi.org/10.1098/rsif.2017.0387
[9] Dunbar, K. (2019). New guidance for AI in
screening. Public Health England Screening Blog. [18] Norgeot, B., Glicksberg, B. S., Trupin, L., Lituiev,
Retrieved from D., Gianfrancesco, M., Oskotsky, B., ... & Butte, A.
https://phescreening.blog.gov.uk/2019/03/14/new- J. (2019). Assessment of a deep learning model
guidance-for-ai-in-screening based on electronic health record data to forecast
clinical outcomes in patients with rheumatoid
[10] UK Department of Health and Social Care (2019). arthritis. JAMA network open, 2(3), e190606-
New code of conduct for artificial intelligence (AI) e190606.
systems used by the NHS. GOV.UK. Retrieved from https://doi.org/10.1001/jamanetworkopen.2019.0606
https://www.gov.uk/government/news/new-code-of-
conduct-for-artificial-intelligence-ai-systems-used- [19] Meiring, C., Dixit, A., Harris, S., MacCallum, N. S.,
by-the-nhs Brealey, D. A., Watkinson, P. J., ... & Singer, M.
(2018). Optimal intensive care outcome prediction
[11] Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A., over time using machine learning. PloS one, 13(11),
Ciompi, F., Ghafoorian, M., ... & Sánchez, C. I. e0206862.
(2017). A survey on deep learning in medical image https://doi.org/10.1371/journal.pone.0206862
analysis. Medical image analysis, 42, 60-88.
https://doi.org/10.1016/j.media.2017.07.005 [20] Nemati, S., Holder, A., Razmi, F., Stanley, M. D.,
Clifford, G. D., & Buchman, T. G. (2018). An
[12] Shen, D., Wu, G., & Suk, H. I. (2017). Deep Interpretable Machine Learning Model for Accurate
learning in medical image analysis. Annual review of Prediction of Sepsis in the ICU. Critical care
biomedical engineering, 19, 221-248. medicine, 46(4), 547-553.
https://doi.org/10.1146/annurev-bioeng-071516- https://dx.doi.org/10.1097/CCM.0000000000002936
044442
[21] Strodthoff, N., & Strodthoff, C. (2019). Detecting
[13] Esteva, A., Robicquet, A., Ramsundar, B., and interpreting myocardial infarction using fully
Kuleshov, V., DePristo, M., Chou, K., Cui, C., convolutional neural networks. Physiological
Corrado, G., Thrun, S. & Dean, J. (2019). A guide to measurement, 40(1). https://doi.org/10.1088/1361-
deep learning in healthcare. Nature medicine, 25(1), 6579/aaf34d
24. https://doi.org/10.1038/s41591-018-0316-z
[22] Hannun, A. Y., Rajpurkar, P., Haghpanahi, M.,
[14] Setio, A. A. A., Traverso, A., De Bel, T., Berens, M. Tison, G. H., Bourn, C., Turakhia, M. P., & Ng, A.
S., van den Bogaard, C., Cerello, P., ... & van der Y. (2019). Cardiologist-level arrhythmia detection
Gugten, R. (2017). Validation, comparison, and and classification in ambulatory electrocardiograms
combination of algorithms for automatic detection of using a deep neural network. Nature medicine,
pulmonary nodules in computed tomography 25(1), 65. https://doi.org/10.1038/s41591-018-0268-
images: the LUNA16 challenge. Medical image 3
analysis, 42, 1-13.
https://doi.org/10.1016/j.media.2017.06.015
– 73 –
2019 ITU Kaleidoscope Academic Conference
[23] Semigran, H. L., Linder, J. A., Gidengil, C., & [34] Trump, D. J. (2019). Executive Order on
Mehrotra, A. (2015). Evaluation of symptom Maintaining American Leadership in Artificial
checkers for self diagnosis and triage: audit study. Intelligence. Retrieved from
BMJ, 351, h3480. https://doi.org/10.1136/bmj.h3480 https://www.whitehouse.gov/presidential-
actions/executive-order-maintaining-american-
[24] Rivenson, Y., Ceylan Koydemir, H., Wang, H., Wei, leadership-artificial-intelligence/
Z., Ren, Z., Günaydın, H., ... & Ozcan, A. (2018).
Deep learning enhanced mobile-phone microscopy. [35] U. S. National Institute of Standards and Technology
Acs Photonics, 5(6), 2354-2364. (2019a). U.S. Leadership in AI: A Plan for Federal
https://doi.org/10.1021/acsphotonics.8b00146 Engagement in Developing Technical Standards and
Related Tools. Prepared in response to Executive
[25] He, J., Baxter, S. L., Xu, J., Xu, J., Zhou, X., & Order 13859.
Zhang, K. (2019). The practical implementation of Retrieved from
artificial intelligence technologies in medicine. https://www.nist.gov/sites/default/files/documents/2
Nature medicine, 25(1), 30. 019/08/10/ai_standards_fedengagement_plan_9aug2
https://doi.org/10.1038/s41591-018-0307-0 019.pdf
[26] United Nations (2015). United Nations Sustainable [36] U. S. National Institute of Standards and Technology
Development Goals. Retrieved from (2019b). Plan Outlines Priorities for Federal Agency
https://www.un.org/sustainabledevelopment/sustaina Engagement in AI Standards Development.
ble-development-goals/ Retrieved from https://www.nist.gov/news-
events/news/2019/08/plan-outlines-priorities-
[27] World Health Organization. WHO guidelines federal-agency-engagement-ai-standards-
approved by the Guidelines Review Committee. development
Retrieved from
https://www.who.int/publications/guidelines/ [37] Ding, J. and Triolo, P. (2018). Translation: Excerpts
from China’s ‘White Paper on Artificial Intelligence
[28] World Health Organization. (2015). Guidelines for Standardization’. New America DigiChina Blog.
the treatment of malaria. World Health Retrieved from
Organization. Retrieved from https://www.newamerica.org/cybersecurity-
https://www.who.int/malaria/publications/atoz/9789 initiative/digichina/blog/translation-excerpts-chinas-
241549127/en/ white-paper-artificial-intelligence-standardization/
[29] Wiegand, T., Krishnamurthy, R., Kuglitsch, M., Lee, [38] China Electronics Standards Institute (2018).
N., Pujari, S., Salathé, M., Wenzel, M. & Xu, S. Artificial Intelligence Standardization White Paper.
(2019). WHO and ITU establish benchmarking Retrieved from
process for artificial intelligence in health. The www.cesi.ac.cn/images/editor/20180118/201801180
Lancet, 394(10192), 9-11. 90346205.pdf
https://doi.org/10.1016/S0140-6736(19)30762-7
[39] Han, M. (2018). China Aims to Get the Jump on AI
[30] IEEE Artificial Intelligence Medical Device Standardization. Synced. Retrieved from
Working Group (EMB/Stds Com/AIMDWG). https://syncedreview.com/2018/01/25/china-aims-to-
Retrieved from https://sagroups.ieee.org/aimdwg/ get-the-jump-on-ai-standardization/
[31] ISO/IEC JTC 1/SC 42 Artificial intelligence. [40] Briard, I. (2019). CEN and CENELEC launched a
Retrieved from new Focus Group on Artificial Intelligence. CEN
https://www.iso.org/committee/6794475.html CENELEC News. Retrieved from
https://www.cencenelec.eu/news/articles/Pages/AR-
[32] Consumer Technology Association (2019). CTA 2019-001.aspx
Brings Together Tech Giants, Trade Associations to
Improve Efficiencies in AI and Health Care. CTA [41] Independent High-Level Expert Group On Artificial
Press Releases. Retrieved from Intelligence Set Up By The European Commission
https://www.cta.tech/News/Press- (2019). Ethics Guidelines For Trustworthy AI.
Releases/2019/April/CTA-Brings-Together-Tech- European Commission Digital Single Market News.
Giants,-Trade-Association.aspx Retrieved from https://ec.europa.eu/digital-single-
market/en/news/ethics-guidelines-trustworthy-ai
[33] Consumer Technology Association (2018). CTA
Releases White Paper on Use Cases in Artificial
Intelligence. CTA Standards News. Retrieved from
https://standards.cta.tech/kwspub/home/
– 74 –
ICT for Health: Networks, standards and innovation
[42] DIN Deutsches Institut für Normung e. V. (2019a). [52] Collins, G. S., & Moons, K. G. (2019). Reporting of
DIN to develop AI standardization roadmap. DIN artificial intelligence prediction models. The Lancet,
Press Releases. Retrieved from 393(10181), 1577-1579.
https://www.din.de/en/din-and-our- https://doi.org/10.1016/S0140-6736(19)30037-6
partners/press/press-releases/din-to-develop-ai-
standardization-roadmap-330542 [53] Khoram, E., Chen, A., Liu, D., Ying, L., Wang, Q.,
Yuan, M., & Yu, Z. (2019). Nanophotonic media for
[43] DIN Deutsches Institut für Normung e. V. (2018). artificial neural inference. Photonics Research, 7(8),
Interdisciplinary DIN Working Committee 823-827. https://doi.org/10.1364/PRJ.7.000823
“Artificial Intelligence”. Retrieved from
https://www.din.de/en/innovation-and- [54] International Medical Device Regulators Forum
research/artificial-intelligence/ai-working-committee (2017). Software as a Medical Device (SaMD):
Clinical Evaluation. Retrieved from
[44] DIN Deutsches Institut für Normung e. V. (2019). http://www.imdrf.org/docs/imdrf/final/technical/imd
DIN SPEC 92001-1:2019-04 Artificial Intelligence - rf-tech-170921-samd-n41-clinical-evaluation_1.pdf
Life Cycle Processes and Quality Requirements -
Part 1: Quality Meta Model. [55] The English National Institute for Health and Care
https://dx.doi.org/10.31030/3050203 Excellence (2019). Evidence Standards Framework
for Digital Health Technologies. Retrieved from
[45] DIN Deutsches Institut für Normung e. V. (2019). https://www.nice.org.uk/Media/Default/About/what-
DIN SPECs on Artificial Intelligence. Retrieved we-do/our-programmes/evidence-standards-
from https://www.din.de/en/innovation-and- framework/digital-evidence-standards-
research/artificial-intelligence/ai-din-spec framework.pdf
[46] Partnership on AI to Benefit People and Society [56] Lemm, S., Blankertz, B., Dickhaus, T., & Müller, K.
(2019). Retrieved from R. (2011). Introduction to machine learning for brain
https://www.partnershiponai.org/ imaging. Neuroimage, 56(2), 387-399.
https://doi.org/10.1016/j.neuroimage.2010.11.004
[47] Askell, A., Brundage, M., & Hadfield, G. (2019).
The Role of Cooperation in Responsible AI [57] Jacobs, C., & van Ginneken, B. (2019). Google’s
Development. arXiv preprint. lung cancer AI: a promising tool that needs further
https://arxiv.org/abs/1907.04534 validation. Nature Reviews Clinical Oncology, 1.
https://doi.org/10.1038/s41571-019-0248-7
[48] Askell, A., Brundage, M., & Clark, J. (2019). Why
Responsible AI Development Needs Cooperation on [58] Couzin-Frankel, J. (2019). Medicine contends with
Safety. OpenAI Blog. Retrieved from how to use artificial intelligence. Science,
https://openai.com/blog/cooperation-on-safety/ 364(6446), 1119-1120.
https://doi.org/10.1126/science.364.6446.1119
[49] Collins, G. S., Reitsma, J. B., Altman, D. G., &
Moons, K. G. (2015). Transparent reporting of a [59] Editorial (July 2019). Walking the tightrope of
multivariable prediction model for individual artificial intelligence guidelines in clinical practice.
prognosis or diagnosis (TRIPOD): the TRIPOD The Lancet Digital Health, 1(3), Pe100.
statement. BMC medicine, 13(1), 1. https://doi.org/10.1016/S2589-7500(19)30063-9
https://doi.org/10.1186/s12916-014-0241-z
[60] Saez-Rodriguez, J., Costello, J. C., Friend, S. H.,
[50] Moons, K. G., Altman, D. G., Reitsma, J. B., Kellen, M. R., Mangravite, L., Meyer, P., ... &
Ioannidis, J. P., Macaskill, P., Steyerberg, E. W., ... Stolovitzky, G. (2016). Crowdsourcing biomedical
& Collins, G. S. (2015). Transparent Reporting of a research: leveraging communities as innovation
multivariable prediction model for Individual engines. Nature Reviews Genetics, 17(8), 470.
Prognosis or Diagnosis (TRIPOD): explanation and https://doi.org/10.1038/nrg.2016.69
elaboration. Annals of internal medicine, 162(1),
W1-W73. https://doi.org/10.7326/M14-0698 [61] Maier-Hein, L., Eisenmann, M., Reinke, A., Onogur,
S., Stankovic, M., Scholz, P., ... & Feldmann, C.
[51] Heus, P., Damen, J. A., Pajouheshnia, R., Scholten, (2018). Why rankings of biomedical image analysis
R. J., Reitsma, J. B., Collins, G. S., ... & Hooft, L. competitions should be interpreted with care. Nature
(2019). Uniformity in measuring adherence to communications, 9(1), 5217.
reporting guidelines: the example of TRIPOD for https://doi.org/10.1038/s41467-018-07619-7
assessing completeness of reporting of prediction
model studies. BMJ open, 9(4), e025611. [62] Michaelis, C., Mitzkus, B., Geirhos, R., Rusak, E.,
https://doi.org/10.1136/bmjopen-2018-025611 Bringmann, O., Ecker, A. S., ... & Brendel, W.
– 75 –
2019 ITU Kaleidoscope Academic Conference
(2019). Benchmarking Robustness in Object [70] Bender, D., & Sartipi, K. (2013). HL7 FHIR: An
Detection: Autonomous Driving when Winter is Agile and RESTful approach to healthcare
Coming. arXiv preprint. information exchange. In Proceedings of the 26th
https://arxiv.org/abs/1907.07484 IEEE international symposium on computer-based
medical systems (pp. 326-331). IEEE.
[63] Filos, A., Farquhar, S., Gomez, A. N., Rudner, T. G. https://doi.org/10.1109/CBMS.2013.6627810
J.. Kenton, Z., Smith, L., Alizadeh, M., de Kroon, A.
& Gal, Y (2019). Benchmarking Bayesian Deep [71] Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J.,
Learning with Diabetic Retinopathy Diagnosis. Appleton, G., Axton, M., Baak, A., ... & Bouwman,
Preprint. Retrieved from J. (2016). The FAIR Guiding Principles for scientific
http://www.cs.ox.ac.uk/people/angelos.filos/publicat data management and stewardship. Scientific data, 3.
ions/diabetic_retinopathy_diagnosis.pdf https://doi.org/10.1038/sdata.2016.18
[64] Parikh, R. B., Obermeyer, Z., & Navathe, A. S. [72] Blum, A., & Hardt, M. (2015). The ladder: A
(2019). Regulation of predictive analytics in reliable leaderboard for machine learning
medicine. Science, 363(6429), 810-812. competitions. arXiv preprint.
https://doi.org/10.1126/science.aaw0029 https://arxiv.org/abs/1502.04585
[65] Finlayson, S. G., Bowers, J. D., Ito, J., Zittrain, J. L., [73] Anderson-Cook, C. M., Myers, K. L., Lu, L.,
Beam, A. L., & Kohane, I. S. (2019). Adversarial Fugate, M. L., Quinlan, K. R., & Pawley, N. (2019).
attacks on medical machine learning. Science, How to Host a Data Competition: Statistical Advice
363(6433), 1287-1289. for Design and Analysis of a Data Competition.
https://doi.org/10.1126/science.aaw4399 arXiv preprint. https://arxiv.org/abs/1901.05356
[66] Lapuschkin, S., Wäldchen, S., Binder, A., [74] Binder, A., Bockmayr, M., Hägele, M., Wienert, S.,
Montavon, G., Samek, W., & Müller, K. R. (2019). Heim, D., Hellweg, K., ... & Treue, D. (2018).
Unmasking Clever Hans predictors and assessing Towards computational fluorescence microscopy:
what machines really learn. Nature communications, Machine learning-based integrated prediction of
10(1), 1096. https://doi.org/10.1038/s41467-019- morphological and molecular tumor profiles. arXiv
08987-4 preprint. https://arxiv.org/abs/1805.11178
[67] Voosen, P. (2017). The AI detectives. Science, [75] Klauschen, F., Müller, K. R., Binder, A., Bockmayr,
357(6346), pp. 22-27. M., Hägele, M., Seegerer, P., ... & Michiels, S.
https://doi.org/10.1126/science.357.6346.22 (2018, October). Scoring of tumor-infiltrating
lymphocytes: From visual estimation to machine
[68] Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, learning. In Seminars in cancer biology (Vol. 52, pp.
J. W., Wallach, H., Daumeé III, H., & Crawford, K. 151-157). Academic Press.
(2018). Datasheets for datasets. arXiv preprint. https://doi.org/10.1016/j.semcancer.2018.07.001
https://arxiv.org/abs/1803.09010
– 76 –
REDESIGNING A BASIC LABORATORY INFORMATION SYSTEM FOR THE GLOBAL
SOUTH
The technological changes in the global south are exemplified Figure 2 – One of the issues of the current interface -
in a recent study conducted in Africa. Adedeji et al. surveyed inconsistent layouts of the query builder
206 nurses in Nigeria to investigate current and preferred
documentation methods and tools for the use of electronic customizable workflow - Lab administrators can determine
health records in a hospital [5]. Although 94% of the nurses their preferred workflows with configurable user interfaces.
in the survey managed patient records using paper-based It means that the system does not require additional training
processes, 80% wanted to change to electronic methods. for new users by adopting their work procedures. and 3)
Forty-two% of the nurses surveyed preferred to use mobile Flexible database - This feature allows to add and modify
devices such as iPad, Tablet PC, or Android Phones, while data fields as labs evolve. For example, as the specimen types
desktop computers were the least preferred devices in this and results may vary widely across laboratories and change
survey. This is interesting in light of the fact that more of over time, even within the same lab, the system can handle the
them used computers compared to mobile phones or tablets, diversity and transition smoothly. Despite many advantages
79.6%, 35.4%, and 14%, respectively. of the system, there were also some problems with its user
interface.
3. C4G BLIS
3.2 Problems of the Current Interface
In this section, we present the overview of C4G BLIS and
analyze the problems of the current interface of the system. In this section, we highlight problems within two core
Then, we propose a more advanced interface to resolve the modules of the system - search and registration. Figure
problems. 1 shows the user interface for a registration function. By
selecting one of the search fields (e.g., Patient Name and
3.1 Overview Patient ID) and entering keywords, a user can retrieve
patients’ information. The problem of this page is that the
C4G BLIS is an open-source web-based system to track size of the actionable components is not responsive to its
patients, specimens and laboratory results. It has been working environment. For example, the system cannot adjust
developed and managed by the Georgia Institute of the size of the drop-down list for a small-size smartphone.
Technology, the Centers for Disease Control and Prevention Similarly, the tiny Search button in the current interface will
(CDC) in the United States, and Ministries of Health of also be difficult to touch or press in the same environment.
several countries in Africa since 2011 [12]. Unlike other
laboratory information systems, C4G BLIS was designed to In the Reports page of C4G BLIS, a user can generate various
address the challenges of resource-constrained settings such reports by filtering on specimen type, test type, period, and so
as computing infrastructure, variability in lab practices, and on. As shown in Figure 2, similar search options and buttons
difficulty of record-keeping. The system provides three key are inconsistently located across the sub-pages of Reports.
features: 1) Robustness - To guarantee the stable operation It means that the user needs to remember all the different
under limited access of the Internet, it does not rely on locations of the same or similar components in each page to
online-based libraries for any of its operation. It only requires perform the same task; otherwise, the user should skim over
a simple network router, which can locally interconnect each page for every trial. Consequently, this interface is likely
computers in a hospital. 2) Fully configurable and to result in a high cognitive load or low task performance. We
– 78 –
ICT for Health: Networks, standards and innovation
4.1 Participants
Figure 3 – The issue of the interface - no distinction between
regular and critical operations For the first part of the user study, we recruited 30
participants from Ghana, Cameroon, and Nigeria through
system administrators of C4G BLIS in each country. For
the second part, we recruited 21 participants from the three
countries through the same method. To be eligible to
participate in both parts of the study, participants were
required to have a prior experience of using C4G BLIS.
Personal information such as age, education, and gender were
not collected in accordance with the research guidelines of
the Institutional Review Board of Georgia Tech. The system
administrators and participants were not compensated for
their participation. Part 1 took 17 weeks to complete and
Part 2 took 7 weeks.
Figure 4 – The proposed interface of the Results.
4.2 Data Collection Tool
can resolve this issue by placing the same components in the
same positions across all the pages. Another problem with For both parts of the user study, we used the same data
the current interface is that there are no visual clues to inform collection tool, HotJar, which is an advanced logging and
the users of critical task execution (see Figure 3). The users analysis system that reveals the online behavior of users
should be cautious while performing important tasks such as [14]. Particularly, the visitor recording feature allows us
deleting a user or a test result. For such an issue, modern to eliminate guesswork by recording of users’ actions while
user interfaces tend to highlight the buttons related to crucial using C4G BLIS. By observing the participants’ clicks, taps,
tasks with noticeable colors or shapes. and mouse movements, we could identify usability issues and
compute the execution time of given tasks. Figure 6 presents
3.3 Design Proposals a captured image of the HotJar system interface.
In order to resolve the interface problems noted in the previous 4.3 System Setup
section, we redesigned the user interface with the focus
on visibility, efficiency, consistency, and adaptability of the To protect personally identifiable information of real patients
system. Figure 4 shows the consistent style and location in the participating labs, we decided not to run the evaluation
of the search options and button in different pages of the on the systems in use; instead we set up the latest version of
new user interface. The size of actionable components C4G BLIS in a Google Cloud server with dummy data set
is also increased. The left side of Figure 5 presents the and dummy login credentials. Since the access of the Internet
Registration page of the new user interface. There is a clear and the supply of electric power are sometimes unreliable,
visual distinction among all the different function blocks: regional administrators were asked to check whether they
Search, Tips, and Results. In the Search block, we kept could access the system before participating in the study.
the same style and layout of search options and button like
other pages. In the Result block, we distinguished the data 4.4 User Study Procedure
and buttons by providing a visual clue, a rounded rectangle,
and highlighted critical actions (e.g., Delete) with a red color. Since the research team is based in the US, and the target
Most importantly, the proposed interface is responsive, which users are based in Africa, we were not able to visit the
means it can adjust its layout and elements to the user’s laboratories where the user studies were conducted. Thus,
screen setting. The right side of Figure 5 shows the same we trained the system administrators in the target countries
registration page accessed from a smartphone. The three through conference calls and documents shared over e-mails.
blocks are vertically re-arranged, and the components of each Later, the instructed administrators conducted the user study
block are also re-configured (i.e., their size and position are on-site. The training process for the administrators took
different). We implemented the proposed interface using about one hour per person. As shown in Table 1, our user
– 79 –
2019 ITU Kaleidoscope Academic Conference
Figure 5 – The proposed interface of the registration page: The user interface on a desktop (Left) and a smartphone (Right).
Figure 6 – The interface of HotJar, the data collection tool For the desktop environment, we were able to collect 20 and
used in our user study. (Left) The red-line indicates the 11 datapoints in Study 1 and 2, respectively. For a smartphone
moving trajectory of the mouse cursor. The color dots on the setting, we gathered five and four datapoints in Study 1 and
bottom side illustrate events such as clicking, typing, and so 2, respectively. For both studies in Cameroon, participants
on. (Right) The list shows the actions performed sequentially. used a two-in-one computer, which is both a laptop and a
tablet combined in one lightweight, portable device. After
finishing the user studies in a laptop setting, they flipped its
screen, switched to the tablet mode, and conducted another
study contains two interfaces, two devices, and three jobs. part of the study.
In this study, laptop and desktop were not distinguished and
considered as “Desktop”. First, we attempted to examine
the usability of the current interface of C4G BLIS. Among We excluded some datapoints for a number of reasons. First,
many factors (e.g., learnability, efficiency, memorability, and the datapoints from the tablet setting were removed because
errors) used to test the usability, we chose the efficiency; the we did not provide precise guidelines for that environment.
time a user takes to complete a task (The rationale for this We also excluded the log data for people that did not follow
being the fact that the primary objective of the system is to directions. For example, we asked a participant to find a
manage patient data as quickly as possible [15]). patient on the registration page. However, he or she moved
to the search page and found the patient. In this case, the
two menus were different for the two pages. In Nigeria,
In the training sessions with the regional administrators, participants were not allowed to access patient data on the
we ensured that they randomize the order of desktop and smartphone. Thus, they decided not to participate in the
smartphone studies for each participant. After a regional study with the smartphone setting. Due to the tablet issue in
administrator explained the goal and details of the user study Cameroon and the security policy issue in Nigeria, we were
to a participant, he/she was asked to perform three tasks - able to capture fewer datapoints on smartphones compared to
1) Find a patient with a given dummy name and view the desktops.
– 80 –
ICT for Health: Networks, standards and innovation
User Study
Study 1 - Current Interface Study 2 - Proposed Interface
Device Desktop Smartphone Desktop Smartphone
Task 1 Finding an existing patient using a given name
Task 2 Finding an existing patient using a given patient ID number
Task 3 Registering a new patient using a given name and additional information (e.g., name, age)
5. DATA ANALYSES AND RESULTS how much the proposed interface has improved, we adopt
one of the the usability metrics proposed by Jakob Nielsen
5.1 Data Analysis [16]. Since adding up the task times may be misleading if the
given tasks are unevenly performed, Nielsen recommended
The three tasks presented in Table 1 contain the same or
to compute the scores of improvement for each of the
similar actions such as clicking the search button and moving
tasks and take the geometric mean of these scores later.
the cursor to the view button in the result block. For the sake
For example, the relative score of the proposed interface
of data analysis, we categorized the tasks into seven actions
for Action 1 in the desktop setting can be computed as
as follows:
((2.5 − 1.7)/1.7 ∗ 100) + 100 = 147% (improvement of
• Action 1: Open the Registration page and select the text 47%). The percentages of improvement of each action are
input box in the search bar. presented in Table 2. The geometric means of the desktop and
smartphone environments are 132% and 134%, respectively.
• Action 2: Open the Registration page and select the In other words, it indicates that the proposed interface
Patient ID in the option of the search field. was perceived as having 32% and 34% higher usability in
the desktop and smartphone settings, respectively than the
• Action 3: Enter the given patient name and click the current one. Nielsen suggested to utilize a user satisfaction
search button score to formulate an overall conclusion if the target website
• Action 4: Enter the given patient identification number is about entertainment or rarely used. However, we did
and click the search button not normalize the usability results with users’ subjective
satisfaction as C4G BLIS is informative and frequently used
• Action 5: After checking the search result, click the in hospitals.
profile view button
6. DISCUSSION
• Action 6: After checking the search result, click the new
patient button 6.1 Design
• Action 7: Fill out the patient information and submit the
In this study, we compared the usability of desktop versus
form.
smartphone interfaces for C4G BLIS. This study was carried
As a quantitative metric for each action, we measured its out at three different sites with actual users. We were able
execution time, the difference between the start and end time, to recruit local system administrators to carry out the user
and excluded the loading time such as the delay from clicking studies. We then collected the log data for three tasks that
the search button to retrieving the inquiry data. were based on dummy data sets. We found improvements
of up to 32% usability for desktop and up to 34% for
5.2 Results smartphone settings for all three tasks. Although our results
are promising, we found some areas where we could improve.
Participants in our study used various devices and software
as follows: Devices (Desktop, Laptop, 2-in-1 Laptop,
Smartphone), Operating Systems (Windows, MacOS, First, the system should allow end-users to customize more
Android, iOS), and Browsers (Firefox, Chrome, Edge, configurations. For instance, we can consider the location of
Safari). Their screen resolution ranged from 320x432 to the tips block (See Figure 5). Since the system is designed for
1600x786. medical professionals, the usage tips might not be useful after
several sessions of using the system. Thus, we need to provide
Table 2 describes the average execution time of each action, an option to let the user hide the tips block. This preference
and the whisker plots in Figure 7 presents the minimum, can be stored in the cache so that we can keep it hidden. By
first quartile, median, third quartile, and maximum values doing so, the user can view more search results on the same
of task execution time in seconds. In order to determine screen without scrolling. Secondly, we can streamline other
– 81 –
2019 ITU Kaleidoscope Academic Conference
Table 2 – The average execution time for each action in seconds and the percentage of improvement
tasks. The actionable items in the rows of the result table need 6.3 Needs in the Near Future
to be effectively reorganized. As shown in the bottom-right
of Figure 5, the user has to swipe right to access the view, A system administrator in Cameroon reported that 75% of
update, and delete buttons in the result table, which are the the participants preferred working with tablets if the screen
essential functions of the page. Overall, we recommend that was large enough, and 25% of them were approved the use
designers and developers of LIS should carefully consider of smartphones to access the laboratory data. We also found
their specific target users, goals, and workflows. One way to similar needs in Section 2; 42% of nurses in Nigeria preferred
do so is to consult with end-users and system administrators to use mobile devices such as iPad, Tablet PC, or Android
before designing the user interface. Phones. Future studies should collect qualitative data to
see where these preferences are coming from – aspirational
desires (i.e., preference for a modern device) or needs (better
6.2 Computing Environments portability throughout the clinic).
As of May 2019, the most widely used browser in Africa 6.4 Interface Standard and Usability
is Chrome for Android (35.09% in the market share), and
the latest Chrome (Version 74.0) is also broadly adopted Medical data exchange standards have been considered as
(17.65%) [17]. In sharp contrast, there was one dominant a central issue of hospital information systems, thus many
operating condition when C4G BLIS was designed in 2011 - researchers and organizations have attempted to develop
Windows (81%), About 1024x768 screen resolution (51%), efficient, inter-operable standards and protocols such as
and Internet Explorer (15%) or Opera (11%) browsers. This Health Level Seven (HL7), Clinical Document Architecture
fact suggests that the user interfaces designed for the global (CDA) and Continuity of Care Document (CCD), and
south should be revisited so that they are stable, adaptive, and Systematized Nomenclature of Medicine (SNOMED).
responsive to their heterogeneous operating environments. Several studies found that adopting such a standard could
– 82 –
ICT for Health: Networks, standards and innovation
simplify communication interfaces and improve the quality the coordinators kept their eyes on the participants during
of patient care [18, 19]. Another critical issue of the clinical the entire user study. In this study, we only focused on the
information systems is the difficulty of use, which is one of the task execution time, but future studies should evaluate task
dissatisfaction factors expressed by healthcare professionals success rate, error rate, and user satisfaction since these are
[20]. Specifically, the complex interfaces and the lack of also important usability measures.
intuitiveness causes usability problems. However, this issue
has not been treated as necessary as the data exchange
standards [21]. 8. CONCLUSION
In the evaluation report of Electrical Medical Record (EMR), In this paper, we have examined the usability issues of an
Belden et al. said that “Usability is one of the major factors – existing LIS and proposed a new, responsive user interface
possibly the most important factor – hindering widespread to resolve them. We were interested in the extent to
adoption of EMRs. Usability has a strong, often direct which the proposed interface could improve usability in
relationship with clinical productivity, error rate, user fatigue heterogeneous environments. Results indicated an average
and user satisfaction ... [22].” Although there have been improvement of about 30% across various metrics. Based on
some efforts to resolve the usability issues such as the Health the results, we highlighted the current status of computing
Information Technology for Economic and Clinical Health environments and user needs in the near future. Additionally,
(HITECH) Act in 2009 and an incentive program by the U.S. we discussed several factors which can improve the quality
Centers for Medicare & Medicaid Services, the improvement of laboratory information systems and recommend adding
of the usability has still been slow [21]. As shown in usability specifications to international standards.
our study, we were able to improve the usability by up to
34% by applying a responsive, simple, open-source website REFERENCES
framework to the existing LIS. It should not be a challenging
[1] I. E. Asangansi, O. O. Adejoro, O. Farriab, and
task in terms of cost and development effort. We highly
O. Makinde, “Computer use among doctors in
encourage that international standards organizations dealing
africa:survey of trainees in a nigerian teaching hospital,”
with health informatics should pay attention to the usability
Journal of Health Informatics in Developing Countries,
standards for laboratory information systems as they include
vol. 2, no. 1, pp. 10–13, 2008.
diverse devices, platforms, and browsers, as important as they
did for medical record interoperability. [2] M. Shekar and K. Otto, “Information and
communication technologies for health in africa,”
Since the International Organization for Standardization The World Bank Group, Tech. Rep., 01 2014. [Online].
(ISO) proposed the definition of usability (ISO 9241-11) Available: http://documents.worldbank.org/curated/en/
and its metrics (ISO/IEC 9126-4), a few usability models 553151468009030957/ICTs-for-health-in-Africa
and evaluation techniques have been studied in the hospital
information system domain [23, 24]. Among them, [3] C. L. Ventola, “Mobile devices and apps for health
a prime example would be the empirical guideline for care professionals: uses and benefits,” P T, vol. 39,
safety-related usability of EHR (NISTIR 7804-1) proposed no. 5, pp. 356–364, May 2014, 24883008[pmid].
by the National Institute of Standards and Technology (NIST) [Online]. Available: https://www.ncbi.nlm.nih.gov/
[25]. The guidelines described three critical implications – pubmed/24883008
1) Consistently present patient identification in a reserved
[4] S. Wallace, M. Clark, and J. White, “‘it’s on my
area to prevent wrong-patient errors, 2) Provide visual cues
iphone’: attitudes to the use of mobile computing
to minimize the risk of entering information and writing
devices in medical education, a mixed-methods study,”
orders in the wrong patient’s chart, and 3) Support efficient
BMJ Open, vol. 2, no. 4, 2012. [Online]. Available:
and easy identification of incorrect items in lists of grouped
https://bmjopen.bmj.com/content/2/4/e001099
information by presenting clear, well-organized evidence.
Similar to the standardization efforts in NIST, more practical [5] P. Adedeji, O. Irinoye, R. Ikono, and A. Komolafe,
and applicable standards are needed for safer and effective “Factors influencing the use of electronic health records
hospital information systems. among nurses in a teaching hospital in nigeria,”
Journal of Health Informatics in Developing Countries,
vol. 12, no. 2, pp. 1–20, 2018. [Online]. Available:
7. LIMITATIONS http://www.jhidc.org/index.php/jhidc/article/view/174
We were in the US, and our target users were in Africa. [6] M. Jones, G. Marsden, N. Mohd-Nasir, K. Boone, and
We opted to use log data for the analyses, which meant that G. Buchanan, “Improving web interaction on small
we were not able to observe participants’ behavior in person displays,” in Proceedings of the Eighth International
while they completed the tasks. Thus, it was challenging Conference on World Wide Web, ser. WWW ’99.
to understand unexpected behaviors such as inactive mouse New York, NY, USA: Elsevier North-Holland,
cursors and random delays in typing. While we did our Inc., 1999, pp. 1129–1137. [Online]. Available:
best to train the administrators, we cannot guarantee that all http://dl.acm.org/citation.cfm?id=313234.313014
– 83 –
2019 ITU Kaleidoscope Academic Conference
[7] V. Roto and A. Oulasvirta, “Need for non-visual to improve communication between healthcare
feedback with long response times in mobile information systems,” in 2011 IEEE 9th International
hci,” in Special Interest Tracks and Posters of Symposium on Intelligent Systems and Informatics, Sep.
the 14th International Conference on World Wide 2011, pp. 453–457.
Web, ser. WWW ’05. New York, NY, USA:
ACM, 2005, pp. 775–781. [Online]. Available: [20] ASC Communications, “Survey: Providers investing
http://doi.acm.org/10.1145/1062745.1062747 heavily in ehrs, but dissatisfied with purchase,” Becker’s
Hospital Review, Jun 2014. [Online]. Available:
[8] Georgia Institute of Technology, “C4G Basic https://www.beckershospitalreview.com/healthcare-
Laboratory Information System.” [Online]. Available: information-technology/survey-providers-investing-
http://blis.cc.gatech.edu/ heavily-in-ehrs-but-dissatisfied-with-purchase.html
[9] BIKA Lab Systems, “Bika lims, open source laboratory [21] P. Sengstack and C. Boicey, Mastering informatics: a
information management system.” [Online]. Available: healthcare handbook for success. Sigma Theta Tau
https://www.bikalims.org International, Honor Society of Nursing, 2015.
[10] GNU Solidario, “Occhiolino, gnu laboratory [22] J. L. Belden, R. Grayson, and J. Barnes, “Defining
information management system.” [Online]. Available: and testing emr usability: Principles and proposed
http://lims.gnu.org methods of emr usability evaluation and rating,”
HIMSS EHR Usability Task Force, pp. 1–38, Jun.
[11] Internet Society, “History of the internet in africa.” 2009. [Online]. Available: https://www.himss.org/file/
[Online]. Available: https://www.internetsociety.org/ 1268216/download?token=L4oSWWi1
internet/history-of-the-internet-in-africa/
[23] “Ergonomics of human-system interaction - Part
[12] S. Vempala, N. Chopra, A. Rajagopal, J. Nkengasong, 11: Usability: Definitions and concepts, ISO
and S. Akuro, “C4g blis: Health care delivery via 9241-11:2018,” International Organization for
iterative collaborative design in resource-constrained Standardization, Geneva, CH, Standard, 2018.
settings,” in Proceedings of the Eighth International
Conference on Information and Communication [24] “Software engineering - Product quality - Part 4:
Technologies and Development. New York, NY, USA: Quality in use metrics, ISO/IEC TR 9126-4:2004,”
Association for Computing Machinery, 2016. [Online]. International Organization for Standardization, Geneva,
Available: https://doi.org/10.1145/2909609.2909657 CH, Standard, 2004.
[13] Codecalm.net, “Tabler ui kit.” [Online]. Available: [25] Svetlana Z. Lowry, Mala Ramaiah, Emily S. Patterson,
https://github.com/tabler/tabler Sandra S. Prettyman, Debora Simmons, David Brick,
Latkany A. Paul, Michael C. Gibbons, Sheryl
[14] Hotjar, “Hotjar: Visitor recording platform.” [Online]. L. Taylor, “(NISTIR 7804-1)Technical Evaluation,
Available: https://www.hotjar.com/ Testing, and Validation of the Usability of Electronic
Health Records: Empirically Based Use Cases for
[15] J. Nielsen, “Usability metrics: Tracking interface Validating Safety-Enhanced Usability and Guidelines
improvements,” IEEE Software, vol. 13, no. 6, for Standardization,” NIST Interagency/Internal Report,
pp. 12–13, Nov. 1996. [Online]. Available: http: pp. 1–44, 2015. [Online]. Available: https:
//dl.acm.org/citation.cfm?id=624616.625647 //doi.org/10.6028/NIST.IR.7804-1
[16] Nielsen Norman Group, “Usability metrics,” Nielsen
Norman Group, Jan. 2001. [Online]. Available:
https://www.nngroup.com/articles/usability-metrics/
[18] C.-H. Lin, I.-C. Lin, J.-S. Roan, and J.-S. Yeh,
“Critical factors influencing hospitals’ adoption of
hl7 version 2 standards: An empirical investigation,”
Journal of Medical Systems, vol. 36, no. 3,
pp. 1183–1192, Jun 2012. [Online]. Available:
https://doi.org/10.1007/s10916-010-9580-2
– 84 –
#RINGINGTHEALARM: CHRONIC “PILOTITIS” STUNTS
DIGITAL HEALTH IN NEPAL
1
George Washington University School of Public Health
2
George Washington University School of Medicine and Health Sciences
1. INTRODUCTION 2. METHODS
The health system in Nepal is fraught with systemic The study uses a scoping design to obtain information on
challenges due to factors such as the country’s status as an digital health initiatives in Nepal. Research activities were
economically least developed country, inaccessible conducted between June 2017 to September 2018. The
mountainous terrains and sociological and topographical George Washington University Institutional Review Board
diversity. Economic and demographic transitions, migration and Nepal Health Research Council (NHRC) approved this
and unplanned urbanization adversely influence the health of study. Scoping review is a method of synthesizing
the population as well [1,2]. A decade-long civil conflict and knowledge on studies when: it is difficult to employ a narrow
political turmoil has also contributed to worsening mental review question; synthesizing information from studies that
health outcomes, disruption of service delivery in impacted have used a wide variety of data collection and analysis
areas, and a compromised health policy and governance techniques; there is a scarcity of prior synthesis on the topic;
system [3,4,5]. Despite these setbacks and challenges, Nepal or a quality control mechanism of the reviewed sources will
has made significant progress in reducing under-five not be conducted [12,14,15]. This study replicated a scoping
– 86 –
ICT for Health: Networks, standards and innovation
(?) Smart Health Nepal 2017 convenience by being able to analyze and visualize data,
Private (3, 15%) monitor data collection and management in real time, and
˟ HealthNet Nepal Telemedicine Program 2004 more importantly improve data quality, have been well
received by impacted teams and stakeholders. The ability to
* Health at Home Nepal 2007
make informed decisions will soon be available at multiple
√ Safe SIM 2015
levels of the health system in Nepal due to these efforts.
*Ongoing √ Completed ˟ Discontinued (?) Unknown Ultimately, these efforts, if sustained, will strengthen health
sector governance and improve the availability and quality
Table 2 identifies the 20 projects, affiliated organizations, of health service delivery [32]. However, doing so will
implementation launch year and their current require forethought and development of data strategies with
implementation status. More projects had been completed feedback loops into implementation strategic plans. Our
(45%) compared to ongoing (40%) ones based on our review. review did not find any mention of either of these activities
At least one identified program, the earliest telemedicine happening in Nepal to date.
program offered in Nepal (HealthNet Nepal) has been
discontinued citing financial viability as their primary barrier 3.3 Health workforce
[26]. To understand these initiatives in greater depth, we
have categorized them per WHO health system building Frontline workers were the target audience for over half of
blocks [16]. the identified projects (55%) especially female community
health volunteers (FCHVs) [24,28,33-37]. Mid-level
3.1 Health service delivery medical personnel and health staff at remote locations were
also other beneficiaries of digital health initiatives in Nepal
Telehealth programs are the most prevalent form of service [19,23,29,38,39,42]. Scarcity of a readily available
delivery format for digital health solutions implemented in workforce trained and skilled in digital health
Nepal. Focus areas for digital health solutions ranged from implementation result in projects relying on internal training
health surveillance, adolescent, sexual, and reproductive to sensitize project staff in the use of technological solutions
health, detection of disease outbreaks, maternal and child [18,28]. Our review did not find existing academic digital
health, and mental health issues [25,26,35,36]. Services health programs in Nepal. Significant issues at the workforce
provided by digital health solutions focused on bolstering level cited by the initiatives reviewed in our study were poor
health information systems, increasing access and utilization coordination between ground level and higher-level staff and
of health services, increasing access to health information, educational, digital and data illiteracy within the workforce.
providing health education and generating health awareness, Readiness to utilize developed solutions was assessed
developing referral mechanisms, providing specialized infrequently.
healthcare via telemedicine, and improving technical
capacity among health professionals and frontline workers 3.4 Technology
[27,28,29,30].
Telecommunication-based digital health solutions were most
3.2 Health information prevalent. These solutions ranged from smart-phone
applications for patient monitoring, care or diagnosis, data
Health information systems developed for population health collection via a simple SMS-based reporting, telemedicine
surveillance, data collection and management, and patient [21] or population-health surveillance via health information
monitoring and support are the major focus areas for digital systems [38]. Challenges that hindered technological
health solutions implemented in Nepal. According to a implementation were weak mobile networks [40], slow
country profile on HMIS capacity developed by Measure Internet service [27], unreliable power [27,40], technical
Evaluation, the existence of a national health information difficulties such as unanticipated system errors and reliance
system (HIS) policy is unknown [31]. Currently, there is an on undeveloped ICT infrastructure [28].
HIS strategic plan and a national HIS coordinating body was
established in the past, but their current activities are also 3.5 Leadership and governance
unknown [31]. In 2016, the Nepalese MoH transitioned from
HMIS to DHIS-2. The DHIS-2 serves as a subnational level There are a few notable policy and strategic frameworks in
electronic system for aggregating routine facility or Nepal that speak towards the utility and uptake of digital
community service data. Its roll-out is in an early phase, health solutions. In 2014, the Nepal Health Sector Strategy
starting at the national level followed by the district level (NHSS) 2015-2020 was published by the Government of
then extending to health posts and primary care facilities in Nepal [41]. It functions as Nepal’s guide to graduate from
a few districts. Remote locations, underdeveloped “Least Developed Country” to “Middle Income Developing
information communication technology (ICT) infrastructure, Country” by 2022. Additionally, it articulates the nation’s
difficult terrain, digital data and technological literacy, and a commitment towards achieving universal health coverage
lack of data standards and interoperability have challenged (UHC) by placing health at the center of the overall
the implementation of health information systems [32,38]. socioeconomic development efforts. Per the NHSS, the GoN
However, substantial time savings in data collection and in collaboration with its development partners, aspires to
entry have helped mitigate these challenges. The increase in leverage novel technologies to address health challenges in
– 87 –
2019 ITU Kaleidoscope Academic Conference
the country. Recognizing that health is a critical indicator of alignment with the results of a global e-health survey
economic advancement, NHSS stipulates ICT such as digital conducted by the WHO which note that financial
health tools as essential in improving access and quality of management in e-health and m-health is largely fulfilled by
health services delivery [43]. Nepal's eHealth Strategy was donors [50]. Among the identified projects, 85% were
published in 2017 which documents the vision of using e- conducted in multi-sectoral collaborative partnership
health to facilitate the delivery of accountable, equitable and between donors, non-governmental organizations, or the
high-quality health service delivery [18,37]. GoN. Some of these actors collaborated directly with the
GoN at the national level. Other collaborations were with the
Evidence supports the notion that investments made to regional or local governing bodies. Direct financial
bolster ICT infrastructure result in tangible improvements in assistance from the government was not apparent in our
economic growth and achievement of human development review aside from the allocated funding for the Hello Health
milestones especially for developing countries [39,44]. telemedicine program. Financial stability was an important
However, mirroring the economic divide between developed, indicator of the sustainability of identified projects. Irregular
developing and least developed nations is the digital divide funding has been noted as a significant hindrance to the
defined as the difference between the uptake and utilization effective implementation of projects [27]. Assessing
of ICT in the daily lives of citizens [45]. Reflecting this economic feasibility was also a rarity among identified
digital divide, Nepal ranks 165th in the e-government initiatives despite most projects indicating monitoring and
development index [38,46]. The ICT policy environment in evaluation frameworks at the project level. One identified
Nepal does not meet modern demands mired by endemic project ended their feasibility trial without plans for scaling
corruption, political instability and civil conflict [6,47]. A up despite improvement in project indicators after
detailed strategic framework has been put forward to performing an economic evaluation which determined the
improve access to quality healthcare using ICT and modern program was not worth the financial and human costs
technologies in the latest update to the national ICT policy. incurred by their organization [28].
This will be achieved through an increased investment in
ICT-based healthcare systems with a special emphasis on Very few projects discussed the potential of sustainability.
telemedicine programs. These systems and networks will be Identified projects were primarily feasibility trials in early-
developed and implemented through a collaborative phases (85%) conducted in rural areas (65%). Since most of
approach involving public, private and civil society actors. these projects were short-term, the results ended with no
Regulatory frameworks will be developed to guide and plans to continue the initiative after the end of the funding
govern health information and ensure security measures period. The GoN was identified as the appropriate lead actor
curtail any privacy and ethical concerns [48,49]. So far, three in taking ownership of the space and assessing the
policies have been established that address digital health effectiveness of digital health solutions in Nepal then
efforts in Nepal: Nepal E-governance Interoperability working towards scaling them up through integration and
Framework, Electronic Transaction Act and institutionalization with national-level systems [27].
Telecommunication Policy [18]. The Nepalese Ministry of Leadership turnover, especially when multiple partners were
Health (MoH) has identified the formulation and involved, was identified as a contributing factor for the
institutionalization of a national e-health steering committee failure of digital health solutions [23,28,40].
and task force to govern strategic planning,
institutionalization of an e-health unit at the MoH, 4. DISCUSSION
procurement of resources and technical assistance,
development of prioritized e-health plans. This is reflected in Due to a recent constitutional mandate, Nepal is currently in
annual work plans and budgets, development of legal the process of transitioning to a federal state. Local
provisions such as a Health Information Act to address data municipal governing bodies will soon fully assume the
use, privacy and confidentiality issues, and development of responsibility of health-sector planning, budgeting and
monitoring and evaluation frameworks embedded into oversight of health service delivery. In parallel, the country’s
NHSS as prerequisites to achieving the goals set forth in the commitment to UHC and its quest to develop high-quality
national eHealth strategy of 2017. health systems to meet Sustainable Health Development
Goals (SDGs) by 2025 place Nepal at an opportune juncture
3.6 Financing and sustainability to map out its strategic course of action within the health
sector. This is an apt time to assess where digital health
More than half of the 20 initiatives were implemented by solutions fit within the current health system and how they
international or national-level non-profit organizations can be leveraged in impactful ways. The implementation
(I/NGOs) such as RTI International, USAID, GIZ, Possible timeline and trajectory represented by our review indicate
Health, Nick Simon’s Institute, and One Heart Worldwide that the momentum for digital health solutions has been
(Table 2). This contrasts the findings of Ahmed et al. [12] in sporadic but continuous in Nepal. Its evolution is still in
Bangladesh where the private sector leads the development nascent stages thus presenting opportunities to influence it
and implementation of digital health solutions. In Nepal, the with evidence-based and data-driven strategies. In the past
private sector is the least prolific in this space (15%). Digital 25 years, there has been an absence of diversity in the
health solutions in Nepal appear to be mainly donor (55%) application, utilization and uptake of digital health solutions
and government driven (30%). Our findings, however, are in in Nepal demonstrated by the limited scope, focus areas,
– 88 –
ICT for Health: Networks, standards and innovation
target audiences, sustainability potential and proliferation in workforce capacity and enriching academic programs to
certain sectors only. meet increasing demands for digital health expertise is
another area that requires investment and sustained
Much like the overall course of the health system evolution commitment by stakeholders. Engaging the private sector,
in Nepal [51], digital health solutions are frayed, issue and especially telecommunication companies, is an underutilized
organization-centric, and primarily driven by donor or non- strategy to address current barriers related to implementation,
governmental organizations. Most solutions have focused on technical, financial and sustainability challenges [12].
telemedicine to enhance gaps in healthcare access in rural Existing pioneers in this space (GoN and I/NGOs) must
areas with frontline workers heavily burdened with multiple engage in strategic planning on how to increase collaborative
e-health or m-health tools and solutions for different partnerships with the private sector or incentivize
programs or organizations. There are missed opportunities to independent commercial ventures [55].
increase service delivery beyond providing basic or
specialized healthcare in rural areas and addressing health 5. STRENGTHS AND LIMITATIONS
challenges related to maternal and child health or sexual and
reproductive health. The rise of non-communicable diseases, Our scoping study was successful in retrieving information
unplanned urbanization and demographic transition to a regarding 20 digital health solutions implemented via formal
higher number of elderly populations juxtaposed with a channels in Nepal to date. To the best of our knowledge, this
continuing flow of migrant youth working abroad warrant is the first paper to capture national level digital health efforts
the exploration of digital health solutions beyond its current implemented in Nepal and analyze the space utilizing two
limited scope and application [1,2,27,52,53,54]. Informal rigorous evidence-based frameworks [15,16]. Informal
consultations with key informants have shed light on the fact consultations with experts suggest the potential of additional
that digital health solutions are being implemented in Nepal digital health solutions implemented via both formal and
in these areas through informal channels on an as-needed informal channels that were not captured in our study since
basis and are yet to be documented or formalized due to they are not publicly accessible via the Internet or published
regulatory, financial/compensation or resource gaps. Our in academic journals. There is currently an effort on our part
team is currently analyzing findings of key informant to analyze findings from key informant interviews to bridge
interviews with stakeholders on the ground to share our this documentation gap. Evaluating the effectiveness of
review findings, attain insights regarding facilitators and identified solutions was not feasible with the limited
barriers, and explore the utilization and potential of digital resources available to achieve the scope of this study. This
health solutions in Nepal beyond what has been found line of investigation is a fruitful avenue for future studies to
through this review. explore and research.
– 89 –
2019 ITU Kaleidoscope Academic Conference
interventions in 2012 [63]. In contrast, the adoption of digital [10] L. Wallist. “Integrating mHealth at point of care in
health solutions sparked into existence in China when low- and middle-income settings: the system
combating SARS in late 2002. In both countries, however, perspective.” Global Health Action. 10(3).
there has been substantial challenges related to buy-in from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC564
stakeholders to adopt digital innovations. The reasons cited 5717/
mirror our findings. In the same vein, the Indian MoH
published a National Digital Health blueprint for public [11] WHO Department of Health Statistics and
comment and review in 2019 to address this endemic issue Information of the Innovation, Information,
of fragmentation and “pilotitis” within the digital health Evidence, and Research Cluster in collaboration
space [64]. To counter the documented stunting of digital with WHO technical programmes and regional
health interventions nationally, regionally and globally, we offices: World Health Statistics 2011. France: World
wholeheartedly agree with Huang and colleagues that the Health Organization, 2011.
key to moving beyond digital health “pilotitis” is in
considering the governmental, social, political and historical [12] T. Ahmed., H. Lucas, A.S. Khan, R. Islan, A.
contexts so that the attitudes, perceptions and needs of Bhuiya, M. Iqbal. “eHealth and mHealth initiatives
stakeholders are accounted for just as much as the design and in Bangladesh: A scoping study”. BMC Health
development of digital health solutions [61,62]. Services Research, vol 14, pp 260, 2014.
[4] Nepal Ministry of Health and Population. National [16] WHO Department of Health Statistics and
Mental Health Policy. Kathmandu: Ministry of Information of the Innovation, Information,
Health and Population; 1997. Evidence, and Research Cluster. Monitoring the
Building Blocks of Health Systems: A Handbook of
[5] S. Singh. “Impact of long-term political conflict on Indicators and their Measurement Strategies,
population health in Nepal”. Can Med Assoc J, Geneva, Switzerland, 2010:
171(12), pp 1499–502. doi: 10.1503/cmaj.1040777, http://www.who.int/healthinfo/systems/WHO_MBH
2004. SS_2010_full_web.pdf
[6] A. Sharma, Y.S. Kim. “Information Communication [17] D. Colaci, S. Chaudhri, A. Vasan. “mHealth
Technology Development in Nepal”. JRSD, pp 101- interventions in low-income countries to address
141, 2018. maternal health: A systematic review”. Ann Glob
Health, vol 82(5), pp 922-935, 2016.
[7] Nepal Telecommunications Authority. MIS Report,
vol 177(129), 2019: https://nta.gov.np/wp- [18] Nepal Ministry of Health. “eHealth initiatives in
content/uploads/MIS-2076-Shrawan.pdf Nepal”, 2017: http://nhsp.org.np/wp-
content/uploads/2017/08/eHealth-Initiatives.pdf
[8] J.G. Kahn, J.S. Yang, & J.S. “Mobile” health needs
and opportunities in developing countries. Health [19] P.R. Lama. “Scope and challenges of Telemedicine
Affairs, vol 29, pp 252 – 258, 2010. in Nepal: A look towards future. A case study at Om
Hospital and research Centre Kathmandu, Nepal.”,
[9] V. P. Gurupur, Wan T.H. “Challenges in 2010:
implementing mHealth interventions: a technical http://nrl.northumbria.ac.uk/11374/1/lama.tshering_
perspective”. mHealth, 3, 32. phd.pdf
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC558
3043/
– 90 –
ICT for Health: Networks, standards and innovation
[20] M. Pun, R. Shields, Paudel R., P. Mucci. Nepal [31] Measure Evaluation. Nepal: HIS Indicators, 2018:
Wireless Networking Project: Case Study and https://www.measureevaluation.org/his-
Evaluation Report, 2006. strengthening-resource-center/country-profiles/nepal
[21] B. Bhandari, B. “Telemedicine: Healthcare in Times [32] K. Birdsall. “Bringing the power of digital data to
of High Technology”. My Republica, 2010: rural Nepal. Healthy Developments”. Federal
http://archives.myrepublica.com/portal/index.php?ac Ministry of Economic Cooperation and
tion=news_details&news_id=1636 Development (BMZ), 2018:
https://health.bmz.de/events/In_focus/bringing_pow
[22] V. Patterson, M. Singh, H. Rajbhandari, S. er_digital_data_rural_nepal/index.html.
Vishnubhatla. “Validation of a phone app for
epilepsy diagnosis in India and Nepal”. Elsevier, [33] Ekantipur. “Ministry of health pilots mHealth,
30:46-49, 2015. eHealth services”. Kathmandu Post, 2017:
http://kathmandupost.ekantipur.com/news/2017-07-
[23] J. Morrison, N.R. Shrestha, B. Hayes, & M. 04/ministry-of-health-pilots-mhealth-ehealth-
Zimmerman. “Mobile phone support for rural health services.html.
workers in Nepal through ‘celemedicine’”. Journal
of Nepal Medical Association, vol 52(191), 2013. [34] Johns Hopkins. “mHealth internships build lasting
partnerships to help prevent human trafficking in
[24] United Nations Population Fund Nepal. Harnessing Nepal”, n.d.:
Mobile Technology to Improve ASRH Knowledge https://www.jhsph.edu/departments/international-
in Nepal, 2014: health/news/student-internships-nepal.html
https://nepal.unfpa.org/en/news/harnessing-mobile-
technology-improve-asrh-knowledge-nepal [35] One Heart World-Wide. “Maternal health in Nepal”,
2016:
[25] USAID. Reaching Pregnant Women in Marginalized https://give.onedayswages.org/campaign/maternal-
Communities to Increase Utilization of Maternal, health-in-nepal-or-one-heart-worldwide/c104911
New Born, and Family Planning Services. Health for
Life: Technical Brief, RTI International, [36] One Heart World-Wide. One heart world-wide –
Kathmandu, Nepal, 2016: quarterly report, 2016:
https://www.rti.org/sites/default/files/mobile_trackin https://static1.squarespace.com/static/5700f29260b5
g_technical_brief_final.pdf e92eb865e179/t/591b31b329687f883309ae10/14949
54456275/OHW+2016+Annual+Report.pdf
[26] A. Dhakal, A. “Telemedicine and Consultation in
Nepal (Unpublished bachelor’s thesis)”. Turku [37] S.M. Van Buren. “Evaluating the viability of a
University of Applied Sciences, Turku, Finland, mobile-device-based intervention to improve
2017. maternal health in Achham, Nepal”, 2013:
https://bir.brandeis.edu/bitstream/handle/10192/2503
[27] R. Bhatta. “Telemedicine and eHealth in Nepal.” 1/VanBurenThesis2013.pdf?sequence=3&isAllowed
Journal of Nepal Public Health Association, vol =y
6(1), pp 41-46.
[38] P. Rana. Two exciting new government open data
[28] D.J. Meyers, M. Filkins, A.H. Bangura, R. Sharma, initiatives presented at Kathmandu Open Data Day
A. Baruwal, S. Pande & D. Maru. “Management 2017: http://archive.opennepal.net/blog/two-
challenges in mHealth: failures of a mobile exciting-new-government-open-data-initiatives-
community health worker surveillance programme presented-kathmandu-open-data-day-2017.
in rural Nepal”. BMJ Innovations, 2017.Nepal”,
2015: http://ictpost.com/telemental-health-support- [39] S. Thapa. “Empowering Health Workers in Nepal
eclinics-in-nepal/ with mHealth”, 2016:
https://www.linkedin.com/pulse/empowering-health-
[29] A.K. Jha & S. Karki. “Reaching the Unreached: workers-nepal-mhealth-samita-thapa
Integrated Approach at Grass Root Population by
Tele-Health”. C.H.E.S.T. Nepal, 2017: [40] A. Harsha, D.S. Maru, E. Okada, P. Acharya, R.
http://smgebooks.com/telemedicine/chapters/TELM Sharma, S. Bhatta, S. Dahal, T. Yangdol. “Pilot and
ED-17-06.pdf evaluation of the feasibility of SafeSIM-a mobile
technology platform for maternal health care
[30] Inmarsat. “Inmarsat enables emergency mHealth in coordination in Nepal”, 2015:
Nepal”, 2015: http://medicmobile.org/assets/downloads/Safe-SIM-
https://www.inmarsat.com/news/inmarsat-enables- preliminary-evaluation_findings-07.2015.pdf.
emergency-mhealth-in-nepal/
– 91 –
2019 ITU Kaleidoscope Academic Conference
[41] Nepal Ministry of Health and Population. “National [52] M.J. Allsop. “The State of mHealth development
Health Sector Strategy 2015-2020.”, 2015: and use by palliative care services in sub-Saharan
http://www.nationalplanningcycles.org/sites/default/ Africa: a systematic review of the literature”. BMJ
files/planning_cycle_repository/nepal/nhss-english- Supportive and Palliative Care, vol 8(2), pp 155-
book-final-4-21-2016.pdf 153, 2015.
[42] ICT Post. “Telemental health support eClinics in [53] M. Staeheli, R.H. Aseltine, E. Schilling, D.
Nepal”, 2015: http://ictpost.com/telemental-health- Anderson & B. Gould. “Using mHealth technologies
support-eclinics-in-nepal/. to improve the identification of behavioral health in
urban primary care settings”. SAGE Open Med,
[43] S. Lange & S. Vollmer. “The effect of economic 5:2050312117712656, 2017.
development on population health: a review of the
empirical evidence”. British Medical Bulletin, vol [54] L.T.H. Vu, N.T.K. Nguyen, H.T.D. Tran, N.
121(1), pp 47-60, 2017. Muhajarine. “mHealth information for migrants: an
eHealth intervention for internal migrants in
[44] M. Kuppusanmy. & S. Santhapparay. “Investment in Vietnam”. Reproductive Health, 13:55, 2016.
Information and Communication Technologies
(ICT) and its Payoff in Malaysia”. Perspectives on [55] K4Health. “The mHealth Planning Guide: Key
Global Development and Technology, vol 4(2), pp Considerations for Integrating Mobile Technology
147-16, 2005. into Health Programs”. K4Health, 2014:
https://www.k4health.org/toolkits/mhealth-planning-
[45] International Telecommunications Union. ICT facts guide
and figures, 2015: https://www.itu.int/en/ITU-
D/Statistics/Documents/facts/ICTFactsFigures2015. [56] Al-Shorbaji, N. “The World Health Assembly
pdf Resolutions on eHealth: eHealth in Support of
Universal Health Coverage”. Methods Inf Med, vol
[46] United Nations. United National E-Government 52, pp 463-466, 2013.
Survey 2014. New York: United Nations:
https://publicadministration.un.org/egovkb/en- [57] The Rockefeller Foundation. “From Silos to
us/Reports/UN-E-Government-Survey-2014 Systems: An Overview of eHealth’s Transformative
Power”. In Rockefeller Foundation Report. Making
[47] R. Shields. “Information access and communications the eHealth Connection: Global Partnerships, Local
policy in Nepal”. Telecommunications policy, vol Solutions, Bellagio Conference Series. New York,
33 (3-4), pp 207-214, 2009. 2010.
[48] S. Sigdel. (2017). “eHealth in Nepal, Policies and [58] World Health Organization & International
Strategies: a quick overview”, 2017: Telecommunications Union. National eHealth
https://medium.com/hisn-health-informatics-society- strategy Toolkit, 2012. Geneva.
of-nepal/e-health-in-nepal-policies-and-strategies-a-
quick-overview-by-mr-giriraj-subedi-5874d2e22d19 [59] USAID & DIAL. “Principles for Digital
Development”, 2017:
[49] K. Regmi. “Mainstreaming mHealth in Nepal: https://digitalprinciples.org/principles/
Highlights of a One-Day Workshop”, 2016:
https://webcache.googleusercontent.com/search?q=c [60] World Health Organization. “Recommendation on
ache:CPUZ-PE- digital interventions for health system
MPMJ:https://www.herd.org.np/blogs/118+&cd=3& strengthening”. Geneva: World Health Organization;
hl=en&ct=clnk&gl=us 2019. Licence: CC BY-NC-SA 3.0 IGO, 2019.
[50] M. Kay, J. Santos, M. Takane. “WHO Global [61] ITU & DIAL. “SDG Digital Investment Framework:
Observatory for eHealth: mHealth new horizons for A whole-of-government approach to investing in
health through mobile technologies”, vol 3, 2011. digital technologies to achieve the SDGs”, 2019.
[51] J. Sharma, A. Aryal, G.K. Thapa. “Envisioning a [62] F. Huang, S. Blaschke, & H. Lucas. Beyond
high-quality health system in Nepal: if not now, pilotitis: taking digital health interventions to the
when?” The Lancet Global Health, vol 6(11), pp national level in China and Uganda. Globalization
1146-1148, 2018. and Health, vol 13, pp 49, 2017.
– 92 –
ICT for Health: Networks, standards and innovation
[63] D. McCann. “A Ugandan mHealth moratorium is a [64] Ministry of Health & Family Welfare Government
good thing”, 2012: of India. “National Digital Health Blueprint”, 2019.
https://ictworks.org/2012/02/22/ugandan-mhealth-
moratorium-good-thing/
– 93 –
DESIGNING NATIONAL HEALTH STACK FOR PUBLIC HEALTH:
ROLE OF ICT-BASED KNOWLEDGE MANAGEMENT SYSTEM
1
Indian Institute of Public Administration, New Delhi
2
Department of Telecommunication, Government of India
3
Indian Institute of Public Administration, New Delhi
Stack can be understood as a set of APIs and systems that countries for shaping its healthcare services to be more
allow various stakeholders including governments, inclusive [5].
businesses, developers, etc. to utilize this digital
infrastructure towards achievement of common goals. At present, emerging technologies such as Internet of things
Creation of a stack on any vertical at a national level would (IoT), artificial intelligence (AI) and others are being
enable paperless, cashless and presence-less secure access to employed to design disruptive health innovations so that
the system. Such a holistic approach fosters innovation in quality healthcare may be accessible to a larger population.
service delivery. Literature is replete with application of wireless technology
almost two decades ago in the area of telemedicine in select
In the present scenario, the data that is needed to build “One countries of Europe including Athens (Greece), Cyprus, Italy
Public Healthcare System” is available in standalone systems, and Sweden [6] and the use of artificial intelligence and
only some of which are connected. The most notable artificial neural networks (ANNs) for maintaining and
attribute of PH is that it aims to widen the health system goals analyzing electronic health records (EHRs) by University of
from the one that is narrowly focused on curing diseases in California and University of Chicago for a period ranging
hospitals by health professionals, to a system that is focused from the year 2009 till 2016 [7]. Even in developing
on keeping populations/communities healthy by providing countries, drones are assisting in providing medicines and
advance information to stakeholders. Therefore, for medical aid in difficult and inaccessible areas of India [8].
improving the PH scenario in a developing country like India, The wearable devices are being popularly used to provide
an integrated knowledge management system (KMS) that remote and continuous monitoring of each heartbeat,
could interconnect the standalone, existing healthcare moment-to-moment blood pressure, oxygen concentration in
applications/other related applications into a holistic blood, body temperature, level of glucose, human activities
integrated national health stack (stack is a data structure used and emotions [9]. IoT-based solutions have the potential to
to store a collection of objects) by employing information reduce the required time for remote health provision and
communication technologies (ICTs) will go a long way in increase the quality of care by reducing costs with enriched
achieving PH goals. user experience. Similarly, robotics process automation
(RPA)-based bots can ‘advise’ primary care patients;
The aim of the present study is to propose a conceptual machine learning (ML)-based systems can help to identify
design and an actionable implementation strategy for diseases early that indeed constitutes an important step
building a KMS, referred to as a national health stack (NHS). towards preventive healthcare [10]. The National Health
Introduction of the paper establishes the ‘multidisciplinary’ Service (NHS), UK is planning to extensively harness the
and ‘preventive’ nature of PH. This is followed by a ‘Review potential of AI to make 30 million outpatient visits
of Literature’,section 2, on core subject areas viz. ICT and unnecessary so that the resources saved can be used for
KMS implementation in the health domain of various frontline care [11].
countries, with special reference to emerging technologies,
PH and best practices related to the subject under However, apart from these various ways in which the
consideration. Section 3 examines the national health advances in communication technology, computing, storage,
scenario of India with special reference to its policies on analytics etc. are helping in achieving the health goals, ICT
information systems pertaining to ICT implementation in can also play a more fundamental role. ICT can be used to
health, the current status and future vision. Once both the link the different sources of data, collate the information
theoretical propositions as well as the ground reality of the available, provide tools for analysis and make it available to
Indian health ecosystem have been elucidated, the paper the stakeholders for predictive analysis. Knowledge
moves on to state the goal, objectives and design of the management (KM) of this data pertaining to different aspects
proposed conceptual model,section 4: Proposed Conceptual of healthcare would help to provide deeper insights into the
Model of NHS. The implementation strategy is described in various aspects of organizational learning and community
the subsequent section, section 5: Implementation of the wellbeing, as indicated subsequently.
Proposed Model), followed by ‘Issues and Opportunities’ in
section 6, and‘Conclusion’ insection 7. 2.2 Applying principles of knowledge management
(KM) in public health
2. REVIEW OF LITERATURE
Knowledge management (KM) refers to a multidisciplinary
2.1 Application of ICT in health sector process of creating, sharing, using and managing the
information from different systems for achieving
Rapid proliferation of ICTs has catalyzed its application in organizational objectives. ICT can provide knowledge
the health domain since the 1990s by implementation of discovery through integrated data mining of health data that
mobile collaboration technologies, hospital management is provided by heterogeneous sources. With special reference
systems, online patient information systems, and so on to developing countries, Blayaet al., (2010) assert health
[4].Several related challenges including, but not limited to, solutions that emanate from well-designed ICT-based KMS
infrastructure concerns that are more often sighted in in resource-poor environments have a tangible impact on the
developing countries like India, did not deter researchers quality of health care [12].
from examining the merits of employing ICTs in these
– 96 –
ICT for Health: Networks, standards and innovation
However, the experience validates that the data that is needed Disease Control Programme (NVBDCP) &National
in the first place for the knowledge management system to Tuberculosis Control Programme (https://nikshay.in/)
work is either not available or recorded and stored in c. Electronic Medical Records (EMRs) including
isolation. Addressing the various issues in the public health the Mother Child Tracking System and Reproductive Child
sector, amongst other factors, include the need for more Health Register
collaborative inter-sectoral engagement, ‘buy-in’ from the d. Clinical decision support system (CDSS):radio-
political authorities and decision makers etc. To resolve diagnostics and laboratories
these concerns, Thailand, in the 1990s, had adopted a e. Computerized physician order entry:secondary
paradigm called "Triangle that Moves the Mountain" [13]. and tertiary care private institutions
‘The Mountain’ means a big and complicated problem, f. Online registry system for patients’
usually unmovable. ‘The Triangle’ consists of: creation of appointments:e-hospitals (ehospital.gov.in)
relevant knowledge through research, social movement or g. Applications based on electronic health records:
social learning, and political involvement. The three PM JAY, national health insurance
components of the triangle have to work together in tandem (https://www.pmjay.gov.in/)
to achieve the goals of resolving the health issues. This has
resulted in the near elimination of the uninsured with its The most popular of these initiatives is the HMIS, which is
universal health coverage, elevated almost a million Thai an online portal that provides information on the human
citizens from poverty while strengthening the capacity of health indicators in the country. It is a tool that provides a
knowledge generation and management [14]. This approach framework for gathering the raw data from primary care
can form a remarkable basis for evolving a structure and health institutions upwards at state level from primary health
systematic approach for building a KM system for any centers (PHCs) onwards. This data is then compiled at block
vertical including health. Some instances of best application level, district level and finally at the state HQ level before
of knowledge management using ICT in public health with feeding into the national level database. Data aggregation
varying degrees of implementation have been observed in a units are at block and district level. The flow of data is
number of countries, notably Canada, Germany, New upwards as well as downwards. HMIS also compiles data
Zealand, South Korea and the U.S.A [14]. from the National Family Health Survey (NFHS), the
District Level Household Survey (DLHS), and the Office of
3. A CASE STUDY FROM INDIA: HMIS AND the Registrar General and Census Commissioner, among
OTHER SYSTEMS HAVING LINKAGES WITH other sources [17]. The information generated from this
PUBLIC HEALTH analysis is then used for taking actions that help in improving
health outcomes. The Online Registry System (ORS) for
India is a country with a vast requirement for a stable public patients is a framework that links various hospitals across the
healthcare system due to its complex health needs of over a country for the Aadhaar-based (biometric digital identity)
billion people with diverse social, economic, geographical online registration and appointment system, where the
and cultural context. Despite its rapid economic growth, it counter-based outpatient department (OPD) registration and
has been ranked 143rdin a list of 188 countries in the ‘Health appointment system through HMIS has been digitized. At the
Related SDG’ index that aims to assess each country’s end of June 2019, 230 hospitals across the country are using
performance across 33 indicators in a global burden of ORS [18].
disease (GBD) study [15]. Also, according to another GBD
study published in the medical journal The Lancet, India has The Integrated Disease Surveillance Programme (IDSP) was
finished 154th among 195 countries on the healthcare index, launched in 2004 by the National Centre for Disease Control
which is based on death rates for 32 diseases that can be (NCDC), India. The program continues under National
avoided or effectively treated with proper medical care [16]. Health Mission with the objective to strengthen/maintain the
These rankings are a cue enough for revamping existing decentralized laboratory-based information technology (IT)-
approaches towards public healthcare systems. Before the enabled disease surveillance system for epidemic-prone
study moves ahead to propose a KMS conceptual model to diseases to monitor disease trends and to detect and respond
address these public health concerns, it might be more to outbreaks in early rising phase through trained rapid
prudent to first enumerate strides that the country has made response teams (RRTs). Under the program, surveillance
in this direction. units have been established in all districts of the country and
it collects data on disease outbreaks for the country as a
At present, a number of health management systems (HMS) whole, excluding non-communicable diseases.
by the public sector are operational in the country that can be
categorized as follows. Similarly, there are other systems being maintained by
different government institutions that are collecting data,
a. Performance reporting portal including the which have implications on PH but are not connected to the
Health Management Information System (HMIS), National HMIS. A case in point is the All India Network Project on
Health Portal (https://www.nhp.gov.in/) andState Health Pesticide Residues (AINPPR). The laboratories under the
Programportals network collect the samples from the nearby Agriculture
b. Disease surveillance portal including Integrated Produce Marketing Corporation (APMC) markets and
Disease Surveillance Programme, National Vector Borne analyze for the possible residues of the pesticide and since
– 97 –
2019 ITU Kaleidoscope Academic Conference
2007-08 and till about 2016 about more than 1 lakh samples research laboratories working in the area of
of various food commodities have been analyzed [19]. Food- pesticides/contaminants/toxicology/pollution, nutrition and
borne diseases (FBDs) not only directly impact human health total diet studies/surveys, etc.). It is therefore of paramount
but also impede socioeconomic development by straining importance that countries like India implement an integrated
healthcare systems and harming national economies, tourism national health stack (NHS) built on the principles of
and trade [20]. Thus, information on pesticide residues in knowledge management systems (KMS) connecting all the
food, as well as food-borne diseases, should be integrated relevant sources. The proposed conceptual model (Figure 1)
with the HMIS. Similarly, information on air pollution, henceforth referred to as the national health stack (NHS) is a
waterpollution, soil pollution etc. being collected by various multilayered and multi-stakeholder model. The main
government authorities (pollution control boards) that has objectives of NHS (Figure1) would be to:
direct or indirect impact on human health should also be
available to PH authorities. However, data being collected by 1. facilitate inter-sectoral involvement and collaboration
these various authorities is “stove piped” into standalone of various stakeholders including government
databases that are not accessible within and across authorities and citizens (Figure 1: referred to as ‘Block1
government agencies. Non-standardized data collection, Data Providers’);
varied data formats, incompatible data IT systems, a sense of 2. develop an integrated KM platform using appropriate
ownership by the group that collects the data are the factors technologies (Figure 1, referred to as ‘Block2 KM
that further worsen the problem [21]. Platform’);
3. evolve policies, regulations and health standards, based
To surmount these issues, India has also renewed its focus on the (public) health predictions made by the related
on the implementation of the proposed ‘National Digital decision making/decision support bodies of the country
Health Blueprint’, the precursor of which can be traced back (Figure 1, referred to as ‘Block3 Decision makers and
to NITI Aayog’s vision document in 2018 laying out the R&D’);
strategy and approach for a national health stack [22]. In 4. spread social learning; NHS to also serve as a tool for
“India’s Trillion-Dollar Digital Opportunity” (pp 122), a spreading awareness on the creation and usage of this
report by the Ministry of Electronics and Information unified approach (Figure 1, Block3).
Technology, Government of India (GoI) [23], there are
examples of actions that are required for improving PH. The The aforementioned four layers (S. No. 1-4) form the basic
need to build an integrated health information platform to design structure of the proposed NHS. However, depending
create and provide access to electronic health records (EHR) on the implementation context of the respective countries,
for every Indian has been highlighted, which would be using more layers can be added to the NHS. Irrespective of the
open APIs. Emphasizing the need of public private number of layers or building blocks in each of these layers,
partnerships (PPP) in the health domain, suggests the the underlying system design principle weaving all of them
development of the PPP model for setting up digital remains the same. This principle is that all data/ information
infrastructure and training for health workers in primary flows emanating to/from each of these blocks/layers of NHS
health centers and other medical care facilities. The report have strong bidirectional feedback loops (Figure 1). Only
also highlights the need of finalization and implementation then the proposed NHS would be able to serve as a common
of the ‘Digital Information Security in Healthcare’ Act repository of data for multiple
(DISHA Act) to provide a framework for the sharing of agencies/authorities/stakeholders. The multisectoral data
health information digitally. The need to frame policy to would be collated, analyzed using advanced ICT techniques
mandate EHR adoption was also highlighted. To catalyze and presented as a ‘visualization layer’, pictorial, user-
implementation of these health aspirations of the nation, the friendly information presented to the decision makers for
Satyanarayana Committee (2018-19), setup by Government strategic planning. Collecting data from different authorities
of India, has been recently tasked with the purpose of will require inter-sectoral cooperation, interoperability and
suggesting a National Digital Health Blueprint so that adequate digital standards and ‘openness’ (to share). Such a
continuum of care could be provided to the citizens. national level KMS will essentially consist of the technology
layer viz.hardware (servers for storage of data, routers for
Thus, it is the right time to design and implement an communication, etc.), software (database for storage of
integrated, comprehensive and effective ICT-based system information, interfaces, etc.), Connectivity (telecom/data
with real-time linkages not only between various public connectivity and related protocols for connecting the various
health authorities but also other authorities that impact public databases and exchange of information) and the application
health. This can be achieved by adopting a holistic layer that would adequately employ a decision support
knowledge management system, as proposed below. system to run data analytics, open APIs etc. (Figure 1,
Block2). The output of these analyses will then be shared
4. PROPOSED CONCEPTUAL MODEL OF NHS: with the related stakeholders (Figure 1, Block3) such as
ICT-BASED KMS FOR PUBLIC HEALTH government, regulatory bodies, at various levels for policy
making, setting standards, regulation making, prospective
As already stated in section 1, PH is a multidisciplinary field planning and building synergies with government plans in an
that requires data/information from multiple sources integrated manner. This synergy can be assured only when
(disease surveillance, health systems, food testing as well as adequate adherence is done to international health standards
– 98 –
ICT for Health: Networks, standards and innovation
such as health level seven (HL7) clinical document adhering to KPIs across identified dimensions and Figure 1,
architecture (CDA), and Integrating the Healthcare Block1).
Enterprise (IHE), Cross-enterprise Document Sharing-b
(XDS.b) profile to build a nationwide secure NHS. The In order to create a robust operational execution strategy for
strategic information emanating from Block3 information communication between Block1 and Block2 (Figure 1), the
can also be shared for research purposes with research journey of the patient from even before the time their
institutions, academics etc. for providing solutions to various exposure might happen to contract a particular disease (the
problems, as well as for supporting social learning activities stage at which prevention is possible) till the time they have
including spreading awareness and conduct of sensitization been completely healed, including diagnosis, treatment,
drives. recovery and follow-up, needs to be thoroughly studied and
mapped with touchpoints of data collection about their health.
5. IMPLEMENTATION OF THE PROPOSED
MODEL
– 99 –
2019 ITU Kaleidoscope Academic Conference
prevention and research side, the stack would also receive Through data collection from a variety of sources, the
data from R&D repositories on health and nutrition of India knowledge management platform will be a treasure trove of
like the Indian Council of Medical Research (ICMR); big data. With the help of artificial intelligence tools, big data
national research institutions like the National Institute of analytics as well as information systems like GIS, this data
Nutrition (NIN); National Centre for Disease Control can be analyzed to extract important insights needed for
(NCDC) implementing the disease surveillance programs of answering relevant questions like reasons for a disease
the government; radio and diagnostic systems, laboratories; outbreak, areas involved, etc. as well as mapping and
and various government departments that have linkages with predicting outbreaks, triggering response mechanisms and
human health including academia. To enable forecasting for taking preventive action. However, there may be a need to
prevention of diseases, effective linkages established democratize this data in a way so as to make it available for
between food control agencies and the public health systems use of machine learning (ML) and AI.
including epidemiologists and microbiologists can provide
information on food-borne diseases, which may be linked to Decision support algorithms employing quantitative data
food monitoring data and lead to appropriate risk-based superimposed on qualitative understanding of local contexts
policies. This information includes annual incidence trends, would help to undertake risk assessments of public health
identification of susceptible population groups, domains. Predictive modeling would help to improve
identification of hazards, identification and tracking of estimates and thereby allow quantification of health risks and
causes of diseases and the development of early warning also find applications for assessing prevention strategies in
systems for disease outbreaks. Therefore, IDSP, HMIS, risk management. The processed data from the stack can be
AINPPR and similar other data emerging from various made available to various stakeholders through open
sources will have to be collected and analyzed centrally by application programming interfaces (APIs).
the knowledge management system (KMS). Once this data
mapping and feeding mechanism is strategized, implemented STEP 4: Use of NHS information for evidence-based
and executed over an extended period of time, the NHS shall decision making, forecasting, planning and research by
act as a centralized health record repository for all citizens. different stakeholders
Once the sources of data are identified by mapping the The insights generated based on the analysis of data can
patient journey, the next step would be to focus on the data provide not only straightforward information that is useful to
formats/databases, and then connecting them. The need for the health functionaries directly but also enable cross-
uniform standards to make multiple EMR systems functional collaboration between various stakeholders
compatible and the information interoperable is paramount (Figure 1, Block 3).
as it will tie up isolated pools of data. A consortium can be
setup consisting of representatives from various consenting For example, information on the immunization status in a
data-sharing stakeholders to identify and list the various particular area can help the health officer to plan resource
current formats being used, come up with short-term allocation of both staff and material for those areas that are
interoperability solutions and envisage long-term data lagging in immunization coverage. On the other hand, cases
sharing standards on common agreed formats. Effective of nicotine toxicity in tobacco harvesters or cases of silicosis
change management would play a pivotal role in aiding the from mining may require collaboration with research
stakeholders to adopt the new agreed formats to process and institutions that can provide technological solutions like
share the data being collected at their end. The costs involved suitable nylon gloves for tobacco farmers or well-designed
in the change can be managed in a way that is offset by the masks for the miners.
overall commercial gains incurred due to the implementation
of the NHS. In terms of channel usage, high speed STEP 5: Social learning: awareness, sensitization and
communication technology is proposed to facilitate data training
collection, analysis and reduce reaction time as well as
enable effective sharing. This digitized data will then be The implementation of a project with an all-encompassing
stored in a central place like a cloud. It will be accessed vision would be meaningful only if stakeholders’ capabilities
remotely by all stakeholders. Also, standards of data security are augmented at all levels ranging from the top till the
need to be strengthened with the use of blockchain ‘bottom of the pyramid’. Political leaders and policy makers
technology so as to protect this data from cyber threats. at the highest level must be encouraged to stay aligned to the
successful culmination of the ‘Health for All’ goal.
This is the most critical step towards building the KMS as it Awareness is equally critical amongst patients whose public
strives to bring together “stove piped” data and needs health data and the related socioeconomic indicators are the
substantial investment of resources not only in terms of funds mainstay of the system. In addition, health data may also be
but also manpower. Here, buy-in from the decision-making crowdsourced from citizens, therefore the citizens need to be
authorities is important as it will drive the project. sensitized about the ‘principle of consent’ with regard to
their health data and personal health records (PHR). Equally
STEP 3: Applying data analytics relevant is capacity building drive for every constituent. As
an example, the capabilities of the grass-root level public
health worker, who is expected to input the information at
– 100 –
ICT for Health: Networks, standards and innovation
the PHC level to the highest level of political leaders and related indicators, as well as their targets. It also provides a
policy makers, all need to be trained according to the federated structure of detailing out interrelated services, their
functions they perform. process flows and the process rules governing these services.
Another differential aspect of this model is clear segregation
6. ISSUES AND OPPORTUNITIES OF THE of the service model and ICT layer which defines this health
PROPOSED MODEL stack as technology-agnostic and relevant at all times.
However, the study is still theoretical rumination by the
Though ICT-based national KMS has a disruptive potential authors and needs to be implemented as a proof of concept
to mainstream preventive public health care, there are before being deemed as a visionary concept.
substantial technical, legal and socioeconomic challenges
that remain to be addressed. Some of the most critical issues 7. CONCLUSION
involved are standardization of the collection methodology,
collection of data, its verification and the identification of Establishment of NHS for public healthcare delivery is a
KPIs across various dimensions. In addition, issues like novel approach to tackle healthcare disparity. New models
interoperability between various databases need to be of data-driven interpretation, forecasting and decision
resolved [24]. Adoption of DSS would permit consistent making facilitated by an ICT-driven KMS can go a long way
dataexchange, robust measurement processes and also lead in establishing evidence-based health systems. The most
to the creation of healthy feedback loops [25]. Similarly, notable attribute of such a KMS would be to enable the
enabling infrastructure that includes robust telecom transformation of the health system from one that is narrowly
connectivity, particularly last-mile, in the context of focused on curing diseases in hospitals by health
developing countries is very important. professionals to a more holistic integrated KMS focused on
examining other aspects that impact human health, like food
Data protection, data privacy, confidentiality and data safety, environmental pollution etc. However, optimal
security are other important issues. The security utilization of ICT-based KMS in healthcare delivery systems
ramifications of personal health data misuse are several and requires overcoming barriers at multiple levels including
cannot be ignored; therefore, the regulatory framework of the standardization of KPIs, databases, processes, technology
country must be stringent. In fact, national/international and policy/regulatory levels. This entails that right from the
consortia and global data communities must ensure that there outset of its design phase, a synergetic cooperation must be
are adequate rights/instruments and related institutions to assured amongst all the disciplines related to public health.
redress grievances if citizens’ personal sensitive (health) data Such a synchronized multi- stakeholder and
is misused anywhere, without their consent. Learning from multidisciplinary collaboration shall provide an increased
the German example, the citizens must be empowered to level of citizens’ confidence in public healthcare systems,
decide, to hide or block any entry in the health record. This which in turn can go a long way to improve the quality of life
can be achieved through awareness and sensitization. ITU-T (QoL) and achieve “Good Health and well being” for all.
Focus Group on Artificial Intelligence 4 Health (FGAI4H)
can also work on the identification of KPIs, collection REFERENCES
methodology and regulatory/policy aspects arising from the
[1] UN SDG 3: United Nations Sustainable
generation of health data and its usage.
Development Goals Good Health and Well-being.
(n.d.). Retrieved from http://in.one.un.org/page/
In terms of opportunities in this domain, wearable sensors
sustainable-development-goals/sdg-3-2
and implantable medical devices (to monitor and transmit
health recorded data) will feature prominently in the future [2] Winslow, Charles-Edward, Amory (1920). "The
of wireless telemedicine systems. There are already many Untilled Field of Public Health". Modern
examples of wearable/implantable medical devices (e.g., Medicine. 2: 183–191.
cochlear implants, cardiac defibrillators/pacemakers, insulin
[3] Van Teijlingen, E. R., Regmi, P. R., Adhikary, P.,
pumps). In the future, interfacing these wireless sensing
Aryal, N., &Simkhada, P. (2019). Interdisciplinary
devices with 5G will present unprecedented opportunities of
Research in Public Health: Not quite straightforward.
crowdsourcing EHR directly into NHS by the patient herself.
Health Prospect, 18(1), 4-7.
Further, the tactile Internet, an ultra-responsive and ultra-
reliable network connectivity, that is envisioned to transmit [4] Zain Abbas, Nadeem Mahmood and Aqil S M
touch and actuation in real time, will pave the way for real- Burney. Article: Information and Communication
time interactive systems being used directly by the citizens Technology in Healthcare Management Systems:
who would be able to access the NHS by employing agile Prospects for Developing Countries. International
APIs [26].The health sector is experiencing technological Journal of Computer Applications 4(2):27–32, July
advancements which are enhancing the government’s ability 2010. Foundation of Computer Science.
in providing a quality health care in an affordable and
[5] Jarosławski, S., Saberwal, G., 2014. In eHealth in
accessible manner.
India today, the nature of work, the challenges and
the finances: an interview-based study. BMC Med.
The unique contributions of this KM model include clear
Inform. Decis. Mak. 14, 1.
identification of service portfolios to meet health goals and
https://doi.org/10.1186/1472-6947-14-1
– 101 –
2019 ITU Kaleidoscope Academic Conference
[6] Pavlopoulos, S., Kyriacou, E., Berler, A., [18] (n.d.). Retrieved July 13, 2019, from
Dembeyiotis, S., &Koutsouris, D. (1998). A novel https://ors.gov.in/index.html
emergency telemedicine system based on wireless
[19] D. (n.d.). Monitoring of Pesticide Residues at
communication technology-AMBULANCE. IEEE
National Level. Retrieved from
Transactions on Information Technology in
http://agricoop.nic.in/sites/default/files/MPRNL
Biomedicine, 2(4), 261-267.
Guidelines_0.pdf
[7] Rajkomar, A., Oren, E., Chen, K., Dai, A. M., Hajaj,
[20] Food safety. (2019, June 04). Retrieved from
N., Hardt, M., ... & Sundberg, P. (2018). Scalable
http://www.who.int/mediacentre/factsheets/fs399/en
and accurate deep learning with electronic health
/
records. NPJ Digital Medicine, 1(1), 18.
[21] Taylor, M.R., and M.B. Batz. 2008. Harnessing
[8] Subhan, I., Ghazi, S. S., & Nabi, S. (2019). Use of
knowledge to ensure Food Safety: Opportunities to
drones (unmanned aerial vehicles) for supporting
Improve the Nation’s Food Safety Information
emergency medical services in India. Apollo
Infrastructure. Gainesville, FL: Food Safety
Medicine, 16(1), 61.
Research Consortium
[9] Adhikary, T., Jana, A. D., Chakrabarty, A., & Jana,
[22] NitiAyog, GoI (n.d.). National Health Stack-
S. K. (2019, January). The Internet of Things (IoT)
Strategy and Approach. Retrieved June,
Augmentation in Healthcare: An Application
https://niti.gov.in/writereaddata/files/document_
Analytics. In International Conference on
publication/NHS-Strategy-and-Approach-Document
Intelligent Computing and Communication
- for-consultation.pdf
Technologies (pp. 576-583). Springer, Singapore.
[23] GoI (n.d.). India's Trillion-Dollar Digital
[10] Davenport, T. H., &Ronanki, R. (2018). Artificial
Opportunity. Retrieved from
intelligence for the real world. Harvard business
https://meity.gov.in/writereaddata/files/india_trillion
review, 96(1), 108-116.
-dollar_digital_opportunity.pdf
[11] (n.d.). Retrieved Spring, 2019, from
[24] Mathur, P., Srivastava, S., Lalch, A., ani, Mehta,
https://www.england.nhs.uk/2019/06/nhs-aims-to-
J.L., 2017. Evolving Role of Telemedicine in Health
be-a-world-leader-in-ai-and-machine-learning-
Care Delivery in India. Prim. Healthc. Open Access
within-5-years/
7, 1–6. https://doi.org/10.4172/2167-1079.1000260
[12] Blaya, J.A., Fraser, H.S.F., Holt, B., e2010. E-
[25] J. Satyanaraya& Charru Malhotra. (2018).
Health Technologies Show Promise In Developing
Universalization and Replication:-Towards a
Countries. Health Aff. (Millwood) 29, 244–251.
consistent service experience- The role of a Digital
https://doi.org/10.1377/hlthaff.2009.0894
Service Standard (DSS) in Citizen Centric
[13] Wasi, P (2000). “Triangle that moves the mountain” Governance. In, DARPG and NASSCOM
and Health System Reform Movement in Thailand”. Publication, ‘Technology for Accelerating
Human Resources for Health Development Journal Development’, Government of India: New Delhi.
(HRDJ) Vol. 4 No. 2 May-August 2000 URL:https://community.nasscom.in/docs/DOC-
1696.
[14] Srivastava, S. K. (2016). Adoption of electronic
health records: a roadmap for India. Healthcare [26] Mittal, M. & Charru Malhotra (2017). e-Health
informatics research, 22(4), 261-269. Initiatives in India - An Analysis from the
Technological Problem. Indian Institute of Public
[15] Lim, S. S., Allen, K., Bhutta, Z. A., Dandona, L.,
Administration. Unpublished.
Forouzanfar, M. H., Fullman, N., ... &Kinfu, Y.
(2016). Measuring the health-related Sustainable
Development Goals in 188 countries: a baseline
analysis from the Global Burden of Disease Study
2015. The Lancet, 388(10053), 1813-1850.
[16] Barber, R.M. et al, 2017. Healthcare Access and
Quality Index based on mortality from causes
amenable to personal health care in 195 countries
and territories, 1990–2015: a novel analysis from
the Global Burden of Disease Study 2015. The
Lancet 390, 231–266. https://doi.org/
10.1016/S0140- 6736(17)30818-8
[17] India Health Management Information System
(HMIS). (n.d.). Retrieved from
http://ghdx.healthdata.org/series/india-health-
management-information-system-hmis
– 102 –
SESSION 5
S5.1 Elderly health monitoring system with fall detection using multi-feature based person
tracking
S5.2 A healthcare cost calculator for older patients over the first year after renal transplantation
S5.3 Automatic plan generating system for geriatric care based on mapping similarity and global
optimization
ELDERLY HEALTH MONITORING SYSTEM WITH FALL DETECTION USING MULTI-
FEATURE BASED PERSON TRACKING
1
Department of Information Technology, Anna University, MIT Campus, Chennai, India
2
National Institute of Information and Communications Technology, Nukui-Kitamachi, Koganei, Tokyo, Japan
– 106 –
ICT for Health: Networks, standards and innovation
ANALYTICS SERVER
VPFD
Angle based
HoG LSTM Fall
Person - Fall
CLIENT Feature Detection
association
Elderly Homes Frame Extraction
selection
H.264 Video based on
SSIM MFPT TARGET ASSOCIATION
through IP
network Person Siamese CNN LSTM for
Detection using for image prediction
YOLO Model similarity
Confidence Score
Existing New
Medical Centre Person Update Person Add new
Appearance, Information Person
Fall detected alert Location Storage
generation
CLOUD STORAGE AND RETRIEVAL
family member, to provide location information of the confidence score has been used as a threshold to eliminate
persons who has fallen. false positive detections. The YOLO model trained on the
COCO dataset [20] has the ability to detect many different
3.1 Key Frame Selection object classes. The output of the model is then filtered to
contain bounding box values of class ‘person’ only.
In the server end, key frame selection has been carried out to
improve the overall processing speed of the system without 3.3 Target Association
degrading its performance. This is achieved by comparing
the previous processed frame and incoming frame using the Target association is the process of mapping the existing
structural similarity index (SSIM), which is calculated by objects with newly detected objects from the current frame.
Equation 1. After preprocessing and person detection stages, one of two
possibilities are tested, either any of the previously moving
(2𝜇𝑋 𝜇𝑦 + 𝑐1 )(2𝜎𝑥𝑦 + 𝑐2 ) persons could have moved to the new position or a new
𝑓(𝑥, 𝑦) = (1) person could have started moving. Using this fact, tracking
(𝜇𝑥2 + 𝜇𝑦2 + 𝑐1 )(𝜎𝑥2 + 𝜎𝑦2 + 𝑐2 )
can be carried out for all persons, who enter and exit the
scene in the video. This architecture performs tracking of the
Where 𝜇𝑋 , 𝜎𝑥2 are the mean and variance of pixels detected persons from the CNN using two distinct features,
in image x respectively and 𝜇𝑦 and 𝜎𝑦2 are the namely the visual feature and motion feature. The visual
mean and variance of pixels in image y, feature denotes the image similarity that helps find whether
respectively. the currently found person matches with the appearance of
one of the existing persons. The motion feature denotes the
The SSIM index value is subjected to a custom threshold to possibility of the existing person moving from his/her
process only dissimilar images by the system. Similar images previous location to the location of the currently detected
are simply skipped for faster video processing. object. Visual and motion features are obtained using
Siamese CNN and LSTM respectively. The utilization of
3.2 Person Detection dual features allows the handling of sudden entry and exit of
persons in the given video.
The decoded frames from the preprocessing stage are given
to the object detector to accurately classify and localize 3.4 Image Similarity
different objects present in each frame. This is achieved with
the help of the CNN-based YOLO model [19]. The given Siamese CNN, shown in Figure 2, is a neural network model
frame is fed as input to the YOLO model, which divides it that operates on a pair of images and produces a score
into segments and finds objects in each segment, along with denoting the appearance similarity between the two images.
their bounding box coordinates and confidence score. The Bounding box coordinates obtained in the previous step
– 107 –
2019 ITU Kaleidoscope Academic Conference
have been used to extract the person’s image from the frame. 3.5 Motion Similarity
The previously stored person’s images are then compared
with the extracted image. The feature extraction layers in Motion prediction has been used in the proposed system in
this CNN network are made the same for both images, thus order to associate objects based on their recent movements.
making it vertically symmetrical. The resultant features of This has been implemented using LSTM on the basis of the
both images are merged by finding the element-wise previous 12 center coordinates of stored persons. This vector
squared difference. It is then fed into fully-connected layers is fed as input to the Motion LSTM model which performs
for dimensionality reduction and finally to obtain the temporal processing and predicts the new center for each
similarity score. stored person. This center indicates the next possible position
– 108 –
ICT for Health: Networks, standards and innovation
– 109 –
2019 ITU Kaleidoscope Academic Conference
– 110 –
ICT for Health: Networks, standards and innovation
CNN 77.1% 15.4% 7.01% 7.5% 70.1% Although fall detection methods based on curvelets and
HMM [7] produce higher accuracy than the optical flow
LSTM 78.96% 14% 8.1% 7.1% 70.8% technique with CNN [8], the proposed technique employing
HoG features in LSTM achieves significantly higher
4.4 VPFD Results accuracy due to an enhanced learning technique.
The UR Fall dataset has been utilized for the training and 5. CONCLUSION
validation phase of the VPFD model. The fall dataset
consists of 30 Fall event videos and 40 normal videos The proposed system is based on the combination of two
containing daily life activities. The ground truth specifies models: MFPT and VPFD to monitor an elderly person’s
whether fall has occurred in each and every frame of the health related activities and report any falls detected through
– 111 –
2019 ITU Kaleidoscope Academic Conference
video surveillance. The system is designed to work within a [8] A. Núñez-Marcos, G. Azkune, and I. Arganda-
confined location such as hospitals, indoor rooms and public Carreras, "Vision-based fall detection with
places. The system has been tested on two different datasets convolutional neural networks", Wireless
of MOT and UR Fall and evaluated the performance of both Communications and Mobile computing, Volume
models. The MFPT model’s precision and accuracy denote 2017, Article ID 9474806, 2017.
the fact that multiple feature-based models help in achieving
higher efficiency. The proposed system achieved 94.67% [9] H. S. Parekh, D. G. Thakore, and U. K. Jaliya, "A
precision in tracking and 98.01% accuracy in elderly fall survey on object detection and tracking methods,
detection. The usage of LSTM model in both the models has " International Journal of Innovative Research in
aided in representing time-series data effectively. The Computer and Communication Engineering vol 2, no.
proposed system for elderly healthcare in homes and 2 pp. 2970-2979, 2014.
hospitals can be standardized in ITU-T Study Group 16,
which is the parent group of Focus Group on Artificial [10] A. Aggarwal, S. Biswas, S. Singh, S. Sural, and A. K.
Intelligence for Health (FG-AI4H). The proposed work can Majumdar, "Object tracking using background
be extended to detect different activities apart from fall subtraction and motion estimation in MPEG videos,"
detection, and recognize and report in the cases of anomalies. Asian Conference on Computer Vision, Springer, pp.
The fall detection module consisting of a HoG feature-based 121-130, 2006.
LSTM training network is the standardization item.
[11] S. Aslani,, and H. M. Nasab, "Optical flow based
REFERENCES moving object detection and tracking for traffic
surveillance," International Journal of Electrical,
[1] World Population Aging Report, United Nations Electronics, Communication, Energy Science and
2017, Engineering” vol. 7, no. 9 pp. 789-793, 2013.
https://www.un.org/en/development/desa/population
/publications/pdf/ageing/WPA2017_Highlights.pdf [12] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner,
"Gradient-based learning applied to document
[2] WHO Guidelines on Integrated Care for Older recognition," Proceedings of the IEEE, vol. 86, no.
People: 11, pp. 2278-2324, 1998.
https://www.who.int/ageing/publications/guidelines-
icope/en/ [13] R. Girshick, J. Donahue, T. Darrell, and J. Malik,
"Rich feature hierarchies for accurate object detection
[3] S. Van der Spek, J. Van Schaick, P. De Bois, and R. and semantic segmentation," IEEE conference on
De Haan, "Sensing human activity: GPS Computer Vision and Pattern Recognition, pp. 580-
tracking," Sensors vol. 9 no. 4, pp. 3033-3055, 2009. 587, 2014.
[4] S. K. Opoku, "An indoor tracking system based on [14] D. Held, S. Thrun, and S. Savarese, "Learning to track
bluetooth technology," arXiv preprint arXiv at 100 fps with deep regression networks," European
1209.3053, 2012. Conference on Computer Vision, pp. 749-765., 2016.
https://arxiv.org/ftp/arxiv/papers/1209/1209.3053.pd
f [15] X. Liu, Y. Zhou, J. Zhao, R. Yao, B. Liu and Y.
Zheng, "Siamese Convolutional Neural Networks for
[5] M. Shankar, J. B. Burchett, Q. Hao, B. D. Guenther, Remote Sensing Scene Classification," IEEE
and D. J. Brady, "Human-tracking systems using Geoscience and Remote Sensing Letters. vol. 16, no.
pyroelectric infrared detectors," Optical 8, pp. 1200-1204, Aug. 2019.
Engineering, vol 45, no. 10, 2006.
[16] ITU focus group on “Artificial Intelligence for
[6] F. Wu, H. Zhao, Y. Zhao, and H. Zhong, Health” https://www.itu.int/en/ITU-
"Development of a wearable-sensor-based fall T/focusgroups/ai4h/Pages/default.aspx
detection system," International Journal of
Telemedicine and Applications, Vol 2015 Article ID [17] N. Dalal, and B. Triggs, "Histograms of oriented
576364, 2015 gradients for human detection,” IEEE Conference on
Computer Vision & Pattern Recognition (CVPR'05),
[7] N. Zerrouki, and A. Houacine, "Combined curvelets vol. 1, pp. 886-893. IEEE Computer Society, 2005.
and hidden Markov models for human fall
detection", Multimedia Tools and Applications vol [18] S. Hochreiter, and J. Schmidhuber, "Long short-term
77 no. 5 pp. 6405-6424, 2018. memory," Neural Computation vol. 9, no. 8 pp. 1735-
1780, 1997.
– 112 –
ICT for Health: Networks, standards and innovation
[19] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, [21] B. Kwolek, and M. Kepski, "Human fall detection on
"You only look once: Unified, real-time object embedded platform using depth maps and wireless
detection," Procs. of the IEEE Conferece on accelerometer," Computer Methods and Programs in
Computer Vision and Pattern Recognition, pp. 779- Biomedicine vol. 117, no. 3 pp. 489-501, 2014.
788. 2016.
– 113 –
A HEALTHCARE COST CALCULATOR FOR OLDER PATIENTS OVER THE FIRST
YEAR AFTER RENAL TRANSPLANTATION
1
Institute of Health Policy, Management and Evaluation, University of Toronto, Canada
We focused on total healthcare costs of transplant recipients We included sensitization indicated by level of peak panel
during the first year after transplantation. In Ontario, Canada, reactive antibodies (PRA) of 0% (not sensitized) or > 0%
renal transplantation is covered for all residents by universal (sensitized), primary cause of ESRD (glomerulonephritis /
public health insurance. Costs were calculated at the patient autoimmune, diabetes, renal vascular, cystic / genetic, or
level across healthcare sectors using a validated, micro- others), and blood type (O, A, B, or AB).
costing, algorithm [15]. We reported costs in Canadian
Dollars (CAD) that were adjusted to 2019 (April) values 2.4.5 Transplant information
using the monthly Consumer Price Index [16], where $1.00
CAD = $0.75 USD [17]. We considered dialysis vintage (pre-emptive transplant or
transplantation without initiating dialysis, or transplant
2.4 Predictors following dialysis duration of <6 months, 6-12 months or >
12 months) and graft number (first graft or re-graft).
We considered a similar set of patient-level predictors as
those examined by Patzer et al. [8] and Tan et al. [9] in their 2.4.6 Pre-transplant healthcare use
respective iChoose Kidney models. Five categories of patient
attributes were collected at transplantation that we have For each patient, we calculated the total healthcare costs for
listed below. during a 6-month workup period before transplant and for
the 12-month period before the start of workup (i.e., pre-
2.4.1 Demographics workup). Costs were measured in 2019 (April) CAD.
We included patient sex (female or male), age (61-70, 71-80, There were missing values found in our dataset for race
or 81+) and race (Caucasian, African American, Asian or (N=98, 7.4%), sensitization (N=237, 17.8%) and primary
Pacific Islanders, or others). cause of ESRD (N=296, 22.3%). In our primary regression
analysis, we imputed Caucasian for those with missing race,
not sensitized (peak PRA = 0%) for those with missing
sensitization, and glomerulonephritis / autoimmune for
– 116 –
ICT for Health: Networks, standards and innovation
patients with missing primary cause of ESRD. We addressed blood type AB and shorter dialysis durations [23]), we
the two variables with significantly missing values (peak applied three regularized linear regression methods.
PRA and primary cause of ESRD) using multiple imputation
methods in sensitivity analysis (see Sensitivity Analysis). 3.1.1 Ridge regression
Reference-based characteristics were male; age 71-80;
Caucasian; from LHIN A; comorbidity-free; not sensitized Ridge regression aims to minimize the sum of the original
(peak PRA = 0%); ESRD caused by glomerulonephritis / loss function 𝑙0 and a regularized term, known as the L2
autoimmune; blood type A; and having received a first graft Norm [24] that we described below. As the parameter 𝜆
pre-emptively. increases, 𝑤𝑗 is shrunk towards 0. Optimal 𝜆 is determined
by ten-fold CV within the training set such that the averaged
3. DATA ANALYSIS test RMSE is minimized.
where y is the log of one-year heathcare costs, 𝑤0 the The regression tree partitions the feature space recursively to
intercept of the equation (i.e., mean log cost for patients with create a tree-like structure [27]. At each split in the tree a
reference-level characteristics), and 𝑤𝑗 the weight associated node is created to ensure maximum homogeneity of the data
being partitioned to the two regions. To train a full tree, we
with predictor 𝑥𝑗 . Conventional ordinary least square (OLS)
selected features and the corresponding thresholds at each
methods search for estimates of 𝑤𝑗 that minimize a loss node such that the squared loss function is minimized:
function that is equal to the total sum of square errors:
𝑁 𝑝 2 ∑ ̂ 𝑥𝑖 )2 + ∑
(𝑦𝑖 − 𝑦 ̂ 𝑥𝑖 )2
(𝑦𝑖 − 𝑦
𝑥𝑖 ∈𝑅1 𝑥𝑖 ∈𝑅2
𝑙0 = ∑ (𝑦𝑖 − ∑ (𝑤0 + 𝑤𝑗 𝑥𝑖𝑗 ))
𝑖=1 𝑗=1
where R1 and R2 denote the two regions separated by the
To overcome potential model overfitting due to node. To avoid overfitting, tree pruning was performed on
multicollinearity among candidate predictors (e.g. having the full tree for a parsimonious tree (T) such that the
following loss function is minimized:
– 117 –
2019 ITU Kaleidoscope Academic Conference
I 177 (13.3%)
|𝑇|
J 82 (6.2%)
∑ ∑ ̂ 𝑥𝑖 )2 + 𝛼|𝑇|
(𝑦𝑖 − 𝑦
𝑚=1 𝑥𝑖 ∈𝑅𝑚 K 68 (5.1%)
L 86 (6.5%)
The size of the final tree, |𝑇|, is penalized by the complexity M 104 (7.8%)
parameter 𝛼 which is determined using ten-fold CV in the N 134 (10.1%)
training set to minimize the averaged RMSE. Comorbidities
CADG 1: Acute minor 906 (68.2%)
3.3 Sensitivity analysis CADG 2: Acute major 1220 (91.9%)
CADG 3: Likely to recur 809 (60.9%)
We conducted analysis to examine the impact of missing CADG 4: Asthma 75 (5.6%)
values on the primary results. Multiple imputation using
CADG 5: Chronic medical, unstable 1310 (98.6%)
Markov-chain Monte Carlo (MCMC) methods, assuming
CADG 6: Chronic medical, stable 873 (65.7%)
missing at random [28], was conducted to address patients
CADG 7: Chronic specialty, stable 51 (3.8%)
with missing peak PRA (N=237, 17.8%) and primary cause
of ESRD (N=296, 22.3%), respectively. We repeated the CADG 8: Eye/dental 218 (16.4%)
imputation procedure ten times [28] and performed the CADG 9: Chronic specialty, unstable 185 (13.9%)
regression analysis on each of the ten newly imputed dataset. CADG 10: Psychosocial 228 (17.2%)
Analyses were performed using R (version 3.5.1). CADG 11: Preventive / administrative 614 (46.2%)
Sensitized (peak PRA > 0%) 548 (41.3%)
4. RESULTS Primary cause of ESRD
Glomerulonephritis / autoimmune 696 (52.4%)
4.1 Baseline characteristics Diabetes 246 (18.5%)
Renal vascular 158 (11.9%)
There are 1328 older deceased-donor renal transplant Cystic / genetic 136 (10.2%)
recipients who survived for at least a year after transplant Others 92 (6.9%)
(Table 1). The majority of these patients received a transplant
Blood type
when aged 61-70 (N=1081, 81.4%), more than three-quarters
O 539 (40.6%)
(N=999, 75.2%) are Caucasian, and over half are male
(N=894, 67.3%). Distribution of older transplant recipients A 541 (40.7%)
amongst the 14 LHINs is imbalanced. Notably, one LHIN B 167 (12.6%)
(LHIN A) is responsible for performing 14.3% (N=190) of AB 81 (6.1%)
all transplantations while another one (LHIN F) accounts for Graft number (first graft) 1261 (95.0%)
less than one-per cent (N=8, 0.6%). Dialysis vintage
Pre-emptive transplant 10 (0.8%)
Table 1 – Characteristics of older transplant recipients at < 6 months 60 (4.5%)
the time of transplantation (N=1328) 6-12 months 72 (5.4%)
> 12 months 1186 (89.3%)
Characteristics Total (N=1328)
Transplant year
Sex (female) 434 (32.7%) 2002 55 (4.1%)
Age 2003 64 (4.8%)
61-70 1081 (81.4%) 2004 57 (4.3%)
71-80 244 (18.4%) 2005 80 (6.0%)
81+ 3 (0.2%) 2006 96 (7.2%)
Race 2007 116 (8.7%)
Caucasian 999 (75.2%) 2008 124 (9.3%)
Asian or Pacific islander 189 (14.2%) 2009 144 (10.8%)
African American 94 (7.1%) 2010 148 (11.1%)
Others 46 (3.5%) 2011 150 (11.3%)
Membership of LHIN (censored) 2012 145 (10.9%)
A 190 (14.3%) 2013 149 (11.2%)
B 160 (12.5%) Costs during transplant workup (CAD)
C 58 (4.4%) Mean ± SD 45460 ± 31271
D 42 (3.2%) Median (IQR) 48971 (55185)
E 50 (3.8%) Costs during pre-workup year (CAD)
F 8 (0.6%) Mean ± SD 75608 ± 71855
G 106 (8.0%) Median (IQR) 70550 (101154)
H 63 (4.7%)
– 118 –
ICT for Health: Networks, standards and innovation
The optimal λ value that minimized test RMSE for elastic net
regression is 0.01024383. Summary of regression results of
the three models is presented in Table 3. Predictors that were
deemed non-significant were denoted by “NS”.
– 119 –
2019 ITU Kaleidoscope Academic Conference
Table 3 – Coefficients of predictors estimated by the three unsensitized patients, sensitized patients were expected to
regularized linear regression models (N=1034) incur 5.0% higher costs during the first year after transplant.
Furthermore, for a one-percent cost increase during the six-
Variables Ridge Lasso Elastic net month workup period and the one-year pre-workup period,
Female 0.0420 NS 0.0198 post-transplant costs at year one are expected to increase by
Age 81+ -0.531 -0.299 -0.350
0.38% and 0.09%, respectively. The lasso model also
identified age 61-70, age 81+, membership of LHIN K, and
Age 61-70 -0.129 -0.0574 -1.717
blood type B as predictors of lower costs at year one.
Asian -0.0399 NS -0.0156 Compared with patients who received a transplant at ages 71-
African American 0.0567 NS 0.0259 80, younger recipients aged 61-70 and older recipients
Other races 0.136 NS 0.101 beyond 80 years of age were found to cost 5.7% and nearly
LHIN B 0.00909 NS NS 30% (29.9%) less at year one, respectively. Furthermore,
LHIN C 0.0493 NS NS patients with blood type B were found to incur 1.6% less
LHIN D 0.0993 NS 0.0320 costs compared to type A patients during the first post-
LHIN E 0.0514 NS NS transplant year.
LHIN F -0.183 NS -0.0850
The elastic net regression model concluded a total of 33
LHIN G 0.00247 NS NS
significant predictors of one-year costs, including all of the
LHIN H 0.0607 NS 0.00906 nine predictors identified by the lasso regression.
LHIN I 0.0769 NS 0.0347
LHIN J 0.130 0.00242 0.0833 4.2.2 Regression tree
LHIN K -0.0996 -0.0317 -0.0962
LHIN L 0.0557 NS NS Figure 3 shows the regression tree model trained by patients
LHIN M -0.0393 NS -0.0178 who underwent transplantation between 2002 and 2011. The
LHIN N -0.0233 NS -0.0143 log of pre-workup (logpre) and workup (logwork) costs were
CADG 1 0.0110 NS NS
identified as the only two predictors of the mean log of costs
during the first year after transplantation (logtarget). The
CADG 2 -0.0424 NS -0.0176
regression rules used are as follows: (1) patients who
CADG 3 0.0259 NS 0.0169 incurred at least $10938 during workup (logwork >= 9.3, i.e.,
CADG 4 -0.00101 NS NS having logged workup costs of at least 9.3) were expected to
CADG 5 -0.208 NS -0.146 cost an average of $59874 during the first post-transplant
CADG 6 -0.00439 NS NS year (logtarget = 11); (2) patients who incurred less than
CADG 7 -0.0432 NS 0.0000955 $10938 during workup (logwork < 9.3) but at least $4024
CADG 8 0.0322 NS 0.0151 during pre-workup (logpre >= 8.3, i.e., having logged pre-
CADG 9 0.0344 NS 0.00798 workup costs of at least 8.3) were expected to cost an average
of $22026 over the first post-transplant year (logtarget = 10);
CADG 10 -0.00708 NS NS
(3) patients who incurred less than $10938 during workup
CADG 11 -0.0310 NS -0.0171
(logwork < 9.3) and less than $4024 during pre-workup
Peak PRA > 0% 0.0862 0.0501 0.0870 (logpre < 8.3) were expected to cost an average of $2981
ESRD: Diabetes 0.0833 0.00617 0.0600 over the first year post-transplant (logtarget = 8).
ESRD: Renal vascular 0.0645 NS 0.0367
ESRD: Cystic/genetic -0.0737 NS -0.0485
ESRD: Others 0.0346 NS 0.00333
Blood type B -0.0895 -0.0155 -0.0713
Blood type AB 0.0870 NS 0.0473
Blood type O 0.0122 NS NS
Re-graft 0.0926 NS 0.0610
Dialysis < 6 months 0.0912 NS 0.00205
Dialysis 6-12 months 0.0224 NS NS
Dialysis > 12 months -0.0377 NS -0.00224
Log of workup cost 0.333 0.382 0.393
Log of pre-workup cost 0.139 0.0887 0.0983
Intercept 6.248 6.037 6.010
Figure 3 – Regression tree
The lasso model identified nine predictors of the log of post-
transplant costs at year one. Membership of LHIN J,
sensitized (peak PRA > 0%), having diabetes as the primary
cause of ESRD, higher pre-workup and workup costs were
found to heighten the cost. Notably, compared with
– 120 –
ICT for Health: Networks, standards and innovation
4.3 Model validation calculators for older adults with chronic diseases. A further
implication of this finding points to the importance of
We validated the four models using ten-fold CV on continuously monitoring an individual’s spending on
transplant recipients during the years 2012 and 2013 (N=294, healthcare. In this way, our calculator serves as a simple
22.1%). Results are summarized in Table 4. Ridge regression, prototype for a more advanced algorithm that is capable of
lasso regression, elastic net regression and regression tree continuous cost prediction.
achieved an averaged test RMSE of 0.618, 0.604, 0.610 and
0.630, respectively, while reaching averaged test R2 values There are limitations associated with our study. First, unlike
of 0.255, 0.258, 0.251 and 0.0101. Hence, we concluded Patzer et al. and Haddad et al., we did not consider patient
lasso regression to have the best performance. ethnicity (i.e., Hispanic or non-Hispanic) in our regression
analysis. However, as Tan et al. pointed out, less than 0.5%
Table 4 – Testing results of the four models (N=294) of Canadians identify as Hispanic [9]. Second, we did not
have access to donor-level information, including donor age
Models RMSE R2 and cause of death, as well as facility-level factors, including
Ridge regression 0.618 0.255 hospital bed size and type of admission. This may explain
the relatively low test R2 (0.258) achieved by our model.
Lasso regression 0.604 0.258
Future investigators with a more comprehensive tracking of
Elastic net regression 0.610 0.251
patients may provide additional insights on predictors of
Regression tree 0.630 0.0101 post-transplant costs. Third, accuracy of our primary analysis
is limited by variables with significantly missing values,
4.4 Sensitivity analysis especially peak PRA (N=237, 17.8%) and primary cause of
ESRD (N=296, 22.3%). However, through extensive
We iterated the multiple imputation methods ten times to sensitivity analysis based on multiple imputation methods
impute values for patients with missing peak PRA (N=237, we were able to rule out such potential bias caused by these
17.8%) and primary cause of ESRD (N=296, 22.3%). missing values.
Training and testing procedures were repeated for each
newly imputed dataset. The lasso regression model had the Our study has some key strengths. First, use of a linked
best performance in each iteration, with an averaged test administrative dataset has enabled us to have comprehensive
RMSE of 0.611 (SD, 0.144) and R2 of 0.257 (SD, 0.100) over tracking of older renal transplant recipients from the year
the ten iterations. The same set of eight predictors were before transplant workup to death. Second, we used CADGs
identified, with the weights of having diabetes as the primary to characterize comorbidities at the time of transplantation,
cause of ESRD and sensitized (peak PRA > 0%) being which enabled us to arrive at conclusions that are specific to
enlarged. Specifically, the estimated coefficient of diabetes disease type. Third, we were able to construct person-level
increased from 0.00617 in the original analysis to an healthcare costs across healthcare sectors, which gave us
averaged 0.0125 (SD, 0.0123) over the ten iterations. transplant recipients’ precise use of healthcare, both before
Meanwhile, the estimated coefficient of sensitized also rose and after transplantation. Fourth, through our use of
from 0.0501 to 0.0823 (SD, 0.032). machine-learning techniques, we were able to identify
predictors of post-transplant healthcare use while
5. DISCUSSION overcoming potential overfitting due to multicollinearity, a
common threat to conventional multivariate regression
In the present study, we used machine-learning methods to analysis.
develop a cost calculator for deceased-donor renal transplant
recipients aged above 60 over the first post-transplant year. 6. ACKNOWLEDGEMENT
The final calculator was based on a lasso linear regression
model and required the following inputs to be collected at the This study was supported by the Institute for Clinical
time of transplantation: age, membership of one of 14 Evaluative Sciences (ICES), which is funded by an annual
regionalized LHINs, blood type, sensitization, having grant from the Ontario Ministry of Health and Long-Term
diabetes as the primary cause of ESRD, and healthcare costs Care (MOHLTC). The opinions, results and conclusions
during the six-month transplant workup and during the year reported in this paper are those of the authors and are
before workup. This cost calculator minimized test RMSE at independent from the funding sources. No endorsement by
0.604 while achieving an acceptable test R2 of 0.258. The ICES or the Ontario MOHLTC is intended or should be
results are robust to missing values found in our dataset. inferred.
– 121 –
2019 ITU Kaleidoscope Academic Conference
[2] A. Laupacis et al., "A study of the quality of life and [13] K. D. Yoo et al., "A machine learning approach
cost-utility of renal transplant," Kidney using survival statistics to predict graft survival in
International, vol. 50, no. 1, pp. 235–242, 1996. kidney transplant recipients: a multicentre cohort
study," Scientific Reports, vol. 7, no. 1, p. 8904,
[3] R. A. Wolfe et al., "Comparison of mortality in all 2017.
patients on dialysis, patients on dialysis awaiting
transplantation, and recipients of a first cadaveric [14] Canadian Institute for Health Information,
transplant, " New England Journal of Medicine, vol. "Canadian Organ Replacement Register Metadata
341, no. 23, pp. 1725–1730, 1999. (CORR)," [Online]. Available:
https://www.cihi.ca/en/canadian-organ-replacement-
[4] American Kidney Fund, "Waiting for a kidney register-metadata-corr
transplant," [Online]. Available:
http://www.kidneyfund.org/kidney-disease/kidney- [15] W. P. Wodchis, K. Bushmeneva, M. Nikitovic, and
failure/treatment-of-kidney-failure/kidney- I. McKillop, "Guidelines on person-level costing
transplant/transplant-waitlist/ using administrative databases in Ontario," Health
System Performance Research Network Working
[5] Trillium Gift of Life network, "Trillium Gift of Life Paper Series, vol.1, 2013. Available:
Network annual report 2017/18," Available: http://www.hsprn.ca/uploads/files/Guidelines_on_Pe
https://www.giftoflife.on.ca/resources/pdf/TGLN_A rsonLevel_Costing_May_2013.pdf.
nnual_Report_17_18_EN_WA_F.pdf .
[16] Statistics Canada, "Table 18-10-0005-01 Consumer
[6] J. F. Whiting et al., "Economic cost of expanded Price Index, annual average, not seasonally
criteria donors in cadaveric renal transplantation: adjusted,"[Online]. Available:
analysis of Medicare payments," Transplantation, https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pi
vol. 70, no. 5, pp. 755–760, 2000. d=1810000501.
[7] S. V. Jassal et al., "Kidney transplantation in the [17] Bank of Canada. "Monthly exchange rates,"
elderly: a decision analysis," Journal of the [Online]. Available:
American Society of Nephrology, vol. 14, no. 1, pp. https://www.bankofcanada.ca/rates/exchange/monthl
187–196, 2003. y-exchange-rates/
[8] R. Patzer et al., "iChoose Kidney: a clinical decision [18] Local Health Integration Network. "Local Health
aid for kidney transplantation versus dialysis Integration Network (LHIN)," [Online]. Available:
treatment," Transplantation, vol. 100, no. 3, pp. http://www.lhins.on.ca/
630–639, 2016.
[19] The Johns Hopkins University, "The Johns Hopkins
[9] V. S. Tan et al., "Predicting 3-year survival in ACG® System Excerpt from Version 11.0 Technical
patients receiving maintenance dialysis: an external Reference Guide, " 2015. [Online]. Available:
validation of iChoose Kidney in Ontario, Canada," https://www.healthpartners.com/ucm/groups/public/
Canadian Journal of Kidney Health and Disease, @hp/@public/documents/documents/cntrb_035024.
vol. 5, pp. 1–11, 2018. pdf
– 122 –
ICT for Health: Networks, standards and innovation
[23] R. Fu, S. J. Kim, C. de Oliveria, P. C. Coyte, "An [26] H. Zou and T. Hastie, "Regularization and variable
instrumental variable approach confirms that the selection via the elastic net," Journal of the Royal
duration of pretransplant dialysis has a negative Statistical Society. Series B (Methodological),
impact on the survival of kidney transplant vol.67, no. 2, pp. 301-320, 2005.
recipients and quantifies the risk," Kidney
International, vol. 96, no. 2, pp. 450-459, 2019. [27] L. Breiman, J. Friedman, C. J. Stone, R. A. Olshen,
Classification and Regression Trees, 1st ed.
[24] A.E. Hoerl and R. W. Kennard, "Ridge regression: Chapman and Hall/CRC, 1984.
biased estimation for nonorthogonal problems,"
Technometrics, vol. 12, no. 1, pp. 55-67, 1970. [28] J. Sterne et al., "Multiple imputation for missing
data in epidemiological and clinical research:
[25] R. Tibshirani, "Regression shrinkage and selection potential and pitfalls," BMJ, vol. 338, p. b2393,
via the Lasso," Journal of the Royal Statistical 2009.
Society. Series B (Methodological), vol. 58, no. 1,
pp. 267-288, 1996.
– 123 –
AUTOMATIC PLAN GENERATING SYSTEM FOR GERIATRIC CARE
BASED ON MAPPING SIMILARITY AND GLOBAL OPTIMIZATION
• Self-repairing artificial fish swarm algorithm: Owing to multi-objective knapsack problem, a global optimization
the stronger optimization ability and faster convergence algorithm is urgently needed to search for an optimized SDSP,
speed, we proposed the self-repairing artificial fish which promotes the accuracy of smart services and reduces
swarm algorithm(SAFSA) to search for an optimized the cost of sensing devices. For a multi-objective knapsack
SDSP by limiting artificial fish moving near the problem, people have put forward valuable methods, mainly
constraint boundary by self-repairing behavior. divided into two types: (1) heuristic algorithms, such
as greedy algorithm, dynamic programming algorithm,
The rest of this paper is organized as follows: we briefly simulated annealing algorithm and so on; (2) swarm
review the past research on the in-home health-care system intelligent optimization algorithm: genetic algorithm [17],
in the smart home, and intelligent optimization algorithms particle swarm optimization algorithm [18], ant colony
in section 2. We propose the framework of the APGS in algorithm [19] and so on. Because of the low efficiency
section 3. The Smart-desire mapping method is described in and slow convergence rate for large-scale problems, heuristic
section 4 and self-repairing artificial fish swarm algorithm is algorithms are replaced by swarm intelligent optimization
presented in section 5. We perform a series of experiments algorithm. However, classical intelligent optimization
to verify the scientificity and validity of SDSP in section 6. algorithms will usually plunge into local optimization and
The conclusion and future work are discussed in section 7. are sensitive to the initial parameters. For example, the
artificial fish swarm may plunge into local optimization if
2. RELATED WORKS the visual of artificial fish is too small; additionally, changes
of crossover and mutation probabilities of genetic algorithm
Recently, owing to the development and advancement of the
will result in different genetic speeds. Therefore, in order to
Internet of things and digital health, research in the smart
generate the optimized SDSP for geriatric care, we present a
home becomes increasingly important. A smart home, in
self-repairing artificial fish swarm algorithm, which introduce
which artificial intelligence techniques control home settings,
self-repairing behavior to limit artificial fish searching for an
collects data by sensors when residents perform their normal
optimized solution near the constraint boundary.
daily routines. Since sensors can collect data in a naturalistic
way without modifying an individual’s behavior, the smart
home provides a new way for automated health care. The 3. THE FRAMEWORK OF AUTOMATIC PLAN
survey in [10] showed that all participants had positive GENERATING SYSTEM
attitudes towards the technology of the smart home and were
willing to accept the installation of sensing devices in their As shown in Figure 1, we formalize user care demands
homes. Afterwards, more researchers were committed to and expert knowledge into digital description based on
providing health care services for users of various ages. Portet expert diagnosis, medical literature, and clinical diagnosis.
et al. [11] showed that inexpensive smart home technologies Additionally, smart services for geriatric care are extracted
could be used for the purpose of self-monitoring of safety, from recent researches. In this framework, Automatic Plan
health and functional statuses in existing homes, and are Generating is the key module including two submodules: (1)
urgently required. Nehmer et al. [12] used the smart home Smart-desire module is proposed to extract required smart
to provide a better assistant system in health monitoring and services for geriatric care by decomposing care demands
to improve the quality of life of elderly and disabled people. into atomic demands, calculating functional similarity and
Skubic et al. [13] provided passive sensor networks to capture non-functional similarity between atomic demands and smart
patterns representing physical and cognitive health conditions service; (2) Global optimization module, in which the SAFSA
in an aging in place elderly-care facility. Additionally, Mario is proposed to search for the optimized solution of SDSP
et al. [14] proposed a software architecture that modeled the for geriatric care based on cost evaluation and performance
functionalities of a smart home platform to deploy sensitive evaluation.
services into the digital home for health care.
However, due to the high costs and untrusted design, the Sensing Device
Selection Plan
Decoding Optimized
Solution
smart home failed to make headway in the field of health Elderly User
Literature and Clinical Diagnosis
Smart-Desire
Intuitionistic Fuzzy
Acquisition
TF-IDF Statistics
Cosine similarity
Calculation
Researches
Multi-objective
Knapsack Probem
the decision process of human experts [16], the expert Smart Services Automatic Plan Generating
knowledge learned in geriatric diagnosis was adopted to
decompose user demands into atomic demands. Thereafter, Figure 1 – The framework of APGS
we extracted smart services based on functional similarity and
QoS similarity between atomic demands and smart services. More narrowly, user demands are decomposed into atomic
As the selection of sensing devices is regarded as a care demands automatically by traveling expert knowledge
– 126 –
ICT for Health: Networks, standards and innovation
demands and smart services. In order to make the selection Service Layer
Smart Smart Smart
of sensing devices unambiguous, we convert the problem Service Service
S2
Service
S3
S1
of selecting suitable sensing devices to realize selected smart
services into a multi-objective knapsack problem of searching
Device Layer
Pressure Sound Infrared Smart Switch
ĊĊ ĊĊ
for the optimized solution. In this paper, SAFSA is selected Sensor Sensor Sensor Bracelet Sensor
• User layer: The concept of the user layer is used De = {I D, Label, Description, Children, QR}
to sort and formalize key information of diseases
description, natural language care needs, and other
living requirements of elderly.
4.1.2 Service Layer
• Demand layer: After we extract expert knowledge
in expert diagnosis, medical literature and clinical In order to obtain the suitable smart services for geriatric
diagnosis of geriatric diseases, the demand layer fuses care automatically, we need to formalize smart services into
user demands and expert knowledge. Specifically, care digital description in the service layer, including ID, Label,
demands are divided into composite care demands and Description and Quality of Service(QoS). Therefore, smart
atomic care demands. services in this paper are formalized with three modules.
• Service layer: It is proposed to formalize technologies 1. Definition: The definition of the smart service indicates
and services in recent research into smart services with a the functional description for geriatric care, including
fixed format, including name, label, description, quality ID, Label, Description.
of service and input requirement.
2. QoS: This represents the non-functional parameters of
• Device layer: This layer formalizes digital descriptions the smart service, including accuracy, response-time,
of sensing devices, including price, precision, measuring smart-level, availability and reliability.
range and so on, which are important in analyzing the
ability and performance of sensing devices. 3. Input requirements: It is proposed to summarize the
hardware requirements of the smart service.
Generally, care demands are extracted in expert diagnosis Knowledge excavation is the mining of potential patterns
of geriatric diseases, which represents the real care needs and behaviors by highly automated analysis of legacy data
of elderly for 24-hour geriatric care in the smart home, to help people make the right decisions. The expression of
including diet care, sport care, daily care, danger warning expert knowledge directly affects the efficiency of knowledge
– 127 –
2019 ITU Kaleidoscope Academic Conference
reasoning and the ability to acquire new knowledge. However, similarity between keyword xi and y j .
due to the complex care demands of various geriatric
diseases, the traditional knowledge expression methods are x1 y1 x1 y2 · · · x1 ym
not suitable for care demands decomposition, such as logical x2 y1 x2 y2 · · · x2 ym ®
© ª
representation, semantic network, rule-based system and so SXY = . .. .. .. ®® (3)
..
. . . ®
on. Therefore, we designed the expert knowledge forest to
« xn y1 xn y2 · · · xn ym ¬
combine care demands and expert knowledge based on the
decision tree structure and the frame knowledge expression. According to [20], the semantic similarity between keywords
In the expert knowledge forest, every frame consists of five xi and y j is calculated based on path length, depth and local
expert knowledge components: (1) name, a unique name density as shown in Equation (4), where l represents the path
that can be any constant; (2) slot, a combination of expert length and h represent the path depth. Simultaneously, α
knowledge and care demand; (3) slot value, the attribute is a constant and β is a smoothing factor where β > 0. In
value, which can be 0 or 1; (4) relation, the knowledge this paper, we set α = 0.2 and β = 0.6 to generate optimal
associations between frames; (5) slot constraints, related semantic similarity, as details are shown in [20].
constraints contributed to the corresponding slot value.
eβh − e−βh
s xi , y j = e−αl βh
(4)
4.2 Smart-desire model e + e−βh
In order to normalize the fuzzy similarity matrix, we
Due to the limitations of family space and expenditure, compress it into one dimension by taking the maximum value
designers need to extract suitable smart services to provide for each row of the matrix, and average these maximum values
geriatric care for the elderly in the smart home. In this paper, by Equation (5). However, s_sem (X,Y ) only represents the
we proposed a Smart-desire model to extract smart services average semantic similarity between the vector B and every
based on expert knowledge forest and mapping similarity. word in A. Thus, we use Equation (6) to calculate the semantic
As shown in Definition 2, Sdesir e is the output of SDMM, similarity between vector X and vector Y .
and PU E D is proposed to decompose care demands(D)
n
by traveling expert knowledge forest (E) based on user 1 Õ
s_sem (X,Y ) =
× max xi y j ; j ∈ [1, m] (5)
information(U). Additionally, in the Smart-desire model, SS n i=1
is a set of all smart services in the service layer, and ϕ (x∗, y∗ )
is the mapping method in SDMM. (s_sem (X,Y ) + s_sem (Y, X))
s_sem (|X,Y |) = (6)
2
Definition 2 Smart-Desire is presented to obtain Sdesire Therefore, the functional similarity between care demand c
based on user information and expert knowledge. and smart service s is calculated by Equation (7), where ϕ1
and ϕ2 are weight coefficients of service name and service
Smart_Desire = {SS, PU E D , ϕ (x∗, y∗ ) , Sdesir e } . description, with ϕ1 +ϕ2 =1.
Õ
Sdesire = ϕ (PU E D , SS) = Si .
f _sim(|c, s|) =ϕ1 × s_sem (|c_Description, s_Description|)
ϕ2 × s_sem (|c_Label, s_Label |)
4.3 Mapping similarity calculation method (7)
– 128 –
ICT for Health: Networks, standards and innovation
Table 1 – The fuzzy number of scoring set Table 2 – Unknown data for global optimization
Then, the cosine similarity is used to calculate the As opposed to web services and cloud services, the cost
non-functional similarity between care demand ci and smart of smart services depends on the price of sensing devices.
service s j , as shown in Equation (12), where Q 0 represents However, two different services may require the same sensing
the normalized QR and Q is the normalized QoS. device. For example, the Activity Monitoring service requires
the acceleration data and the Falling Detection service
requires the same data. Therefore, we need to reduce
5 repeatable sensing devices of selected services.
q 0 k × qk
Í
Q Q0 k=1 Definition 3 As a sensing device in smart service work in a
q_sim(ci , s j ) = =s (12)
||Q|| × ||Q 0 ||
s
5 5 specific position, we formalize every required sensing device
2
0 qk2
Í Í
qk× with its t ype and position, as di = {t ype, position}.
k=1 k=1
In Definition 3, t ype represents the type of the sensing device
Finally, the mapping similarity between atomic care demand and position is the installation position. Therefore, we count
ci and smart service s j is calculated by Equation (13), based the quantity of every type of sensing device with two rules:
on functional similarity and non-functional similarity, where
γ1 and γ2 are the weight coefficients with γ1 + γ2 = 1. 1. If di and dj from different services in Sdesir e have
the same d_t ype and d_position, we suppose that one
sensing device is enough for both smart services. Thus,
m_sim(ci , s j ) = γ1 × s_sim(ci , s j ) + γ2 × q_sim(ci , s j ) (13) the quantity of this type of sensing devices is unchanged,
while the weight is increased by one.
5. GLOBAL OPTIMIZATION ALGORITHM 2. If di and dj from different services in Sdesir e have the
same d_t ype but different d_positions, only one sensing
5.1 Transformation of sensing devices selection
device is not enough. Hence, the quantity and weight of
In order to improve service performance and reduce the cost this type of sensing devices are both increased by one,
of sensing devices, we first need to count the requirement of and two installation locations are added into the L.
sensing devices for geriatric care. Suppose that there are A In this paper, we adopt Formula (15) to evaluate the
types of sensing devices and B commodities for each type of performance of sensing devices, where ω is the compensation
sensing devices. The problem of selecting sensing devices coefficient and ξ is the general error. Additionally, y is the
can be converted into a multi-objective knapsack problem, service life of sensing devices, r is the measuring range and
if we can calculate the unknown data in Table 2. Then, we r is the average measuring range. η (η ∈ (0, 1)) indicates the
establish a global optimization selection model to maximize effect of sensing devices in smart services.
the total performance and to minimize the total cost of sensing
devices, as shown in Equation (14). Therefore, this problem ωy r
r
is a NP-complete problem like the knapsack problem. R=η (15)
ξ r
– 129 –
2019 ITU Kaleidoscope Academic Conference
5.3 Self-repairing artificial fish swarm algorithm objective functions Y1 , as shown in Equation (17):
Definition 4 The artificial fish X f is located near the Y = Max{λ1 w1Y1 + λ2 w2Y2} (18)
constraint boundary of the problem, if X f is a feasible
solution and it would become infeasible when we find a 1
w1 = (19)
commodity k in type i and assign k to xi , where i ∈ Max{Y1}
{1, 2, 3...A}, k ∈ {1, 2, 3...B} and the Performance/Price of 1
kth commodity is bigger than xi ’s in type i sensing devices. w2 = (20)
Max{Y2}
The research [22] proves that an optimized solution and 6. EXPERIMENTS AND RESULTS ANALYSIS
constraint boundary of the knapsack problem are usually
symbiotic. Thus, we proposed a self-repairing strategy In order to verify the accuracy and validity of the proposed
to repair artificial fish return to the constraint boundary. APGS, we formalized 40 care demands and 200 smart
For any infeasible solution X f , there are two self-repairing services for geriatric care, including hypertension, heart
strategies, where Operation (1) on xi with the least value of diseases and diabetes. Firstly, we need to verify the
Performance/Price and Operation (2) on xi with the biggest scientificity and validity of selected services for geriatric care
value of Price: in experiment 1; then, the global optimization capability and
Operation (1): Replacing xi with k, if Performance/Price of convergence rate of SAFSA should be verified in experiment
kth commodity is biggest in the same type of sensing devices 2; finally, we built the evaluation indicator system of SDSP
whose Price is less than the Price of xi , where i ∈ {1, 2, 3...A} to verify the scientificity and validity of SDSP of the smart
and k ∈ {1, 2, 3...B}. home, based on evaluations of smart home researchers.
Operation (2): Replacing xi with k, as the Price of kth In the first experiment, we listed care demands for geriatric
commodity is least in the same type of sensing devices whose experts to select needed care demands for 100 sample elderly,
Performance/Price is bigger than xi , where i ∈ {1, 2, 3...A} suffering from hypertension, diabetes and heart diseases.
and k ∈ {1, 2, 3...B}. Weight parameters in SDMM are set as: ϕ1 =ϕ2 =0.5, γ1 =0.7
and γ2 =0.3. Then, we recommended smart services, whose
Proof 1 If the infeasible artificial fish X f becomes a feasible mapping similarity are bigger than 0.8, for geriatric experts to
solution after the last Operation (1), we can conclude that manually select suitable smart services
Ñ to meet care Ñ demands.
the Price of kth sensing device is less than xi . Assume that Ultimately, we use Precision( C C D ), Recall( C D D ) and
the artificial fish X f is not near the constraint boundary. F1( 2Pr ecision+Recall ) to verify the validity of selected
∗Precision∗Recall
According to Definition 4, there is a sensing device whose services, where C is the service set calculated by proposed
Performance/Price is bigger than k and Price is less than SDMM and D is the service set selected by experts.
xi . However, due to the description of Operation (1), the We compared the performance of mapping similarity method
value of Value/Price of k is biggest in the candidate sensing (MS), keyword mapping method (KM) and variable precision
devices whose Price is less than Price of xi . Therefore, the rough set method (VPRS). As shown in Figure-3(a), the
assumption is not true and X f is near the constraint boundary precision is more than 94.4%, which indicates that selected
after Operation (1) of self-repairing behavior. smart services can excellently cover care demands of the
sample elderly. Compared to KM and VPRS methods, our
As the above Proof 1 reveals, the artificial fish X f approach has higher accuracy and coverage, since precision,
become feasible solution after Operation (1). Additionally, recall and F1 of the proposed MS are almost the biggest.
Operation (2) can be similarly proved with the same method. The similar function and description between smart services
Therefore, we properly invoke self-repairing behavior to are important reasons causing 4.5% of the mismatch. In
search for the optimal solution, when the artificial fish X f Figure-3(b), the performance of MS for elderly suffering
become infeasible after four basic behaviors. from diabetes is better than hypertension and heart diseases.
Since two objective functions of the multi-objective knapsack In particular, the performance of MS for a single geriatric
problem have opposite targets, we make a conversion for the disease is usually better than two or more geriatric diseases.
– 130 –
ICT for Health: Networks, standards and innovation
(a) (b)
30%
20%
10%
0%
Figure 3 – (a)The results of MS, KM and VPRS; (b) the EC UR Acc RS IT AR
40 33.9
0.86
27.8 26.7 28.2
30
0.81
20
0.76
SAFSA AFSA PSO GA 8 9
Average Solution 0.899 0.893 0.873 0.851 10 3 4
Optimal Solution 0.904 0.904 0.899 0.897
0
Worst Solution 0.894 0.878 0.846 0.798
SAFSA AFSA PSO GA
(a) (b)
Figure 6 – The prototype system of APGS.
Figure 4 – (a) The optimal values of optimization algorithms;
(b) The cycle times for obtaining the optimal solutions.
7. CONCLUSIONS AND FUTURE WORK
Finally, we built an evaluation indicator system of SDSP to
verify the scientificity and validity of SDSP of the smart Due to the complexity and waste of resources in traditional
home, based on evaluations of smart home researchers, SDSP design, the smart home is not widely used in health
as shown in Table 3. Then, we presented 20 SDSP to care. In order to promote the role of the smart home in the
conduct surveys of 20 smart home researchers. As shown field of health care, we proposed an APGS to generate the
in Figure 5, the average score is bigger than ‘9’ of the SDSP for the smart home automatically and efficiently based
three secondary indexes of smart services and showed that on the Smart-desire mapping method and Self-repairing
SDSP has better accuracy, faster response times and higher artificial fish swarm algorithm. Ultimately, experiments
intelligence levels. However, more than half of researchers of elderly suffering from hypertension, diabetes and heart
gave a ‘4-6’ score to ‘Energy Consumption’, mainly because diseases verified the scientificity and validity of SDSP. Our
the energy consumption was not taken into account to evaluate work provides a novel approach to designing a user-oriented
the performance of sensing devices. smart home efficiently and automatically, which could reduce
the labor costs and make the design pattern more transparent
Table 3 – The evaluation indicator system of SDSP
and reliable. Furthermore, an accurate SDSP is very helpful
Primary Index Secondary Index Scores for promoting the use of the smart home in geriatric care. In
Energy Consumption(EC) 1-10 future work, many standards in the field of geriatric care and
Sensing Devices smart services will be considered to standardize the design
Utilization Rate(UR) 1-10
Accuracy(Acc) 1-10 of the smart home, since the standards can not only help
Smart the elderly understand smart services, but also provide an
Services Response Speed(RS) 1-10
Intelligence Level(IL) 1-10 aid for the application and promotion of our method. Then,
Installation a new method of assigning intelligence to sensing devices
& Deployment Accuracy Rate(AR) 1-10 by software-defining intelligence is necessary to improve the
coding efficiency in smart home design. We will commit
Summarily, we implemented a prototype system to extract the to providing user-oriented geriatric care for the elderly in the
– 131 –
2019 ITU Kaleidoscope Academic Conference
smart home to greatly improve the quality of life of the elderly [12] Jürgen Nehmer, Martin Becker, Arthur Karshmer, and
in our future work. Rosemarie Lamm. Living assistance systems: an
ambient intelligence approach. In Proceedings of the
REFERENCES 28th international conference on Software engineering,
pages 43–50. ACM, 2006.
[1] Christopher B Newgard and Norman E Sharpless.
Coming of age: molecular drivers of aging and [13] Marjorie Skubic, Gregory Alexander, Mihail Popescu,
therapeutic opportunities. The Journal of clinical Marilyn Rantz, and James Keller. A smart home
investigation, 123(3):946–950, 2013. application to eldercare: Current status and lessons
learned. Technology and Health Care, 17(3), 2009.
[2] Sharon R Kaufman. Old age, disease, and the discourse
on risk: geriatric assessment in us health care. Medical [14] Mario Vega-Barbas, Iván Pau, María Martín-Ruiz, and
Anthropology Quarterly, 8(4):430–447, 1994. Fernando Seoane. Adaptive software architecture based
on confident hci for the deployment of sensitive services
[3] Igor Akushevich, Julia Kravchenko, Svetlana
in smart homes. Sensors, 15(4):7294–7322, 2015.
Ukraintseva, Konstantin Arbeev, and Anatoliy I
Yashin. Age patterns of incidence of geriatric [15] Moeness G Amin, Yimin D Zhang, Fauzia Ahmad, and
disease in the us elderly population: Medicare-based KC Dominic Ho. Radar signal processing for elderly
analysis. Journal of the American Geriatrics Society, fall detection: The future for in-home monitoring. IEEE
60(2):323–327, 2012. Signal Processing Magazine, 33(2):71–80, 2016.
[4] Jan Vanus, Jana Belesova, Radek Martinek, Jan [16] Xuewei Wang, Haibin Qu, Ping Liu, and Yiyu Cheng. A
Nedoma, Marcel Fajkus, Petr Bilik, and Jan Zidek. self-learning expert system for diagnosis in traditional
Monitoring of the daily living activities in smart chinese medicine. Expert systems with applications,
home care. Human-centric Computing and Information 26(4):557–566, 2004.
Sciences, 7(1):30, 2017.
[17] Paul C Chu and John E Beasley. A genetic algorithm
[5] Frank G Miskelly. Assistive technology in elderly care. for the multidimensional knapsack problem. Journal of
Age and ageing, 30(6):455–458, 2001. heuristics, 4(1):63–86, 1998.
[6] Ahmad Alaiad and Lina Zhou. Patients’ adoption [18] Jagdish Chand Bansal and Kusum Deep. A modified
of wsn-based smart home healthcare systems: binary particle swarm optimization for knapsack
An integrated model of facilitators and barriers. problems. Applied Mathematics and Computation,
IEEE Transactions on Professional Communication, 218(22):11042–11061, 2012.
60(1):4–23, 2017.
[19] Min Kong, Peng Tian, and Yucheng Kao. A new ant
[7] Diane J Cook, Maureen Schmitter-Edgecombe, and
colony optimization algorithm for the multidimensional
Prafulla Dawadi. Analyzing activity behavior and
knapsack problem. Computers & Operations Research,
movement in a naturalistic environment using smart
35(8):2672–2683, 2008.
home techniques. IEEE journal of biomedical and
health informatics, 19(6):1882–1892, 2015. [20] Yuhua Li, Zuhair A Bandar, and David McLean. An
approach for measuring semantic similarity between
[8] Boyuan Sun, Qiang Ma, Shanfeng Zhang, Kebin Liu,
words using multiple information sources. IEEE
and Yunhao Liu. iself: Towards cold-start emotion
Transactions on knowledge and data engineering,
labeling using transfer learning with smartphones. ACM
15(4):871–882, 2003.
Transactions on Sensor Networks (TOSN), 2017.
[9] Jens Lundström, Eric Järpe, and Antanas Verikas. [21] Liangzhao Zeng, Boualem Benatallah, Anne HH
Detecting and exploring deviating behaviour of smart Ngu, Marlon Dumas, Jayant Kalagnanam, and
home residents. Expert Systems with Applications, Henry Chang. Qos-aware middleware for web
55:429–440, 2016. services composition. IEEE Transactions on software
engineering, 30(5):311–327, 2004.
[10] George Demiris, Marilyn J Rantz, Myra A Aud, Karen D
Marek, Harry W Tyrer, Marjorie Skubic, and Ali A [22] Qing Liu, Tomohiro Odaka, Jousuke Kuroiwa,
Hussam. Older adults’ attitudes towards and perceptions Haruhiko Shirai, and Hisakazu Ogura. A new
of ‘smart home’technologies: a pilot study. Medical artificial fish swarm algorithm for the multiple knapsack
informatics and the Internet in medicine, 29(2), 2004. problem. IEICE TRANSACTIONS on Information and
Systems, 97(3):455–468, 2014.
[11] François Portet, Michel Vacher, Caroline Golanski,
Camille Roux, and Brigitte Meillon. Design and
evaluation of a smart home voice interface for the
elderly: acceptability and objection aspects. Personal
and Ubiquitous Computing, 17(1):127–144, 2013.
– 132 –
SESSION 6
S6.1 Invited paper - Preparing for the AI era under the digital health framework
S6.2 Operationalizing data justice in health informatics
PREPARING FOR THE AI ERA UNDER THE DIGITAL HEALTH FRAMEWORK
1
China Academy of Information and Communication Technology (CAICT), China
ABSTRACT and limitations of AI for health has become more and more
necessary.
Information and communication technology (ICT) for health
has shown great potential to improve healthcare efficiency, 1.1 Health industry changes
especially artificial intelligence (AI). To better understand
the influence of ICT technology on health, a framework of The developing requirement for the health industry has
the digital health industry has been proposed in this paper. changed over the decades in terms of demand and supply.
Factors from the health industry and the ICT part are The gap between supply and demand is now increasingly
extracted to study the interaction between two groups of growing, waiting for new technological productivity to fill
component factors. Health factors include service and this gap. From the demand side, an aging population is
management; and ICT factors include sensors, networks, poised to become one of the most significant social
data resources, platforms, applications and solutions. The transformations in the twenty-first century, with implications
interaction between ICT and health can be traced through for nearly all sectors of society. According to data from
the development history, from the stage of institutional World Population Prospects: the 2019 Revision [3], by 2050,
informationization to regional informationization, and one in six people in the world will be over age 65 (16%), up
finally to service intelligentization. Following such a from one in 11 in 2019 (9%). This is especially directly
developmental roadmap, AI was chosen as one of the most reflected in changes of the population's disease spectrum.
powerful technologies to study the penetration effect and key Take China for example, the prevalence of chronic diseases
development trends from the perspectives of data, computing in the population over 65 years old is 539.9‰ [4], which is
power and algorithms. The health industry will be much much higher than for the entire population. For the future,
improved or redefined in the coming AI era. To better chronic diseases, such as cardiovascular and cerebrovascular
understand the strengths, weaknesses and limitations of AI diseases, cancer, diabetes, chronic respiratory diseases, etc.
for health, exogenous factors are discussed at the end of the will become the biggest threat to public health. However, the
paper; preparations on collaboration mechanism; growth of supply resources lags much behind demand [6],
standardization and regulation have been proposed for the the situation is particularly serious for imaging, pathology
sustainable development of digital health in the AI era. and general practitioners. For example, the growth rate of
medical imaging data in China is about 30% per year, while
Keywords – Artificial intelligence, digital health, the annual growth rate of radiologists is only 4.1%, the gap
framework, information and communication technology, of pathologists is estimated to be 100 000[6][7], and it is not
interaction only limited to radiology. Training doctors takes a long time,
which means that the gap cannot be solved in the short term.
1. INTRODUCTION
1.2 ICT penetration into health
Health, as defined by the World Health Organization (WHO),
is "a state of complete physical, mental and social well-being The good news is that the arrival of the fourth industrial
and not merely the absence of disease or infirmity."[1][2]. revolution has brought us ICT technology, which has
After decades of development, the demand and supply of the enormous potential to help overcome this socioeconomic
health industry have quietly changed. Complying with the challenge. Digital health refers to the use of ICT to help
industrial development requirements, information and address the health problems and challenges faced by people
communication technology (ICT) was introduced as a new under treatment [8]. Technologies such the fifth generation
technology which has great potential to improve healthcare (5G) communications, machine-to-machine (M2M)
efficiency, thus the digital health era has begun. Among communications, cloud computing, Internet of things (IoT),
those ICT technologies, artificial intelligence (AI) is big data, AI and machine learning(ML) etc. [9] will
regarded as the most powerful one with an unpredictable inevitably penetrate theinto health industry and lead to new
developing rate. To understand the strengths, weaknesses improvements in digital health services, from medical device
manufacturing to healthcare delivery, from medical research
– 136 –
ICT for Health: Networks, standards and innovation
different scenarios and requirements, such as various contributed to the digitization of data, and they started to be
wearable devices. applied in 1960s, when the United States began research on
HIS in military hospitals, and then Japan and some European
➢ Network: ensures effective transmission of information countries followed and promoted HIS in the late 1970s [13].
collected by sensors. 5G is an innovation on networks, Additionally, data played an invisible key role in the process.
combined with upper factors, giving birth to new In particular, the replacement of paper electronic data is a
applications such as telemedicine. huge innovation for the traditional industry. Electronic
➢ Data resource: refers to how the data transmitted by information is more efficiently processed, analyzed and
networks is effectively integrated and further processed. calculated than paper records, with an improvement in
Data center construction is an example. efficiency, as well as a reduction in the operational costs [14].
However, there is still limited improvement when only
➢ Platforms: integrate various component and computing considering the simple point of institution; the full value of
capabilities to achieve integration and modularity. For informationization is still waiting to be amplified.
example, cloud computing provides the foundation
support for the upper applications of digital health.
➢ Applications: promote the service in specific scenarios
and meet specific needs. AI and ML are examples
closely integrated into the service process to provide a
comprehensive service.
➢ Solutions: provide an overall service including
technical support, consulting, design, operation, etc.
– 137 –
2019 ITU Kaleidoscope Academic Conference
– 138 –
ICT for Health: Networks, standards and innovation
Processors are major computing units in AI for health laboratory test, medical treatment, monitor & recovery. The
systems. Performance could be evaluated by the metrics of mainstream applications of AI are listed as below.
calculation speed, data bandwidth and power consumption
per unit of time. Processors used in AI usually include central ➢ AI Virtual assistant: Logistics such as online
processing unit (CPU), graphics processing unit (GPU), reservation, intelligence triage, payment and
field-programmable gate array (FPGA), application specific repairment monitoring could be effectively completed
integrated circuit (ASIC) and system-on-a-chip (SoC) by AI virtual assistants. Information input could be in
accelerators, etc. The Google Brain was once based on the various formats, such as audio, pictures, EHR scan and
CPU, but the general calculators have limited abilities for questions answering. By speech/image recognition and
floating point calculations, and was unable to meet deep natural language processing technology, it could
learning requirements, especially in model training. Though understand the patient's description of symptoms,
GPU is currently the primary choice due to its high- automatically provides intelligent consultation, triage
bandwidth caches and strong parallel computing power, suggestions and payment assistance. An intelligent
customized chips have more potential than these general- voice product “Yun Medical Sound” has been applied
purpose chips. FPGA is very flexible to achieve a high in more than 40 hospitals in China, with a voice
degree of customization, and ASIC even has a better transcription accuracy rate of over 97%. Additionally,
performance, with a computing speed of over 5-10 times these kinds of products’ functions can also be expanded
faster than FPGA. High R&D costs and production cycles to service rating, doctor matching, in-hospital
are two main obstacles for customized processors. A scale navigation, medical insurance reimbursement, pre-
effect may reduce the cost in the long term. Tractica forecasts diagnosis data collection, post-diagnosis follow-up, re-
that the market for deep learning chipsets will increase from examination reminder, health knowledge teaching, etc.
$1.6 billion in 2017 to $66.3 billion by 2025, and ASIC The application forms of the virtual assistant are very
market will be the largest by 2025 [26]. flexible and adaptive to certain scenarios, including
APP, websites and embedded programs, etc.
Network architecture is also customized to support AI
services. Continuous health condition monitoring and ➢ Medical imaging aided diagnosis: The core steps in
complex health management scenarios require flexible health service, such as symptom check and image
computing abilities. Architecture with a combination of inspection are currently penetrated with AI in the form
cloud and edge computing will be increasingly suitable for of medical imaging aided diagnosis. Based on
growing health needs. Cloud-computing solutions offer a computer vision and pattern recognition technology, AI
pay-per-use model that provides on-demand access to could achieve image classification and retrieval, 3D
computing resources. The cloud platform for deep learning reconstruction, image segmentation, feature extraction,
can be customized on TensorFlow, Caffe, MXNet, Torch, lesion identification, target area delineation and
etc., and provides developers with common models to reduce automatic annotation, etc. Various application
R&D costs. Algorithm training, assessment, visualization scenarios include fundus screening [27], breast
tools and API services are also available for customization. pathology diagnosis [28], X-ray reading, brain CT
Because of the convenience and low-cost operation, AI segmentation, bone injury identification, bone age
training tasks are gradually deployed on the cloud instead of analysis, organ delineation, dermatological auxiliary
the device. Meanwhile, edge computing developed on the diagnosis, etc. Some research even shows a better
devices is designed to be adaptive to application scenarios. It performance and efficiency than that provided by
is a blue sea with diverse forms and low competitiveness. IoT humans [29].
or wearables such as intelligent watches, headphones and
wristbands, and mobile phones are currently major drivers of ➢ Clinical decision support system (CDSS): Key steps
the edge market. AI inference tasks are increasingly such as laboratory test judgement and medical
deployed on devices to support the diversified scenarios and treatment are integrated with CDSS. Traditional CDSS
needs. builds on a top-down approach, with expertise and rules
based on expert systems to simulate the clinical
4.3 Closely integrated algorithm with health process decision-making process. AI based CDSS, without the
reliance on predefined rules, could ensure the
With large databases, high-performance computing, AI timeliness of evidence updates. Advanced natural
algorithms could strongly support and achieve personalized language processing, cognitive computing, automatic
medicine. The close combination between AI algorithms and reasoning and deep learning, etc. are used. AI-based
traditional health processes is the key to success. As is shown CDSS could greatly take full advantage of digital
in Figure 1, the framework of the health industry consists of medical data accumulated on a large scale in clinical
service and management. The integration can also be seen work in recent years, and overcome the weakness of
from these two perspectives. inefficiency in knowledge construction and limited
information coverage for traditional decision making,
Service process usually includes reservation, check-in & thus eventually accelerating industry development.
triage, payment/ pre-pay, symptom check, image inspection,
– 139 –
2019 ITU Kaleidoscope Academic Conference
The management process in the health industrial framework These preparation factors sit alongside the health direct
is divided into two parts: equipment and staff. The workflow in healthcare institutions; thus, they can be defined
corresponding applications of AI are also listed as below. as exogenous factors that play an external effect on the health
industry. Figure 6 shows the complete framework of digital
➢ Staff and institution management: Personnel in health health with the consideration of exogenous factors including
institutions usually includes physicians of various collaboration mechanisms, standardization and regulation,
specialties, nurses, technicians specializing in specific etc.
equipment, administrative financial clerks and other
support personnel. Intelligent institution management
application could either refer to specific problems like
scheduling the nurse personnel, performance appraisal,
workload distribution, task assignment, patient
feedback collection and analysis, etc. [12]. Besides,
auxiliary talent training is another direction for AI to
improve staff management. Royal Philips' annual
health survey shows that in Singapore, about 37% of
medical professionals can use artificial intelligence to
support administrative tasks, only 28 % of them have
the digital literacy to use it for diagnosis. Auxiliary
talent training could be a customized education and
collaboration platform. InferScholar Center released in
March 2019 [30] is equipped with advanced models and
visualization tools for the clinical research. Ali Health
is trying to break down various clinical case data into a
three-dimensional “virtual patient” in the physician
training system of Ali ET Medical Brain [31]. Figure 6 – Effects from exogenous factors
– 140 –
ICT for Health: Networks, standards and innovation
financing. It includes major R&D project budget agencies need to answer. Recently, the International
support, profit distribution and national health Telecommunication Union (ITU) established an
insurance management, etc. Especially with the ITU/WHO Focus Group on artificial intelligence for
progress of aging, how to effectively reduce the cost of health (FG-AI4H), which works in partnership with
medical insurance has become a common problem that WHO to especially establish a standardized assessment
needs to be solved in various countries. framework for the evaluation of AI-based methods for
health, diagnosis, triage or treatment decisions [42]
5.2 Standardization
➢ Security : It includes the concerns of data security,
Under the overview control of the collaboration mechanism, network security and product security. Data security
standardization could technically act as an accelerator for refers to data ownership, data handling policy and
integration innovation. Consideration on standardization privacy protection during usage. Network security
could be prepared as the following part. refers to cybersecurity, avoiding products being
attacked by illegal cyberattacks. Product security refers
➢ Data format and interface: Much previous efforts have to the safe use of the product. Currently with the rapid
been made to meet the demand of medical system development of AI for health application, most of the
interconnection and compatibility. Personal health applications investigated could not provide relevant
device standards (ISO / IEEE 11073) and Digital evidence or peer-reviewed research to support their
Imaging and Communications in Medicine (DICOM) products. According to a study published in Nature
are known as addressing the interoperability. Though Digital Medicine, only two of the 73 applications in
AI for health focuses more on the application layer, their survey provided evidence of research [43].
updates on data formats and interfaces should also be
considered to meet the development requirement. ➢ Ethics:Ethics are important to consider especially for
the health field. Major countries and international
➢ Data quality: It refers to the standardization of content organizations have established AI ethics institutions
requirement input to AI algorithms, and it is a new focusing on the discussion of ethical guidelines and
demand due to new technology of AI. Medical images standards. In June 2019 the US Food and Drug
used by AI may contain undesirable artefacts (e.g. Administration released a discussion paper of proposed
background noise), lack focus, exhibit uneven ‘Regulatory Framework for Modifications to Artificial
illumination or under/overexposure, etc. [35]. Intelligence/Machine Learning (AI/ML)-Based
Moreover, the quality of the annotation for AI training Software as a Medical Device (SaMD)’ and also
is also critical. To form a unified understanding and requested feedback including on ethical aspects [44].
workflow on annotation among different groups of
clinicians is a difficult but necessary task. Several 6. CONCLUSION
public datasets are released for research, including
Kaggle, ImageNet, Messidor database [36-40], but for In this work, we choose the perspective of interaction of ICT
long-term development and scaled application, on the health industry. An industrial framework of the digital
standardization on data content and annotation are very health industry was proposed to better understand the
necessary requirements for sustainable development. interaction between component factors from the health and
ICT sides. We extracted and reconstructed different
5.3 Regulation component factors to expand the framework from the
traditional health industry to digital health. The traditional
With collaboration mechanism acting as a macro-control, health industrial framework is divided into service and
standardization as a technical accelerator, regulation is to management parts, and ICT factors are listed as sensors,
define the bottom line of the industry and maintain its networks, data resources, platforms, applications and
legality. The International Medical Device Regulators solutions. This paper also tracks the interaction through the
Forum (IMDRF) was established to discuss the common development history of the digital health industry, from
problems in international medical device regulation, with institutional informationization to regional
representatives from regulatory authorities in Australia, informationization, and finally to service intelligentization.
Brazil, Canada, China, European Union, Japan and the Following such a developmental roadmap, AI was chosen as
United States, as well as WHO [41]. From their working one of the most powerful technologies to discuss the key
group setting, main concerns on regulation can be divided trends from data, computing power and algorithms. Service
into reliability, security and ethics parts. and management processes in the health industry were
observed on the effects of ICT penetration. In the end,
➢ Reliability: The performance of AI algorithms can be exogenous factors such as a collaboration mechanism,
evaluated in the metrics of accuracy, precision, ROC, standardization and regulation were proposed and discussed
F-measures, interpretability, robustness, generalization, to better prepare for supporting the sustainable development
etc. In the face of such an emerging technology, how to of digital health in the AI era.
evaluate AI/ML-based software as a medical device
(SaMD) is a problem that all national regulatory
– 141 –
2019 ITU Kaleidoscope Academic Conference
[3] United Nations: World Population Prospects 2019: [15] Qian W, Jie W, Yao D etal. Ontology-Based design
Highlights: and information sharing for medical information[C].
https://population.un.org/wpp/Publications/Files/WP In: IEEE Int Conf on Software Engineering and
P2019_Highlights.pdf Service Sciences. Piscataway, NJ, USA: IEEE,
2010, 663-666.
[4] World Bank East Asia and Pacific Regional Report:
https://openknowledge.worldbank.org/bitstream/han [16] Kilic O, Dogac A, Eichelberg M. Providing
dle/10986/23133/32945.pdf?sequence=4&isAllowed interoperability of eHealth communities through
=y Peer-to-Peer networks[J]. IEEE Transactions on
Information Technology in Biomedicine, 2010,
[5] World Health Organization, World Health Statistics 14(3): 846-853.
2018:
https://apps.who.int/iris/bitstream/handle/10665/272 [17] Edwards, Michael. Lippeveld,Theo. Decision
596/9789241565585-eng.pdf?ua=1 support systems for improving the analytic capacity
of decentralized routine health information systems
[6] Annual Report on Healthy City Construction in in developing countries. Proceedings of the Hawaii
China (2018), ISBN:978-7-5201-3936-6 Intemadonal Conference on Sciences[J]. v 37, 2004,
P 2375-2383.
[7] 2018-2024 Artificial Intelligence + Medical Imaging
Industry Market Research and Investment Prospect: [18] Gajanayake R, Iannella R, Sahama T. Sharing with
https://www.chyxx.com/research/201710/578114.ht care: an information accountability perspective[J].
ml IEEE Internet Computing, 2011, 15(4): 31-38.
[8] Bhavnani, Sanjeev P.; Narula, Jagat; Sengupta, [19] Tiina,Suominen,Tarjal, Asikainen, Paula. The
Partho P. (7 May 2016). "Mobile technology and the outcomes of regional healthcam information systems
digitization of healthcare". European Heart Journal. in health care-A review of the research literature
37 (18): 1428–38. doi:10.1093/eurheartj/ehv770. International Journal of Medical Informafics[J]. V
PMC 4914890. PMID 26873093. 78, nil, 2009(9) ,P 757-771.
[9] Svetlana Sicular, Kenneth Brant, Hype Cycle for [20] Stolyar,Anna. Lober, William B. Drozd,Daniel R. A
Artificial Intelligence, 2018[J], Gartner. patient-centered health record in a demonstration
regional health information network Conference
Proceedings* 1st Transdisciplinary Conference on
– 142 –
ICT for Health: Networks, standards and innovation
[21] Haux,. Health information systems-Past, present, [31] Ali ET Medical Brain.
future. International Journal of Medical https://www.alibabacloud.com/et/medical
Informatics[J]. V 75,n 3-4 SPEC. ISS. March / April
2006,P 268-281. [32] Joseph A. DiMasi, Henry G. Grabowski, Ronald W.
Hansen (2016). "Innovation in the pharmaceutical
[22] Winter Alfred. Brig Birgitl. Funkat Gert. 3LGM2- industry: New estimates of R&D costs". Journal of
Modeling to support management of health Health Economics.
information systems. International Journal of
Medical Informatics [J]. V 76, n2*3, February / [33] California Biomedical Research Association. "New
March 2007, P145-150. Drug Development Process" https://ca-
biomed.org/wp-content/uploads/2018/02/FS-
[23] Centers for Medicare & Medicaid Services (CMS), DrugDevelop.pdf
and Office of the National Coordinator for Health
Information Technology (ONC), HHS. Medicare [34] Mercom Capital Group report,
and Medicaid programs; modifications to the https://www.fiercehealthcare.com/tech/venture-
Medicare and Medicaid Electronic Health Record funding-for-digital-health-reached-9-5b-2018
(EHR) Incentive Program for 2014 and other
changes to EHR Incentive Program; and health [35] A. Galdran., T. Araújo, A. Mendonça, A. Campilho.
information technology: revision to the certified Retinal Image Quality Assessment by Mean-
EHR technology definition and EHR certification Subtracted Contrast-Normalized Coefficients, in:
changes related to standards. Final rule[J]. Federal European Congress on Computational Methods in
Register, 2014, 79(171):52909-33. Applied Sciences and Engineering Springer 2017 pp.
844–853.
[24] Doyle-Lindrud, Susan, DNP, AOCNP, DCC.
Watson Will See You Now: A Supercomputer to [36] Kaggle diabetic retinopathy competition
Help Clinicians Make Informed Treatment forum. https://www.kaggle.com/c/diabetic-
Decisions. Clinical Journal of Oncology Nursing; retinopathy-detection/forums/t/15605/human-
Pittsburgh Vol. 19, Iss. 1, (Feb 2015): 31-32. performance-on-the-competition-data-set . Accessed
June 30, 2016.
[25] Julia Powles, Hal Hodson. Google DeepMind and
healthcare in an age of algorithms, Health and [37] Kaggle Diabetic Retinopathy Detection
Technology, 2017, Volume 7, Number 4, Page 351 competition. https://www.kaggle.com/c/diabetic-
retinopathy-detection . Accessed June 30, 2016.
[26] Deep Learning Chipsets, CPUs, GPUs, FPGAs,
ASICs, SoC Accelerators, and Other Chipsets for [38] Krizhevsky A, Sutskever I, Hinton GE. ImageNet
Training and Inference Applications: Global Market classification with deep convolutional neural
Analysis and Forecasts: networks. Adv Neural Inf Processing Syst.
https://www.tractica.com/research/deep-learning- 2012:1097-1105.
chipsets/
[39] Russakovsky O, Deng J, Su H, et al. ImageNet large
[27] Mookiah MRK, Acharya UR, Chua CK, Lim CM, scale visual recognition challenge. Int J Comput Vis.
Ng EYK, Laude A. Computer-aided diagnosis of 2015;115(3):211-252.
diabetic retinopathy: a review. Comput Biol Med.
2013;43(12):2136-2155. [40] Decencière E, Etienne D, Xiwei Z, et al. Feedback
on a publicly distributed image database: the
[28] Elmore JG, Longton GM, Carney PA, et al. Messidor database. Image Anal Stereol.
Diagnostic Concordance Among Pathologists 2014;33(3):231-234. doi:10.5566/ias.1155
Interpreting Breast Biopsy Specimens. JAMA.
2015;313(11):1122–1132. [41] The International Medical Device Regulators Forum
doi:10.1001/jama.2015.1405 (IMDRF):http://www.imdrf.org
[29] Gulshan V, Peng L, Coram M, et al. Development [42] Wiegand Thomas, Krishnamurthy Ramesh,
and Validation of a Deep Learning Algorithm for Kuglitsch Monique, et al. (2019). WHO and ITU
Detection of Diabetic Retinopathy in Retinal Fundus establish benchmarking process for artificial
Photographs. JAMA. 2016;316(22):2402–2410. intelligence in health. The Lancet. 394.
doi:10.1001/jama.2016.17216 10.1016/S0140-6736(19)30762-7.
– 143 –
2019 ITU Kaleidoscope Academic Conference
[43] Digital mental health tools won't see clinical [44] US Food and Drug Administration, proposed
adoption without validation, industry players say. Regulatory Framework for Modifications to
https://www.mobihealthnews.com/content/digital- Artificial Intelligence/Machine Learning (AI/ML) -
mental-health-tools-wont-see-clinical-adoption- Based software as Medical Device (SaMD):
without-validation-industry-players https://www.fda.gov/media/122535/download
– 144 –
OPERATIONALIZING DATA JUSTICE IN HEALTH INFORMATICS
Mamello Thinyane
United Nations University Institute on Computing and Society, Macau SAR, China
monitoring; secondly, to understand the attitudes and values management of their health, including to inform behavior
around data sharing and social sense-making. The survey change and track progress on specific health goals.
consisted of 14 questions on demographics, personal health
informatics practice, sustainable development goals, data 2.1 Personal health informatics
sharing, and an open-ended question on the current practice
(i.e. framed as “What information and data do you use in Li et al. [4] formally define personal health informatics
your everyday life that you find relevant for your simply as a class of applications “that help people collect and
wellbeing?”). The 981 respondents in the survey, the reflect on personal information.” The field has gained
majority of whom are from North America, were recruited increasing popularity due to several developments, including
via virtual snowballing, social media channels and a research the rise of quantified-self movement [5], the availability of
panel via an online survey platform. affordable self-tracking technology, and the proliferating
phenomenon of datafication of individuals and societies [6].
The paper is structured as follows: the next section provides The promise of the self-tracking devices to offer individuals
a broad overview of digital health, paying attention to health a non-subjective and unambiguous assessment of their
informatics and the value proposition of data for health. This physical wellbeing and the state of their bodies has been part
is followed by an introduction of the notion of data justice of society for over a century; the weigh scales have played a
and its relevance to the digital health domain in general, but predominant role in this regard [7]. Beyond the development
also to health informatics specifically. Various formulations of new technologies used for personal health informatics, the
of data justice are discussed, after which is distilled a list of 21st century self-tracking landscape has also introduced new
requirements to inform technology designs. This is followed considerations, including the commoditization of personal
by a proposal of a health informatics architecture that is data, new value dimensions associated with aggregate data,
informed by the data justice principles. Lastly, the merits of and the wide sharing of data beyond the individuals who the
this architecture are discussed, juxtaposed to other related data is about [7], [8]. Therefore, while personal informatics
technologies. fundamentally regards the use of own data by individuals for
their benefits, the contemporary reality is that personal data
2. DIGITAL HEALTH AND HEALTH and its use exists within a broad, multifaceted ecosystem.
INFORMATICS
One of the core elements within digital health is health Figure 1 - Motivation and uses of personal informatics
informatics, comprising the technologies for the
management of electronic health records, medical data, The use of data towards the achievement of health outcomes
health indicators and personal health data. Traditionally, the has traditionally been premised on the argument that more
bulk of health data collection and processing was undertaken and better data leads to better health choices and decisions,
by health service providers, with individuals as the primary and that the increasing availability of health information on
sources of health data, as well as the primary beneficiaries of the Internet would lead to the emergence of ‘informed
the health outcomes associated with the use of the data. This patients’ [9] and ‘digitally engaged patients’ [10]. The
data, which represents one of the key resources for the transtheoretical model (TTM) of behavior change [11],
business operation of health providers, typically exists in the which has been the predominant model for the psychology
form of electronic health records. However, with the of intentional behavior change, has also informed the
growing ubiquity of health technology tools, individuals are formulation of personal health informatics models such as
increasingly also participating in the collection and the stage-based model of personal informatics [4] and the
management of their health data. In the context of personal lived informatics model of personal informatics [12]. In our
health informatics, individuals are collecting data for self- research, we have identified, through the thematic coding of
– 146 –
ICT for Health: Networks, standards and innovation
the open-ended survey question, the predominant pathways Beyond the risks of social control, this has the potential to
to impact as well as the motivations and current practice of open up individuals to the risks of exploitation through
the participants with regards to personal informatics. Figure surveillance capitalism and commoditization of personal
1 highlights the main “motivation and use” themes with their data, as has been demonstrated, for example, in the cases of
corresponding coding reference frequencies. 23andMe, Facebook, and Cambridge Analytica [17]–[19].
Table 1 - Motivations and styles of personal informatics Thirdly, the empowerment narrative echoes the
technological determinism sentiments, which are not
Motivation & use / Tracking universally valid and consistent. In our survey results, on the
Documentary
style investigation of the participants’ use of personal health
Diagnostic
Fetishised
Directive
Rewards
informatics towards health outcomes and the attitudes
towards data sharing, we coded 18 references that expressed
both a strong resistance and refusal to use and / or to share
personal health data. For example: “I don’t use information
Awareness and monitoring X
or data. I take my medicines and vitamins, and see my doctor
Benefit for others
often” (GIS_806); “How I feel, do not use data” (GIS_504),
Compare and reflect X “Mindfulness of my moods and stress level; awareness that I
Curiosity and information am the major actor in my life, but that I can’t control
Dealing with an ailment X X anything outside myself” (GIS_379), where the participants
Informing action X emphasized the reliance on self-awareness as opposed to on
Maintaining health and X technology devices and data; “Actually none because
wellbeing everything changes and everyone has their viewpoint to
Reach new goals and improve X make you believe what they are telling is true” (GIS_596),
expressing the lack of trust in the system stakeholders.
The motivations and uses of personal health informatics
identified in our research correspond to three of five of Health informatics tools and technologies are employed to
Rooksby et al. [13] style styles of personal information empower patients to achieve better health and to improve
tracking. These observations support the position that as health service delivery by health service providers. The
individuals engage in the collection and use of personal impact pathways from these digital resources to specific
health data, through various impact pathways and a health outcomes are non-trivial and need to be critically
combination of personal conversion factors [14], they are understood, taking into consideration the situations of the
empowered to pursue and achieved desired health outcomes. different actants, the contextual factors, as well as the overall
This empowerment narrative of personal health informatics digital health ecosystem.
has informed many digital health programs and projects
around the world. It has, however, been criticized and shown 2.2 Health data ecosystem
to present an overly simplified techno-utopian perspective
that fails to consider the nuanced complexities of personal Individuals are the primary unit of attention within the health
health informatics. domain, as far as being the main beneficiaries of the targeted
health outcomes. They, however, exist as one of the actants
Firstly, while the importance of the informational and within a complex ecosystem consisting of a variety of
technology resources cannot be denied, the empowerment stakeholders, including health service providers, health
narrative fails to recognize the varying agency as well as the industry stakeholders, public sector entities, households, and
endowments of conversion factors, such as underlying data communities, as well as other civil society stakeholders. The
and digital illiteracies, as well as general illiteracy, for use of data towards the achievement of health outcomes,
different individuals and population groups [14]. In an therefore, permeates this complex ecosystem and needs to be
empirical research investigating “informational practices” of considered when taking into consideration the interactions
32 mid-life women on the use of hormone replace therapy with and the data exchanges between the different
(HRT) for relief of menopausal symptoms, Henwood et al. stakeholders.
[9] found that there was a strong reluctance on the part of the
participants to take on the implied responsibilities of data Firstly, in the context of the sustainable development data
management; they observed problems with the information ecosystem or that of future data-driven societies, sharing of
literacy of the participants; and there were also challenges personal data needs to be considered not only with
associated with information-sharing in medical encounters individuals’ personal social circles but also with other
with health professionals. stakeholders within the wider data ecosystem. For example,
the role of citizen-generate data to support the monitoring of
Secondly, the proliferation of personal health informatics progress towards the sustainable development goals, through
technologies that track and monitor our everyday functioning direct contributions to the indicators or via proxy indicators,
has the potential to unleash Orwellian techno-dystopia of has been recognized and well highlighted in the literature.
panoptic surveillance assemblages that extend paternalistic
social control by the strong and the powerful [15], [16].
– 147 –
2019 ITU Kaleidoscope Academic Conference
Secondly, while deriving relevant insights from health justice. As noted by Taylor [27], the ends of various data
informatics primarily ensues through the individual’s justice formulations is to achieve both specific outcomes and
engagement with their data, research has found that also specific configurations of the associated data
individuals also engage in sharing of their data with others assemblages towards the achievement of those outcomes: in
for sense-making purposes [20]–[22]. Thus, the collection the case of Johnson’s [29] framework, the end goal is
and use of data by individuals also comprise the social embedding anti-discrimination principles and features in the
dimension. design of database systems; for Heeks and Renken [28], the
focus is on data distribution in a way that achieves fair access,
Thirdly, personal health data also gets shared to support participation and representation; and lastly Dencik et al. [30]
external pursuits such as biomedical research, where data on are interested in the means of limiting data collection and
health profiles, cohort data, as well as physical activity data distribution in contexts of surveillance capitalism.
can support projects such as the Global Alliance for
Genomics and Health [23]. The sharing of data in this In the work of Mortier et al. [31], in which they formalize
context can be motivated from the perspective of the the notion of human-data interaction (HDI), they explicate
Universal Declaration of Human Rights, which recognizes the interaction between humans and data systems in a way
the “right of everyone to share in scientific advancement and that places “the human at the center of the flows of data, and
its benefits” [23]. Thus sharing of data can be towards these providing mechanisms for citizens to interact with these
goals, which are associated with citizen science, as well as systems and data explicitly”. While the formulation of HDI
increased participation and engagement in advancing is not explicitly from a social justice nor ethics perspective,
scientific research [24]. it gives recognition to the fact that the underlying issues in
HDI sit at the intersection of “the various disciplines
In all these cases of external sharing of personal health data, including computer science, statistics, sociology,
there is, however, the persistent risk of “Googlization of psychology, and behavioral economics” [31]. Further, it
health research,” which is associated with the increasing gives recognition to the fact that human-data interaction
data-driven encroachment and involvement of the major happens in the context of complex data ecosystems, which
technology companies within the health and biomedical are constitutive of the global data-driven society. In this
sectors [17]. The potential benefits of the application of these complex interaction of different stakeholders with different
technological developments on issues of health and capabilities, interests and agendas, there is an ongoing
wellbeing are immense; they include major improvements in contestation for the voices of humans and human-centric
disease diagnosis, improving access to services through perspectives not to be marginalized and excluded. Some of
telehealth solutions and advancing the developmental the powerful and key actants within the health informatics
aspirations of achieving universal health coverage. The ecosystem include health-service providers, the health
challenges, however, are equally immense and are associated industry, as well as the non-human technology-related
not only with adverse health outcomes but also with negative actants, as has been highlighted by Sharon [17] regarding the
sociocultural and economic consequences. These challenges influence of the technology companies in the health data
are related to issues of bias, privacy [25], informed consent, research. Further highlighting the complexity, Morley and
context transgressions [26], health data commoditization, Floridi [16] offer a poignant critique of the techno-utopian
new power asymmetries and discriminations [27], data formulation of mHealth technologies as empowering devices
valorization and benefit-sharing, and the importation of and warn against the risk of medical paternalism. Privileging
digital capitalism practices into the health realm [17]. the position of the humans within the health informatics
ecosystem, as has been done in the HDI framework, allows
3. DATA JUSTICE IN HEALTH INFORMATICS for the critical investigation of issues towards an explicit goal
of enhancing the substantive freedoms of individuals to
Numerous definitions of “data justice” have been advanced achieve their desired health outcomes and enhancing their
in the literature, which fundamentally recognize the social health capabilities [32].
justice dynamics and impacts of the use of data in society.
Taylor [27] defines data justice as the “fairness in the way In this paper, the HDI framework has been adopted to frame
people are made visible, represented and treated as a result the discussion of the outworking of data justice in health
of the production of digital data.” In her formulation of data informatics systems. The paper expands on the imperatives
justice she decomposes the concept to three notions of of legibility, agency and negotiability to identify specific
(in)visibility – associated with access to representation, and considerations and non-functional requirements to inform
informational privacy; (dis)engagement with technology, the design of health informatics systems.
which is linked to sharing in data benefits as well as
autonomy in data choices; and to antidiscrimination, which 3.1 Legibility
is linked with the ability to challenge bias and preventing
discrimination. Heeks and Renken [28] define data justice Legibility is summarily defined as “being able to be
simply as “the primary ethical standard by which data- understood by people they concern, as a precursor to
related resources, processes and structures are evaluated.” exercising their agency” [31]. This is defined with regards to
They, however, expand this to formulate three notions of the data, as far as individuals understanding what data has
instrumental, procedural and distributive rights-based data been collected, how it is being used, by whom, and when it
– 148 –
ICT for Health: Networks, standards and innovation
is being used; but legibility is also defined with regards to 3.3 Negotiability
the algorithms that process the data, towards ensuring that
algorithms are understood and that the various forms of Negotiability is defined in terms of “active and engaged
algorithm opacity are reasonably mitigated [33]. While at a interaction with data as contexts change.” This makes
simple level the “concerned” people could be understood to recognition of the fact that not only do situations and
refer to the people who the data is about, in reality, the people contexts change, but also do individuals’ desires, attitudes
who are impacted by collected health data, which Loi [34] and preferences. The use of personal health data is tightly
terms as digital phenotypes, and the nature of the impact are coupled to and contingent on the context; individuals need to
very diverse. In the case of health informatics, there are the retain the legibility and agency in different contexts. This
identified individuals who the data is about; there are further decomposes into the following considerations:
individuals who collect the data and who are involved in the
creation and shaping of the digital phenotypes, and there are 1. (Perpetual) Control: the continued ownership and
also people who are impacted by generalizations that control of personal health data and digital phenotypes,
emanate from health informatics [34]. In this paper, the the digital traces that have value towards specific health
notion of “ownership” of data is used in the first sense, which outcomes, in perpetuity [34].
regards health informatics as the self-extension of and as 2. Data provenance: with the changing contexts and the
being constitutive of the individual who the data is about. evolution of data, it is vital to maintain the genealogy of
personal health data.
From the analysis of Mortier et al.’s [31] description and
discussion of “legibility,” supported by the investigations
undertaken in this research, the following health informatics
systems requirements and considerations are formulated:
3.2 Agency
– 149 –
2019 ITU Kaleidoscope Academic Conference
4. PRELIMINARY ARCHITECTURE PROPOSAL the functionality (e.g. access control, auditing, context
integrity checks) to use the data to be shared with the health
The architecture proposed in this paper is framed for a very service providers; this is achieved through the cloning and
specific digital health scenario, specific requirements and migration of the DataAgent from the PHIX main container
specific context. The scenario is that of sharing personal to the service provider containers, and provides inter-agent
health information data (e.g. health indicators collected on communication and synchronization between the associated
personal monitoring devices, historical health records and DataAgents. The data owner has control of his community
digital phenotypes [34]) with a health service provider, and of DataAgents with the ability to gain visibility of where his
ensuring legibility, agency and negotiability in the data has been shared, to understand the specific utilization of
interaction between the individual and their data. A subset of their data and to control the use of specific DataAgents, for
the requirements detailed in the previous section can be met example, updating permissions and access control, revoking
and implemented with standard techniques and solutions. and killing shared agents.
For example, some of the requirements around data privacy
can be handled using information security techniques, such
as public cryptography systems [35], as has been the practice
for say HIPAA compliance and, more recently, GDPR
compliance. However, there are specific requirements
associated with the HDI imperatives, specifically
negotiability, that give motivation for the architecture
proposed in this paper. In particular, the architecture
addresses the requirement for enhancing the control that
owners of data retain over their data once the data is shared,
and ensuring that the dynamic contextual constraints are
enforced on the subsequent use of the data.
– 150 –
ICT for Health: Networks, standards and innovation
– 151 –
2019 ITU Kaleidoscope Academic Conference
[13] J. Rooksby et al., “Personal tracking as lived [25] B. Jacobs and J. Popma, “Medical research, Big
informatics,” in Proceedings of the 32nd annual Data and the need for privacy by design,” Big Data
ACM conference on Human factors in computing Soc., vol. 6, no. 1, p. 205395171882435, 2019.
systems - CHI ’14, 2014, pp. 1163–1172.
[26] H. F. Nissenbaum, Privacy in context : technology,
[14] S. Nambiar, “Capabilities, conversion factors and policy, and the integrity of social life. Stanford Law
institutions,” Prog. Dev. Stud., vol. 13, no. 3, pp. Books, 2010.
221–230, Jul. 2013.
[27] L. Taylor, “What is data justice? The case for
[15] K. D. Haggerty, Richard V. Ericson, “The connecting digital rights and freedoms globally,”
surveillant assemblage,” Br. J. Sociol., vol. 51, no. Big Data Soc., vol. 4, no. 2, p. 205395171773633,
4, pp. 605–622, 2000. Dec. 2017.
[16] J. Morley and L. Floridi, “The Limits of [28] R. Heeks and J. Renken, “Data justice for
Empowerment: How to Reframe the Role of development,” Inf. Dev., vol. 34, no. 1, pp. 90–102,
mHealth Tools in the Healthcare Ecosystem,” Sci. Jan. 2018.
Eng. Ethics, no. 0123456789, 2019.
[29] J. Johnson, “The question of information justice,”
[17] T. Sharon, “When digital health meets digital Commun. ACM, vol. 59, no. 3, pp. 27–29, 2016.
capitalism, how many common goods are at stake?”
Big Data Soc., vol. 5, no. 2, p. 205395171881903, [30] L. Dencik, A. Hintz, and J. Cable, “Towards data
2018. justice? The ambiguity of anti-surveillance
resistance in political activism,” Big Data Soc., vol.
[18] S. Sterckx, J. Cockbain, H. Howard, I. Huys, and P. 3, no. 2, p. 205395171667967, Dec. 2016.
Borry, “‘Trust is not something you can reclaim
easily’: patenting in the field of direct-to-consumer [31] R. Mortier, H. Haddadi, T. Henderson, D.
genetic testing,” Genet. Med., vol. 15, no. 5, pp. McAuley, and J. Crowcroft, “Human-Data
382–387, May 2013. Interaction: The Human Face of the Data-Driven
Society,” SSRN Electron. J., Oct. 2014.
[19] S. Zuboff, “Big other: Surveillance Capitalism and
the Prospects of an Information Civilization,” J. Inf. [32] J. P. Ruger, “Toward a Theory of a Right to Health:
Technol., vol. 30, no. 1, pp. 75–89, Mar. 2015. Capability and Incompletely Theorized
Agreements.,” Yale J. law Humanit., vol. 18, no. 2,
[20] A. Puussaar, A. Clear, and P. Wright, “Better p. 3, 2006.
together: Reciprocal sharing and social
sensemaking of data,” in Workshop on “Quantified [33] J. Burrell, “How the machine ‘thinks’:
Data and Social Relationships” -CHI2017, 2017. Understanding opacity in machine learning
algorithms,” Big Data Soc., vol. 3, no. 1, p.
[21] R. Fleck and D. Harrison, “Shared PI: Sharing 205395171562251, Jan. 2016.
personal data to support reflection and behaviour
change,” in Workshop on" Beyond Personal [34] M. Loi, “The Digital Phenotype: a Philosophical
Informatics: Designing for Experiences of Data"- and Ethical Exploration,” Philos. Technol., vol. 32,
CHI2015, 2015. no. 1, pp. 155–171, Mar. 2019.
[22] S. Stals, M. Smyth, and O. Mival, “Sharing and [35] H. S. Gardiyawasam Pussewalage and V. A.
exploring Quantified-Self Data on In-Place Oleshchuk, “Privacy preserving mechanisms for
experiences and emotions,” in ACM CHI 2017 - enforcing security and privacy requirements in E-
Quantified Data and Social Relationships, 2017. health solutions,” Int. J. Inf. Manage., vol. 36, no.
6, pp. 1161–1173, 2016.
[23] B. M. Knoppers and A. M. Thorogood, “Ethics and
Big Data in health,” Curr. Opin. Syst. Biol., vol. 4, [36] F. Bellifemine, A. Poggi, and G. Rimassa, “JADE –
pp. 53–57, Aug. 2017. A FIPA-compliant Agent Framework,” Fourth Int.
Conf. Pract. Appl. Intell. Agents Multi-Agent
[24] J. P. Woolley et al., “Citizen science or scientific Technol. (PAAM 1999), pp. 97–108, 1999.
citizenship? Disentangling the uses of public
engagement rhetoric in national research [37] E. K. Adams, M. Intwala, and A. Kapadia,
initiatives,” BMC Med. Ethics, vol. 17, no. 1, p. 33, “{MeD}-Lights: A Usable Metaphor for Patient
Dec. 2016. Controlled Access to Electronic Health Records,”
in Proceedings of the 1st {ACM} International
Health Informatics Symposium, 2010.
– 152 –
SESSION 7
Phillip H. Griffin
ABSTRACT "universal access to health care for all a reality – across the
globe" [3]. With over "95% of the world population" being
Identity authentication techniques based on password- "covered by mobile networks" as of December 2018 and
authenticated key exchange (PAKE) protocols rely on weak over "7 billion mobile subscriptions in the world" [3], ICT
secrets shared between users and host systems. In PAKE, a is poised to connect patients to the "social services, health
symmetric key is derived from the shared secret, used to workers, and care agencies" that can help them overcome
mutually authenticate communicating parties, and then their healthcare challenges [2].
used to establish a secure channel for subsequent
communications. A common source of PAKE weak secrets Though there have been notable improvements in achieving
are password and passphrase strings. Though easily SDG outcomes, there is still much more work to be done.
recalled by a user, these inputs typically require keyboard ICT promises to play an increasingly important roll in this
entry, limiting their utility in achieving universal access. work, as it is the "technology with the greatest impact in
This paper describes authentication techniques based on promoting the inclusion of persons with disabilities" [4],
weak secrets derived from knowledge extracted from and it has the ability to eliminate isolation of the elderly by
biometric sensors and brain-actuated control systems. The "connecting them to the world around them" [5]. With the
derived secrets are converted into a format suitable for use growing availability of smart phones, wireless and mobile
by a PAKE protocol. When combined with other computing, ICT can deliver a new age, "not only of
authentication factors, PAKE protocols can be extended to information sharing in general, but of the proliferation of
provide strong, two-factor identity authentication that is web-based services" and mobile access that can help bring
easy to use by persons living in assistive environments. health and wellbeing to both " disabled and non-disabled
communities alike" [4].
Keywords – assistive environments, authentication,
biometrics, key exchange, security It is especially important to remediate security risk for those
people requiring assistive living services, and for those who
1. INTRODUCTION depend on telemedicine. The delivery of ICT "services
provided through cloud and web-based systems over
In 2017, the World Health Organization (WHO) reported unsecured public networks exposes this vulnerable
that more than "one billion people worldwide - about 15% population to increased security risk" [5]. Authentication
of the world's population" are persons with some form of and secure communications are crucial security controls for
disability [1]. Earlier United Nations (UN) and WHO those who must rely on telemedicine, which uses
reports predicted a tripling of the number of "people aged "telecommunications to, remotely, provide medical
65 or older" in 2010 "to 1.5 billion in 2050, 16 % of the information and services” and to reliably “transfer medical
entire world population" [2]. As the numbers of elderly and information and services from one place to another" [6].
disabled people continue to grow, more of them are striving
to retain their autonomy and remain in their homes. As the Providing vulnerable populations and their caregivers who
cost of healthcare continues to rise, governments have rely on these systems with security assurance begins with
struggled to find ways of providing care to these vulnerable reliable mutual authentication that is accessible by everyone.
populations. A user-centric approach guided by the design goals of
universal access can help to ensure that inclusive outcomes
Ambient assisted living (AAL) aims to achieve the UN are achieved. Providing data confidentiality and secure
Sustainable Development Goal (SDG) of ensuring healthy communications solutions that combat man-in-the-middle
lives and promoting the wellbeing of all people, regardless and phishing attacks is also critical. These goals can be met
of their age, location or income. At its core, AAL relies on by extending the capabilities and scope of an existing
the use of information and communications technology protocol used for secure authentication, Recommendation
(ICT) innovation, networks and standards to deliver ITU-T X.1035.
services that increase "the life quality of patients" and "their
relatives" [2]. ICT and "specifically mHealth solutions"
provide new opportunities to bring access to healthcare and
AAL services "to people in remote areas" and to make
2. PAKE PROTOCOL STANDARDIZATION challenge. In this case, the protocol will end without the
user credentials being exposed to the attacker.
Password-authenticated key exchange (PAKE) protocols
have been defined internationally in Recommendation ITU- When the client authentication-attempt message in a PAKE
T X.1035 [7] and ISO/IEC 11770-4 [8]. PAKE is a protocol is augmented with a user's biometric sample, the
“cryptographic protocol that allows two parties who share PAKE protocol can be extended to provide both mutual
knowledge of a password to mutually authenticate each authentication, and two-factor user identity authentication.
other and establish a shared key, without explicitly The biometric sample included by the user in their
revealing the password in the process” [9]. PAKE protects authentication-attempt message enjoys the same protection
users from phishing and man-in-the-middle attacks, so that against phishing and man-in-the-middle attacks afforded by
users can authenticate with an easily recalled password that PAKE. The user still benefits from mutual authentication,
is never exposed to an attacker. gaining assurance that the intended server has been
accessed instead of an attacker's server.
PAKE protocols achieve mutual authentication without
requiring that users possess digital certificates. By not 3. BIOMETRIC EXTENDED PAKE PROTOCOL
requiring certificates, the cost and operational complexity
of providing mutual authentication solutions can be reduced Biometric authenticated key exchange (BAKE) is an
compared to solutions that rely on a public key extension of the PAKE protocol that provides strong, two-
infrastructure (PKI). By design, PAKE protocols never factor user identity authentication [10]. BAKE extends
expose "the user password to a server impersonation or PAKE by including a user biometric sample, a something-
eavesdropping attack" [5] during a user authentication you-are authenticator, in the PAKE authentication-attempt
attempt. message sent by a user to a server [10]. A claimed user
identity (i.e., an account name) is sent to the server in the
This characteristic of PAKE “prevents off-line dictionary clear. Transfer of the user biometric sample is protected by
attacks, a common password authentication problem.” [9]. encryption under the symmetric key derived from a PAKE
The user's password is input to a Diffie-Hellman key user password, a something-you-know authenticator.
exchange process to derive a symmetric key. This derived
key is used as the basis for ensuring the confidentiality of ICT innovations have led to increased availability and
communications between a user and a server during sophistication of "inexpensive mobile computing devices"
operation of a PAKE protocol. that incorporate "wide varieties of biometric sensors" [5].
"Face, voice, gesture and touch biometric sensors are
The operation of a PAKE protocol, as depicted in Figure 1, becoming commonplace" [5]. This makes it practical for
begins with the user providing a password to a browser or system designers to offer users greater choice that serves
user agent. The password must be preregistered, a value more users. Designers "no longer need to settle on just one
known to the server, so that the user and server can derive biometric technology for authentication" [5]. The ubiquity
the same cryptographic key. The user can assert an identity of sensor-rich ICT devices presents opportunities "to create
claim by presenting an account name to the server in the designs that provide secure authentication and access to
clear, along with their authentication-attempt message web-based services to a greater number of elderly and
encrypted using their password-derived key. disabled users" [5]. ICT innovation is an important enabler
of universal access.
– 156 –
ICT for Health: Networks, standards and innovation
knowledge, a spoken password that can be used to operate X.894 [11], can transfer encrypted content of any type or
PAKE. format using any symmetric encryption algorithm and a
named key. When this CMS type is used with BAKE or
Figure 2 describes the steps required to operate a BAKE PAKE, this key name can be set to the user account
protocol. These steps illustrate that BAKE operations differ associated with the password known to the user and server.
little from those of PAKE. These differences are in the
collection of a user biometric sample, inclusion of the 4. BRAIN-ACTUATED AUTHENTICATION
sample in the authentication-attempt message, and in the
matching of the biometric sample by the server required by Data sources other than biometric sensors can be mined for
BAKE. user knowledge. Researchers have shown that
"noninvasively recorded electric brain activity can be used
to voluntarily control switches and communication
channels" [12]. Using brain accentuated techniques can
allow "near-totally paralyzed subjects the ability to
communicate" using "brain-actuated control" (BAC)
devices [12]. Electroencephalogram (EEG) data collected
from a human brain through a scalp sensor array can be
filtered to reduce noise, and then further decomposed into
discrete, independent components.
– 157 –
2019 ITU Kaleidoscope Academic Conference
Knowledge-based cryptographic techniques such as PAKE At present, information exchange in the ITU-T X.1035
combined with signals capable of "noninvasive brain- protocol is defined only in prose. An ITU-T X.1035
actuated control of computerized screen displays or revision should augment this prose with an ASN.1 schema
locomotive devices" could allow even "motor-limited and defined in terms of ITU-T X.894 CMS type
locked-in subjects" to securely authenticate their identities NamedKeyEncryptedData. This schema should associate a
to an information system and to establish a secure channel PAKE OID with this CMS type in a message. The
for subsequent communications [12]. NamedKeyEncryptedData type provides a standardized
way for applications to encrypt content of any type or
The key to making this approach to identity authentication format with a cryptographic key that uses any encryption
and secure communication viable relies on the realization algorithm specified by a message sender.
that human intentions manifested as electrical signals that
emanate from the human brain can be used as something- This key can be identified using the keyName field of type
you-know authentication factors. If a user's intentions can NamedKeyEncryptedData. This field can be transferred
be treated as weak secrets that are represented in the form unencrypted by a sender to indicate the name of their user
of character strings, they are in a format suitable for input account on a target server. The indicated account name can
to a PAKE protocol. At present there are no standardized then be used by the server to identify that user's password.
techniques for mapping the results produced by a neural Type NamedKeyEncryptedData can be associated with an
network model to the weak secrets needed to operate a OID that identifies any PAKE protocol version as follows:
PAKE protocol.
PAKExchange ::= SEQUENCE {
5. FUTURE STANDARDIZATION type OBJECT IDENTIFIER,
pake NamedKeyEncryptedData
5.1 Focus areas -- The keyName field is a UserID
}
ITU-T Study Group 17 (SG17) has developed a wide range
of ICT standards. Their expertise spans many different The account name indirectly identifies the user password on
areas of technology, including telebiometrics, cryptography, the server. The server uses this password to derive the key
identity management, security architecture, modeling and needed to decrypt the user message. If decryption succeeds,
formal definition languages for information exchange. This identity authentication of the user has also succeeded and
breadth of expertise makes it possible for SG17 to "bridge the user challenge recovered.
multiple domains, bringing them together in standards with
a cross industry focus that benefit multiple communities" The server can encrypt its response to the user's challenge
[15] and makes SG17 well suited to developing the cross- with their shared symmetric key, and send the response to
domain standards required to address the needs of elderly the user in another NamedKeyEncryptedData message.
and disabled populations. When the user receives a correct response from the server,
mutual authentication is achieved, and a secure channel for
These populations often include underserved people that subsequent communications is established.
could benefit from remote services provided to AAL and
other healthcare environments. To enhance the ability of An ASN.1 schema for the content encrypted for exchange
these users to securely access remote resources, SG17 between the user and server should be defined and
should revise its 2007 version of the ITU-T X.1035 PAKE standardized. At a minimum, the encrypted payload of
protocol. Following revision, standardization efforts that NamedKeyEncryptedData must contain components for a
leverage ITU-T X.1035 to create new PAKE-based user challenge and a server response. These components
mechanisms for identity authentication and access control should be optional but constrained so that at least one
should be undertaken. A core focus of this standardization component is present in an exchange. This would allow the
effort should be on achieving the goals of universal access client and the server to exchange the same schema payload
to enable more inclusive authentication solutions. during PAKE operation.
A first step in an ITU-T X.1035 revision should enable The encrypted payload schema should contain an optional
PAKE use with the secure information exchange messages component to support a BAKE extension to the PAKE
approved recently in the ITU-T X.894 CMS protocol. This allows two-factor user authentication to be
Recommendation. This effort should define an information supported but not required. The payload schema should also
object identifier (OID) in ITU-T X.1035 that include an optional extensibility mechanism for use by any
unambiguously identifies its processing requirements in an implementation for any purpose. This mechanism should be
instance of communication. An ITU-T X.1035 OID will defined as a series of one or more authenticated attributes,
allow the ITU-T version of PAKE to be distinguished from each uniquely identified by an OID. These attributes are
the other standardized versions of PAKE defined in protected in the encrypted payload and authenticated by
ISO/IEC 11770-4, which already assigns a unique OID PAKE processing.
value to each of its PAKE versions.
– 158 –
ICT for Health: Networks, standards and innovation
In their 2016 paper on password authentication in the SG17 should create a new standard that provides a strong,
transport layer security (TLS) protocol, Manulis, Stebila, two-factor identity authentication solution based on PAKE.
Kiefer and Denham noted that password authentication is The new standard should expand the current ITU-T X.1035
"perhaps the most prominent and human-friendly user protocol processing to include a step for matching a user
authentication mechanism widely deployed on the Web" biometric sample to a reference template associated with
[16]. The authors described the many threats associated their server account and password. For purposes of
with user reliance for the protection of their credentials on biometric matching, the user could be enrolled in a
secure server-authenticated TLS channels established using biometric system local to the server, or they could be
a public key infrastructure (PKI) [16]. They attribute these enrolled in a separate system that provides a remote
threats to PKI-related problems including that "security matching service. The later case could enable 'biometric
fully relies on a functional X.509" PKI that in practice may portability', allowing a user to enroll one time in a biometric
be flawed, and on "users correctly validating the server’s system, then subsequently to be matched from any device.
X.509 certificate" without being phished by an attacker [16].
These assumptions about PKI implementations have been In current ITU-T X.1035 protocol processing, a user
shown not to be unreliable. attempting authentication sends the server an encrypted
message along with their account name. The server locates
The authors note that many PKI failures in TLS are due to the password associated with the account and derives the
the "problems with the trustworthiness of certification key needed to decrypt the message and authenticate the user.
authorities (CAs), inadequate deployment of certificate When a biometric sample is included by the user in the
revocation checking, ongoing threats from phishing attacks, encrypted authentication attempt, the server can use this
and the poor ability of the users to understand and validate biometric sample to further authenticate the user with a
certificates" [16]. Rather than rely on the rare case where second authentication factor.
users possess the personal certificates needed to benefit
from mutual authentication, the authors propose using The confidentiality of the authentication-attempt message is
PAKE as "part of the TLS handshake protocol" [16]. provided using a symmetric key derived from the user
Following the execution of PAKE in the TLS handshake, password. The user can safely include their biometric
"the key output by PAKE" would be used as "the TLS pre- sample in the encrypted message, since the PAKE protocol
master secret" for deriving "further encryption keys protects the confidentiality of their personally identifiable
according to the TLS specification" [16]. information (PII) from phishing and man-in-the-middle
attack. Only the intended message recipient, the server that
Though PAKE techniques have been standardized for years shares the user account password, can derive the key
in Recommendation ITU-T X.1035 and in ISO/IEC, there needed to decrypt the message and gain access to the user
has been no PAKE standard "agreed upon and implemented biometric sample.
in existing web browser and server technologies" [16].
SG17 should standardize PAKE for use as an option in the When biometric matching is performed local to the server,
TLS handshake. This would broaden the use of PAKE as a at a minimum, the user biometric sample must be included
standalone authentication technique to its use in a protocol in the encrypted user message. When more than one
widely used to conduct online electronic commerce biometric technology type is supported, an identifier of the
transactions and to provide secure communications between type of sample being presented for authentication must also
internet applications. be included. It is possible for a biometric matching system
to support multiple technology types, so more than one
Adding PAKE to TLS would enable all users to benefit sample and type may be presented by the user for
from "secure password authentication" in "any application authentication. The format and processing of these values
that makes use of TLS", without requiring users to possess should be standardized by SG17 to promote vendor
X.509 certificates [16]. ITU standardization of PAKE usage interoperability.
in the handshake would allow "standard TLS mechanisms
for key derivation and secure record-layer communication" Biometric matching may be performed on a system remote
to continue being used [16]. An ITU-T standard for using to the server authenticating the user. In this case, the
PAKE in TLS would provide users the convenience and encrypted user authentication-attempt must also identify the
low cost of passwords and the security benefits of mutual location of the remote matching service for each biometric
authentication. By making PAKE available to users as a type being presented for authentication. The unique
PKI alternative, the threats to users from phishing and man- biometric reference template identifier associated with the
in-the-middle attacks that are known to plague TLS could user enrolled in a biometric system, and the type of the
be addressed. biometric sample should also be included. A standardized
schema for exchanging this information as an encrypted
attribute should be standardized by SG17.
– 159 –
2019 ITU Kaleidoscope Academic Conference
– 160 –
ICT for Health: Networks, standards and innovation
type and sample, and the user-enrolled biometric reference images" that can "represent movements completely
template identifier at the BSP. This feature would provide unrelated to any language" [10].
user portability of their biometric credential across multiple
devices with only a single user enrollment. As described by Fong, Zhuang and Fister, these types
include footsteps, "finger positions and hand posture" [19].
Recommendation ITU-T X.1080.0 provides an informal Some biometric technology types are considered to be
CMS specification for data protection based on IETF RFC ‘weak’ for general use. For a constrained population living
5652. SG17 should revise Recommendation ITU-T in an in-home healthcare environment, who may have been
X.1080.0 to reference ITU-T X.894 CMS, whose syntax authenticated on entry, these types may offer value for user
complies with the current ASN.1 standards. This change identification and authentication, especially when
will allow ITU-T X.1080.0 adopters to eliminate the use of telemedicine and telemonitoring services are assisted by
RFC 5652 syntax that is “based on X.208, the deprecated robotics.
1988 version of ASN.1 that was withdrawn as a standard in
2002” [15]. Adoption of ITU-T X.894 will allow any of the REFERENCES
ASN.1 encoding rules to be used, removing the IETF one-
rule restriction. ITU-T X.894-based ITU-T X.1080.0 [1] Astbrink, G., Shabbir, M., Giannoumis, G.A. (2018)
implementations will also gain new options for 1.3 Billion Reasons for Making Technology
telebiometric data protection, including field level Accessible. Internet Governance Forum Workshop
tokenization and a SigncryptedData type that can be used to Session (WS) #45.
replace SignedData and EnvelopedData.
[2] Calvaresi, D., Cesarini, D., Sernani, P., Marinoni, M.,
The TLS protocol is widely used and well suited for server et al. (2017). Exploring the ambient assisted living
to server mutual authentication, since both communicating domain: a systematic review. Journal of Ambient
parties are likely to possess digital certificates. However, Intelligence and Humanized Computing, 8(2), 239-
when individuals must authenticate to a server using TLS, 257.
they must often rely on a password. The lack of users with
certificates makes mutual authentication with TLS rare and [3] ITU-D. (2019). mHealth for NCD - BeHe@lthy
successful phishing attacks on users likely. BeMobile. ICT Applications.
ITU should standardize a profile of the TLS handshake [4] ICT Consultation. (2013). The ICT opportunity for a
protocol that ensures secure access for mobile device users. disability-inclusive development framework.
This profile should support mutual authentication based on
user passwords protected by PAKE. An ITU standard for [5] Griffin, P. H. (2015). Security for ambient assisted
TLS should enable the use of PAKE extensions that provide living: Multi-factor authentication in the internet of
multifactor user authentication with samples collected from things. In 2015 IEEE Globecom Workshops (GC
the biometric sensors readily available on smart mobile Wkshps) (pp. 1-5). IEEE.
devices. This new TLS standard would allow mobile users
to enjoy the benefits of strong, two-factor user [6] Hamdi, O., Chalouf, M. A., Ouattara, D., & Krief,
authentication and mutual authentication without the cost of F. (2014). eHealth: Survey on research projects,
digital certificates and the risk of being phished. comparative study of telemonitoring architectures
and main issues. Journal of Network and Computer
A revision of Recommendation ITU-T X.1081 should Applications, 46, 100-112.
include consideration of non-biometric telemedicine
devices. This revision could be achieved through a new [7] Recommendation ITU-T X.1035 Password-
normative annex or as a separate standard. ITU-T X.1081 authenticated key (PAK) exchange protocol, 2007.
security aspects could incorporate work being proposed in
ITU-T X.tas, since, as this paper has shown, it is possible [8] ISO/IEC 11770-4 Key Management – Mechanisms
for a biometric sensor to also collect user knowledge for Based on Weak Secrets, 2017.
use as an authentication factor. The X.tas: Telebiometric
authentication using speaker recognition standardization [9] Engler, J., Karlof, C., Shi, E., Song, D. (2009). Is it
should especially consider the BAKE protocol extension to too late for PAKE? In Web 2.0 Security and Privacy
PAKE. (W2SP) 2009.
The X.tas work should be broadened to include other types [10] Griffin, P. H. (2015). Biometric knowledge
of biometrics for which "knowledge extraction can mine extraction for multi-factor authentication and key
something-you-know information from biometric sensor exchange. Proceedings of the 2015 Complex
data" [10]. These biometric types could be used to support Adaptive Systems Conference, Missouri Institute of
the operation of PAKE and its extensions. These types of Science and Technology, in San Jose, California.
biometrics include collected "user gestures as binary video
– 161 –
2019 ITU Kaleidoscope Academic Conference
[11] Recommendation ITU-T X.894 Information [16] Manulis, M., Stebila, D., Kiefer, F., & Denham, N.
technology - Generic applications of ASN.1: (2016). Secure modular password authentication for
Cryptographic message syntax, 2018. the web using channel bindings. International
Journal of Information Security, 15(6), 597-620.
[12] Makeig, S., Enghoff, S., Jung, T. P., & Sejnowski, T.
J. (2000). A natural basis for efficient brain-actuated [17] Recommendation ITU-T X.1081 The telebiometric
control. IEEE Transactions on Rehabilitation multimodal model - A framework for the
Engineering, 8(2), 208-211 specification of security and safety aspects of
telebiometrics, 2011.
[13] LaFleur, K., Cassady, K., Doud, A., Shades, K.,
Rogin, E., & He, B. (2013). Quadcopter control in [18] Recommendation ITU-T X.1080.0 Telebiometrics -
three-dimensional space using a noninvasive motor Access control for telebiometrics data protection,
imagery-based brain–computer interface. Journal of 2017.
neural engineering, 10(4), 046003.
[19] Fong, S., Zhuang, Y., & Fister, I. (2013). A
[14] Bhattacharya, S. (2016). Optimization of biometric authentication model using hand gesture
Electroencephalograph-Based Classification for images. Biomedical engineering online, 12(1), 111/.
Imaginary Motion Brain Computer Interface Study.
– 162 –
CYBER-SAFETY IN HEALTHCARE IOT
Duncan Sparrell1
1
sFractal Consulting, United States
IoHT is the merging of the information technology (IT) Looking at real-world non-cyber failures and disasters gives
world with the operational technology (OT) world to bring great insights into what an attacker could do. As an example,
about increased innovation, efficiency and quality of there were 8 failures that led to the Deepwater Horizon / BP
healthcare [5-11]. oil spill in the Gulf of Mexico [24]. Seven of the 8 problems
were with actuators, sensors or decision algorithms, all
As IoT pervades the healthcare industry, cybersecurity in failures that could also be caused by a cyberattack. Even the
IoHT must evolve both to recognize new threats but also to 8th cause, faulty cement, could be caused by a supply chain
recognize different consequences i.e. impact on patient attack. The analysis does not have to be of mega-disasters. A
health. This paper will discuss trends that when combined similar analysis could be done of the typical failures in
will make IoHT safer and demonstrate the important role almost any manufacturing process, or to any medical
cybersecurity standards will play as cybersecurity evolves simulation. This points out the need for domain-specific
knowledge when assessing IoHT cybersecurity. This is not because the new technology itself becomes a low-probability,
something that could be accomplished by the IT department high-impact threat vector. A better alternative would be to
or outside cybersecurity experts as it requires healthcare increase the upfront costs to mitigate the catastrophic case
domain-specific knowledge. with other controls and/or insurance.
3. RISK ANALYSIS
4. AUTOMATION
Figure 1– Loss Exceedance Curve The defense has not kept pace with the offense. Attackers
can, on average, breach a system in seconds or minutes
The real advantage comes from comparing two alternatives, whereas it takes defenders weeks, months or even years to
as in Figure 2. The existing base case is shown in blue. The respond [39]. The attacker utilizes sophisticated, adaptive,
red alternative costs $500k to implement and decreases the automated tools and the defender reactively responds with
50% loss by over $4M. But note this particular alternative manual, slow, uncoordinated tools and processes. The
also increases the chance of a catastrophic loss by a factor of defense must automate to operate at the speed of the offense.
10. Although this may seem counterintuitive, it is common Automation is a win/win, cheaper AND better.
– 164 –
ICT for Health: Networks, standards and innovation
Integrated Adaptive Cyber Defense (IACD) is a research Threat sharing has not been as effective as envisioned and
effort jointly funded by the US Department of Homeland has run into obstacles [44]. The initial STIXTM version 1
Security (DHS) and the US National Security Agency (NSA), specification was not readily accepted by industry. Version
in collaboration with The Johns Hopkins University Applied 2 is now available with significant improvements addressing
Physics Lab (JHU/APL) and industry. IACD seeks "to many of the industry concerns and is gaining broader
revolutionize cybersecurity ... through the universal adoption. In the US, the passage of the Cybersecurity
automation and interoperability of cyber defenses" [40]. Information Sharing Act [45] incentivized sharing by
IACD is an effort to get humans from ‘in the loop’ to ‘on the removing certain liabilities. Significant progress has been
loop’. The focus is product agnostic interoperability by made in this area, particularly in certain industry groups
decoupling functions and standardizing interfaces. IACD working with their Information Sharing and Analysis
seeks to create an adaptable, extensible ecosystem “to Centers (ISACs).
dramatically change the timeline and effectiveness of cyber
defense via integration, automation, and information To maximize the benefit of STIXTM involves not only
sharing.” This can be accomplished by decoupling the sharing ‘what happened’ but also deciding ‘what to do’,
functions and standardizing the interfaces between functions. called Courses of Action (CoAs). To effectively share CoAs,
IACD categorizes security functionality into: standards for both atomic actions and for a playbook
including the decision points and the flow of the atomic
• sensing: gathering all the data actions are required.
• sense-making: correlating and analyzing data,
transforming into information, knowledge and Another IACD objective is the standardization of the
intelligence command and control (C2) language for security
• decision-making: deciding what to do technologies, the atomic actions in a CoA. For example,
• acting: sending the actual commands. firewalls have existed for over 25 years yet the ‘word’ used
to prevent a packet passing through the firewall could be:
‘drop’, ‘deny’, ‘reject’, ‘block’, ‘blacklist’, etc. depending on
Gap analysis of the IACD work has led to standards activities which implementation is used. This is compounded across
in sharing threat intelligence, sharing courses of action, and many security technologies with new ones being
in a common command and control (C2) language. continuously added. This poses several problems for the user
community. It is hard to share CoAs when two organizations
One of the IACD objectives is the sharing of threat data use different vendor products. The cost of retooling
among interested, trusted parties. DHS started an industry disincentivizes changing vendors or adding alternative
forum on threat sharing that evolved and moved into OASIS, vendor products. This was less of an issue when security
a non-profit standards development organization. The technology was physical appliances. In today’s cloud/IoT
OASIS Cyber Threat Intelligence (CTI) Technical environment, switching vendors and/or adding new
Committee (TC) [41] created the Structured Threat technologies should be quicker and easier.
Information Expression (STIXTM) [42] and the Trusted
Automated Exchange of Intelligence Information (TAXII TM) The OASIS Open Command and Control (OpenC2)
specifications [43] to address the need to model, analyze and Technical Committee was founded to standardize machine-
share cyberthreat intelligence. Figure 3 shows an example of to-machine command-and-control to enable cyber defense
STIXTM ontology. system interoperability at machine speed [46,47]. Just as
automation has a fundamental impact on attacker economics,
OpenC2 will have a fundamental impact on defender
economics [48,49].
– 165 –
2019 ITU Kaleidoscope Academic Conference
REFERENCES
– 166 –
ICT for Health: Networks, standards and innovation
[4] USFDA, “FDA informs patients, providers and [14] J. Corman, et al. “Hacking Healthcare Live”, RSA
manufacturers about potential cybersecurity Conference USA, San Francisco, April 25, 2018.
vulnerabilities for connected medical devices and Available:
health care networks that use certain https://www.rsaconference.com/videos/live-
communication software”, USFDA News Release, simulation-a-medical-device-hack-and-patient-
1-Oct-2019. Available https://www.fda.gov/news- codes-rsac-2018 .
events/press-announcements/fda-informs-patients-
providers-and-manufacturers-about-potential- [15] A. Jenna, et al, “Delays in Emergency Care and
cybersecurity-vulnerabilities . Mortality during Major US Marathons”, New
England Journal of Medicine, Vol 376, 2017, pp.
[5] A. Martinez-Balleste, “Exploring IoT Industry 1441-50. available:
Applications: Paving the Way for Future IoT https://www.nejm.org/doi/pdf/10.1056/NEJMsa161
Applications in Healthcare”, IEEE. Available 4073 .
https://ieeexplore.ieee.org/servlet/opac?mdnumber=
EW1485 . [16] A. Morse, “Investigation: Wannacry cyber attack
and the NHS”, UK National Audit Office, April 25,
[6] A. K. Alharam and W. El-madany, “Complexity of 2018. Available: https://www.nao.org.uk/wp-
Cyber Security Architecture for IoT Healthcare content/uploads/2017/10/Investigation-WannaCry-
Industry: A Comparative Study”, 5th International cyber-attack-and-the-NHS.pdf .
Conference on Future Internet of Things and Cloud
Workshops, 2017. Available [17] K. Chinthapalli, “The hackers holding hospitals to
https://ieeexplore.ieee.org/document/8113801 . ransom”, BMJ 2017;357:j2214. Available
https://doi.org/10.1136/bmj.j2214 .
[7] H. Breivold and K. Sandstrom, "Internet of Things
for Industrial Automation -- Challenges and [18] C. Sienko, “Ransomware Case Studies”, Available
Technical Solutions", 2015 IEEE International https://resources.infosecinstitute.com/category/healt
Conference on Data Science and Data Intensive hcare-information-security/healthcare-attack-
Systems, Sydney, Australia, 2016, pp. 532-539. statistics-and-case-studies/ransomware-case-studies-
hollywood-presbyterian-and-the-ottawa-hospital/ .
[8] L. Probst, et al., “Digital Transformation
Scoreboard 2018”, European Commission, [19] "Inside the New York hospital hackers took down
Luxembourg 2018. Available: for 6 weeks". 18-Aug-2917. Available
https://ec.europa.eu/growth/tools- https://www.cbsnews.com/news/cbsn-on-
databases/dem/monitor/sites/default/files/Digital%2 assignment-hackers-targeting-medical-industry-
0Transformation%20Scoreboard%202018_0.pdf . hospitals/ .
[9] K. Pretz, “How the Industrial Internet of Things is [20] B. Lovelace, L. Gurdis, “Hospital CEO forced to
Changing the Face of Manufacturing”, The pay hackers in bitcoin now teaches others how to
Institute, IEEE, Feb. 23, 2018. available: prepare for the worst”. Available
http://theinstitute.ieee.org/technology-topics/smart- https://www.cnbc.com/amp/2018/04/06/hosptial-
technology/how-the-industrial-internet-of-things-is- ceo-forced-to-pay-hackers-in-bitcoin-now-teaches-
changing-the-face-of-manufacturing. others.html .
[10] ”Understanding the Industrial Internet of Things”, [21] C. O’Gara, "Would You Have Surgery in a Hacked
Transmitter, IEEE, May 8, 2018. available: Hospital?", 7-Oct-2019. Available
https://transmitter.ieee.org/understanding-industrial- https://www.secureworldexpo.com/industry-
internet-things/. news/would-you-agree-to-surgery-in-a-hacked-
hospital .
[11] “Revolutionizing the IIoT Industry One Streetlight
at a Time”, Spectrum, IEEE. Available: [22] L. Beers, "Multiple Victorian hospitals hacked in
https://spectrum.ieee.org/telecom/internet/revolutio suspected ransomware attack". 1-Oct-2019.
nizing-the-iiot-industry-one-streetlight-at-a-time . Available
https://7news.com.au/news/health/multiple-
[12] J. Corman, “Bits & Bytes Meet Flesh & Blood”, victorian-hospitals-hacked-in-suspected-
presented at LASCON, Austin, TX, Oct 23, 2015. ransomware-attack-c-481074 .
[13] E. Perakslis, “Cybersecurity in Healthcare”, New [23] D. Sparrell, “Let it be hacked,” Code Beam SF
England Journal of Medicine 371(5):395-7, July- 2018, San Francisco, USA, 15-17-March-2018,
2014. available: https://youtu.be/NYkwvVKlbU8 .
– 167 –
2019 ITU Kaleidoscope Academic Conference
[24] J. Mullins, “The eight failures that caused the Gulf [33] G. Zheng, G.Zhang, W.Yang, C.Vali, R. Shankaran,
oil spill”, New Scientist, Sep. 8, 2010 available: M.Orgun, “From WannaCry to WannaDie: Security
https://www.newscientist.com/article/dn19425-the- Trade-offs and design for implantable medical
eight-failures-that-caused-the-gulf-oil-spill . devices”, IEEE, 17th International Symposium on
Communications and Information Technologies,
[25] S. Morgan, “2018 Cybersecurity Market Report”, 25-27 Sept 2017. Available
Cybersecurity Ventures, May 31, 2017,\. Available: https://ieeexplore.ieee.org/document/8261228 .
https://cybersecurityventures.com/cybersecurity-
market-report/ . [34] J. Freund, J. Jones, “Measuring and Managing
Information Risk: A FAIR Approach”, Butterworth-
[26] S. Morgan, “Cyber Crime Costs Projected to Reach Heinemann, 2014. ISBN-13: 978-0124202313.
$2 Trillion by 2019”, Forbes, Jan. 17, 2016.
available: [35] Open Group, “Risk Taxonomy (O-RT)”, Open
https://www.forbes.com/sites/stevemorgan/2016/01/ Group Technical Standards, Version 2.0, 18-Oct-
17/cyber-crime-costs-projected-to-reach-2-trillion- 2013. Available
by-2019 . https://publications.opengroup.org/c13k .
[27] S. Morgan, “One Million Cybersecurity Job [36] Open Group, “Risk Analysis (O-RA)”, Open Group
Openings in 2016”, Forbes, Jan. 2, 2016. Available: Technical Standards, 18-Oct-2013. Available
https://www.forbes.com/sites/stevemorgan/2016/01/ https://publications.opengroup.org/c13g .
02/one-million-cybersecurity-job-openings-in-
2016/. [37] DoCRA Council, “Duty of Care Risk Analysis
Standard”, Version 0.5, April, 2018. Available
[28] P. Sawers, “Global cybersecurity workforce to be https://docraprod.wpengine.com/wp-
short 1.8 Million by 2022”, Venture Beat, June 7, content/uploads/2018/04/Duty-of-Care-Risk-
2017. Available: Analysis-Standard-v05.pdf .
https://venturebeat.com/2017/06/07/global-
cybersecurity-workforce-to-be-short-by-1-8- 38] C. Cronin, “The Questions a Judge will ask you
million-personnel-by-2022-up-20-on-2015/ . after a Data Breach”, Borderless Cyber, Oct 2019.
Available https://us19.borderlesscyber.org/program-
[29] C. Bryant, “How to justify your IT Security Budget”, schedule/program/108/plenary-session-keynote-the-
17-Oct-2016. Available questions-a-judge-will-ask-you-after-a-data-breach .
https://www.risklens.com/blog/how-to-justify-your-
it-security-budget/ . [39] Verizon, “2016 Data Breach Report”, available:
https://www.verizonenterprise.com/resources/report
[30] D. Hubbard and R. Seiersen, “How to Measure s/rp_DBIR_2016_Report_en_xg.pdf .
Anything in Cybersecurity Risk”. Hoboken, NJ:
Wiley, 2016. [40] “Integrated Adaptive Cyber Defense” available:
https://www.iacdautomate.org/learn/ .
[31] C. Valasquez, et al., “Hybrid Loss Exceedance
Curve (HLEC) for risk assessment”, International [41] “OASIS Cyber Threat Intelligence (CTI) Technical
Symposium on Computational Civil Engineering, Committee (TC)”. Available: https://www.oasis-
Iasi, Romania, May, 2013. available: open.org/committees/tc_home.php?wg_abbrev=cti .
https://www.researchgate.net/profile/Alex_Barbat/p
ublication/259609767_Hybrid_loss_exceedance_cu [42] R. Piazza, J. Wunder, B. Jordan, “STIXTM Version
rve_HLEC_for_risk_assessment/links/0deec52ce53 2.0”, OASIS Committee Specification, 19 July
c507de4000000/Hybrid-loss-exceedance-curve- 2017. Available https://www.oasis-
HLEC-for-risk- open.org/standards#stix2.0 .
assessment.pdf?origin=publication_detail .
[43] J. Wunder, M. Davidson, B. Jordan, “TAXIITM
[32] M. Nederlof (1994) "Comparing probabilistic Version 2.0”, OASIS Committee Specification 01.
predictions with outcomes in petroleum exploration Available http://docs.oasis-
prospect appraisal". Nonrenewable Resources,v.3, open.org/cti/taxii/v2.0/cs01/taxii-v2.0-cs01.html .
No.3, 1994, p.183-189.
– 168 –
ICT for Health: Networks, standards and innovation
[44] C. Douris, “Cyber Threat Data Sharing Needs [54] C. Douris, “As Cyber Threats to the Electric Grid
Refinement”, Lexington Institute white paper for Rise, Utilities and Regulators Seek Solutions”,
Energy Industry, Aug 2017. Available: Forbes, Jan 16, 2018. Available:
https://www.lexingtoninstitute.org/cyber-threat- https://www.forbes.com/sites/constancedouris/2018/
data-sharing-needs-refinement/ . 01/16/as-cyber-threats-to-the-electric-grid-rise-
utilities-regulators-seek-solutions/#624559c1343e .
[45] U.S. Senate. 114th Congress, (2015, Oct. 27). “The
Cybersecurity Information Sharing Act”. Available: [45] W. Peters, “IACD Overview and IACD
https://www.congress.gov/bill/114th- Framework”, IACD Community Day, Laurel,
congress/senate-bill/754 . Maryland, 23-March-2017.
[46] T. Spring, “NSA-Backed OpenC2.org Aims to [56] P. Royer, “Orchestration and Automation”,
Defend Systems at Machine Speed”, Threatpost, OpenC2 Forum, 29-Sep-2016.
June 22, 2017. Available https://threatpost.com/nsa-
backed-openc2-org-aims-to-defend-systems-at- [57] A. Bradbury, “OpenC2 and OrchID – Using
machine-speed/126454/ . OpenC2 in a Managed Security Services Provider”,
OpenC2 Forum, 29-Sep-2016.
[47] J. Romano, D. Sparrell, “Open Command and
Control (OpenC2) Language Specification Version [58] S. Yu, “The Future of Orchestration” Integrated
1.0”, OASIS Committee Specification, 11-July- Cyber, Laurel, Maryland, 1-2-May-2018.
2019. Available https://www.oasis-
open.org/standards#oc2-ls-v1.0 . [59] “Symantec Integrate Cyber Defense Exchange”,
Symantec, 2019. Available
[48] D. Sparrell, “Defense at Machine Speed,” https://www.symantec.com/content/dam/symantec/d
Borderless Cyber USA Conference, New York, NY ocs/data-sheets/integrated-cyber-defense-exchange-
USA, June 21-22, 2017. en.pdf .
[49] J. Hiner, “The four types of cybersecurity threats [60] M. Sherif, D. Sparrell, “Standards and innovation
and a formula to fight them”, June 22, 2017. in telecommunications”, Communications
Available: Magazine, IEEE Volume 30, Issue 7, Jul 1992
https://www.techrepublic.com/article/the-4-types- Page(s):22- 29.
of-cybersecurity-threats-and-a-formula-to-fight-
them/ . [61] J. Gurney, “OpenC2 AWS Actuator”. Available:
https://github.com/newcontext-oss/openc2-aws-
[50] “OASIS Collaborative Automated Course of Action actuator .
Operations (CACAO) for Cyber Security Technical
Committee”. Available https://www.oasis- [62] W. Trost, “OpenC2 AWS Reference
open.org/committees/cacao/charter.php . Implementation”. Available:
https://github.com/att/openc2-aws .
[51] “The Cyber-Physical Systems Revolution” in IEEE
Computing Edge, Vol. 4, No. 5, May, 2018, pp. 10- [63] M. Stair, D. Hamilton, “OpenC2 Lycan for Python”.
13. Available: Available: https://github.com/oasis-open/openc2-
https://www.computer.org/cms/computing-edge/ce- lycan-python .
may18-final.pdf .
[64] E. Ortiz, “A Web Self Service Application for
[52] R. Corrigan, “California Energy Systems for the Generating OpenC2 Commands”, available:
21st Century (CES-21)”, presented at IACD https://github.com/netcoredor/openc2-cmdgen .
Community Day, Laurel, Maryland, March 23,
2017. Available: [65] E. Ortiz, “GUI Tool for OpenC2 Command
https://drive.google.com/file/d/1NWtwUN4Uv2LFk Generation”, DefCon 26 Demolabs, Las Vegas,
TUdTfL5q-TfWATDgUMi/view . Nevada, 12-Aug-2018. Available:
https://www.defcon.org/html/defcon-26/dc-26-
[53] J. Van Randwyk, et al., ‘California Energy Systems demolabs.html .
for the 21st Centruy 2016 Annual Report”,
Lawrence Livermore National Laboratory, March [66] D. Sparrell, “OpenC2 API Simulator (OCAS)”.
29, 2017. Available: https://e-reports- Available: https://github.com/oasis-open/openc2-
ext.llnl.gov/pdf/878504.pdf . ocas .
– 169 –
2019 ITU Kaleidoscope Academic Conference
[67] D. Kemp, “JSON Abstract Data Notation (JADN)”, [71] “Healthcare Proof of Concept”, NTIA Software
Available: https://github.com/oasis-open/openc2- Component Transparency Working Group, 27-June-
jadn . 2019. Available
https://www.ntia.doc.gov/files/ntia/publications/wor
[68] B. Essing, “OpenC2 Lycan for Java”. Available: king_group_healthcare_proof_of_concept_-_2019-
https://github.com/oasis-open/openc2-lycan-java . 06-27_-_v1_002.pdf .
– 170 –
SESSION 8
S8.1 Technical and legal challenges for healthcare blockchains DQG smart contracts
S8.2 Design of a credible blockchain-based e-health records (CB-EHRs) platform
S8.3 The GDPR transfer regime and modern technologies
TECHNICAL AND LEGAL CHALLENGES FOR HEALTHCARE BLOCKCHAINS
AND SMART CONTRACTS
Steven A. Wright1
1
Georgia State University
and legal risks of healthcare blockchains are then presented and operations of the blockchain are truly immutable, then it
in section 5. cannot evolve to reflect the changing needs from its users
and the commercial / legal environment. A blockchain is
2. HEALTHCARE BLOCKCHAINS simply a distributed data structure that does not create value
until it is applied in a particular economic context. In their
A variety of healthcare applications have been proposed survey, [17] categorized the health data referenced by the
[3],[11],[12],[13] including drug counterfeiting prevention3, blockchain as financial, database queries, transaction records,
clinical trials, public healthcare management, longitudinal ambient temperature, consent forms, clinical trial records,
healthcare records, automated health claims adjudication, personal records, medical records, sensor data and/ or
online patient access, sharing patients’ medical data, user- location data. Medical records, sensor data and personal
oriented medical research, precision medicine and smart records were the most frequently identified data categories.
contracts to improve the credibility of medical research. From 19 companies, [18] identified healthcare data being
Healthcare applications are being created; [14] identified managed as electronic medical records, electronic health
nine different healthcare applications on Ethereum and two records, and personal health records. electronic medical
applications on Hyperledger. [15] points out that healthcare records (EMRs) contain clinical data related to a specific
applications must balance patient care with information patient stored by the responsible healthcare provider [19].
privacy, access, completeness and cost. The designers of Personal health records (PHR), store data collected by
healthcare information systems may have a number of patients monitoring their health conditions, using their smart
different requirements associated with the systems they are phones or wearable devices [20]. Electronic health records
designing, and the criteria for applying blockchain are not (EHRs), for example, are designed to allow patient medical
always clear [16]. Applications may be a good fit for history to move with the patient or be made available to
blockchain according to [15] if: multiple stakeholders are multiple healthcare providers [21]. [22] proposed a
contributing; more trust is required between parties than blockchain-based EHR as a mechanism to share data
currently exists; an intermediary could be removed or between PHRs and EHRs; but it did not resolve the trust and
omitted to increase trust or efficiency; there is a need for access control mechanisms required. The blockchain PHR
reliable tracking of activity and there is a need for data to be feasibility study in [23] revealed some challenges due to the
reliable over time. In their survey, [17] categorized size of the data records (which impacted both the
healthcare blockchain application areas as clinical trials, performance of the system and operational costs) as well as
biomedical databases, health records, medicines supply, privacy aspects. A blockchain as a database of personalized
medical insurance, wearables and embedded or mhealth, records is likely structured significantly differently than a
with the majority of papers on health records. [14] also noted cryptocurrency blockchain (e.g. separate blockchains for
electronic medical records as the most common area with an personal records c.f. common blockchains for fungible
increasing numbers of papers. In their survey, [17] identified currencies of commodities). The data lifecycle in the
the following rationales for using blockchains in healthcare healthcare blockchain would depend on the use cases that are
applications: access control, non-repudiation, data needed to support the healthcare actors.
versioning, logging, data provenance, data auditing and data
integrity. Access control, data integrity and logging were the 2.2 Actors in healthcare blockchains
most prevalent rationales. [14] identified the benefits of
blockchains for healthcare applications as decentralization, In considering the role of blockchain in a value chain, [24]
improved data security and privacy, health data ownership, identified the need to decide not just what valuable
availability/robustness, transparency and trust, and data information to be captured in the blockchain, or what
verifiability. The data stored in and the actors operating on a operations need to be performed by the blockchain, but also
healthcare blockchain lead to some differences (c.f. for whom the blockchain should be accruing this value. In
cryptocurrencies) in required blockchain features. EMR applications of blockchains, as an example, the value
of the recorded data in the blockchain is not intrinsic; rather
2.1 Data in healthcare blockchains it comes from the association with a particular user e.g. a
patient or a medical practitioner. As a result, the transactions
Blockchains maintain timestamped and cryptographically adding data to the blockchain are most likely not anonymous
signed blocks of transaction data. The data integrity transactions as in bitcoin; rather using the hash as an
mechanisms of the blockchain provide limitations in electronic signature to establish the provenance of the data.
operational flexibility and governance. If the data structures While medical actors may have need of the broader range of
– 174 –
ICT for Health: Networks, standards and innovation
blockchain applications such as those identified by [3], in against dimensions of time to achieve consistency, system
this paper the focus is on application areas that are uniquely availability, failure tolerance, scalability, latency,
related to healthcare, rather than application areas like auditability, liveliness, denial of service resistance and
payments for services by bitcoin that may be broadly system complexity. Standardized benchmarks and targets for
applicable across a number of industries. [25] identified the healthcare blockchain performance have not yet been
major stakeholders in digital health systems as patients, the identified.
public, healthcare professionals and health administrators;
however, regulatory agencies and legal systems may also Software engineering has developed tools and methods to
need to be able to operate on or interact with the healthcare support the development and operation of software systems,
blockchains. Provisions would also be needed for minors, but to date these are not optimized for blockchain systems.
access under a healthcare power of attorney, and in some [32] identifies the features and implementation challenges of
cases access after death by heirs. interoperability for healthcare blockchain applications and
proposes foundational software patterns to help address them.
3. TECHNOLOGY ISSUES [33] identifies blockchain oriented software engineering
challenges as new professional roles, security and reliability,
Technology issues can be seen as risks impeding design and software architecture, modeling languages and metrics, and
deployment of healthcare blockchains. Ethereum and proposes new directions for blockchain oriented software
Hyperledger were the most frequently mentioned blockchain engineering related to enhancement of testing and debugging
implementation technologies, found by [17] but only 2% of for specific programming languages and the creation of
the papers surveyed were reporting on implementations, so software tools for smart contract languages. [34] echoes the
healthcare blockchains are still at the early stages of adoption. call for further development of blockchain oriented software
There is not one blockchain but a variety of implementations engineering best practices and design patterns.
with different characteristics [26],[27] (even bitcoin has
forked). Identified technology challenges to the development 3.2 Identity and trust issues in healthcare
of healthcare blockchains include interoperability, security blockchains
and privacy, scalability, speed and patient engagement [14].
Interoperability, scalability and speed are characteristics of Many of the benefits (e.g. improved data security and
the software implementation of healthcare applications on privacy, health data ownership, transparency and trust, data
the blockchain. The degree of patient engagement can be verifiability, non-repudiation, data provenance) sought from
significantly impacted by not just the implementation and healthcare blockchains rely on some form of trust. To
trust issues, but also the usability of the system and the achieve their healthcare objectives, patients need to trust
overall user experience with the healthcare blockchain. healthcare providers. Patients and healthcare professionals
Security, privacy and trust issues reflect concerns about not need to trust the validity of data used for diagnosis and
just the implementation, but the processes for assuring the treatment. Trust has been defined in many different ways by
users can trust the blockchain and its associated software, as different researchers. [35] proposed an interdisciplinary
well as the organizational and legal context. [28] points out model of trust involving components for disposition to trust,
that health information technology in general needs to institution-based trust, trusting beliefs and trusting intentions.
consider not just clinical information, but also socio- Since literally everyone is potentially a patient, and patients
technical concepts of value and trust concepts to be are actors in most healthcare blockchains, addressing all of
successful. those trust components may be necessary for the broad
adoption of healthcare blockchains; not all of them, however,
3.1 Implementation issues in healthcare blockchains are directly solved by blockchains. Disposition to trust and
institution-based trust lie more in the realm of psychological,
Healthcare blockchain applications, whether directly on the sociological and economic concepts. Trusting beliefs and
blockchain, or smart contracts or DAOs, are all software; and intentions may be more manageable for healthcare
software bugs impact the functionality and quality of blockchains that are explicit about what actors can rely on
blockchain systems. [29] performed an empirical study of and for what purposes in the healthcare blockchain use cases.
over one thousand bugs identified from 19 open source
blockchain systems categorizing them and studying their Because of the use of blockchain technology in the financial
resolution to determine that the frequency distributions of industry, and the associated loss risks, the security of
bug types share similar trends across the studied projects, blockchains and related smart contracts have received
implying that these would apply to healthcare blockchain significant attention. In a survey on the security of
applications also. They noted that security bugs took the blockchain systems [36] proposed a taxonomy of the targets
longest median time to fix and that more than 35% of of security attacks. For Blockchain 1.0 (cryptocurrency)
performance bugs took more than a year to fix. While not blockchains the targets were the blockchain consensus
providing specific metrics, [30] identified dimensions for the mechanism, the public key encryption scheme, the
quality of blockchain implementations as including security, cryptocurrency application criminal activity (e.g. money
privacy, throughput, size and bandwidth, performance, laundering ransomware), the transaction verification
usability, data integrity and scalability. [31] surveyed the mechanism, and transaction design flaws that could lead to
performance characteristics of six different blockchains privacy leakage. For Blockchain 2.0 (smart contract)
– 175 –
2019 ITU Kaleidoscope Academic Conference
blockchains, the targets were the smart contract application data contained or referenced in the blockchain record is
(criminal smart contracts), program design flaws, program protected by some form of intellectual property a smart
implementation flaws, smart contract virtual machine (e.g. contract associated with the blockchain could provide an
Ethereum Virtual machine) design flaws. These same automated market for efficiently licensing such content.
components would be risks for healthcare blockchains, Where the blockchain acts as a substitute for the law, there
though the incentives for exploitation would be different is no backstop of traditional legal enforcement. This may be
than for fungible commodities or currencies. attractive in regions where there is no rule of law, or legal
enforcement is weak. As an example of a substitute, the UN
[31] analyzed six types of blockchains to identify the conducted a successful trial using blockchain to track food
mechanisms that they used to implement traditional aid to refugees [40]. The challenge for such systems is the
information security principles of confidentiality, human actors interfacing with the blockchain system, and
information availability, integrity, non-repudiation, their incentives (or the lack of them) for participation.
provenance, pseudonymity and selective disclosure, with
confidentiality and selective disclosure being the least 4.1 Legal entities in healthcare blockchain
supported principles. Data security and privacy, however, architectures
have been identified as key objectives for healthcare
blockchains, and the lack of support for these features would The law covers relations among people and the things they
reduce trust in these systems. Confidentiality features can be own. At least since the industrial revolution, the law will
built on top of the blockchain using smart contracts. [37] consider human beings (or other legal persons) responsible
proposed a system for sharing medical records using for their machines’ acts. While blockchains may be more
permissioned blockchains for access control and smart secure than other approaches, courts can apply existing legal
contracts for monitoring and logging access violations but mechanisms to decide which parties bear the losses and
did not encrypt the underlying records for confidentiality. responsibility for damages. Legal risks do not vanish if
[38] proposed a mechanism for secure storage of medical healthcare services are provided or supported through
records for use with blockchains. Most blockchains require blockchains and smart contracts, etc. Whether DAOs could
some entity in the role of a “miner” to maintain the operation eventually rise to the status of a legally recognized person
of the blockchain through consensus decisions for remains to be seen.
blockchain consistency, blockchain checkpointing, etc. but
simple blockchains do not assure confidentiality of The actors that control the governance of the blockchain are
blockchains during mining operations. While basic not necessarily those using the blockchain. Disruptive
blockchain functionality excels at assuring integrity, evolution could strand users on an unsupported fork of the
additional capabilities (different to cryptocurrencies) will blockchain. The Ethereum and Hyperledger blockchain
likely be required to monitor and assure actors’ requirements systems used in a number of healthcare blockchain
for confidentiality and selective access. Confidentiality and applications are both open source projects that have some
privacy considerations in healthcare use cases may require form of governance through the open source community;
additional emerging crypto-technologies to enable patients’ open source, however, is a gift economy which may be
control of their data. challenged to timely respond to some users’ needs for
evolution and support of the blockchain. Private blockchains
4. LEGAL ISSUES whether organized for profit, or as non-profit consortiums
can provide an entity to control the evolution of the private
Legal issues can be seen as risks impeding design and blockchain, but at the cost of centralizing the function on that
deployment of healthcare blockchains. Legal systems have entity (e.g. what happens if that entity fails?). Decred 4 and
geographic boundaries, but the distributed nature of Tezos5, in contrast, build in governance mechanisms for
blockchains can cross those boundaries. Participants in evolution of their blockchains.
blockchains that cross the boundaries of different legal
systems may be subject to foreign jurisdiction. Both the legal A healthcare blockchain application could rise to the level of
system and blockchains can promote trust or undermine it. a smart contract; with autonomous (workflow) actions
[39] notes that blockchain can act as supplement, triggered by transactions as programmed in the terms of the
complement or substitute for the law. Where the existing smart contract running on the blockchain. A regular contract
trust architecture is generally functional, the blockchain would identify the parties involved and their roles or actions
application can act as an additional (supplementary) layer required as part of the contract and similarly a smart contract
subject to established legal rules, e.g. by enhancing existing defines the actors and roles associated with the contract. [41].
messaging or transaction systems with authenticated While the roles and responsibilities of actors in a smart
messages or transactions. Where the existing trust based on contract can be changed at design time, they cannot be
the legal system is insufficient or breaking down, then the changed during operation. The entities designing the smart
distributed ledgers of the blockchain could complement and contract may not be the same as those creating instances of
extend the existing trust architecture. As an example, if the
4 https://decred.org . 5 https://tezos.com .
– 176 –
ICT for Health: Networks, standards and innovation
the smart contract; nor supporting the execution environment corporations to use distributed ledgers to maintain their share
of that smart contract. The parties are identified with their ownership registry. Vermont (12 V.S.A. § 1913) explicitly
blockchain accounts and transactions record obligations identified blockchain records as being admissible as
fulfilled under the smart contract. Smart contracts can evidence in court. Wyoming (§17-206) exempts open
replace error-prone human judgements with specific rule- blockchain tokens from registration as securities in contrast
based actions capturing best practices and by automating with views from regulatory agencies like CFTC and SEC
workflows eliminate the need for acknowledgements by treating cryptocurrencies as securities or commodities.
healthcare professionals [42]. While healthcare data may not seem like a commodity or
security, healthcare blockchain advocates may need to care
Perhaps inspired by Barlow’s declaration of the whether emerging regulatory language is over-inclusive.
independence of cyberspace [43], the decentralized,
anonymous and autonomous nature of some early blockchain Much of the existing legal precedents are based on criminal
implementations lead to proponents of DAOs which behavior around blockchain 1.0 cryptocurrency/ fintech, but
purported to have the operating software of the blockchain the data underlying healthcare blockchains is not a fungible
be an independent legal entity. Other automated trading financial asset. Market participants involved in distributed
systems have made automated transactions on behalf of their ledger systems like blockchain also must keep in mind
account owners for some time, but here the software itself conduct-related legislation implementing public policy
was purported to be the account owner. The law has a history including Antitrust, data protection, copyright, property and
of recognizing fictitious entities (e.g. corporations). tax, but, in comparison with cryptocurrencies, these areas are
Ownerless corporations have been proposed almost 30 years not anticipated to be of particular concern for healthcare
ago [44] and the enabling acts of several states would seem blockchains.
to permit zero-member LLCs [45] though such entities
would raise a number of social and political concerns [46]. The recent European Union General Data Protection
An early implementation of a DAO based on bitcoin did not Regulation (GDPR) creates additional legal protections for
fare well [47]. Despite efforts to transition governmental personal information in general, and other jurisdictions may
services to electronic form, from service of process to be considering similar regulations. Blockchains operated
judgement enforcement a purely software entity would be within the scope of those regulations may need additional
difficult to interface with a human and paper-driven legal design features to meet the GDPR requirements [49].
system. Beyond general data privacy regulation, healthcare
blockchains and smart contracts would be impacted by
Distributed ledger technologies could be considered by healthcare specific regulations (e.g. HIPPA [50] which has
courts in several legal systems as joint ventures or obligations for data privacy in contrast with many
partnerships between participants [48]. If a partnership were blockchain implementations that rely on a publicly visible
determined to exist, then joint and several liability would blockchain).
extend to all the partners. Joint and several liability means
the plaintiffs can collect any damages award from any one of Consider a healthcare smart contract executing on a
a group of partners. The extent of the partnership would be blockchain accepting data from an oracle reporting on a
determined by the court given the facts and circumstances of physiological condition through a smart phone, making some
the case. analysis of the data and reporting exceptional health
conditions as an alarm to a healthcare professional. The
4.2 Public law definition of a medical device [51] is sufficiently broad that
this healthcare smart contract could be considered a medical
Because blockchain technology is relatively new, there is not device and subject to medical device regulation. An error in
yet a lot of blockchain specific laws and regulation in place, such a smart contract medical device could create product
and what there is has been driven by cryptocurrencies rather liabilities.
than healthcare blockchain applications. Blockchain related
legislation is under consideration in a number of 4.3 Private law
states6.Arizona explicitly recognized electronic signatures
secured by blockchains as valid signatures and defined smart Private law differs in different legal systems, but generally
contracts secured through blockchains as valid electronic liabilities can arise through contracts, torts, partnerships or
records. A.R.S §44-7061 defines a smart contract as: “an specific legislation. Tort claims are particularly important
event-driven program, with state, that runs on a distributed, where there is no contractual liability. Joint tortfeasors are
decentralized, shared and replicated ledger and that can take two or more individuals with joint and several liability in tort
custody over and instruct transfer of assets on that ledger”. for the same injury to the same person or property. Whether
Delaware enacted legislation (D.C. §8-224) enabling healthcare smart contracts implemented on blockchain
6 http://www.ncsl.org/research/financial-services-and-commerce
/the-fundamentals-of-risk-management-and-insurance-viewed-
through-the-lens-of-emerging-technology-webinar.aspx.
– 177 –
2019 ITU Kaleidoscope Academic Conference
would be recognized as legitimate contracts would depend (see e.g. [58] using the wisdom of the crowd rather than
on the facts and circumstances in a particular jurisdiction, but arbitrators, but ethical issues may limit the applicability of
it appears that some jurisdictions may recognize smart this approach in healthcare smart contract disputes. While
contracts as legally binding [52] [53]. Contract law typically legislation typically directs courts to respect private
recognizes a contract established through a variety of arbitration decisions, such legislation might need extensions
different mechanisms (implied agreements, click through to support dispute resolution by smart contracts.
licenses etc.), however currently, there is not a lot of specific
legislation, regulation or legal precedent related to smart 5. CONCLUSIONS AND RECOMMENDATIONS
contracts. In the case of a private blockchain, there may be a
number of ancillary documents that users of the blockchain Emerging technologies often present a challenge or gap
may be required to agree to before using the service that can between technology advancements and the law. The gap
establish the legitimacy of the legal effect desired by the creates uncertainty that limits commercial investments and
smart contract terms. adoption of the technologies. The gap can be closed from
both sides, by designing technological solutions considering
Smart contracts are good at setting forth, ex ante, the existing legal issues and/or by changing legal or regulatory
anticipated conditions and consequences and then ensuring regimes to consider aspects of the technology solutions.
the consequences occur upon fulfillment. Legal contracts are Open source and standards can help eliminate some barriers
good at cleaning up the mess ex post, when, inevitably, to wider deployment.
things do not go according to plan [39]. Smart contracts, as
proposed by Szabo, have no explicit linkage between the Many of the technological risks of blockchain
smart contract and any external legal contracts; Ricardian implementations are not unique to healthcare use cases.
contracts explicitly link them [54]. Many smart contracts Healthcare specific blockchain design patterns and
need to interface with the outside world for information and performance benchmarks may eventually emerge through
those external data sources are called oracles. Oracles can be the open source communities as the healthcare use cases
humans though perhaps more typically some external entity evolve. Evolution of blockchain capabilities and design
makes its data feed available to the smart contract. patterns to support confidentiality and selective disclosure
may be particularly helpful for healthcare applications. The
Smart contracts can include “self-help” enforcement blockchain work on zero knowledge proofs (e.g. zcash) and
mechanisms where breaches of the contractual terms can be privacy preserving computation (e.g. [59]) seem promising
identified, and remedies enforced, through execution of the directions.
smart contract [55]. The completeness of such approaches is
one concern (e.g. did the parties identify all the possible ways While the evolution of public law and regulation will typically
in which the contract could be breached; can they agree on wait for action by deliberative bodies, private law related to
reasonable valuations for liquidation of the damages if the contracts may be more amenable to innovative approaches
form of the breach is not foreseen). Another concern arises from open source and standards. [39] identified safe harbors
when the contract is impacted by external events not foreseen and sandboxes as well as modularization or standardized terms
in the contract terms (e.g. a bankruptcy stay that impacts a in smart contracts as legal initiatives to reduce risk in
blockchain transaction, frustration or force majeure). Courts blockchain adoption. Healthcare is typically a regulated
may void contracts that exceed the limits of contract law, e.g. industry and the creation of safe harbors or sandboxes for
if the contract is unconscionable or violates some public healthcare blockchain applications would typically require
policy. Smart contracts would not provide mechanisms to actions by regulatory agencies. There are already examples of
escape the limits of contract law. Blockchains have been regulatory agencies such as the FDA experimenting with
hacked and forked which can impact the owners of assets blockchain technology to build expertise [60].
administered by the blockchain. Smart contracts may be
better positioned than the legal system to respond to such Contracts are private law, and as lawyers build expertise in
events assuming that they are foreseeable and detectable. particular transactions, they typically reuse standard terms that
experience has indicated as being enforceable to meet their
Smart contracts and oracles have been proposed as a objectives. Standardized forms may be used for particularly
mechanism for dispute resolution with courts or private common transactions. Several consortia and commercial
actors interfaced with the smart contract [56]. More entities and standards bodies7 are working on standardized
speculative and controversial approaches would be to use a terms for smart contracts though these seem to be focused on
computational mechanism for dispute resolution. The commercial transaction terms rather than healthcare
DAMN proposal [57] envisions a dispute resolution smart transaction terms. The availability of standardized healthcare
contract because arbitration is often easier to enforce transactions and processes should facilitate the development of
internationally than local court decisions. Prediction markets smart contract terms for the healthcare blockchains. Healthcare
have also been proposed for smart contract dispute resolution smart contracts could be designed to include dispute resolution
– 178 –
ICT for Health: Networks, standards and innovation
clauses, but standardized smart contract support for [12] T-T.Kuo, et .al."Blockchain distributed ledger
confidentiality and selective disclosure (e.g. access logging) technologies for biomedical and health care
may be more tractable in the near term. applications." J. of the Am. Med. Informatics Assn. 24.6
(2017): 1211-1220.
Reducing risks to design and deploy healthcare blockchains
[13] M. Mettler, "Blockchain technology in healthcare: The
and smart contracts will require different approaches for the revolution starts here." 18th Int’l Conf. on e-Health
different types of risk. Interoperability risks can be reduced Networking, App. and Serv. (Healthcom). IEEE, 2016.
through the development of standardized data formats for the
secure interchange of different types of health records. This [14] C.Agbo, et.al.,"Blockchain Technology in Healthcare: A
may not be the first risk to tackle though as it assumes there Systematic Review," Healthcare, vol. 7, no. 56, 2019.
are existing systems that needs to interoperate. The security
toolbox already provides a number of mechanisms to support [15] M. Englehardt, "Hitching healthcare to the chain: An
fine grained distinctions in authentication and authorization to introduction to blockchain technology in the healthcare
support the privacy and security requirements of healthcare sector.," Tech. Innovation Mgmt Rev., vol. 7, no. 10, 2017.
blockchains, though the specific use cases and role definitions
may still be evolving. Reducing legal uncertainties may be a [16] S. Lo, "Evaluating Suitability of Applying Blockchain," in
longer process due to the timescales required to accumulate IEEE Int’l Conf. on Engg. of Complex Comp. Sys., 2017.
legal precedents or evolve laws and regulations through
consensus mechanisms. Private law arrangements can provide [17] G.Drosatos, et. al., “Blockchain Applications in the
Biomedical Domain: A Scoping Review”, Computational
some protections while public laws evolve.
and Structural Biotech. J., Vol 17, pp 229-240, 2019
6. REFERENCES [18] D.Dimitrov, "Blockchain Applications for Healthcare
Data Management," Healthcare Information Res., vol. 25,
[1] A. Narayanan, "Bitcoin's Academic Pedigree," Comm. of no. 1, pp. 51-56, 2019.
the ACM, vol. 60, no. 12, pp. 36-45, 2017.
[19] M.Steward, "Electronic Medical Records," J. of Legal
[2] K. Christidis, et. al. "Blockchains and smart contracts for Medicine, vol. 26, no. 4, pp. 491-506, 2005.
the internet of things," IEEE Access, vol. 4, pp. 2292-
2303, 2016. [20] P. Tang. et.al, "Personal Health Records: Definitions,
Benefits, and Strategies for Overcoming Barriers to
[3] F.Casino, "A systematic literature review of blockchain- Adoption," J. of the American Medical Informatics Assoc.,
based applications: Current status, classification and open vol. 13, no. 2, pp. 121-126, 2006.
issues," Telematics and Informatics, vol. 36, pp. 55-81,
2019. [21] K. Hayrinena, et.al., Impacts of Electronic Health
Records: A Review of the Research Literature,," Int'l J. of
[4] UN, "Global Opportunity Explorer," 2019. Medical Informatics, vol. 77, no. 5, pp. 291-304, 2008.
https://www.unglobalcompact.org/docs/publications/GOE
_2019_Insights.pdf. [22] C.Esposito, et.al., "Blockchain: A Panacea for Healthcare
Cloud-Based Data Security and Privacy?," IEEE Cloud
[5] N. Szabo, “Smart Contracts”, Virtual School, 1994 Computing, vol. 5, no. 1, pp. 31-37, 2018.
[6] N. Szabo, “The idea of smart contracts”, Nick Szabo’s [23] Y. Park, et.al.,"Is Blockchain Technology Suitable for
papers and concise tutorials Vol. 6, 1997 Managing Personal Health Records? Mixed-Methods
Study to Test Feasibility," J. Med. Internet Res., vol. 21,
[7] M. Swan, “Blockchain: Blueprint for a new economy", no. 2, pp. 1-8, 2019
O'Reilly Media, Inc., 2015.
[24] T.Felin,et.al., "What problems will you solve with
[8] D. Osterwalder, "Trust Through Evaluation and blockchain?," MIT Sloan Mgmt Rev., Fall, pp. 32-38, 2018
Certification?," Soc. Science Comp. Rev., vol. 19, no. 1,
pp. 32-46, 2001. [25] A. Adjekum, et.al.,"Elements of Trust in Digital Health
Systems: Scoping Review," J. Med. Internet Res., vol. 20,
[9] S.Wright, et., al., "Open source and standards: The role of no. 12, 2018
open source in the dialogue between research and
standardization." IEEE Globecom Workshops 2014. [26] M. Macdonald, M., et.al. "The blockchain: a comparison
of platforms and their uses beyond bitcoin." COMS4507-
[10] Y. Ho, "Trust and distrust in open source software Adv. Computer and Network Security(2017).
development.," J. of Comp. Inf. Sys. , vol. 54, no. 1, pp.
84-93, 2013. [27] T-T. Kuo, et.al. "Comparison of blockchain platforms: a
systematic review and healthcare examples." J. of the Am.
[11] S. Angraal, et.al,"Blockchain technology: applications in Med. Informatics Assn. 26.5 (2019): 462-478.
health care." Circulation: Cardiovascular quality and
outcomes 10.9 (2017): e003800.
– 179 –
2019 ITU Kaleidoscope Academic Conference
[28] M. Wong, et.al.,"Socio-Technical Considerations for the [44] K. Iwai, "Persons, Things and Corporations: The
Use of Blockchain Technology in Healthcare," Studies in Corporate Personality Controversy and Comparative
Health Tech. and Informatics, vol. 247, pp. 636-640, 2018. Corporate Governance," AM. J. COMP. L., vol. 49, pp.
583-600, 1999.
[29] Z.Wan, et.al., "Bug characteristics in blockchain systems:
A largescale empirical study," in IEEE/ACM 14th Int’l [45] S.Bayern, "Of Bitcoins, Independently Wealthy Software,
Conf. on Mining Software Repositories (MSR), 2017. and the Zero-Member LLC," Northwestern U. Law Rev.,
vol. 108, pp. 257-270, 2014.
[30] B. Koteska, et.al., "Blockchain Implementation Quality
Challenges: A Literature Review," in 6th Workshop on [46] L. Lopucki, "Algorithmic Entities," Washington U. Law
Software Quality, Analysis, Monitoring, Improvement, and Rev., vol. 95, no. 4, pp. 887-953, 2018.
Applications, 2017.
[47] Q. DuPont, "Experiments in Algorithmic Governance: A
[31] A. Anjum, et.al.,"Blockchain standards for compliance history and ethnography of ‘The DAO,’ a failed
and trust.," IEEE Cloud Computing, vol. 4, no. 4, pp. 84- Decentralized Autonomous Organization.," in Bitcoin and
90, 2017. Beyond: Cryptocurrencies, Blockchains and Global
Governance, Routledge, 2017.
[32] P.Zhang, et.al.,"Applying Software Patterns to Address
Interoperability in Blockchain-based Healthcare Apps," [48] D. Zetsche, et.al. "The distributed liability of distributed
arXiv preprint arXiv:1706.03700, 2017. ledgers: Legal risks of blockchain.," U. Ill. L. Rev., 2018.
[33] S.Porru, et.al., "Blockchain-oriented Software [49] M.Beberich, et.al., "Blockchain Technology and GDPR-
Engineering: Challenges and New Directions," in How to reconcile privacy and distributed ledgers.," Eur.
IEEE/ACM 39th Int’l Conf. on Software Engg., 2017 Data Prot. L. Rev., no. 2, p. 422, 2016
[34] G. DeStefanis, "Smart contracts vulnerabilities: a call for [50] B.Fiedler, "Challenges of New Technology: Securing
blockchain software engineering?," in IEEE Int’l Medical Devices and Their Software for HIPPA
Workshop on Blockchain Oriented Software Engg. , 2018. Compliance.," in Managing Medical Devices Within a
Regulatory Framework, Elsevier, 2017, pp. 315-329.
[35] D. McKnight, et.al.,"What is Trust? A Conceptual
Analysis and an Interdisciplinary Model," in AMCIS 2000 [51] World Health Organization, "Medical device regulations:
Proc., 2000. global overview and guiding principles," 2003.
[36] X.Li, et.al. "A Survey on the Security of Blockchain [52] K. Lauslahti, "Smart Contracts – How will Blockchain
Systems," in Future Gen. Computer Systems, 2017. Technology Affect Contractual Practices?," ETLA Reports
No 68, 2017.
[37] Q.Xia, et.al.,"MeDShare: Trust-Less Medical Data
Sharing Among Cloud Service Providers via Blockchain," [53] A. Savelyev, "Contract law 2.0:‘Smart’contracts as the
IEEE Access, vol. 5, pp. 14757-14767, 2017. beginning of the end of classic contract law," Information
& Communications Technology Law, vol. 26, no. 2, pp.
[38] Y. Chen, et.al.,"Blockchain-based medical records secure 116-134, 2017.
storage and medical service framework," J. of Medical
Systems , vol. 43, no. 1, 2019. [54] I. Grigg, "The Ricardian Contract," in Proc. 1st IEEE
Workshop on Electronic Contracting, 2004.
[39] K. Werbach, "Trust, but Verify: Why the Blockchain
Needs the Law," Berkeley Technology Law J., vol. 33, no. [55] M. Raskin, "The Law and Legality of Smart Contracts,"
2, pp. 487-550, 2018. Georgetown Law and Tech. Rev., vol. 1, no. 2, pp. 305-
341, 2017.
[40] L. Cuen, "UN Using Blockchain Technology to Help
Refugees, Fight World Hunger,," Int’l Bus. Times, 4 May [56] A. Wright, et.al., "Decentralized Blockchain Technology
2017. and the Rise of Lex Cryptographia," 2015. [Online].
Available: https://perma.cc/3YUP-PNM4.
[41] I. Karamitsos, et.al.,"Design of the Blockchain Smart
Contract: A Use Case for Real Estate," J. of Information [57] P. Morgan, "Decentralized Arbitration and Mediation
Security, vol. 9, pp. 177-190, 2018. Network (DAMN)," 23 May 2016. [Online]. Available:
https://github.com/thirdkey-
[42] P.Cuccuru, "Beyond bitcoin: an early overview on smart solutions/damn/blob/master/proposal.asciidoc
contracts.," Int’l J. of Law and Information Technology.,
vol. 25, pp. 179-195, 2017. [58] A. Koulu, "Blockchains and Online Dispute Resolution:
Smart Contracts as an Alternative to Enforcement,"
[43] J. Barlow, "A Declaration of the Independence of SCRIPTed - A J. of Law, Tech. & Soc., vol. 13, no. 1, pp.
Cyberspace," Electronic Frontier Foundation, [Online]. 40-69, 2016.
Available: https://perma.cc/SF3L-Y7PX.
– 180 –
ICT for Health: Networks, standards and innovation
[59] A. Kosba, et al. "Hawk: The blockchain model of [60] J. Scott, "Booz Allen Hamilton Helps FDA Test
cryptography and privacy-preserving smart contracts." Healthcare Blockchain Technology," Govcon Wire, 26
Symp. on Security and Privacy. IEEE, 2016. June 2018.
– 181 –
DESIGN OF A CREDIBLE BLOCKCHAIN-BASED E-HEALTH RECORDS (CB-EHRS)
PLATFORM
Lingyu Xu1; Antoine Bagula1; Omowunmi Isafiade1; Kun Ma1; Tapiwa Chiwewe2
1University of the Western Cape
2IBM Research Africa
ABSTRACT
designed to promote EHRs’ successful development and added to the Internet due to the growing popularity
sustainability across the globe. This paper revisits the issue of EHRs, medical genetic sequencing and wearable
of healthcare management to describe the design of a credible medical devices. Therefore, the protection of user
blockchain-based e-health records (CB-EHRs) platform and privacy and medical information security becomes more
its performance evaluation. The proposed platform can urgent and important.
be used to secure transactions through anonymity and
traceability of health data in cyber-healthcare systems, such 2.2 Blockchain-based EHRs related research
as those proposed in [6–10].
The rest of the paper is structured as follows: section 2 In order to address the aforementioned problems, many
presents related study on blockchain and a summary of researchers have proposed relevant solutions through strategy
challenges in the EHRs platform, highlighting the gaps and research, architectural frameworks and model designs [16–
opportunities that have been identified in the current system. 18]. For example, in 2011 Sebastian Haas [16] et
The key contribution of this paper is described in sections 3 al. proposed a privacy protection system based on data
and 4, detailing the components and related functions of the services and patient service models. Yarmand et al. [17]
proposed CB-EHRs. Section 5 details the performance of in 2013 proposed a behavior-based access control for a
the choice consensus algorithm. Finally, section 6 concludes distributed healthcare model to solve the user’s privacy
the paper and presents a possible extension to the CB-EHRs problem. Recently, with the development of blockchain
platform. technology, blockchain has not only been used in the
financial industry, but has also been adopted as a protective
2. BLOCKCHAIN-BASED RELATED WORK AND mechanism in the medical domain. In 2016, Drew lvan [18]
CHALLENGES OF EHRS PLATFORM proposed a blockchain approach to securely store patient
medical records. While the above-mentioned models used
This section presents an expository detail on blockchain, blockchain technology to facilitate data sharing and privacy
the challenges of EHRs, and related research on protection, the framework it adopted presents a large amount
blockchain-based EHRs. The section concludes by of network resource wastage or consumption. For users
highlighting the contribution of CB-EHRs platform proposed who need treatment, functionalities of these models are not
in this paper. complete and need to be improved. Hence, this research
presents a credible blockchain-based electronic health records
2.1 Challenges of EHRs platform (CB-EHRs) management platform, which is characterized by
decentralization, data tamper-proof, collective maintenance
The EHRs platform is widely recognized as an enabling mechanisms, security and credibility. This platform cannot
platform, which promotes telemedicine in the current only realize the capabilities of medical data sharing, but also
technology age [11]. However, despite its wide acceptance ensures the privacy of users.
and use, there are some challenges that remain pertinent to
the platform. Some of the known key challenges are:
3. BLOCKCHAIN TECHNOLOGY
• Difficulty in medical data sharing: In order to advance
medical research and facilitate patients, researchers 3.1 The concept of blockchain
are committed to the sharing of medical data. At
present, there are some trust issues among medical There are two basic definitions of blockchain [19, 20].
institutions around the world. Hence, the verification, In a narrow sense, blockchain is a non-tampered and
synchronization and storage of medical data are unforgeable distributed ledger that uses cryptographic
hindered. When medical institutions and patients share correlation algorithms. It is a chained data structure that
data, they need to spend a lot of time and resources links data blocks together in time order [19]. Broadly
on identity and data validation. This makes medical speaking, blockchain technology refers to a new distributed
data acquisition very difficult [12]. Moreover, data infrastructure and computing paradigm which is built on
transmission is not secure and data can be tampered with. a peer-to-peer network. It uses a chained data structure
All of these challenges seriously hinder the development to validate and store data, a distributed node consensus
of medical big data and electronic healthcare systems. algorithm to generate and update data, and a cryptography
mechanism to protect data transmission and access security
• Information security and privacy protection concerns: [20]. Alliance chain (adopted in this paper): In an alliance
Firstly, there is no complete scheme of medical privacy blockchain, the consensus process is controlled by preselected
protection in the electronic medical industry [13]. nodes. Only alliance members have permissions to read and
Secondly, big data mining technology has potential write the blockchain [21]. This type of blockchain can be
risks of linkage attacks, which attempt to re-identify seen as "partially decentralized".
individuals in an anonymized data set by merging
information from two or more datasets. Thirdly, the 3.2 PKI and digital signature
user does not fully participate in the access control
policy of the electronic health record [14,15]. However, The comprehensive technology required to provide
more and more personal health information is now being public-key encryption and digital signature services is known
– 184 –
ICT for Health: Networks, standards and innovation
as a public-key infrastructure (PKI) [22]. The basic idea of • Decentralization: The blockchain system is based on
applying PKI to the EHRs platform is to have the trusted a peer-to-peer distributed network, this feature makes
user digitally sign documents certifying that a particular it not to rely on a centralized management agency.
cryptographic key belongs to a particular user. The EHRs Therefore, the operation of the CB-EHRs platform is
platform in the Internet brings convenience to users, but it jointly accomplished and maintained by all medical
also increases the possibility of privacy leaks. Using the PKI institutions in the blockchain. If any medical institution
to generate a public key and identity identifier for him/her node is lost or damaged, the system can still function
during the user registration process. Users are active and effectively.
authenticated in the EHRs platform in this particular state.
This approach not only ensures the user’s legal identity but • Tamper-proof: The chained structure of the blockchain
also protects the user’s privacy. The purpose of PKI is to ensures the tamper resistance of the data in the
manage keys and certificates. blockchain system. If the information of a block is
Digital signature, also known as public key digital signature changed, the block header hash of the next block will
[23], usually defines two corresponding operations, one for also change. Therefore, if the attacking node wants
the signature (private key) and the other for the authentication to successfully change the transaction information, it
(public key). The process of signing is as follows. First, the only recalculates all subsequent blocks of the changed
sender uses the hash algorithm to obtain a digital abstract, block and catches up with the progress of the legal
and the signature private key encrypts the digital abstract to blockchain in the network. In the current state of
obtain a digital signature. The original text and the digital network technology development, such an attack is
signature will be sent to the recipient together. The receiver difficult to achieve [33]. Therefore, applying blockchain
then verifies the signature by decrypting the digital signature technology to the CB-EHRs platform makes EHRs data
with the sender’s public key to obtain a digital abstract. The more secure.
receiver will use the same hash algorithm to get a new digital • Anonymity: Transaction data on the blockchain is
abstract and compare the two abstracts. If the two match, the open and transparent. However, the owner’s identity
digitally-signed electronic file is successfully transmitted. corresponding to the transaction is anonymous. The
CB-EHRs platform can use blockchain encryption to
3.3 Consensus mechanism hash user identity information. The resulting hash value
Consensus refers to the process by which network nodes is used as the unique identifier for the user, similar
adhere to a common rule and achieves consistent results to the bitcoin’s wallet address. The user’s behavior
for certain problems through asynchronous interaction. The on the CB-EHRs platform is associated with the hash
consensus mechanism in the blockchain is mainly used to value obtained earlier, rather than with the user identity
make the network nodes agree on block generation and information. This separation of user identity and user
benefit distribution. The major difference between different data protects patient privacy.
blockchain networks comes from the difference in consensus
In what follows, we present the detail of the proposed
mechanisms. Therefore, the characteristics and performance
CB-EHRs, its associated features and potential benefits.
of different blockchain networks are also different, depending
on the choice of consensus mechanism.
4. DESIGN APPROACH OF CB-EHRS PLATFORM
Currently, Proof of Work (PoW) is used in the bitcoin
network [24]; Ethereum also uses PoW, and tend to gradually This section provides a detailed explanation of related
replace PoW with Proof of Stake (POS) [25]; Ripple uses features and functioning of the proposed credible blockchain
Ripple Proof of Consensus Algorithm (RPCA) [26]; Fabric electronic health records (CB-EHRs) platform.
uses Practical Byzantine Fault Tolerance (PBFT) [27]; in
addition, there are Delegated Proof of Stake (DPoS) [28] 4.1 Platform architecture
and Delegated Byzantine Fault Tolerance (dBFT) [29]. The
CB-EHRs platform uses the dBFT consensus mechanism. The proposed architecture of the CB-EHRs platform is
This consensus mechanism was proposed by NEO [30], which designed as a layered architecture. The entire platform
is an open source blockchain project driven by the community. is divided into three layers, which comprise: (i) the user
A comparison [31, 32] of the characteristics of several major interface layer, which is at the top; (ii) the business logic
consensus mechanisms is shown in Table 1. layer, which is in the middle; and (iii) the data access layer,
positioned at the bottom of the framework. Figure 2 presents
3.4 Blockchain and CB-EHRs platform an overview of the CB-EHRs architectural framework and
details of the architecture is presented hereafter.
• Openness: The blockchain system code is open source.
The system’s interface and data are also open to everyone 4.1.1 User Interface layer
[33]. Each node in the system is allowed to obtain the
complete copy of the database. Based on this feature, The user interface layer is used to display data and receive
the CB-EHRs platform enables medical data to be shared user’s input information. It provides users with a graphic
among medical institutions. interface that can interoperate with the entire blockchain
– 185 –
2019 ITU Kaleidoscope Academic Conference
– 186 –
ICT for Health: Networks, standards and innovation
– 187 –
2019 ITU Kaleidoscope Academic Conference
– 188 –
ICT for Health: Networks, standards and innovation
– 189 –
2019 ITU Kaleidoscope Academic Conference
[11] Dimitrov D V. Medical internet of things and big data [25] Bartoletti M, Lande S, Podda A S. A
in healthcare[J]. Healthcare informatics research, 2016, Proof-of-Stake protocol for consensus on Bitcoin
22(3): 156-163. subchains[C]//International Conference on Financial
Cryptography and Data Security. Springer, Cham,
[12] Lo’ai A T, Mehmood R, Benkhlifa E, et al. Mobile cloud 2017: 568-584.
computing model and big data analysis for healthcare
applications[J]. IEEE Access, 2016, 4: 6171-6180. [26] Schwartz D, Youngs N, Britto A. The ripple protocol
consensus algorithm[J]. Ripple Labs Inc White Paper,
[13] Chen M, Qian Y, Chen J, et al. Privacy protection 2014, 5.
and intrusion avoidance for cloudlet-based medical data
sharing[J]. IEEE transactions on Cloud computing, [27] Castro M, Liskov B. Practical Byzantine fault
2016. tolerance[C]//OSDI. 1999, 99: 173-186.
[14] Reti S R, Feldman H J, Ross S E, et al. [28] ThePiachu’s Bitcoin Blog. Thoughts on Delegated Proof
Improving personal health records for patient-centered of Stake and Bitshares[OL].2014. http://tpbit.blogspot.
care[J]. Journal of the American Medical Informatics com/2014/08/thoughts-on-delegated-proof-of-stake.
Association, 2010, 17(2): 192-195. html
[15] Papoutsi C, Reed J E, Marston C, et al. Patient and [29] Crain T, Gramoli V, Larrea M, et al. DBFT: Efficient
public views about the security and privacy of Electronic Leaderless Byzantine Consensus and its Application
Health Records (EHRs) in the UK: results from a to Blockchains[C]//2018 IEEE 17th International
mixed methods study[J]. BMC medical informatics and Symposium on Network Computing and Applications
decision making, 2015, 15(1): 86. (NCA). IEEE, 2018: 1-8.
[16] Haas S, Wohlgemuth S, Echizen I, et al. Aspects of [30] NEO - An Open Network For Smart
privacy for electronic health records[J]. International Economy[OL].2014. https://neo.org/
journal of medical informatics, 2011, 80(2): e26-e31.
[31] Xia Q, Zhang F J, Zuo C. Review for consensus
[17] Yarmand M H, Sartipi K, Down D G. Behavior-based mechanism of cryptocurrency system[J]. Computer
access control for distributed healthcare systems[J]. Systems & Applications, 2017.
Journal of Computer Security, 2013, 21(1): 1-39.
[32] Viriyasitavat W, Hoonsopon D. Blockchain
[18] Ivan D. Moving toward a blockchain-based method for characteristics and consensus in modern business
the secure storage of patient records[C]//ONC/NIST processes[J]. Journal of Industrial Information
Use of Blockchain for Healthcare and Research Integration, 2019, 13: 32-39.
Workshop. Gaithersburg, Maryland, United States:
ONC/NIST. 2016. [33] Tasca P, Tessone C J. Taxonomy of blockchain
technologies. Principles of identification and
[19] Iansiti M, Lakhani K R. The truth about blockchain[J]. classification[J]. arXiv preprint arXiv:1708.04872,
Harvard Business Review, 2017, 95(1): 118-127. 2017.
[20] Crosby M, Pattanayak P, Verma S, et al. Blockchain [34] Gonzalez A. Unified application, user interface and
technology: Beyond bitcoin[J]. Applied Innovation, data linking: U.S. Patent Application 10/796,422[P].
2016, 2(6-10): 71. 2005-9-8.
[21] Casino F, Dasaklis T K, Patsakis C. A systematic [35] Sandhu R S, Coyne E J, Feinstein H L, et al. Role-based
literature review of blockchain-based applications: access control models[J]. Computer, 1996, 29(2):
current status, classification and open issues[J]. 38-47.
Telematics and Informatics, 2018.
[36] Anderson C. Docker [software engineering][J]. IEEE
[22] Vacca J R. Public key infrastructure: building Software, 2015, 32(3): 102-c3.
trusted applications and Web services[M]. Auerbach
Publications, 2004.
[23] Kaur R, Kaur A. Digital signature[C]//2012
International Conference on Computing Sciences.
IEEE, 2012: 295-301.
[24] Gervais A, Karame G O, Wüst K, et al.
On the security and performance of proof of
work blockchains[C]//Proceedings of the 2016 ACM
SIGSAC conference on computer and communications
security. ACM, 2016: 3-16.
– 190 –
THE GDPR TRANSFER REGIME AND MODERN TECHNOLOGIES
Keywords – Data protection, health data, transfer, transit Legally a lot can be said about modern technologies, their
use, privacy risks, infringements of rights, etc. This paper
1. INTRODUCTION focusses specifically on transfer and modern technologies.
Inherent to the nature of these technologies is that data is
In our rapidly evolving digital world, people use various not bound by borders. Users of modern technologies may
modern technologies to track and measure their health and be located anywhere in the world and data may move
fitness. Modern technologies such as mobile applications across the globe while being processed by companies
and wearables (including watches, bracelets and smart established anywhere in the world. One of the main
fashion) are used to get into shape, keep fit, lose weight, challenges of the borderless nature of data processing is that
reduce stress, manage mental health disorders, test and it is difficult to track the data and as a consequence difficult
diagnose for specific diseases such as malaria, help with to determine jurisdiction, which may lead to difficulties in
family planning and ovulation tracking, etc. The data subjects exercising rights in cases of infringements.
technologies enable people to monitor their own health and
fitness by entering personal health data and using (pressure) Within the European Union (EU) data is protected by the
sensing technologies which measure vital signs (such as General Data Protection Regulation (GDPR) [10]. The
heartrate) and track progress (such as counting steps) [2]. GDPR protects data, among other things, when it is
New health technologies are a key area of 21 st century transferred across borders. This research aims to answer
knowledge societies and economies, offering potential for how the GDPR transfer regime applies to data processing
growth and economic development [3]. It is one of the by modern technologies, if at all, and whether the GDPR
largest growing global markets. According to a recent legal framework as such offers sufficient protection. When
article, there are more than 300 000 health related mobile using modern technologies, the data is collected by a device
device applications [4]. While the use of these technologies (such as a smartphone or wearable) by using applications
may bring benefits to society as they reduce the burden on developed by commercial companies. The applications
‘send’ the data to the servers of the company which owns European Commission, the Article 29 Working Party (now
the app and which then processes the data. What exactly the European Data Protection Board [16]) clarified the
happens technically behind the scenes is unclear. It is scope of the definition of data concerning health in relation
therefore unclear whether ‘sending’ data between the to lifestyle and wellbeing apps and provides criteria to
device and the server of a company can be seen as a determine when data processed by such apps and devices is
transfer within the meaning of the GDPR and whether the health data [17]. According to the Article 29 Working Party,
GDPR transfer regime applies to processing by modern personal data is health data when (1) the data is clearly
technologies. medical data, (2) the data is raw sensor data that can be
used in itself or in combination with other data to draw a
This research argues that the complexity of the GDPR legal conclusion about the actual health status or health risk of a
framework does not offer sufficient protection against person or (3) conclusions are drawn about a person’s health
processing by modern technologies. By taking a technical, status or health risk [18]. This means that, in general, data
behind the scenes perspective and looking at whether the is health data when it is used or can be used to draw
(technical) process of ‘sending’ data from a user’s device to conclusions about a person’s health. However, the Article
the server of a company can be seen as a transfer within the 29 Working Party also acknowledges that in some cases the
meaning of the GDPR, we argue that this process is a mere raw data itself is considered to be health data. It also
transit of data where the device functions only as a tool for acknowledges that presumably simple facts about
the companies to collect data [11]. In coming to this individuals, such as IQ, wearing glasses or lenses, smoking
conclusion, this article first needs to establish what the legal and drinking habits, membership of patient support groups,
basis for processing health data by modern technologies is. etc. are considered to be health data. In our view, the mere
We then look at the technical process used by modern fact that a person uses an app, for example to help quit
technologies and whether the GDPR transfer regime applies smoking or to count calories already says a lot about a
to this process in order to conclude whether the legal basis person. Whether or not true, the conclusion can be drawn
and the GDPR legal framework offer sufficient protection that the person is a smoker or may be obese and that he or
to processing by modern technologies. she may have health issues (such as lung or heart problems)
because of this. The mere fact that a person uses a health
2. LEGAL BASIS FOR PROCESSING HEALTH app already can say a lot about their health, and even more
DATA BY MODERN TECHNOLOGIES so when the data is combined with other health information
about a person. For example, an employer or insurer buying
The GDPR provides rules for the protection of personal health data and combining it with the information already
data and free movement of such data in order to protect the on record not only violates privacy but can also
fundamental rights and freedoms of persons. It applies to discriminate against their employee or the insured. This
the processing of personal data of data subjects who are in could lead to increases in insurance fees, rejection of
the EU, regardless of where the controller or processor are insurance and perhaps even in unemployment. Data
established [12]. This means that the GDPR applies to any generated by modern technologies which can conclude
company around the globe processing data of data subjects something about a person’s health in the broadest sense can
who are in the EU if the processing activities relate to therefore generally be seen as health data.
offering goods or services to data subjects or monitoring the
behavior of data subjects. As such, the GDPR aims at Health data has had a long history of being seen as a special
offering a similar level of protection for EU citizens category of data, also referred to as sensitive data, that
regardless of where the data is being processed [13]. This is requires additional protection. As such, Article 9 of the
particularly important when health data is being processed GDPR prohibits the processing of health data unless there is
by commercial companies who are not under any obligation a legal basis to do so. If there is no legal basis for
of professional secrecy. In previous research we have processing, the processing is considered to be unlawful.
established that many companies deny or at least do not According to the GDPR, explicit consent given by the data
mention the fact that they process health data while in fact subject is the legal basis for processing health data by
they are [14]. modern technologies [20, 21]. The GDPR thus allows
processing of personal health data by companies when a
While we use the more overarching term health data, data subject explicitly consents. Consent of the data subject
Article 4 (15) of the General Data Protection Regulation within the meaning of the GDPR means a clear affirmative
(GDPR) refers to it as ‘data concerning health’ and defines act establishing at least the freely given, informed
it as: indication that the data subject agrees to the processing of
his or her personal data [22]. Consent can also be given by
Personal data related to the physical or mental health of a electronic means, for example by ticking a box when
natural person, including the provision of healthcare visiting a website, choosing certain technical settings or any
services which reveal information about health status [15]. other statement or conduct which clearly indicates in this
context the data subject’s acceptance of the proposed
This is a very broad definition: any information which can processing. Pre-ticked boxes or inactivity by the data
reveal something about a person’s (mental) health is subject do not constitute consent [23]. The request for
considered to be health data. In the annex to its letter to the consent has to be clear, concise, not unnecessarily
– 192 –
ICT for Health: Networks, standards and innovation
disruptive and needs to be presented in a clearly These modern technologies, such as mobile applications
distinguishable form, meaning that it may not be buried and wearables process large amounts of personal (health)
within the fine print of a privacy policy or contract [24]. data. The technologies make it possible to continuously
monitor the user. Most people carry their mobile phone
While at first sight it looks as if the GDPR offers sufficient with them during the day and wearables made tracking even
protection against the processing of health data, the easier. A smart watch or smart glasses for example allow
practical reality is quite different. Previous research has users to track their health and fitness with objects which are
shown that companies offering health apps are by no means easy to carry. While making life and health easy for users,
transparent about their processing activities and whom they large amounts of health data become available to
share the data with [25]. While data subjects to some degree commercial companies who are by no means under any
consent to data processing, some health apps do not even obligation of professional secrecy and what happens behind
recognize the fact that they process health data, resulting in the scenes of these technologies is unknown to many. When
a lack of legal basis. As a result of this, risks of violation of unravelling what happens, behind the scenes, to the data we
rights and freedoms remain, as well as physical and stumbled upon 2 major ways that the technologies function
practical challenges related to the use of modern that are relevant for this article. Many health apps and
technologies to process health data, such as jurisdiction and wearables by default:
exercise of rights.
1. collect data via an app and store it on the device
3. BEHIND THE SCENES OF MODERN itself until the user actively choses to send the data
TECHNOLOGIES to a cloud or server;
2. collect data via an app and store it on a (cloud)
Processing personal data according to the GDPR includes server. In this case the data exists outside of the
‘collection, recording, organization, structuring, storage, app and is accessible to the developer, i.e. the
adaptation or alteration, retrieval, consultation, use, device is used as a tool to collect data, the data can
disclosure by transmission, dissemination or otherwise be seen separately from the app considering that it
making available, alignment or combination, restriction, exists even if the app is deleted.
erasure or destruction’ of data [26]. This very broad
definition means that basically any action performed on If we picture a user in the first situation and we take the
personal data is processing. The one word that is missing example of an app that counts how many steps someone
from the definition is transfer of data. What is however takes during the day, the app counts the steps and stores the
mentioned by the definition in Article 4 (2) GDPR is that data on the device itself by default. The data is stored on the
processing also includes disclosing the data by transmission device for as long as the user does not delete the data or
and dissemination or otherwise making it available. While chooses to store the data somewhere else, for example
it is interesting that transfer is not included in the definition when the storage space of the device is full. In other words,
for processing, disclosing and making data available can be the collected data remain on the user’s device until the user
seen as transfer of data. actively decides to store the data elsewhere, outside of the
app or wearable.
Transfer has an important role in the GDPR. While the free
flow of information has always been promoted by data More importantly for this research is however the second
protection legal frameworks, the major concern was that situation, where data is collected by an app or wearable
data protection legislation could be circumvented by which does not intend to store it on the device. Instead, by
moving processing operations to countries with no or less default, the data is sent to and stored on the (cloud) server
strict data protection laws [27]. European data protection of the app company. Sending the data requires an active
legal frameworks have therefore always been cautious connection between the device and the (cloud) server. If
about transferring data to third countries who are not part of this connection is unavailable, the data is most likely stored
the legal regime. In order to prevent data from being on the device until the connection is available.
transferred to ‘data havens’, the principle of equivalent
protection was introduced, meaning that there should be no There is a significant legal difference between the two
restrictions on transborder data flows to states with legal situations. In the first situation the app is closely related to
regimes which ensure data protection equivalent to data the data and therefore to the user, it is merely a means to an
protection offered by the GDPR. Chapter V of the GDPR is end. In the second situation, the purpose of the app or
dedicated to transfers of personal data to third countries or wearable is mainly to generate data. The device is not used
international organisations. Modern technologies process for storage or not meant to be used for storage. As soon as
data electronically, making it easy to transfer data across an active connection is available, the data is sent to the
the globe. The data can be sent from one actor to another or designated (cloud) server. In this regard, we can make an
made accessible to more than one actor in a blink of an eye. analogy with streaming data. The user might have the app
Modern technologies thus impact the way that personal on their mobile phone or wearable, but the data exists
health data can be collected. separately, outside this app. For example, when watching a
YouTube video, the app is solely used to stream the data
available on the YouTube server. While health apps and
– 193 –
2019 ITU Kaleidoscope Academic Conference
wearables are more of a two-way-street considering that information. Unfortunately, there is not a lot of case law in
they can also generate data, the basic concept and this regard to help further clarify the matter. If one of the
comparison to YouTube streaming is the same. factors determining what transfer is includes the technical
nature by which it takes place, the question that arises is
Processing health data in a way where data is collected by what technical circumstances can facilitate transfer. Council
an app or wearable and sent to a (cloud) server for (further) of Europe Convention 108 for the protection of individuals
processing still leaves the question whether sending the with regard to automatic processing of personal data [31]
data can be seen as a transfer within the meaning of the provides some insight in this regard.
GDPR and is as such protected or whether the device
functions merely as a tool for the companies to collect data Convention 108 includes a chapter on transborder data
where sending the data can be seen as a mere transit of data flows and determines that the provisions apply to the
[28]. The concept of ‘transfer’ will therefore be discussed transfer across national borders by whatever medium [32].
in the next paragraph. It is aimed at the free flow of information, regardless of
frontiers, taking into account the wide variety of factors
4. THE NOTION OF TRANSFER determining the way in which data is transferred. These
factors include: the mode of representation of the data, their
The GDPR aims at offering a similar level of data storage medium, way of transport, interface, the circuit
protection, regardless of where in the world data of data followed and the relations between the sender and recipient
subjects who are in the EU is being processed. Therefore, [33]. According to the explanatory memorandum the way of
Chapter V of the GDPR includes provisions on transfers of transport includes physical transport, mail, and circuit-
personal data to third countries. This section provides rules switched or packet-switched telecommunications links. The
in order to ensure data protection equivalent to the GDPR, interface, i.e. the point where two systems interact, can be,
meaning that data may only be transferred to third countries among other things, computer to terminal, computer to
outside the EU if the conditions of the GDPR are met. In computer, and manual to computer. The circuit followed
short, this means that there needs to be: 1) an adequacy can be direct from the country of origin to the country of
decision (such as the EU-U.S. Privacy Shield) or 2) destination or via one or more countries of transit [34]. The
appropriate safeguards or 3) that the data subject has given explanatory report to the Modernized Convention provides
explicit consent for data processing in the third country. some more clarity in determining that transborder data
With emerging modern technologies, where data may be transfers occur when personal data is disclosed or made
processed anywhere in the world, it is of the utmost available to a recipient subject to the jurisdiction of another
importance to protect the data, in particular health data. In state or international organization. According to Article 2 (e)
order to establish whether sending data, from the app or of the Convention a recipient is ‘a natural or legal person,
wearable onto the (cloud) server of a company for the public authority, service, agency or any other body to
purpose of being processed by that company, can be seen as whom data are disclosed or made available. The GDPR
a transfer within the meaning of the GDPR, it is important definition of recipient is almost the same, determining that
to establish what transfer exactly is in order to determine recipient means a natural or legal person, public authority,
whether or not it falls under Chapter V GDPR and agency or another body, to which the personal data are
consequently whether or not health data in this regard is disclosed, whether a third party or not’ [35]. The recipient
sufficiently protected. In literature transfer is described as thus receives the data or is given access to the data and can
to occur as a part of networked series of processes made to be a controller or a processor [36].
deliver a business result [29].
When it comes to moving data, there are two main ways to
The GDPR is, however, unclear about what transfer is and technically do this, namely by exchanging or sharing data.
does not provide a definition. What is clear is that it is a According to Doan et al. data exchange is the process of
process where data moves between different actors. taking data that is structured within the source database
According to the European Data Protection Supervisor system and transforming it into data structured under a
(EDPS) in its position paper on transfer to third countries target database system [37]. In other words, the data is
and international organizations by EU institutions and transformed so that it becomes compatible with other
bodies, the lack of a definition leads to the assumption that systems which receive an accurate representation of the
the term needs to be used in its natural meaning. As such source data. Exchange thus allows data to be shared
transfer means that data ‘moves’ between different users. between systems and programs. The introductory report for
However, as the EDPS also concludes, this is not always updating Recommendation No. R (97) 5 defines exchange
straight forward. According to the Court of Justice of the as the communication of information to (a) clearly
European Union (CJEU) in the Lindqvist case, it is identified recipient(s) by a known transmitter (such as
necessary to take account of both the technical nature of the secured e-mailing) [38]. When health data is exchanged,
operations carried out and of the purpose and structure of the data is sent from A to B using a transmitter. This can be
the provisions on transfer in EU legislation [30]. Taking an e-mail or other way of sending the data so that it can be
into account the technical nature of processing operations, read and used by B. Figure 1 below shows this process. In
transfer, as such entails, among other things, the this case, A is the original controller of the health data and
automatically or intentionally sending or accessing of
– 194 –
ICT for Health: Networks, standards and innovation
B becomes the new controller of the data and will build on • Transfer has a natural meaning, i.e. data moves
the received data for their own purpose. between users.
• Transfer may be the exchange or sharing of data.
• Data movement takes place by whatever medium.
• Data is disclosed or made available to a recipient.
Exchange
5. TRANSFER OR TRANSIT?
– 195 –
2019 ITU Kaleidoscope Academic Conference
research [43] has shown that there is a gap between the And even if they were to read them, they might not
GDPR and practical reality. There is a general lack of understand the meaning or the risks involved. As such,
transparency from commercial companies about their people do not know what they are consenting to. Therefore,
processing activities, their purposes for processing, the combining the fact that commercial companies are
quantity of health data processed, the location of storage generally not transparent enough about their processing
and recipients the data is shared with. In particular, the activities with the fact that users generally do not know
sharing of data is of a great concern as the data is collected what they are consenting to, results in a weak legal basis.
and shared with actors who are by no means under any As a consequence, violations take place more frequently
obligation of professional secrecy and who sell the data to than we would wish.
the highest bidder which may lead to various forms of
discrimination, violation of fundamental rights and As such, the complexity of the GDPR legal framework does
difficulties with exercising rights in case of infringements. not offer sufficient protection against data processing by
This is even more concerning considering that people modern technologies and commercial companies are not
generally do not inform themselves before giving away taking sufficient responsibility when processing health data.
their data and/ or choose convenience over privacy. It is the Perhaps the solution lies in prohibiting the use of health
responsibility of companies to protect their users’ privacy; data in certain situations as suggested by Frank Pasquale
however, unfortunately they often fail to do so. Consent as [44]. A stricter approach, i.e. prohibiting the use of health
a legal basis for processing health data by modern data in certain situations, would at least be an incentive for
technologies is therefore not enough. As a result of this, the companies not to violate the privacy of a person’s most
complexity of the GDPR legal framework does not offer intimate sphere. This approach will require further research
sufficient protection for processing of health data by on how to limit processing health data by modern
modern technologies. technologies. The situations where it might be limited or
prohibited would have to be defined. It is, however, our
6. CONCLUSION opinion that we need another way of looking at health data
processed by modern technologies that would be beneficial
The multitude of modern technologies that are available to all parties and still protects rights and freedoms.
today process large amounts of health data. When
processing data, controllers and processors need to abide by REFERENCES
the GDPR, which requires that there needs to be a legal
basis for processing. Commercial companies therefore need [1] Council of Europe, Explanatory memorandum to
to request the users of their modern technologies for Recommendation No. R (81) 1 of the Committee of
consent before being allowed to process health data. On Ministers to member states on regulations for
many occasions, these companies collect data via an app automated medical data banks [1981], para. 6.
and store it on a (cloud) server where it is being processed.
The device is used as a tool to collect data and the data can [2] B. Millington, ‘Smartphone Apps and the Mobile
be seen separately from the app considering that it exists Privatization of Health and Fitness’, Critical Studies
outside of the app (even if the app is deleted) where it is in Media Communication, v31 n5, December 2014,
accessible to the company. Taking into consideration that p. 479-493.
the data exists outside the app and that the data subject
cannot be the controller of his or her own data, the transfer [3] M. L. Flear et al., European Law and New Health
regime of the GDPR does not apply when the data is being Technologies, Oxford: University Press, 2013, p. 1.
sent from the device to the (cloud) server. This process is a
mere transit of data. [4] Editorial, An app a day is only a framework away,
Elsevier, The Lancet Digital Health, Volume 1,
Considering that the GDPR transfer regime does not apply, Issue 2, June 2019, Page e45, available at
the question is whether consent as a legal basis is enough. https://www.sciencedirect.com/science/article/pii/S2
While the GDPR applies to the processing of the data of 589750019300317.
data subjects who are in the EU, regardless of where the
controller or processor is established, the reality remains [5] See for example:
that it is more difficult to track data processed by modern https://www.theguardian.com/world/2018/jan/28/fitn
technologies, i.e. where it is stored and with whom is it ess-tracking-app-gives-away-location-of-secret-us-
shared, which may result in discrimination and violation of army-bases and https://decorrespondent.nl/8480/this-
rights. There is a general lack in transparency from fitness-app-lets-anyone-find-names-and-addresses-
companies as regards to their processing operations. for-thousands-of-soldiers-and-secret-
Furthermore, informing people via privacy policies of agents/260810880-cc840165.
modern technologies does not offer sufficient protection
considering that most people do not actually read them [45].
– 196 –
ICT for Health: Networks, standards and innovation
[6] See for example: [17] Article 29 Data Protection Working Party, Letter to
https://www.washingtonpost.com/technology/2019/0 Mr. Timmens and Annex – health data in apps and
4/10/tracking-your-pregnancy-an-app-may-be-more- devices, 2015, available at:
public-than-you- https://ec.europa.eu/justice/article-
think/?noredirect=on&utm_term=.7f91beb5e812. 29/documentation/other-
document/files/2015/20150205_letter_art29wp_ec_h
[7] See for example: https://www.wsj.com/articles/you- ealth_data_after_plenary_en.pdf and
give-apps-sensitive-personal-information-then-they- https://ec.europa.eu/justice/article-
tell-facebook-11550851636?mod=e2tw and 29/documentation/other-
https://www.theguardian.com/technology/2017/may/ document/files/2015/20150205_letter_art29wp_ec_h
01/facebook-advertising-data-insecure-teens>. ealth_data_after_plenary_annex_en.pdf.
[8] See: https://theoutline.com/post/7039/there-is-a- [18] Article 29 Working Party (A29WP), Annex by letter
reason-apps-make-it-so-fun-to-track-your-health. – health data in apps and device, 2015 p. 5.
[9] T. Mulder, M. Tudorica, Privacy policies, cross- [19] According to Article 5 (1, a) GDPR.
border health data and the GDPR, Information &
Communications Technology Law (2019) 28,3. [20] Article 9 (2,a) GDPR.
[10] Regulation (EU) 2016/679 of the European [21] Article 6 (a) GDPR.
Parliament and of the Council of 27 April 2016 on
the protection of natural persons with regard to the [22] Article 4 (11) GDPR.
processing of personal data and on the free
movement of such data, and repealing Directive [23] Recital 32 GDPR. See also L. Golba, ‘Consent for
95/46/EC (General Data Protection Regulation) Personal Data Processing in Digital Environment
[2016] OJ L 119/1. According to GDPR’, Annuals of the Administration
and Law, no. 17 (2), p. 253-265.
[11] See for example: The concept of ‘transfer’ of data
under European data protection law – in the context [24] Article 7 (1) GDPR.
of transborder data flows, Faculty of Law –
University of Oslo, 01-12-2015, available at: [25] See for example:
https://www.duo.uio.no/bitstream/handle/10852/497 https://www.theguardian.com/technology/appsblog/
22/8026_The-concept-of-transfer-of-data-under- 2013/sep/03/fitness-health-apps-sharing-data-
European-data-protection-law---In-the-context-of- insurance; T. Mulder, ‘Health Apps, their Privacy
transborder-data- Policies and the GDPR’, European Journal of Law
flows.pdf?sequence=1&isAllowed=y. and Technology, (10) 1 (2019).
[13] N. Dasko, ‘The General Data Protection Regulation [27] Council of Europe, “Explanatory report to the
(GDPR) – A Revolution Coming of European Data Convention for the protection of individuals with
Protection Laws in 2018. What’s New for Ordinary regard to automatic processing of personal data”
Citizens?’, Comparative Law Review, p. 128, (ETS No 108), para. 9.
available at:
http:h//tdtpx:./d/odix.o.drgo/i.1o0rg.g1/21707.152/7 [28] See for example: The concept of ‘transfer’ of data
C77L7R5./2C0L1R7.2.00156. under European data protection law – in the context
of transborder data flows, Faculty of Law –
[14] T. Mulder, M. Tudorica, Privacy policies, cross- University of Oslo, 01-12-2015, available at:
border health data and the GDPR, Information & https://www.duo.uio.no/bitstream/handle/10852/497
Communications Technology Law (2019) 28,3. 22/8026_The-concept-of-transfer-of-data-under-
European-data-protection-law---In-the-context-of-
[15] Article 4 (15) GDRP. transborder-data-
flows.pdf?sequence=1&isAllowed=y.
[16] The EDPB is the independent European advisory
body on data protection. [29] P.M. Schwartz, ‘Managing Global Data Privacy’
(2009) Privacy Projects, p. 4.
– 197 –
2019 ITU Kaleidoscope Academic Conference
[30] Case C-101/01 Criminal proceedings against Bodil [39] C. Cava et al., Bioinformatics clouds for high-
Lindqvist [2015] ECLI:EU:C:2003:596. In the throughput technologies, in: Handbook of research
Lindqvist case one of the question was whether there on cloud infrastructures for big data analytics, 2014,
was ‘transfer of data’ when personal data is loaded p. 489 – 507.
onto an Internet page which is stored on an Internet
site on which the page can be consulted, thereby [40] J. Bossi Malafosse et. al., Introductory report for
making the data accessible to anyone who connects updating recommendation R (97) 5 of the Council of
to the Internet, including people in a third country. Europe on the protection of medical data, 2015, T-
As regards the technical nature of the operations, the PD (2015)07, available at
Court concluded that the Internet pages in question https://rm.coe.int/introductory-report-for-updating-
did not contain the technical means to send recommendation-r-97-5-of-the-council-
information automatically to people who did not /168073510c.
intentionally seek access to those pages. Internet
users would have to connect to the Internet and [41] Article 29 Data Protection Working Party, Opinion
personally carry out the necessary actions to consult 8/2010 on applicable law, WP179, p. 23, available
those pages. As such, the data was not directly at: https://ec.europa.eu/justice/article-
transferred between the person uploading the 29/documentation/opinion-
information to the website and persons entering the recommendation/files/2010/wp179_en.pdf.
website. The CJEU refers to the provisions on
transfer in Chapter IV of Directive 95/46/EC, which [42] The concept of ‘transfer’ of data under European
has been replaced by Chapter V of the GDPR. data protection law – in the context of transborder
data flows, Faculty of Law – University of Oslo, 01-
[31] Convention for the protection of individuals with 12-2015, available at:
regard to automatic processing of personal data https://www.duo.uio.no/bitstream/handle/10852/497
[1981] ETS No. 108. 22/8026_The-concept-of-transfer-of-data-under-
European-data-protection-law---In-the-context-of-
[32] Convention for the protection of individuals with transborder-data-
regard to automatic processing of personal data flows.pdf?sequence=1&isAllowed=y.
[1981] ETS No. 108, Article 12.
[43] T. Mulder, M. Tudorica, Privacy policies, cross-
[33] Council of Europe, “Explanatory report to the border health data and the GDPR, Information &
Convention for the protection of individuals with Communications Technology Law (2019) 28,3.
regard to automatic processing of personal data”
(ETS No 108), para. 62, 63. [44] F. Pasquale, ‘Redescribing Health Privacy: the
Importance of Information Policy’ (2014) 103
[34] Council of Europe, “Explanatory report to the HJHLP 127.
Convention for the protection of individuals with
regard to automatic processing of personal data” [45] See for example: A M McDonald and L F Cranor,
(ETS No 108), para. 63. ‘The Cost of Reading Privacy Policies’ (2008) 4 A
Journal of Law and Policy for the Information
[35] Article 4 (9) GDPR. Society 543; F Schaub and os, ‘Designing Effective
Privacy Notices and Controls’ (2017) 99 IEEE
[36] Council of Europe, “Explanatory report to the Internet Computing 70.
Protocol amending the Convention for the protection
of individuals with regard to automatic processing of
personal data” (ETS No 223), para. 23.
– 198 –
ABSTRACTS
Session 1: ICT infrastructure for healthcare1
S1.1 5G-enabled health systems: Solutions, challenges and future research trends
Di Zhang and Teng Zhang, Zhengzhou University, China; Yunkai Zhai, Zhengzhou University
and National Engineering Laboratory for Internet Medical Systems and Applications, China;
Joel J.P.C. Rodrigues, Federal University of Piauí, Brazil and Instituto de Telecomunicações,
Portugal; Dalong Zhang, Zhengzhou University, China; Zheng Wen, Keping Yu and Takuro
Sato, Waseda University, Japan
The transition from analog to digital television has availed new spectrum called white space,
which can be used to boost the capacity of wireless networks on an opportunistic basis. One
sector in which there is a need to use white space frequencies is the healthcare sector because
of existent protocols which are using it and the white space frequency is not as crowded as Wi-
Fi. However, design simulations of wireless communication networks in white space
frequencies have revealed dense network topology because of better signal propagation and
penetration properties of white space frequencies. Consequently, communication networks
designed in white space frequencies will require topology reduction for better communication
and routing. Therefore, this paper proposes a link-based topology reduction algorithm to reduce
a dense mesh network topology designed in white space frequencies into a sparse mesh
network topology. The paper also proposes a network optimization function to introduce a
hierarchical backbone-based network topology from the sparse network topology for better
scalability. Performance evaluation on the proposed designs show that the designs can guide
network engineers to select the most relevant performance metrics during a network feasibility
study in white space frequencies, aimed at guiding the implementation process.
1
Papers marked with an “*” were nominated for the three best paper awards.
– 201 –
S1.3 Exploration of the non-intrusive optical intervention therapy based on the indoor smart lighting
facility
Jian Song, Xiaofei Wang, Hongming Zhang and Changyong Pan, Tsinghua University, China
Light, originally the natural light, is one of the important contributing factors to the creation of
life on earth, the evolution of human beings and the development of civilization. With the
emergence of electric light sources, more specifically the LED lighting lamps which are now
being utilized all over the world, the concept of Internet of light (IoL) using the existing LED
illumination network with the combination of ICT technologies was created. It has become
popular recently and is now widely believed to have a long-lasting impact. IoL not only
improves the lighting efficiency, indoor lighting comfort level and other value-added services,
but also provides the possibilities for regulating human physiological rhythm, especially for the
alleviation of degenerative neurological diseases, even for the treatment and service of healthy
lighting in a non-intrusive way. This paper first introduces the concept and the system structure
of IoL, and then gives the preliminary results and considerations on how this integrated
platform can be utilized to carry the life sciences research and potentially the future
applications for the wellness of senior people. More work could be conducted and it would be
quite necessary to take into consideration standardization from the perspectives of
communication, Internet of things applications, and non-intrusive optical intervention therapy.
S1.4 Access technologies for medical IoT systems
Junaid Ahmed Siddiquee, Ericsson, India
ICT technologies are evolving and advances in the technologies hold promise for applications
in diverse domains such as healthcare. Along with the development of access technologies,
rapid advances are also taking place in related areas, machine learning, artificial intelligence,
cloud computing, and big data. Availing healthcare in the developing countries is costly, time-
consuming and, for populations located in remote areas, it also means adding in the cost of
travel to nearby towns and cities where expert healthcare facilities are normally available.
Leveraging ICT technologies, IoT systems for healthcare can bring affordable and quality
healthcare to the population through e-health and m-health applications. The role of ICT
technologies is paramount to the success of IoT applications for healthcare. Two such ICT
access standards are the 3GPP-based 5G technology and IEEE-based Wi-Fi 6. However,
challenges exist in the ecosystem that inhibit the realization of the full potential of these
technologies. Based on current and future requirements, the paper proposes a model
incorporating key factors impacting an IoT communication system and comes up with a set of
recommendations to harness the Internet of things for healthcare.
Several million people around the world live with limb loss. Prosthetics are useful to improve
their quality of life, and some powered prosthetics enable them to walk naturally. However, most
are too expensive for most amputees to afford. We propose a module structure for a foot
prosthetic and standardized interfaces between modules to lower the price of powered ones. The
prosthetic is battery-powered and controlled by data from sensors built into the heel of a shoe for
a healthy foot. Some modules can be applied to people with walking disabilities. Such
standardization can lower the price of such modules, and many amputees and people with
walking disabilities, such as hemiplegia, can easily afford them, which can help improve their
quality of life.
– 202 –
S2.2 Development of hearing technology with personalized safe listening features
Shayan Gupta, Carnegie Mellon University & Audition Technology, LLC, United States; Xuan
Xu, Hongfu Liu, Jacqueline Zhang; Joshua N Bas and Shawn K. Kelly, Carnegie Mellon
University, United States
Noise induced hearing loss (NIHL) is a growing public health concern in the US and globally
due to the emergence of lifestyle preferences and environmental exposures to sound levels
exceeding safe listening limits for extended periods of time. Issuance of the ITU guidelines for
safe listening devices/systems (ITU-T H.870) leading to the 2019 WHO-ITU standard, along
with existing US federal and military standards, provide a framework for developing an
accessible tool for promoting safe listening. Our proposed Hearing Health app, is being
developed for an aggregated assessment of a user's daily sound exposure, through the audio
system and the environment (occupation and beyond) by integrating WHO-ITU and US safe
listening standards, providing real-time alerts, user-centric recommendations and education that
can be integrated into user lifestyles, representing a wide demographic including young adult,
adult, civilian and military populations. The overall goal of the app will be to increase NIHL
awareness and facilitate improvement of user's listening behaviors.
– 203 –
Session 4: Digital health strategies
S4.1 Invited paper - Towards international standards for the evaluation of artificial intelligence for
health
Markus A. Wenzel, Fraunhofer Heinrich Hertz Institute, Germany; and Thomas Wiegand,
Fraunhofer Heinrich Hertz Institute and Technische Universität Berlin, Germany
Healthcare can benefit considerably from advanced information processing technologies, in
particular from machine learning (ML) and artificial intelligence (AI). However, the health
domain only hesitantly adopts these powerful but complex innovations so far, because any
technical fault can affect people’s health, privacy, and consequently their entire lives. In this
paper, we substantiate that international standards are required for thoroughly validating AI
solutions for health, by benchmarking their performance. These standards might ultimately
create well-founded trust in those AI solutions that have provided conclusive evidence to be
accurate, effective and reliable. We give reasons that standardized benchmarking of AI solutions
for health is a necessary complement of established assessment procedures. In particular, we
demonstrate that it is beneficial to tackle this topic on a global scale and summarize the
achievements of the first year of the ITU/WHO focus group on “AI for Health” that has tasked
itself to work towards creating these evaluation standards.
S4.2 Redesigning a basic laboratory information system for the global south*
Jung Wook Park, Aditi Shah, Rosa I. Arriaga and Santosh Vempala, Georgia Institute of
Technology, United States
Laboratory information systems (LIS) optimize information storage and processing for clinics
and hospitals. In the recent past, developers of LIS for the global south have worked under the
assumption that computing environments will be very limited. However, the computing
resources in the area have been rapidly enriched. This has also changed the expectations that
users have about the LIS interface and functionality. In this paper, we provide a case study of
C4G BLIS that has been in operation for nearly a decade in seven African countries. In two
studies that included 51 participants from three African countries, we redesigned the LIS to
better suit the changing technical landscape and user needs and evaluated the new design. The
study procedure, usability metrics and lessons learned from our evaluation provide a model that
other researchers can use. The findings provide empirical insights that can benefit designers and
developers of LIS in the global south. The results also highlight the need for adding usability
specifications for international standard organizations.
S4.3 #RingingTheAlarm: Chronic "Pilotitis" stunts digital health in Nepal*
Ichhya Pant, George Washington University School of Public Health, United States; and
Anubhuti Poudyal, George Washington University School of Medicine and Health Sciences,
United States
Nepal Health Sector Strategy (NHSS) 2015-2020 aspires to leverage digital health to improve
health outcomes for Nepalese citizens. At present, there is a paucity in evidence on digital health
projects that have been implemented in Nepal. This study aims to map past and extant digital
health projects using Arksey and O'Malley's scoping design framework and assess projects using
the World Health Organization (WHO) building blocks of a health systems framework. Our
findings shed light on the current actors in the digital health space, the spectrum of health
services offered, along with opportunities and challenges to move beyond "pilotitis". In total, 20
digital health solutions were identified through our review that were implemented between 1993
to 2017. The momentum for digital health projects in Nepal is sporadic but continuous. Overall,
digital health solutions in Nepal are limited in scope, focus areas, target audiences and
sustainability potential. At the national level, implementation of digital health projects is frayed,
issue and organization-centric, and primarily driven by donor or non-governmental
organizations. Engaging the private sector, especially telecommunications companies, is an
underutilized strategy to move beyond "pilotitis". Existing pioneers in the space must engage in
strategic collaborative partnerships with the private sector or incentivize independent
commercial health technology ventures.
– 204 –
S4.4 Designing national health stack for public health: Role of ICT-based knowledge management
system
Charru Malhotra, Indian Institute of Public Administration, India; Vinod Kotwal, Department of
Telecommunication, India; and Aniket Basu, Indian Institute of Public Administration, India
Public health (PH), as a domain, requires astute amalgamation of the workings of different
disciplines, because its eventual aim is to ‘prevent’ and not just ‘cure’ the health concerns of the
entire community/population under consideration. Public health goals can be achieved more
meaningfully by the application of information communication technology (ICT) that helps in
overcoming the bottlenecks of brick-and-mortar healthcare models. Online consultations, cloud-
based health management solutions, smart service-supported diagnoses are some such examples.
The present study attempts to explore the design and implementation of ICT-based holistic
knowledge management systems (KMS) to address public health concerns at the national level.
At any point in time, different management information systems (MIS) are being used by various
public authorities that directly or indirectly impact PH. However, the data being generated by
these MIS is “stove piped” into standalone, heterogeneous databases. Non-standardized data
formats, incompatible IT systems, an aggravated sense of ownership by the agency that collects
the data are some of the factors that further worsen the problem. To overcome these issues, based
on the study of best practices and literature review, the review paper proposes a conceptual
model, referred to as national health stack (NHS). NHS is a multilayered KMS designed to
support evidence-based decisions of public health and would pave the way towards “Good
Health and well being” (UN SDG 3) for All.
The need for personalized surveillance systems for elderly health care has risen drastically.
However, recent methods involving the usage of wearable devices for activity monitoring offer
limited solutions. To address this issue, we have proposed a system that incorporates a vision-
based deep learning solution for elderly surveillance. This system primarily consists of a novel
multi-feature-based person tracker (MFPT), supported by an efficient vision-based person fall
detector (VPFD). The MFPT encompasses a combination of appearance and motion similarity in
order to perform effective target association for object tracking. The similarity computations are
carried out through Siamese convolutional neural networks (CNNs) and long-short term memory
(LSTM). The VPFD employs histogram-of-oriented-gradients (HoGs) for feature extraction,
followed by the LSTM network for fall classification. The cloud-based storage and retrieval of
objects is employed allowing the two models to work in a distributed manner. The proposed
system meets the objectives of ITU Focus Group on AI for Health (FG-AI4H) under the
category, "falls among the elderly". The system also complies with ITU-T F.743.1 standard, and
it has been evaluated over benchmarked object tracking and fall detection datasets. The
evaluation results show that our system achieves the tracking precision of 94.67% and the
accuracy of 98.01% in fall detection, making it practical for health care system use. The HoG
feature-based LSTM model is a promising item to be standardized in ITU for fall detection in
elderly healthcare management under the requirements and service description provided by ITU-
T F.743.1.
– 205 –
S5.2 A healthcare cost calculator for older patients over the first year after renal transplantation
Rui Fu, Nicholas Mitsakakis and Peter C. Coyte, University of Toronto, Canada
Forecasting tools that accurately predict post-transplantation healthcare use of older end-stage
renal disease (ESRD) patients are needed at the time of transplantation in order to ensure smooth
care delivery in the post-transplant period. We addressed this need by developing a machine-
learning-based calculator that predicts the cost of healthcare for older recipients of a deceased-
donor kidney over the first year following transplantation. Regression tree and regularized linear
regression methods, including ridge regression, lasso regression and elastic net regression were
explored on all cases of deceased-donor renal transplants performed for patients aged over 60 in
Ontario, Canada between March 31, 2002 and April 31, 2013 (N=1328), The optimal model
(lasso) identified age, membership of one of 14 regionalized Local Health Integration Networks,
blood type, sensitization, having diabetes as the primary case of ESRD, total healthcare costs in
the 12-month pre-workup period and the 6-month workup period to be inputs to the cost
calculator. This cost calculator, in conjunction with clinical outcome information, will aid health
system planning and performance to ensure better management of recipients of scarce kidneys.
S5.3 Automatic plan generating system for geriatric care based on mapping similarity and global
optimization
Fei Ma, Chengliang Wang and Zhuo Zeng, Chongqing University, China
The smart home is an effective means of providing geriatric care to increase the ability of the
elderly to live independently and ensure their health in daily life. However, the smart home is not
widely used because it is arduous to obtain a sensing devices selection plan. In this paper, the
accuracy of service selection and cost savings assumes enormous importance. Therefore, we
propose an automatically plan generating system for the elderly based on semantic similarity,
intuitionistic fuzzy theory, and global optimization algorithm, aiming at searching for an
optimized plan. Experiment results indicate that our approach can satisfy care demands and
provide an optimized plan of sensing devices selection.
Information and communication technology (ICT) for health has shown great potential to
improve healthcare efficiency, especially artificial intelligence (AI). To better understand the
influence of ICT technology on health, a framework of the digital health industry has been
proposed in this paper. Factors from the health industry and the ICT part are extracted to study
the interaction between two groups of component factors. Health factors include service and
management; and ICT factors include sensors, networks, data resources, platforms, applications
and solutions. The interaction between ICT and health can be traced through the development
history, from the stage of institutional informationization to regional informationization, and
finally to service intelligentization. Following such a developmental roadmap, AI was chosen as
one of the most powerful technologies to study the penetration effect and key development
trends from the perspectives of data, computing power and algorithms. The health industry will
be much improved or redefined in the coming AI era. To better understand the strengths,
weaknesses and limitations of AI for health, exogenous factors are discussed at the end of the
paper; preparations on collaboration mechanism; standardization and regulation have been
proposed for the sustainable development of digital health in the AI era.
– 206 –
S6.2 Operationalizing data justice in health informatics*
Mamello Thinyane, United Nations University, Macao SAR, China
There is a growing awareness of the need and increasing demands for technology to embed, be
sensitive to, be informed by, and to be a conduit of societal values and ethical principles. Besides
the normative frameworks, such as the Human Rights principles, being used to inform
technology developments, numerous stakeholders are also developing ethical guidelines and
principles to inform their technology solutions across various domains, particularly around the
use of frontier technologies such as artificial intelligence, machine learning, Internet of things,
robotics and big data. Digital health is one of the domains where the convergence of technology
and health stands to have a significant impact on advancing sustainable development
imperatives, specifically around health and wellbeing (i.e. SDG3). As far as digital health is
concerned, what values and ethical principles should inform solutions in this domain, and more
significantly, how should these be translated and embedded into specific technology solutions?
This paper explores the notion of data justice in the context of health informatics and outlines the
key considerations for data collection, processing, use, sharing and exchange towards health
outcomes and impact. Further, the paper explores the operationalization of Mortier et al.'s
Human-Data Interaction principles of legibility, agency and negotiability through a health
informatics system architecture.
Healthcare is becoming more connected. Risks to patient and public safety are increasing due to
cybersecurity attacks. To best thwart cyberattacks, the Internet of health things (IoHT) must
respond at machine speed. Cybersecurity standards being developed today will enable future
IoHT systems to automatically adapt to cybersecurity threats in real time, based on a quantitative
analysis of reasonable mitigations performing triage to economically optimize the overall
healthcare outcome. This paper will discuss cybersecurity threats, risk, health impact, and how
future IoHT cybersecurity systems will adapt to threats in real time.
– 207 –
Session 8: Data protection and privacy in healthcare
S8.1 Technical and legal challenges for healthcare blockchains and smart contracts
Steven A. Wright, Georgia State University, United States
The paper considers the technical and legal challenges impacting recent proposals for healthcare
applications of blockchain and smart contracts. Healthcare blockchain data and actors are rather
different to cryptocurrency data and actors, resulting in a different emphasis on blockchain
features. Technical issues with healthcare blockchain implementation and trust are considered, as
well as a variety of potential legal issues. Conclusions and recommendations are proposed for
open source and standardization efforts to reduce technical and legal risks for healthcare
blockchains and smart contracts.
With the rapid development of electronic health care, the era of medical big data has already
emerged. However, in the global electronic health industry environment, one of the significant
challenges is that the various medical institutions are independent of one another. Patients,
doctors and medical researchers have significant barriers in accessing medical data. As an
intervention strategy using blockchain principle, this paper explores the characteristics of
blockchain which are applicable to the management of electronic health records (EHRs), and
presents a credible blockchain-based electronic health records (CB-EHRs) management
platform. A CB-EHRs platform is characterized by decentralization, data tamper-proof,
collective maintenance mechanisms, security and credibility. This platform cannot only realize
data sharing between medical institutions, but also ensures the privacy of users. This paper
introduces the components of the CB-EHRs platformmodel and the implementation principle of
its related functions. In addition, this paper also reviews and selects the delegated Byzantine
Fault Tolerance (dBFT) consensus mechanism as a viable option for the CB-EHRs platform.
Finally, by comparing with the Practical Byzantine Fault Tolerance (PBFT) consensus
mechanism and our research, we highlight the potential advantages of our proposed CB-EHRs
platform in the medical domain.
Health data comes within a person's most intimate sphere. It is therefore considered to be
sensitive data due to the great impact it could have on a person's life if this data were freely
available. Unauthorized disclosure may lead to various forms of discrimination and violation of
fundamental rights. Rapid modern technological developments bring enormous benefits to
society. However, with this digitization, large amounts of health data are generated. This makes
our health data vulnerable, especially when transferred across borders. The new EU General
Data Protection Regulation (GDPR) legal framework provides for rights for users of modern
technologies (data subjects) and obligations for companies (controllers and processors) with
regard to the processing of personal data. Chapter V of the GDPR protects personal data that are
transferred to third countries, outside the EU. The term 'transfer' itself, however, is not defined
by the GDPR. This paper examines whether transfer within the meaning of the GDPR applies to
health data processed by modern technologies and if the complexity of the GDPR legal
framework as such sufficiently reflects reality and protects health data that moves across borders,
in particular to jurisdictions outside the EU.
– 208 –
INDEX OF AUTHORS
Malhotra, Charru................................................ 95
Index of Authors
Mauwa, Hope....................................................... 9
Min, Dong ........................................................ 135
Arriaga, Rosa I. .................................................77
Mitsakakis, Nicholas........................................ 115
Mulder, Trix..................................................... 191
Bagula, Antoine .......................................... 9, 183 Murata, Yoshitoshi ............................................ 33
Bas, Joshua N. .................................................... 39
Basu, Aniket ....................................................... 95
Nakamura, Akinori .......................................... 59
Ngqondi, Tembisa................................................ 9
Cha, Hongki......................................................49
Chiwewe, Tapiwa ............................................. 183
Pan, Changyong ............................................... 17
Choi, Younghwan ............................................... 49
Pant, Ichhya ....................................................... 85
Coyte, Peter C................................................... 115
Park, Jung Wook................................................ 77
Poudyal, Anubhuti ............................................. 85
Dharmalingham, Vivekanandan .....................105
Ravikumar, Aswin Kumar ............................. 105
Fu, Rui ...........................................................115 Rodrigues, Joel J. P. C. ........................................ 1
– 211 –
Wright, Steven A. ............................................. 173
Zeng, Zhuo ..................................................... 125
Zhai, Yunkai ........................................................ 1
Xu, Lingyu ......................................................183 Zhang, Dalong ..................................................... 1
Xu, Shan ........................................................... 135 Zhang, Di ..............................................................1
Xu, Xuan............................................................. 39 Zhang, Hongming ...............................................17
Zhang, Jacqueline ...............................................39
Zhang, Teng ..........................................................1
Yamato, Tomoki ...............................................33
Yoshikawa, Kentaro ........................................... 59
Yu, Keping ........................................................... 1
– 212 –
11TH ITU ACADEMIC CONFERENCE
ITUKALEIDOSCOPE
ATLANTA 2019
ICT for Health:
Networks, standards and innovation
4-6 December
International
Telecommunication
Union
Telecommunication
Standardization Bureau
Place des Nations
CH-1211 Geneva 20
Switzerland
ISBN 978-92-61-24301-2
ISBN: 978-92-61-28401-5