AEGLE: a reference big data architecture for the healthcare sector

Transcription

1 AEGLE: a reference big data architecture for the healthcare sector jos.dumortier@timelex.eu Amsterdam, 7 June 2016

2 Introduction Horizon 2020 Innovation Action Partners from Belgium, France, Greece, Italy, Netherlands, Portugal, Sweden & UK Started in March 2015 (42 months) Tuesday, 7 June

3 Tuesday, 7 June

4 Objective Reference big-data architecture Covering a large part of the health spectrum Malignant chronic diseases Non-malignant chronic diseases Acute care jos.dumortier@timelex.eu Tuesday, 7 June

5 Access control rules Common representation of data Analytic results to decision makers AEGLE Inventory of available data Syntax structure and semantics Integrated data analysis Informed based decisions AEGLE Principles Alignment with the data value chain Users Stakeholders Providers Hospitals Specialists Patients Researchers DII ICU CLL collect prepare organize integrate analyze visualize decide Product Oriented Research Oriented Tuesday, 7 June

6 Users Challenges Understand the exploitation perspective of real-life big bioclinical data for diverse use cases Provide a framework to accommodate diverse big bioclinical data management and analytics requirements according to the data type and the application domain Enable scientific question answering by exploiting big bioclinical data in a way not possible until now Provide an adaptable and user-friendly working end-user environment Rely on clear and comprehensive end-user agreements for service provision jos.dumortier@timelex.eu Tuesday, 7 June

7 Technical Challenges Address the requirements posed by different types of big bioclinical data: Biomolecular and clinical data (the CLL use case) Real-time streaming and clinical events data (the ICU use case) Large observational healthcare databases (the DII use case) Implement the necessary mechanisms for comprehensive data management and analytics, fully-compliant with privacy, legal and ethical norms (e.g. anonymisation, policies, etc.) Provide efficient response time (when needed) via acceleration technologies Offer a scalable & sustainable IT solution jos.dumortier@timelex.eu Tuesday, 7 June

8 Business Challenges Create sustainable business models taking into account the needs of the three scenarios to which AEGLE is applied Accurately identify the exploitable items Define a strategy for the long-term viability of the platform Create an ecosystem of stakeholders jos.dumortier@timelex.eu Tuesday, 7 June

9 AEGLE Cases Tuesday, 7 June

10 Intensive Care Unit (ICU) Challenges Mechanical ventilation & patient-ventilator interactions Personalization of Patient Care & Early Identification of Deterioration Questions to be answered Ineffective Efforts (IE) characteristics Recognition, Incidence, Significance, Prediction of IE Tools to guide and monitor nutrition Tools to identify early deteriorating trends Tuesday, 7 June

11 Chronic Lymphocytic Leukemia (CLL) Challenges Clinically and biologically heterogeneous Optimal care and treatment decisions depend on the integration of tumor- and host-derived variables Questions to be answered Identification of novel prognostic markers Prediction model for Monoclonal B Lymphocytosis evolution Prediction model for CLL natural course jos.dumortier@timelex.eu Tuesday, 7 June

12 Type 2 Diabetes (T2DM) Challenges Long term condition Increasing in prevelance Increasing in morbidity and mortality Improving disease management Questions to be answered Define why some cases do better than others prognostic indicators Improve patient outcomes methods of intervention Define accurate cohort and feasibility for doing clinical trial Identify potential points for intervention and types of intervention available jos.dumortier@timelex.eu Tuesday, 7 June

13 Where we stand now? User-centered design approach for developing a big bioclinical data analytics platform First release of the AEGLE system architecture Rapid prototyping for proof-of-concept illustration and user engagement: data, analytics and data management mechanisms already in place Beginning of first validation phase Initial legal and ethical assessment First steps taken on the business landscape for AEGLE jos.dumortier@timelex.eu Tuesday, 7 June

14 SQOOP data transfer to HDFS AEGLE Big Data Framework Software Stack WebHDFS HDFS REST API RM RT YARN REST API LIVY SPARK REST API HIVE hadoop sql api PIG scripting workflow mgnt Pydoop Pyhton hadoop api SPARK SQL sql api for spark MLlib machine learning library HADOOP MAPREDUCE distributed engine for batch jobs processing SPARK distributed engine for fast in-memory processing YARN virtual cluster resource manager HDFS2 virtual cluster distributed file system VM Node VM Node VM Node VM Node VM Node VM Node VM Node Tuesday, 7 June

15 Big Data Framework Features Dockerized Hadoop/SPARK cluster Dockerized MySQL cluster REST APIs for web based integration Platform independent automatic deployment Current deployment: 1 Master node, 2 SQL nodes, 4 Data ~okeanos 8 node Hadoop/SPARK ~okeanos jos.dumortier@timelex.eu Tuesday, 7 June

16 Legal Challenges Identify and incorporate all ethical and regulatory issues underlying the realization of the project aims Contribute in the definition of a common ethical and regulatory framework for big bioclinical data management and analytics at the European level jos.dumortier@timelex.eu Tuesday, 7 June

17 Two legal perspectives 1. Short term: legal compliance of research & innovation activities in the scope of the AEGLE action (all phases of the data value chain) 2. Long term: legal framework for the AEGLE start-up (and for other European big data initiatives in the health sector) 17

18 1. Short term Focus on compliance of AEGLE RIA with current data protection law ( avoid doing something illegal ) Core provision (Directive 95/46, article 6.1(b)): Further processing of data for historical, statistical or scientific purposes shall not be considered as incompatible provided that Member States provide appropriate safeguards AEGLE: further processing of clinical data for scientific purposes Main question to examine: which appropriate safeguards of (which) Member States have to be taken into account? (example: need of approval of DPA if re-use is not based on patient consent) 18

19 2. Longer term: EU legal framework for big data processing in the health sector Context: AEGLE start-up Initial legal framework: General Data Protection Regulation Core provision: art. 6.1(b) juncto art. 83 Objective: analyse the (current/developing) legal situation in the 28 Member States + recommend possible EU initiatives 19

20 EU Data Protection Regulation 2016/1679 Art. 6.1(b) : [Personal data must be] collected for specified, explicit and legitimate purposes and not further processed in a way incompatible with those purposes; further processing of personal data for archiving purposes in the public interest, or scientific and historical research purposes or statistical purposes shall, in accordance with Article 83(1), not be considered incompatible with the initial purposes; Art : Where personal data are processed for scientific, statistical or historical purposes Union or Member State law may, subject to appropriate safeguards for the rights and freedoms of the data subject, provide for derogations from Articles 14a(1) and (2), 15, 16, 17, 17a, 17b, 18 and 19, insofar as such derogation is necessary for the fulfilment of the specific purposes (...) Art : The appropriate safeguards referred to in paragraphs 1 and 1a shall be laid down in Union or Member State law and be such to ensure that technological and/or organisational protection measures pursuant to this Regulation are applied to the personal data ( ), to minimise the processing of personal data in pursuance of the proportionality and necessity principles, such as pseudonymising the data, unless those measures prevent achieving the purpose of the processing and such purpose cannot be otherwise fulfilled within reasonable means. 20

21 Co-funded by the Horizon 2020 Framework Programme of the European Union under Grant Agreement nº Partners EXUS AE (Coordinator), ICCS, KINGSTON, CERTH, Maxeler Tecnologies Limited, UPPSALA UNIVERSITET, UNISR, Time.Lex, EUR, CHS, LOBA, PAGNI, GNUBILA FRANCE