eHealth Context Inference

A Review of Open Source Frameworks Initiatives

Arsénio Reis

1,2

, Dennis Paulino

1,2

, Paulo Martins

1,2

, Hugo Paredes

1,2

and João Barroso

1,2

INESC TEC, Porto, Portugal

University of Trás-os-Montes e Alto Douro, Vila Real, Portugal

Keywords: eHealth Framework, Context Awareness, Context Inference, Predictive Models.

Abstract: The collection of health and fitness longitudinal data can be used to model disease progression and shape new

algorithms to diagnose and predict health hazards. Continuously tracking vital signs, in particular heart rate

and skin temperature, can be very informative by using models and algorithms to predict and notify the user

about when he might be falling ill. With the current wearable devices and the proper algorithms, the individual

can be permanently monitored, which might be much more interesting than a one-off reading comparison

with the population average, made by a doctor. It would be possible to intervene earlier and to prevent

somebody from becoming seriously ill. From a broader perspective, the knowledge about a user’s health can

be considered as an element of that user’s context and be used by context aware applications to provide higher

value to the user. After the trivialization of the data acquisition sensors, wearable devices, and raw data, the

next logical step is the development of contained software components that can infer and produce knowledge

from the raw data. These components and the knowledge they produce can be used by all sorts of applications

in order to further customize their usage by a specific user. Customization and context awareness, in regard

to health, is a wide field for research and there are a multitude of proposals for models and algorithms. In this

review work we searched for software components (frameworks, software libraries, etc.), freely available and

that can be used as building blocks for other computer systems by software developers.

1 INTRODUCTION

In this work, we conduct a review of the currently

available software systems, freely accessible to

developers, which can be used as software building

blocks, providing functionalities associated to

information inference and knowledge production,

related the user’s health status.

There are currently multiple wearable devices,

equipped with sensors, which can read vital signs

from a user. Some of these devices are not very

accurate, mainly developed and targeted for the low

end recreational and fitness market, others have

medical grade accuracy and are targeted for the high-

performance sports and medical environments.

For a health-related solutions developer, the raw

data must be interpreted according to the specific

purpose of the desire solution. In that regard, data

analysis and knowledge inference is a specific

research domain from which can be produced

software modules, encapsulating all the research and

deep knowledge necessaries to create the inference

functionalities. These modules can be used as

building blocks for other software solutions than will

then benefit from having knowledge regarding their

user’s health.

This work reviews and analyses the currently

available software systems, designed to provide

specific health knowledge and that can be used as

building blocks for more general software solutions.

In this document, we designate these systems as

frameworks and middleware in same sense as the

terms are used in software engineering, although not

being very strict and using the framework term even

when in fact the system doesn’t have all the elements

of a software framework according to the well-known

concept of software framework.

A framework is a structure or conceptual guide to

build an entity by expanding itself into something

useful. In electronic systems a framework, generally,

describes a layered structure of software programs

and hardware devices and how they are built,

interconnected, and interact, to provide a full system

functionality. The framework is comprehensive and

Reis A., Paulino D., Martins P., Paredes H. and Barroso J.

eHealth Context Inference - A Review of Open Source Frameworks Initiatives.

DOI: 10.5220/0006752707070714

In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (HEALTHINF 2018), pages 707-714

ISBN: 978-989-758-281-3

prescriptive. In computer programming, a software

framework is an abstraction in which a software

system, providing a generic functionality, can be

extended to provide application specific functionality

by building and changing specific elements of the

framework, thus creating a specific solution from a

generic set of elements. The software framework

provides a standard way to build and deploy

applications. It can provide a functionality, as part of

a larger software platform, simplifying the

development of software application and solutions in

a universal and reusable environment, which may

include several types of elements, e.g., programs,

compilers, code libraries, tool sets, and application

programming interfaces (APIs) that can be used for

the development of a system.

Software frameworks have the following key

features that distinguishes them from other software

components. (1) inversion of control, in which is the

framework that controls the overall program's flow

and not the caller library or application; (2)

extensibility, providing mechanisms for the user to

extend the framework, by selective overriding or

addition of specialized user code to provide a specific

functionality; (3) non-modifiable framework code,

meaning the users can’t modify the framework’s code

but must extend it instead.

A software library is a collection of resources used

to develop software and build computer programs. It

includes configuration data, documentation, help

data, message templates, pre-written code and

subroutines, classes, values or type specifications,

etc. Usually these resources have specific behaviors

and are accesses by well-defined interfaces.

In software engineering, an Application

Programming Interface (API) is a set of subroutine

definitions, protocols, and tools for building

application software. It clearly defines the methods of

communication between various software

components. An API provides the methods for the

programmer to develop a computer program by

assembling together different building blocks.

An API usually describes and prescribes the

expected behavior of a software library, while the

library is an actual implementation of the

functionalities. The separation of the API from the

implementation allows programs written in one

language to use a library written in a different one

(Robillard et al., 2012).

Middleware is a software that stands in between

two software layers (in the middle), usually providing

value added services to software applications, on top

of those available from the operating system.

A framework can include a software library,

accessible by means of an API, and freely available

to the programming community as middleware

software, as represented in Figure 1.

Figure 1: Overview of the connections between framework,

middleware and API.

In the context of the software development

elements previously described (frameworks, libraries

and APIs), we are interested in those that can use

health related data, e.g., vital signs and others, and

produce health related knowledge, that can be used as

building blocks by user developed applications.

2 HEALTH DATA AND

PROGNOSTIC MODELS

Currently there are several consumer wearables and

portal services that can monitor some basic vital

signs, e.g., heart rate and skin temperature, allowing

the user to register and view their evolution over time.

Based on this data, models and algorithms can be

developed that will be able to predict the user’s health

status and evolution by computing the longitudinal

data of the basic vital signs. The recent consumer

grade wearables are not really expensive (some

hundred euros per device) and some have good

quality sensors, mainly developed for the fitness

market. These devices have, mostly, an informative

usage. They provide monitoring information to the

users. To have a real impact on the users’ health

management and on the health care system usage,

besides to provide information about the user’s vital

signs, this devices must also be able to analyze and

draw conclusions about the user’s health on his

behalf. It is then necessary to collect data and create

models and algorithms for the interpretation of the

vital signs data.

The abundance of wearable devices, together with

the Internet of Things (IoT) connectivity paradigm,

are very strong drives for the acquisition of real-time

data, particularly, data collection from individuals in

their daily lives, as well as in specific sportive

activities, such as the data of simple vital signs, e.g,

heart rate and skin temperature, as well as, more

complex data, regarding the individual health

evolution, e.g., ECG, surveys (Paulino et al, 2017;

Felisberto et al, 2015). The basic data can be used as

parameters to model specific health phenomena and

the complex data can be used to assess the correlation

of basic data patterns with the health phenomena. It is

then possible to create inference algorithms to predict

the user’s health evolution.

To create health related prediction models, several

techniques can be used, e.g, decision rules, score

systems, Markov processes, decision trees, neural

networks and Baysian networks. To model the

evolution of health phenomena over time, the Baysian

approach is one the most suitable solutions. However,

the adoption of a specific technique must be

considered according to the available data, specific

phenomena, and objective.

There are several techniques currently used to

create prognostic models. The simplest are the

decision rules, which are based on a prognostic score,

classifying the patient in a specific risk category

(Knaus et al., 1991). Also, in decision analysis, the

Markov decision processes, which is a stochastic

process that evolve over time, can be used as basis for

prognostic models (Sonnenberg et al., 1993). More

sophisticated techniques, with support from the

artificial intelligence community, are becoming

popular as prognostic models, including: decision-

trees, neural networks, support vector machines, and

Bayesian networks (Cruz et al., 2006; Ohno-

Machado, 1997). These techniques have much better

performance but require substantial amounts of

quality data.

The Bayesian networks technique have been

successfully used to model health phenomena, and

other natural phenomena, e.g., genomics and ecology

(Jansen et al., 2003), and has become a powerful tool

for the analysis of real-world data, from health and

the environment, leading to the shaping of novel

algorithms (Gerven et al., 2007; Lucas et al., 1995;

Murphy et al., 2002; Pearl et al., 2000). This tool and

the collection of longitudinal data are creating

research challenges for diagnosis and the modelling

of disease progression (Abu-Hanna, 2001).

Wearable devices can be used to track the

individual’s vital signs and detect when he is about to

get hill. In the follow-up of a recent research, by

wearing sensors during a long period (over an year),

Michael Snyder (Snyder, 2017) was able to detect

abnormal readings, showing an increase in heart rate,

when compared with his regular heart rate pattern,

and a rise in skin temperature. On a recent interview,

Snyder would confirm: “A mild fever soon followed,

and Snyder asked a doctor for the antibiotic

doxycycline, which can be used to treat Lyme

disease. His symptoms cleared within a day”.

Subsequent tests confirmed his self-diagnosis. Snyder

and his team are working with more than 40

volunteers, who wore smartwatch devices for up to

two years and have demonstrated a solution to detect

the first signs of illness by continuously monitoring

their pulse and skin temperature (Li et al., 2017;

Klein, 2017).

Several vital parameters can be continuously

monitored using wearable devices. The most frequent

is heart rate, as it is simple to monitor and has been

widely studied, regarding several aspects, e.g., its

levels, patterns, correlation with health, and death risk

(Johansen et al., 2013). Skin temperature can also be

monitor (Nakayama et al., 1977).

Heart rate variability (HRV) is a term that

indicates the level of changes in HR. It was firstly

used in 1846 when Carl Ludwing discovered

Respiratory sinus arrhythmia, through which HR

increases with inspiration and decreases during

exhalation. HR is controlled via the two components

of the autonomic nervous system (ANS):

Sympathetic and parasympathetic indices (Makivic et

al., 2013; Javorka et al., 2001). The sympathetic

component is responsible for the HR acceleration

during physical stress (i.e. exercise) whereas lower

HR (example: Resting HR) is an impact of the vagal

tone (parasympathetic tone) (Makivic et al., 2013).

Consequently, the balance between the two

components affects the time consistency between

Heart beats, which is measured through HRV

(Makivic et al., 2013).

Several studies have shown the correlation

between heart rate data and particular health

conditions. The analysis of heart rate variability

(HRV) and its pattern can provide an insight into the

health of the cardiovascular system (Javorka, 2001).

One of the simplest analysis, using heart rate data,

is the interpretation of the resting pulse rate (RPR),

which can provide useful information. RPR is a key

vital sign measure in clinical practices with widely

available reference data. Resting heart rate is

independently associated with increased risks of all-

cause and cardiovascular mortality, and, as shown by

(Zhang et al., 2016), can be used as a mortality

predictor. Beddhu et al, examined the association of

resting heart rate with insulin resistance,

cardiovascular events and mortality in the moderate

chronic kidney disease (CKD) population and have

demonstrated how higher resting heart is associated

with increased mortality and possibly cardiovascular

events in this population (Beddhu et al., 2009). There

are several other authors using heart rate as a

predictor for several health conditions, e.g., a tool for

risk stratification in primary care (Leistner et al.,

2012), osteoporotic fractures and mortality in older

women (Kado et al., 2012), cardiovascular mortality

in the general population (Hozawa et al, 2004), to

sudden death and all-cause mortality in asymptomatic

men (Adabag et al., 2008), coronary heart disease in

the elderly (Legeai et al., 2011).

In the context of the sports practice, the exercise

intensity assessment is important to inform the athlete

of the correct training effort and prevent overtraining

syndrome (OS) (Dressendorfer et al, 1985; Hedelin et

al., 2000) or sudden cardiac death (SCD) (Savonen,

2006; Batty et al., 2010).

According to (Link and Estes, 2012), SCD rates

in the US reaches 150 cases annually. Such cardiac

complications occur either due to the medical history

of the athlete or because of unmonitored exercising.

If a trainee exercised without exceeding the

endurance limit, then it is said he is in the functional

over-reaching (FO) where the safety is highly assured

and normal stamina can be obtained after hours/day

of recovery. However, ignoring the OR region might

lead to developing Overtraining syndrome (OS),

which is a precursor of SCDs.

The pattern associated with how the heart rate

recovers, after exercising, is a parameter that

represents how fast the heart returns to the normal

state after exercising. Typically, a drop of 12 bpm or

less in the first minute of recovery is considered

abnormal, and greater drop during this period

signifies that the person is fit (Cole et al., 2000; Reis

et al., 2016a).

With the vital signs collected data, e.g., HR and

skin temperature, and with the support of meta data,

it can be created longitudinal models and algorithms,

correlating the health hazards (e.g., physical

exhaustion, developing flu) and the HR patterns.

These models will represent the time progression of

the hazard and the algorithms will position the

individual in a specific time of the model, by

interpreting the patterns on the HR time line.

Although there are, currently available, devices

that can continuously monitor the basic vital signs,

the research, by Bloss et al., suggests: “there are not

large short-term increases or decreases in health care

costs or usage associated with monitoring chronic

health conditions using mobile health or digital

medicine technologies.” (Bloss et al., 2016; Reis et

al., 2016b). That is a disappointing conclusion, and

calls for further work to intelligently analyze the data

and deliver truly valuable information to the user, as

well as make the health-related inference and

knowledge production features available to the

general community of software solutions developers.

Context aware applications should then be able to

use the knowledge about the user’s health as part of

the user’s context. In a very simple example, a context

aware groceries store shopping application, besides

knowing the items that the user needs to buy, could

also know the items the users should avoid buying

due to its health status. Other applications have

proposed the usage of electronic assistants to interact

and accompany elderly people (Reis et al., 2017a;

Marceline et al., 2009; Reis et al., 2017b).

3 METHODOLOGY

For this review we defined the following

methodology in order to identify the software

frameworks to consider and study:

1. An internet search using the terms “software”,

“healthcare”, “health”, “inference”, “system”,

“middleware”, “framework”. The search was

conducted using the “publish or perish” software,

using the google scholar engine data. The search

was limited to 1,000 item and 980 items were

returned.

2. A filter, excluding surveys and literature reviews.

It returned 892 items.

3. A filter, excluding those without the term “health”

on their title. The return list was narrow down to

301 items.

4. A filter, by means of a comprehensive reading by

a researcher, and exclusion of those not related to

a software framework or middleware. The list was

shortened to 45 items.

5. A filter, excluding those not related to this work

theme or not having any bibliographic citation.

This filtered the results to 7 items.

After all the searching and filtering, the items were

carefully analyzed and described, namely regarding:

 The inference algorithms (machine learning,

vector learning, deep learning, Bayesian, Rules,

Markov, etc);

 The system’s learning strategy;

 The types of data necessary for the inference to

work;

 The pathologies for which the system can produce

knowledge;

4 ANALYSIS AND RESULTS

The results are summarized in Table 1, in which are

listed the framework’s titles and their online

information links.

Table 1: Review results.

“Context-aware hybrid reasoning framework for pervasive

healthcare”

(https://link.springer.com/article/10.1007/s00779-013-0696-5)

“Bdcam: Big data for context-aware monitoring-a personalized

knowledge discovery framework for assisted healthcare”

(http://ieeexplore.ieee.org/document/7117389/)

“MediAlly: A provenance-aware remote health monitoring

middleware”

(http://ieeexplore.ieee.org/document/5466985/)

“An integrated multi-sensing framework for pervasive

healthcare monitoring”

(http://ieeexplore.ieee.org/document/5191197/)

“A framework for context-aware home-health monitoring”

(http://www.inderscienceonline.com/doi/abs/10.1504/IJAACS

.2010.030313?journalCode=ijaacs)

“A Middleware Framework for Ambiguous Context Mediation

in Smart Healthcare Application”

(http://ieeexplore.ieee.org/abstract/document/4390866/)

“An Inference System Framework for Personal Sensor Devices

in Mobile Health and Internet of Things Networks”

http://dro.deakin.edu.au/view/DU:30092154)

The following list describes, in more detail, the

analysis of the selected frameworks.

1. “Context-aware hybrid reasoning framework

for pervasive healthcare” (Yuan et al., 2014)

This article main objective is to describe the creation

of a framework to manage the context monitoring of

elderly people on their homes. The inference method

is based on Fuzzy Rules (Wang et al., 1991) and the

system learns by Case Based Reasoning (CBR) (Xu

et al., 1995). The framework uses vital signs data and

infers if the individual has a normal health condition.

2. “Bdcam: Big data for context-aware

monitoring-a personalized knowledge

discovery framework for assisted healthcare”

(Forkan et al., 2015)

This framework is intended to build a system to

monitor a person’s vital signs, inferring the

individual’s current health status from that data. The

system uses the MapReduce Apriori algorithm

(Yahya et al., 2012) that can register the correlations

between distinct variables of the acquired context,

producing a set of rules. It uses data acquired from the

blood pressure and the heart rate.

3. “MediAlly: A provenance-aware remote

health monitoring middleware” (Chowdhury

et al., 2010)

In this framework, the main purpose is to be a

building block of a system based on mobile devices

to monitor the user’s physiological data, which is later

sent to a middleware that infers the user’s health

status by applying a set of rules.

4. “An integrated multi-sensing framework for

pervasive healthcare monitoring” (ElHelw et

al., 2009)

This framework is intended for the implementation of

a system to monitor a person’s vital data as well as

several specific personal activities, e.g., walking,

eating, sleeping, etc. The system uses videos cameras

and wearable devices to collect data. The inference

determines the activity being executed by the person

and is accomplished using Hidden Markov Models

(HMM) (Eddy, 1996).

5. “A framework for context-aware home-health

monitoring” (Esposito et al, 2010)

This article describes the construction of a framework

and a context aware computational system,

implemented in a home-care scenario. The vital data

is collected and used for inference based on a

predefined rule set.

6. “A Middleware Framework for Ambiguous

Context Mediation in Smart Healthcare

Application” (Roy et al., 2007)

In this article it is described the development of a

middleware software system, tailored to deal with the

context ambiguities in health data monitoring. In a

Smart Home scenario it was used wireless sensors

and RFID tags to acquired data, including: location,

activities, and identification of the users. From this

data, inferences are made using Dynamic Bayesian

Networks, with the objective of measuring the

context quality.

7. “An Inference System Framework for

Personal Sensor Devices in Mobile Health and

Internet of Things Networks” (Kang, 2017)

In this thesis the main objective is to create a

framework to monitor vital signs. The framework can

detect and alert when the user’s vital signs have

abnormal readings. The inference is a simple method

based on verification of rules and predefined

thresholds applied to the vital signs data.

5 CONCLUSIONS

Although the health data monitoring subject is widely

spread in the research community, there aren’t many

proposals towards the development and

implementation of health related inference systems

frameworks. There is much more work related to the

acquisition and presentation of vital signs data, which

suggests a gap between the data and the inference and

knowledge production. On that line of reasoning,

knowledge inference is the next logic step to make

sense from all the data.

Comparing the health knowledge inference with

other research areas, (e.g., computer operating

systems, computer vision, statistics analysis, content

management, etc.) for which there are high quality

and freely available frameworks (e.g. linux, android,

java, openCV, R, wordpress, etc), we can’t find

parallel frameworks for health knowledge inference.

That might be because, although there are lots of high

quality research regarding the models and algorithms,

the technology hasn’t mature enough for the arising

of a general software framework.

Unfortunately we did just a review work, but a

deeper analysis, including actual tests would be very

important to verify the quality of the inference

provided by the systems. In a scenario on which the

data acquisition and processing is trivial, it may be the

models and algorithms to determine the quality of a

complete solution.

Reading thought the documentation of the several

frameworks it is clear that to apply some inference

techniques and longitudinal analysis, it is necessary

to have large amounts of data, which might explain

the gap between the available analysis techniques and

their actual usage in actual software frameworks.

ACKNOWLEDGEMENTS

This work was supported by the Project

“NanoSTIMA: Macro-to-Nano Human Sensing:

Towards Integrated Multimodal Health Monitoring

and Analytics/NORTE-01-0145-FEDER-000016"

financed by the North Portugal Regional Operational

Programme (NORTE 2020), under the PORTUGAL

2020 Partnership Agreement, and through the

European Regional Development Fund (ERDF).

REFERENCES

Abu-Hanna, A. and Lucas, P. (2001). Prognostic models in

medicine: AI and statistical approaches. Meth Inform

Med, 40, pp. 1–5

Adabag, A., Grandits, G., Prineas, R., et al. (2008). Relation

of heart rate parameters during exercise test to sudden

death and all-cause mortality in asymptomatic men. Am

J Cardiol; 101: pp. 1437–1443.

Batty, G. Shipley, M., Kivimaki, M., et al. (2010). Walking

pace, leisure time physical activity, and resting heart

rate in relation to disease-specific mortality in London:

40 years follow-up of the original Whitehall study. An

update of our work with professor Jerry N. Morris

(1910–2009). Ann Epidemiol 2010;20:661–9.

Beddhu, S., Nigwekar, S., Ma, X., Greene, T. (2009).

Associations of resting heart rate with insulin

resistance, cardiovascular events and mortality in

chronic kidney disease. Nephrol Dial Transplant.

2009;24(8):2482–8. doi: 10.1093/ndt/gfp057.

pmid:19307229

Bloss, C., Wineinger, N., Peters, M., Boeldt, D., Ariniello,

L., Kim, J., et al. (2016). A prospective randomized trial

examining health care utilization in individuals using

multiple smartphone-enabled biosensors. PeerJ.

2016;4:e1554. doi: 10.7717/peerj.1554.

pmid:26788432

Chowdhury, A., Falchuk B., Misra, A. (2010). MediAlly: A

Provenance-Aware Remote Health Monitoring

Middleware. In Proceedings of Pervasive Computing

and Communications, pp. 124-134, IEEE

Cole, C., Foody, J., Blackstone, E., Lauer, M. (2000). Heart

rate recovery after submaximal exercise testing as a

predictor of mortality in a cardiovascularly healthy

cohort. Annals of internal medicine, 132(7), pp. 552-

555.

Cruz, A. and Wishart, D. (2006). Applications of machine

learning in cancer prediction and prognosis. Cancer

Inform, 2, pp. 59–78

Dressendorfer, R., Wade, E., Scaff Jr, H. (1985). Increased

morning heart rate in runners: a valid sign of

overtraining?. The Physician and Sportsmedicine,

13(8), pp. 77-86.

Eddy, S. (1996). Hidden Markov models, In Current

Opinion in Structural Biology, Volume 6, Issue 3, 1996,

pp. 361-365, ISSN 0959-440X, https://doi.org/10.1016/

S0959-440X(96)80056-X.

ElHelw, M., Mcllwraith, J. et al. (2009) An integrated

multi-sensing framework for pervasive healthcare

monitoring, 3rd Int. Conf. on Pervasive Como.

Technologies for Healthcare, pp. 1-7

Esposito, A., Tarricone, L., Zappatore, M., Catarinucci, L.,

Colella, R. (2010). A framework for context-aware

home-health monitoring. Int. J. of Autonomous and

Adaptive Communications Systems, Vol.3, No.1, pp.75

- 91

Felisberto, F., Laza, R., Fdez-Riverola, F., Pereira, A.

(2015). A distributed multiagent system architecture for

body area networks applied to healthcare

monitoring. BioMed research international, 2015.

Forkan, A., Khalil, I., Ibaida, A., Tari, Z. (2015). BDCaM:

Big Data for Context-aware Monitoring-A Personalized

Knowledge Discovery Framework for Assisted

Healthcare. IEEE transactions on cloud computing,

2015. Retrieved from: http://ieeexplore.ieee.org/

document/7117389/

Gerven, M. (2007) Bayesian networks for clinical decision

support. Radboud University Nijmegen, Nijmegen, the

Netherlands

Hedelin., R., Kenttä, G., Wiklund, U., Bjerle, P.,

Henriksson-Larsén, K. (2000). Short-term overtraining:

effects on performance, circulatory responses, and heart

rate variability. Medicine and science in sports and

exercise, 32(8), pp 1480-1484.

Hozawa, A., Ohkubo, T., Kikuya, M., et al. (2004).

Prognostic value of home heart rate for cardiovascular

mortality in the general population: the Ohasama study.

Am J Hypertens; 17: pp. 1005–1010.

Jansen, R., Yu, H., Greenbaum, D., Kluger, Y., Krogan, N.

J., Chung, S., ... & Gerstein, M. (2003). A Bayesian

networks approach for predicting protein-protein

interactions from genomic data. Science, 302(5644),

pp. 449-453.

Javorka, M., Zila, I., Balharek, T., Javorka, K. (2002). Heart

rate recovery after exercise: relations to heart rate

variability and complexity. Braz J Med Biol Res.;35(8),

pp. 991–1000.

Johansen, C., Olsen, R., Pedersen, L., et al. (2013). Resting,

night-time, and 24 h heart rate as markers of

cardiovascular risk in middle-aged and elderly men and

women with no apparent heart disease. Eur Heart J

2013;34: pp. 1732–1739

Kado, D., Lui, L., Cummings, S. (2002). Rapid resting heart

rate: a simple and powerful predictor of osteoporotic

fractures and mortality in older women. J Am Geriatr

Soc 50: pp. 455–460.

Kang, J. (2017). An inference system framework for

personal sensor devices in mobile health and internet of

things networks, PhD thesis, School of Information

Technology, Deakin University. Retrieved from:

http://dro.deakin.edu.au/view/DU:30092154

Klein, A. (2017). Smartwatches know you’re getting a cold

days before you feel ill. New Scientist. Retrieved from:

https://www.newscientist.com/article/2117854-

smartwatches-know-youre-getting-a-cold-days-before-

you-feel-ill/

Knaus, K., Wagner, D., Draper, E., Zimmerman, J.,

Bergner M. and Bastos, P. (1991). The APACHE III

prognostic system. Risk prediction of hospital mortality

for critically ill hospitalised adults. Chest, 100, pp.

1619–1636

Legeai, C., Jouven, X., Tafflet, M., et al. (2011). Resting

heart rate, mortality and future coronary heart disease

in the elderly: the 3C Study. Eur J Cardiovasc Prev

Rehabil;18: pp 488–97.

Leistner, M., Klotsche, J., Palm, S., Pieper, L., Stalla, G. K.,

Lehnert, H., ... & Zeiher, A. M. (2012). Resting heart

rate as a tool for risk stratification in primary care: does

it provide incremental prognostic information?.

European journal of preventive cardiology, 19(2), pp.

275-284.

Li, X., Dunn, J., Salins, D., Zhou, G., Zhou, W., et al.

(2017) Digital Health: Tracking Physiomes and

Activity Using Wearable Biosensors Reveals Useful

Health-Related Information. PLOS Biology 15(1):

e2001402. doi: 10.1371/journal.pbio.2001402

Link, M., Estes, M. (2012) Sudden cardiac death in the

athlete: bridging the gaps between evidence, policy, and

practice. Circulation 125: pp.2511–2516.

Lucas, P. (1995). Logic engineering in medicine. Knowl

Eng Rev, 10 (2), pp. 153–179

Makivić, B., Nikić, M., Willis, M. (2013). Heart rate

variability (HRV) as a tool for diagnostic and

monitoring performance in sport and physical activities.

Journal of Exercise Physiology, vol. 16, pp.103-131

Marcelino, I., Pereira, A. (2009, September). Elder care

modular solution. In Advances in Human-oriented and

Personalized Mechanisms, Technologies, and Services,

2009. CENTRIC'09. Second International Conference

on (pp. 1-6). IEEE.

Murphy, K. (2002). Dynamic Bayesian networks. UC

Berkeley, Berkeley, CA

Nakayama, T., Ohnuki, Y., Niwa, K. (1977). Fall in skin

temperature during exercise. Jpn J Physiol.;27(4), pp.

423–37. pmid:599738

Ohno-Machado, L. (1997). A comparison of Cox

proportional hazards and artificial neural network

models for medical prognosis. Comput Biol Med, 27,

pp. 55–65

Paulino, D., Reis, A., Barroso, J., Paredes, H. (2017).

Mobile devices to monitor physical activity and health

data. 12th Iberian Conference on Information Systems

and Technologies (CISTI); 06/2017,

DOI:10.23919/CISTI.2017.7975771

Pearl, J. (2000). Causality: models, reasoning and

inference. Cambridge University Press, New York, NY

Reis A., Barroso, I., Monteiro, M., Khanal, S., Rodrigues,

V., Filipe, V., Paredes, H., Barroso, J. (2017a).

Designing Autonomous Systems Interactions with

Elderly People. Universal Access in Human–Computer

Interaction. Human and Technological Environments,

01/2017: pp. 603-611; ISBN: 978-3-319-58699-1,

DOI:10.1007/978-3-319-58700-4_49

Reis, A., Lains, J., Paredes, H., Filipe, V., Abrantes, C.,

Ferreira, F., Mendes, R., Amorim, P., Barroso, J.

(2016a). Developing a System for Post-Stroke

Rehabilitation: An Exergames Approach. Universal

Access in Human-Computer Interaction. Users and

Context Diversity, 1st edited by Margherita Antona,

Constantine Stephanidis, 07/2016: pp. 403-413;

Springer International Publishing., ISBN: 978-3-319-

40237-6, DOI:10.1007/978-3-319-40238-3_39

Reis, A., Morgado, L., Tavares, F., Guedes, M., Reis, C.,

Borges, J., Gonçalves, R., Cruz, J. (2016b). Gestão de

listas de espera para cirurgia na rede hospitalar pública

portuguesa - O sistema de informação dos programas

de recuperação de listas de espera. CISTI 2016, 11.ª

Conferência Ibérica de Sistemas e Tecnologias de

Informação; 06/2016, DOI:10.1109/CISTI.2016.

7521612

Reis, A., Paulino, D., Paredes, H., Barroso, J. (2017b).

Using Intelligent Personal Assistants to Strengthen the

Elderlies’ Social Bonds. Universal Access in Human–

Computer Interaction. Human and Technological

Environments, 01/2017: pp. 593-602; ISBN: 978-3-

319-58699-1, DOI:10.1007/978-3-319-58700-4_48

Robillard, M., DeLine, R. (2012). A field study of API

learning obstacles. Empirical Software Engineering

(EMSE), 16, pp.703-732.

Roy, N., Pallapa, G., Das, S. (2007). A Middleware

Framework for Ambiguous Context Mediation in Smart

Healthcare Application, Proc. IEEE Int'l Conf. Wireless

and Mobile Computing Networking and Comm.

(WiMob '07).

Savonen, K., Lakka, T., Laukkanen, J., et al. (2006). Heart

rate response during exercise test and cardiovascular

mortality in middle-aged men. Eur Heart J

2006;27:582–8.

Snyder, M. (2017). Michael Snyder, STANFORD W.

ASCHERMAN, MD, FACS, PROFESSOR IN

GENETICS. Retrieved from https://med.stanford.edu/

profiles/michael-snyder?tab=bio

Sonnenberg, F. and Beck, J. (1993). Markov models in

medical decision making: a practical guide. Med Decis

Making, 13 , pp. 322–338

Wang, J., Mendel, M. (1991). Generating fuzzy rules by

learning from examples. Proc. 6th Int. Symp. Intelligent

Control, pp. 263-268.

Xu, L. (1995). Case-based reasoning: A major paradigm of

artificial intelligence. IEEE Potentials, vol. Dec 1994-

Jan 1995, pp. 10-13.

Yahya, O., Hegazy, O., Ezat, E. (2012). An Efficient

Implementation of Apriori Algorithm Based on

Hadoop-Mapreduce Model, Proc. of the International

Journal of Reviews in Computing 31st December 2012.

Vol. 12: pp. 59-67.

Yuan, B., Herbert, J. (2014). Context-aware hybrid

reasoning framework for pervasive healthcare. Pers

Ubiquit Comput 18: pp. 865–881. Retrieved from:

https://link.springer.com/article/10.1007/s00779-013-

0696-5

Zhang, D., Shen, X., Qi, X. (2016) Resting heart rate and

all-cause and cardiovascular mortality in the general

population: a meta-analysis. CMAJ: Canadian Medical

Association journal. 88(3), pp.53–63. doi:

10.1503/cmaj.150535. pmid:26598376