160 likes | 317 Views
Documenting the semantics of medical data. Adela Jarolimkova, Petr Lesny, Krystof Slaby, Helena Bouzkova, Jan Vejvalka. MediGrid project. 2005 - 2009 Partners CESNET (Czech national research and education network operator) Faculty Hospital Motol - Prague Masaryk Hospital - Ustí nad Labem
E N D
Documenting the semantics of medical data Adela Jarolimkova, Petr Lesny, Krystof Slaby, Helena Bouzkova, Jan Vejvalka 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
MediGrid project • 2005 - 2009 • Partners • CESNET (Czech national research and education network operator) • Faculty Hospital Motol - Prague • Masaryk Hospital - Ustí nad Labem • Main goal • to propose and test MediGrid – a modular application system for distributed processing of health-related data based on Grid networkas the integrating technology 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
MediGrid structure - data and indicators • Medical data are described as indicators – to underline their subjective nature • Why not data records? • the term data record is quite overloaded while we need a term with a clearly defined meaning • in the term indicator the subject taking the record is also reflected • Transformations • Indicator classes 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
MediGrid structure - modules • Module • Automated tool for indicators tranformation • Relation among indicator classes • Each module implements one relation – algorithm, expert consultation, conference of experts • MediGrid structure • Modules • Sorter • Documentation service 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Description of data and their relations – traditional systems • Terminological and classification systems (MeSH, ICD etc.) • Hierarchical structure • Relations – taxonomic, synonymic, meronymic • Main problem – limited number of relation types, insufficient for describing complex relationships in algorithmic medicine 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Description of data and their relations - ontologies • Ontologies • „explicit specification of conceptualization“ • Medical ontologies • „old“ systems – SNOMED, UMLS • Cover the whole field of medicine, but contain many inconsistencies and inaccuracies as described in literature • „new“ systems – OpenGalen, On9 • Maintain formal ontology rules, but structured from the top, whereas concepts belonging to algorithmic medicine are typically very „low-level“ 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Description of data and their relations - example • Description of BMI module using UMLS MetaThesaurus and Semantic Network • Body Mass Index calculation • Inputs – body weight, body height • Output – body mass index • Corresponding Metathesaurus Concepts • Body mass index calculation – no such concept • Body weight - cui C0005910 • Body height - cui C0005890 • Body mass index - cui C1305855 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Description of data and their relations – example cont. • Corresponding semantic types • Organism attribute • Clinical attribute • Quantitative concept • Relations between semantic types • Is_a • Degree_of • Associated_with • X body_mass_index_calculation • X is_input_of, is_output_of 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
MediGrid Documentation Service • Tasks • to address the problem of unambiguous description of data and algorithms • to prove evidence / credibility for modules • to serve as a mean of scientific communication between individual specialists • to serve as a network for sharing unique knowledge 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
DS metadata • Basic entity categories • Modules • Indicator classes • Citations • Concepts (primarily UMLS, other controlled vocabularies are allowed)/ module title, author code, structured description with relevant citations • Kept in a machine-readable (database) format • Presented to users in application-specific views 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
external resources PubMed modules documentation 1:11 fulltext citations indicator classes shared comments 1:1 DS scheme documentation service catalogue 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
DS ontology • „ad-hoc“ ontology generated from the machine-readable documentation • Contains only entities(indicator classes and modules) currently existing and documented and their relationships • Changes every time a new knowledge element (algorithm) is implemented 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Citations in DS • used to confirm the value and credibility (evidence-base) of implemented algorithms • Application for citation handling similar to simple reference manager • allows to save a record of any published or unpublished information, both manually and by acquiring the record from an external database • to insert the citation into documentation • to interconnect the citation with some external resource 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Citation record format • Based on XML • Allows citing any published or unpublished information in any form (print or electronic) • Supports evaluation of quality and relevance • Conforms with Vancouver requirements • Allows arbitrary extending to meet the needs of DS • Compatible especially with PubMed 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Conclusion • Documentation service in this proposed form is now being implemented within the framework of the MediGrid project • Its functionality, especially the use of ad-hoc generated ontologies for processing of health-related data, will then be tested – as a novel approach to unambiguous representation of medical knowledge, readily available for exploitation in computerized information systems. 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006
Thank you for your attention Any questions? Contacts: http://www.medigrid.cz ajarolimkova@seznam.cz petr.lesny@lfmotol.cuni.cz 10th European Conference of Medical and Health Libraries, Cluj-Napoca, 11th - 16th September 2006