A Semantic Web based Framework for Linking Healthcare Information with Computational Physiology Models
- 1. A Semantic Web based Framework for Linking Healthcare Information with Computational Physiology Models Koray Atalag, Reza Kalbasi, David Nickerson The University of Auckland Koray Atalag MD, PhD, FACHI k.atalag@auckland.ac.nz Senior Research Fellow, ABI Management Board Member, openEHR Foundation Chief Information Officer, The Clinician
- 2. Outline • Problem / Research Question • Intro to Computational Physiology • Linking both Domains • Semantic Web and Ontology Mapping • Results • Discussion
- 3. Problem / Research Question • Computational models have great potential to improve healthcare • Lots of real-world data information stored in EHRs – however discovery and reuse is poor • Linking the two domains requires shared semantics – Ontology/Terminology used in CP are different from Healthcare – Semantic Web has little use in healthcare IT/informatics Can we map knowledge sources from both domains to enable discovery and reuse of clinical data and computational models? Ontology mapping is hard!
- 4. What’s Computational Physiology Governing equations (laws of physics) Anatomy & structure High-performance computing Software Material properties from measurement Observed function Validation Predicted function Mechanistic insight
- 5. Tissue Osteon NephronAcinus Liver lobuleLymph nodeCardiac sheets Organ Heart Lungs Diaphragm Colon EyeKnee Liver Environment Organ system Organism Cell Protein Gene Atom Network
- 6. Human Physiome Project & Virtual Physiological Human (VPH) a methodological and technological framework that will enable collaborative investigation of the human body as a single complex system Descriptive, Integrative And Predictive
- 7. Computational Physiology of the Heart Workflow: cardiac imaging to patient-care Image Segmentation OpenCMISS-Zinc, cmgui, VTK, ITK, CIM, Matlab, 3D Slicer Biomechanics Simulation End-SystoleDiastasis Pressure, contractile force OpenCMISS, Continuity, FEBio Ansys, ABAQUS Parameter ID/Calibration rffrcffc crrrccff EEEEC EEECECQ QexpCW 3 222 3 2 2 1 2 2where 11 Caa TT OpenCMISS, Matlab, Python, Prediction Stress (kPa) Fibre strain OpenCMISS, Continuity, Chaste, Abaqus Diseased Normal Subject-specific modelling OpenCMISS-Zinc, CIM, CAP
- 8. Photo by Jerry Bunkers Diagnosis Measurement Computational Physiology of the Breast Breast image analysis workflow
- 9. Why Linked Health Data?
- 10. Computational Modelling: CellML • CellML includes information about: – Model structure (how the parts of a model are organizationally related to one another); – Mathematics (equations describing the underlying biological processes); – Metadata (additional information about the model that allows scientists to search for specific models or model components in a database or other repository). • CellML includes mathematics and metadata by leveraging existing XML-based languages, such as Content MathML, XML Linking Language (XLink), and Resource Description Framework (RDF). (C. M. Lloyd, M. D. B. Halstead, and P. F. Nielsen, "CellML: its future, present and past" Progress in Biophysics & Molecular Biology, vol. 85, pp. 433-450, June-July 2004)
- 11. (www.cellml.org) Cuellar AA, Lloyd CM, Nielsen PF, Halstead MDB, Bullivant DP, Nickerson DP, Hunter PJ. An overview of CellML 1.1, a biological model description language.SIMULATION: Transactions of the Society for Modeling and Simulation, 79(12):740-747, 2003 Physiome Standards and Tooling
- 12. Semantic web • The meaning of data, or semantics, is the target of semantic web • Subject - Predicate - Object • W3C languages: RDF/RDFS/OWL etc. • Machine processable Web • More enhanced search results, meaning-based data integration, reasoning services • Healthcare semantic annotations • Computational physiology models with embeded semantic annotations Semantic web components defined by (Hebeler et al. 2009)
- 13. Ontologies • Ontologies are formal descriptions of knowledge enabling sharing and reuse. • Ontologies provide: – Classing mechanisms (multiple inheritance, subclassing, domain, range, …); – Class expressions (union, intersection, complement, …); – Class axioms (one of, disjoint with, equivalence, …); – Property characteristics (transitive, symmetric, functional, …); – Cardinality (minimum, maximum, exactly); – Rich set of primitive data types (string, boolean, integer, real, datetime, URI, …); – Management (imports, versions, compatibility, …).
- 14. Computational Models and Ontologies • We are developing ontologies for physiological form and function, and the CellML modelling language. • Other groups are also developing ontologies relevant to biological modelling: – Anatomical (Gene Ontology/GONG, FMA); – Gene regulation pathway (Gene Ontology/GONG, BioPax); – Gene expression (Gene Ontology/GONG); – Common access to bioinformatics sources (TAMBIS/TaO); – Physical and mathematical (SBO, Stanford Knowledge Systems, OPB). • We are linking model entities in the CellML repository to our own and other ontologies.
- 16. List of Ontologies in our OLS • OPB • FMA • CHEBI • Gene Ontology (GO) • Cell Ontology (CL) • Phenotypic quality (PATO) • Protein Ontology (PR) • PRIDE • EDAM • EFO • PROBONTO • OBO relations Ontology • SNOMED CT • LOINC • ICD (TODO)
- 17. • Open access specs & tooling for representing healthcare data, enabling interoperability and building EHR • Supports very elaborate DCM development (=Archetypes) • Scope is full EHR - not just health information exchange • Not-for-profit organisation - established in 2001 • Based on 20+ years of international research and practice • Also an ISO/CEN standard (ISO 13606) • Big international community • All DCMs are available from: http://openehr.org/ckm www.openehr.org
- 19. Semantics in openEHR • Whole-of-model meta-data: – Description, concept references (terminology/ontology), purpose, use, misuse, provenance, translations • Item level semantics (Schema level) – Trees/Clusters (Structure) – Leaf nodes (Data Elements) Formally: different types of terminology bindings: 1) linking an item to external terminology/ontology for the purpose of defining its real-world clinical/biological meaning 2) Linking data element values to external terminology (e.g. a RefSet or terminology query) AlsoInstance level semantic annotations – applies to actual data collected
- 20. mindmap representation of openEHR Archetype 1) Linking items to SNOMED to define clinical meaning
- 21. 2) Linking data element values to external terminology
- 22. openEHR Lab Test Archetype - SNOMED & LOINC encoded
- 23. Realising the Physiome / VPH
- 24. Ontology mapping • Ontology; formal specification of a domain knowledge • Ontology mapping; the definition of corresponding objects (entities) from one domain ontology to the other domain ontology • Ontology matching may appear to be virtually impossible • Solution: semi-automated collaborative ontology mapping via user interaction (Web 2.0)
- 25. Mappings: Biophysical ⇐⇒ Clinical Concepts
- 26. CellML Model > Clinical Data Discovery
- 27. PhysioMedApp
- 28. Acknowledgements David NickersonReza Kalbasi Peter Hunter Tommy Yu