Research Tracks for Intelligent Processing to Support

Cardiovascular Disease Management in Personalized

Healthcare: Results from LiNK Project

Antonio Martinez-Millana

, Alvaro Martinez-Romero

, Jorge Henriques

Anna Maria Bianchi

, Carlos Fernandez-Llatas

, Vicente Traver

, Ricardo Couceiro

and Paulo de Carvalho

ITACA, Universitat Politècnica de València,

Camino de Vera sn, 46022 Valencia, Spain

CISUC, Universidade de Coimbra, Coimbra, Portugal

POLIMI, Politecnico di Milano – DEIB, Milano, Italy

Abstract. The goal of LiNK is to discover and define research tracks in the area

of intelligent processing to support cardiovascular disease management in Per-

sonalized Healthcare (PHC). The strategy was based on a two-fold approach.

First, using a status assessment of the current research on PHC and an interna-

tional research forum. From this roadmap, specific innovative research tracks

were developed by common workgroups. Knowledge transfer methodologies be-

tween partners by using a “learn by doing” approach increased the research ex-

cellence momentum. These research tracks supported concept definition activi-

ties that will be the basis for new project, network and PhD grant proposals, lead-

ing to a continuum of widening. Second, existing links to international leading

organizations and key actors in PHC (academic, industrial and users) were ex-

ploited to launch a research and innovation forum in which the definition of a

research agenda and curricula for advanced training was completed.

1 Introduction

Cardiovascular diseases (CVD) are the deadliest among chronic diseases, and with

more people surviving their first cardiac event, CVD is becoming a chronic disease. It

is estimated that they are responsible for 12 million disability adjusted life years lost

annually1 and that nearly half of all deaths in Europe (48%) and in the EU (42%) are

due to CVD. It is the main cause of the disease burden (illness and death) in Europe

(23%). There is a multitude of etiologies for CVD such as ischemic heart disease, hy-

pertension, valvular heart disease, infection and other primary and secondary myocar-

dial diseases [1]. CVD has a major impact on health expenditure. Overall CVD is esti-

mated to cost the EU27 €192 billion distributed between €109 billion of direct costs

(10% of the EU expenditure) and about €83 billion in indirect costs (€41 billion of lost

in productivity and €42 billion for informal care).

Martinez-Millana, A., Martinez-Romero, A., Henriques, J., Bianchi, A., Fernandez-Llatas, C., Traver, V., Couceiro, R. and de Carvalho, P.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK Project.

DOI: 10.5220/0008862000560069

In OPPORTUNITIES AND CHALLENGES for European Projects (EPS Portugal 2017/2018 2017), pages 56-69

ISBN: 978-989-758-361-2

In order to handle the challenges induced by the chronic disease burden, the EU

health systems are undergoing a paradigm shift from reactive care to preventive care

and from in-hospital to home care. Prevention systems support and motivate users in

adopting healthy lifestyles (e.g., physical activity, nutrition, stress management) in or-

der to prevent or delay manifestations of disabling chronic diseases. Disease manage-

ment systems handle the care of patients with chronic disease, combining expertise

from different areas, and integrating new technologies to offer the patient better and

more cost effective care. In this context, personalizing health and care (PHC) systems

have a central role in supporting the paradigm shift by assisting in the provision of

continuous and personalised services to empower patients and professionals in manag-

ing their health. Although in the last decade there has been an intense and significant

research on developing and deploying PHC services in CVD management (up to 50%

of the PHC market products and 40% of research projects are related to CVD manage-

ment2), there are still some major gaps that need to be addressed [2].

Today’s PHC systems miss adequate integration of clinical evidence and knowledge

from holistic clinical practice and biomedical research required to support truly holistic

management of chronic diseases and their co-morbidities. Current PHC systems are

designed using the “one fits all” principal lacking a truly personalization by capturing

and adapting to the patients’ phenotype (e.g., by linking systems medicine and the vir-

tual physiological patient to tele-monitoring data) and individualized treatment or con-

text needs. Data processing is at the core of PHS where acquired data is turned into

meaning and action. In order to pave the way from personal to personalised systems,

PHC require intelligent algorithms to treat and correct data obtained from uncontrolled

conditions, to efficiently integrate multimodal and multi-scale data, to be self-adapting

(moving from population-based to patient-specific adaptations) and interpretable, and

to integrate clinical and biomedical evidence at their genesis.

In this chapter our goal is to outline the approach developed and implemented in the

LiNK project and to provide an overview of the achieved results in identifying and

creating a research ecosystem to address central scientific and technical challenges for

PHC deployment. The methodology followed is fostering EU impact and leadership in

intelligent processing for CVD management in PHC, led from Coimbra, Valencia and

Milan.

2 Methodology

LiNK project promotes the cooperation between the University of Coimbra, the Uni-

versitat Politècnica de València and the Politecnico di Milano in a long term sustainable

way. The main goal of the project is to link competences in intelligent processing for

CVD (cardiovascular diseases) management in PHC (Personal Health Care) with the

aim of creating a research ecosystem to address two central scientific and technical

challenges for PHC deployment: 1) Infusion of clinical evidence biomedical knowledge

in PHC solutions and 2) Moving PHC solutions from personal to personalized services,

i.e., services adapted to the specific user needs and characteristics. LiNK has estab-

lished an innovation forum involving key international players as well as key regional

players in the PHC area in order to identify gaps, industrial needs and best practices to

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

foster innovation culture in PHC. The main output of this forum is a research agenda to

support international research in the field, but also to guide the consortium in its re-

search strategy and approach, as well as its project proposals.

The LiNK project consists of three main phases (Fig. 1). The work was structured in

four major Work Packages and in the project lifetime the definition of the research

tracks is the major challenge and outcome, which is the focus of the present chapter.

Planning

- Advanced training

- Exchange programs

- Common research activities

- Shared data repository

...

WP1

Research &

Innovation

forum

Phase 1

WP2

Concepts

Phase 2 Phase 3

participation

Implementation

- Advanced training

- Exchange programs

- Common research activities

- Shared data repository

...

WP2

Research tracks

Month 6 Month 15 Month 36

Management, networking and dissemination

International

Societies

WP2 WP2

WP3

WP1

WP4

Fig. 1. LiNK phases and relationships between the Work Packages of the Project.

The research tracks are based on specific responses to selected specific research

questions addressed by Academia, Industry and Healthcare professionals. These re-

search questions are the result from the intersection of the inputs provided by the Re-

search and Innovation Forum (RIF in WP1 – Fig. 1), key competences inside the con-

sortium and shared research interests. This forum has supported the process of identi-

fying the gaps in intelligent processing for PHC, to spot the relevant internal and exter-

nal factors for innovation and high quality research in this field, and highlight innova-

tion barriers and innovation best practices. The main output of this forum is a research

agenda to support international research in the field, but also to guide the consortium

in its research strategy and approach as well as project proposals. This key forum has

also supported structuring curricula for advanced training in the area.

The first challenge was to create and select meaningful research questions to there-

after build inter-institutional teams for establishing the basement of the research tracks.

Such activities included team building activities, training programs for knowledge

transfer, researcher exchange and internships, as well as a set of PhD thesis and post-

docs programs. During this phase knowledge transfer was implemented as well as the

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

identification and setup of a common data repository and joint laboratories (activities

to be coordinated by WP4- Dissemination and training in Fig. 1).

One of the major outcomes of research track activity was the Concepts definition

(WP2 in Fig. 1), implemented in the third phase of the project, and managed in strait

cooperation with WP3-EU inclusion, networking and transference. Concept defini-

tion/development aimed at identifying the scientific, technical, social and economic re-

quirements of relevant project ideas. Such project concepts will be performed and

leaded by LiNK consortium and other institutions in forthcoming years.

To achieve the definition of the research tracks, three main contributions were con-

sidered: 1) Outcomes provided by the RIF – research and innovation forum; 2) Key

competences inside the consortium; and 3) Shared research interests.

The first contribution included a set of stakeholders, composed by experts from dif-

ferent countries (US, Italy, Spain, Portugal, Belgium, etc.) with different profiles (cli-

nician, decision maker, multinational company, researcher, entrepreneur, patients’ as-

sociations). During the first months of the project, a questionnaires and interviews were

performed in a Delphi study, helping the consortium to identify the top research ques-

tions and challenges with a global perspective.

Based on these questionnaires, it was possible to identify hot challenges to face from

the research perspective in the PHC for CVD field. These challenges were also priori-

tized on the basis of the following statements: i) Analysis of the state of the art and

future trends in pHealth solutions for integrated CVD management; ii) Identification of

open research problems and opportunities in algorithms for pHealth solutions for inte-

grated CVD management to support these concepts; and iii) Identification of opportu-

nities for Concept development in pHealth solutions for integrated CVD management.

The last two contributions (key competences inside the consortium and shared re-

search interests) were addressed through a state-of-the-art revision and a systematic

discussion inside the consortium. These discussions were focused around four main

perspectives: data sets, research interests/gaps, health outcomes and CVD applications.

2.1 Setup of the Research and Innovation Forum

Strategic Research Agenda definition should be a heuristic and unbiased process that

provides a privileged view of which are the most and least important topics for next

years. Even though a systematic review or a review of the later 10 years research direc-

tions could provide a picture of the current status and a brush on the future directions,

the aim of LiNK was to extract opinions from key stakeholders. Such key persons were

already identified and engaged through the Research and Innovation Forum (RIF), ac-

counting of high performance professionals riding and conducting clinical, research and

commercial actions in the field of CVD. To this extent, the methodology used to prior-

itize concepts was the consensus analysis throughout a Delphi interview study.

Among the decision making process, one of the most common methodologies used

in projects design and management is the Delphi technique. This approach has been

adopted in several fields from 90s and even though is a well-known world-wide used

technique, still little guidance exists to conduct a rigorous method of data collection.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

Nevertheless, the literature gathers many proposition and lessons learned papers on

which are the mandatory areas that should be tackled for preparing a Delphi survey.

Table 1. Areas for designing the Delphi survey.

Research problem

-Research Agenda for CVD

Rationale

-State of Art and future trends related to PHC

-Opportunities for integrated management

-Open research problems related to other concepts

Methodology

Data Collection

-Structured online form and email request with periodic re-

minders

Sample

-Research and Innovation Forum panel of experts

Data Analysis

-Number of responders and response rate.

-Statistical distributions for priorities and quantitative an-

swers

-Qualitative analysis of open answers

Delphi methodology is appropriate to investigate specific problems with careful con-

siderations and assumptions. The rationale for the study was categorized in three parts,

with a set of statements for each case:

Part 1. Analysis of the State of the Art (SoA) and future trends in pHealth solutions for

integrated CVD management, composed of 13 items (Table 2).

Table 2. Statements for the analysis of the State of the Art in future trends for CVD management.

Statement

Success of PHC for CVD is fully dependent on data availability

Physical robots will act as main catalysers for PHC services either at home or

at the hospital

New algorithms are needed to infer knowledge from data in order to practice

evidence based medicine

Quality of data is a dimension not to be missed when researchers are inferring

knowledge from heterogeneous databases

New data infrastructures, mostly cloud based, are needed to support PHC ser-

vices

Pervasive telemonitoring is key to manage PHC services in an efficient way

It is crucial to have real time telemonitoring to provide PHC services, being

possible thanks to mobile services

Feature selection and patient stratification are key elements for PHC

Ethical issues need to be addressed in order to guarantee a successful provi-

sion of PHC services

Legal issues need to be addressed in order to guarantee a successful provision

of PHC services

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

Clinical process management needs to be technology based to improve use of

clinical resources.

The creation of individualized behavioural models will allow personalised

medicine and patient empowerment

It is key to address and model the Patient context for a proper PHC manage-

ment.

Part 2. Identification of open research problems and opportunities in algorithms for

pHealth solutions for integrated CVD management to support these concepts, com-

posed of 11 elements (Table 3).

Table 3. Open research problems and opportunities.

Statement

How can different sources of information be integrated, especially on large

data-bases?

Based on data exploitation, how will new associations be discovered? How

will new therapeutic strategies be suggested?

How will mobile technologies improve continuous monitoring?

Which techniques will be used for pervasive monitoring?

Which indicators are the key ones for CVD risk assessment?

What are the best strategies to combine data and knowledge-driven learning

and modelling approaches?

What are the key stones in interpretable models for DSS?

How can the assessment of cardiovascular functions and status using wearable

technologies can be improved?

How can multi modal and multi scale data fusion for robust biosignal pro-

cessing be improved?

How to personalize models for diagnosis and prognosis?

How to support the integration, discovery and use of clinical knowledge in

daily clinical practice?

Part 3. Identification of opportunities for Concept development in pHealth solutions

for integrated CVD management (Table 4)

Table 4. Concept development opportunities in integrated CVD management.

Statement

Coronary artery disease (CAD) e.g. physiological models

Heart Failure (HF) – keeping patients diagnosed in a secure path

Sleep – e.g. sleep patterns and correlation with CVD

Stress

Diabetes

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

Experts were asked to prioritize the top five elements for each part in a 5 item Likert

Scale: Low Priority, Low to Medium Priority, Medium Priority, Medium to High Pri-

ority and High Priority. The questionnaire included open answer questions, in which

experts could add new statements from their own opinion and expertise, and the justi-

fication for the statement selected in each priority level.

2.2 Research Tracks Definition

The approach to accomplish the definition of the Research Tracks followed a strategy

based on three different inputs: 1) Experts discussions - where the key competences

inside the consortium and shared research interests were identified; 2) Analysis of the

state of the art and future trends in pHealth solutions for integrated CVD management;

and 3) Outcomes provided by the RIF – research and innovation forum.

The first two actions (internal discussions and state of the art) started during the kick-

off of the Project (Jan 2016), and have been maintained regularly during the progress

of the project through conference calls and face to face meetings. RIF priorities, aiming

to help the consortium in the identification of the top research questions and challenges,

were available at month 7 (July 2016). These inputs were used to complement as well

as to validate the definition of the Research Tracks.

Part 1. Internal Discussions. Aligned with the main goals of LiNK, four main per-

spectives have been considered in the discussions (Fig. 2).

Patient

Research Gaps

Personalization

Data processing

Algorithms

Data fusion (multi-parametric)

Fusion of models

Knowledge/interpretability

Sources

data

Clinical evidence

Datasets: EHR

Telemonitoring

Social networks

Bioinformatics

Genomics

Nutrition

Context

Health outcomes

Stratification

Diagnosis

Prevention

Prediction / early diagnosis

Simulation models

Treatment

Inputs from research forum

Key competences

Shared research interests

Research Tracks

CVD diseases

Fig. 2. Perspectives considered in the definition of the research tracks.

Available Datasets: The identification of possible sources of data, and related available

data sets, was considered by the consortium as a decisive requirement. These include

physiological signals, clinical evidence, bioinformatics, electronic health records, etc.

In fact, all the research, implementation and validation work to be carried out depends

decisively of the existence of useful data sets.

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

Research Interests/Gaps: Moreover, the research tracks should be aligned with the cur-

rent research gaps, recognized as pertinent and relevant by the scientific community, as

well as in line with the research interests of each partner. Examples of such gaps are:

personalized algorithms, predictive decision tools and multi-modal and multi-scale

models.

Health Outcomes: Another relevant perspective is oriented towards the specific prob-

lems to be addressed. In other works, the clinical relevance of each RT should be clearly

defined. Examples of such health outcomes are diagnosis, prognosis, and stratification.

CVD applications: Finally, between the several possible CVD applications, the specific

domain of application has to be identified. Examples of such CVD applications are

related with cardiovascular status, heart failure, and sleep comorbidities.

In a first phase these four perspectives were discussed independently. Then, in a

second phase, the different perspectives were combined in a multi-input matrix (Figure

3), enabling the systematization of the different perspectives, and the identification of

the RTs and a set of associated activities. Basically, the process was guided by the fol-

lowing principle: “Each RTs addresses a particular health outcome (3), and is composed

of several activities, each one addressing a specific research gap (2), for a particular

CVD application (4.), using a suitable data set (1)”.

06/08/2016

Sleep

HFCAD Diabetes

1. Data sets

4. CVD Applications

Health

outcomes

Research

Interests

Fig. 3. Matrix used in the RTs definition combining the different perspectives: sources of data,

health outcomes, research interests and CVD applications.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

3 Results

3.1 Outcomes from the Research and Innovation Forum

24 members of the Research and Innovation Forum (RIF) participated in the Delphi

study during the month of July 2016. Participation into the RIF was approved upon

application acceptance (Table 5)

Table 5. Concept development opportunities in integrated CVD management.

Academic Profile

Eng

PhD

M Sc

Professor

Type of Entity

Corporate

Research Institute

Hospital

Scientific Society

Patient Associations

SME

Policy Maker

University

Area/Continent

Central Europe

North Europe

East Europe

South Europe

North America

Part 1. Analysis of the State of the Art (SoA) and future trends in pHealth solutions for

integrated CVD management. Two analyses were done for the inputs provided by ex-

perts dealing with the state of the art, considering the priorities they gave (allocating

from 5 to 1 points to the different options) or just considering equal the 5 top priorities

(allocating 1 point to all the priorities selected by each expert) as some experts ex-

pressed that for them, there were a set of 4-5 statements with the highest top priority.

Based on answers from experts, it can be inferred especially from Error! Reference

source not found. that there are three big blocks of statements: Top priority statements

(between 37 and 16 points); Middle priority statements (between 11 to 7 points) and

Low priority statements (between 4 and 3 points)

Table 6. Top priority statements in future trends of pHealth Solutions for CVD Management.

Statement

Priority

Points

New algorithms are needed to infer knowledge from data in order

to practice evidence-based medicine

Quality of data is a dimension not to be missed when researchers

are inferring knowledge from heterogeneous databases

Success of PHC for CVD is fully dependent on data availability

It is crucial to have real time telemonitoring to provide PHC ser-

vices, being possible thanks to mobile services

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

To avoid biasing due to high scores given by a single expert, the analysis was also

done considering just 1 point for all the selections of each expert (Fig. 4)

Fig. 4. Statement selection in the priorities of State of the Art.

Part 2. Identification of open research problems and opportunities in algorithms for

pHealth solutions for integrated CVD management to support these concepts. Two

analyses have been done for the inputs provided by experts dealing with the opportu-

nities for development of new concepts around integrated CVD management, consid-

ering the priorities they gave (allocating from 5 to 1 points to the different options) or

just considering equal the 5 top priorities (allocating 1 point to all the priorities se-

lected by each expert) as some experts expressed that for them, there were a set of 4-5

statements with the highest top priority. Based on answers from experts, it can be in-

ferred three big blocks of statements: Top priority statements (between 26 and 19

points); Middle priority statements (between 16 to 10 points) and Low priority state-

ments (between 6 and 4 points).

Being the top priority statements concerning the new research questions to be ad-

dressed for PHC in CVD management the ones listed in Table 7.

Table 7. Top priority research problems and opportunities in algorithms for pHealth Solutions

for CVD Management.

Statement

Priority

Points

How to support the integration, discovery and use of clinical

knowledge in daily clinical practice?

How to personalize models for diagnosis and prognosis?

Which indicators are the key ones for CVD risk assessment?

What are the best strategies to combine data and knowledge driven

learning and modelling approaches?

On the other side, if we just consider the amount of experts that selected each state-

ment as it is reflected in Figure 5, the groups of research questions are almost the same

with the difference of statement 6, which moved to middle priorities.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

Fig. 5. Statement selection in the priorities of research problems and opportunities.

Part 3. Identification of opportunities for Concept development in pHealth solutions

for integrated CVD management. Last part of the survey was focused on the develop-

ment of concepts that engage transversally different research questions and problems

previously raised with a direct focus on the society. In this case, as there were five

themes, the analysis has been performed on the priority levels assigned by each ex-

pert. Figure 5 show graphically the results where Heart Failure (HF), namely, to keep

diagnosed patients in a safe track and Coronary Artery Diseases (CAD) are the top

opportunities identified. That is logical as they are strictly based on heart problems

whereas the rest of options address the issue from a comorbidity point of view. At this

stage, all experts prioritize diabetes versus stress and sleep.

Fig. 6. Opportunities for concept development in pHealth solutions for integrated CVD manage-

ment.

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

3.2 Definition of the Research Tracks for CVD in Personalized Healthcare

RT1 - Stratification and Diagnosis. The first research track proposed to design and

develop models to assess CVD patients in three categories: CVD Status, Cardiorespir-

atory Assessment and Cycling Alternating Pattern (CAP) during sleep. Given that (i)

CVDs are the most prevalent chronic diseases, (ii) which have a very significant impact

both in patient’s quality of life as well as in the financial sustainability of health provi-

sion systems, it is observed that research and development of new methodologies capa-

ble of a continuous and non-invasive assessment of the cardiovascular status of patients

is of crucial importance to implement effective personal health care systems for ade-

quate and preventive cardiovascular diseases (CVD) management. Being blood pres-

sure (BP) a critical risk factor of CVD onset and progression, it is seen that the contin-

uous and cuff-less monitoring of blood pressure is one critical research topic that has

been intensively pursued by researchers during the last decade. Continuous hemody-

namic assessment using BP, is still one of the major open research questions for CVD

status assessment.

RT2 – Personalized Predictions. Cardiovascular diseases are a major public health

concern, and a cause of considerable morbidity and mortality. As a consequence, the

prediction of severe events is of great importance for professionals, since it provides

the adequate tools to diagnose. The main goal of this work is the development of algo-

rithms the early detection of critical events, using multi-parametric approaches that

combine physiological measurements (e.g. ECG, blood pressure, weight) and other

sources of information (e.g. medication).

Two main scientific challenges will be addressed: prediction methods and infor-

mation fusion schemes, based on computational intelligence methodologies. The main

hypothesis is that physiological time series with similar progression have prognostic

value in future clinical states. Possible applications are the prediction of hypertension

or decompensation episodes for heart failure patients.

RT3 – Integrated Care and Process Mining. The application of process mining tech-

niques is highly influenced by the quality of the event logs that are used to discover

models. Since this field of research is relatively new, there are no standardized ways to

create process logs. Additionally, some of the logs used come from direct observation

and manual annotation (especially in healthcare), what can introduce errors in the data

which invalidates the models inferred. Therefore, some efforts are required to standard-

ize the way data is collected and pre-processed to avoid bias in process mining algo-

rithms outcomes. The aim of this research track is the creation of mechanisms that allow

to collect, share and evaluate the data available from patients to ensure the creation of

proper models using process mining techniques.

Currently, there are a big amount of available data coming from very different and

heterogeneous sources and with different qualities. The homogenization of these data

is crucial for the creation of accurate and reliable models. For any data analysis tech-

nique, the quality of the underlying data is important. Otherwise, the data processing

will risk drawing the wrong conclusions.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project

4 Conclusions

One of the main goals of the LiNK project was to identify with a very solid perspective

what are the right questions and challenges to be solved from a research perspective, to

therefore define the current opportunities in the form of research tracks.

To engage representatives from the different stakeholders in the PHC for CVD field,

that will help us to prioritize the key hot topics. Through the creation of a preliminary

research agenda that will enhance S&T excellence of LINK institutions, but also for the

entire community in systems for pHealth solutions for CVD management.

The RIF’s community has been created with 24 participants from Europe and Amer-

ica, with different backgrounds and types of entities. RIF members have actively par-

ticipated in the discussion, prioritising and providing comments about the suggested

statements.

Therefore, based on existing literature and an analysis a set of topics have been iden-

tified, helping us to draft the preliminary research agenda. Concerning the State of the

Art (SoA) and future trends in pHealth solutions for integrated CVD management, the

top priorities identified by experts are dealing with new algorithms to infer knowledge

from data, quality of data when data is coming from heterogeneous sources and the

need for real time telemonitoring for PHC services using mobile technologies. Being

these ones the top ones, there are also a set of trends that have been raised at least by

half of the experts about data availability, robots, pervasive telemonitoring, legal issues,

ICT based clinical management and creation of individualized behavioural models.

About the identification of the open research questions for phealth solutions for in-

tegrated CVD management to support these concepts, the three identified main ques-

tions to be addressed were: How to support the integration, discovery and use of clinical

knowledge in daily clinical practice? ; How to personalize models for diagnosis and

prognosis? And , which indicators are the key ones for CVD risk assessment?

Whereas being not top priority, experts recommend not to forget issues as the inte-

gration of different information sources on large databases, the assessment of cardio-

vascular functions and status using wearable technologies, the improvement of multi-

modal and multi-scale data fusion for robust biosignal processing be improved, the con-

tinuous monitoring through mobile technologies and the selection of strategies to com-

bine data and knowledge driven learning and modelling approaches?

Finally, concerning the identification of opportunities for concept development in

phealth solutions for integrated CVD management, Heart Failure (HF), that is, to keep

diagnosed patients in a safe track and Coronary Artery Diseases (CAD) are the top

opportunities identified. That is logical as they are strictly based on heart problems

whereas the rest of options address the issue from a comorbidity point of view. At this

stage, all experts prioritize also diabetes versus stress and sleep.

Research tracks were defined on the basis of RIF’s priority challenges, key compe-

tences inside the LiNK consortium ad shared research interests. As result of this strat-

egy, three main research tracks have been identified: RT1 – Stratification and diagnosis;

RT2 - Personalized predictions and RT3 - Integrated care and process mining.

Additionally, each one of these RT is composed of several activities, each one ad-

dressing a specific research gap, for a particular CVD application, using a specific data

EPS Portugal 2017/2018 2017 - OPPORTUNITIES AND CHALLENGES for European Projects

set. Basically, the first research track, stratification and diagnosis, addresses the devel-

opment of intelligent algorithms and models for stratification and diagnosis, providing

an effective support to the decision making of professionals. The major data analysis

functionalities include advanced tools for the personalization, integration, analysis and

interpretation of the different sources of clinical information. The second research tack,

personalized predictions, is linked with the concept of clinical prognosis, aiming to

predict the future of a specific variable, a clinical condition or, ultimately, the health

status of a patient. Moreover, predictive methods can be applied to help the planning

and adjusting of treatments, by forecasting the effectiveness of the therapy in the future.

Finally, the third research track, Integrated care and process mining, focus on the de-

velopment of algorithms and tools to facilitate and support the application of the newest

clinical evidence to clinical daily practice, mainly through interactive pattern recogni-

tion techniques.

The lack of patient-specific predictive algorithms with the capacity of estimating the

evolution of the disease and, therefore, to pave the way for optimal, patient specific and

coordinated treatment actions are major gaps to increase PHC solutions’ robustness and

acceptance by professionals and patients. The infusion of clinical evidence and existing

biomedical knowledge while addressing these gaps will play a decisive role in improv-

ing accuracy, quality, personalization and, ultimately, acceptability by patients and pro-

fessionals.

LiNK project has contributed to close these gaps by defining and implementing re-

search tracks involving some of these key issues or others that might be identified inside

the RIF. Research questions identified in these research tracks shall be used as contexts

for shared advanced training such as in PhD thesis development or post-doc programs.

5 Funding Statement

This work was supported by LINK Project, a H2020 Project funded by the European

Commission (GA ref: 692023)

References

1. Steven Allender, Peter Scarborough, Viv Peto and Mike Rayner, European Heart Disease

Statistics, 2008 Ed., Health Economics Research Centre, Department of Public Health, Uni-

versity of Oxford

2. Cristiano Codagnone, Reconstructing the Whole: Present and Future of Personal Health Sys-

tems, PHS 2020, 2009.

Research Tracks for Intelligent Processing to Support Cardiovascular Disease Management in Personalized Healthcare - Results from LiNK

Project