WORKSHOP

AMIA 2005

Ontological Foundations of Biomedical Terminology Systems

organized by AMIA-WG on Formal Biomedical Knowledge Representation

 

Contributors:

Olivier Bodenreider^o, Anita Burgun^b, Werner Ceusters^c, James Cimino^d, James Davis^e, Peter Elkin^f, C. Greg Hagerty^q, Vasant Honavar^r, Ira Kalet^g, Anand Kumar^a, Timothy B. Patrick^s, Alan Rector^h, Jim Riceⁱ, Jeremy Rogers^h, Stefan Schulz^j, Kent Spackman^l, Davide Zaccagini^m, Pierre Zweigenbaumⁿ, Barry Smith^a,k

 

^aIFOMIS, University of Saarland, Saarbruecken, Germany

^bLIM, University of Rennes, France

^cECOR, University of Saarland, Saarbruecken, Germany

^dDepartment of Biomedical Informatics, Columbia Univ. College of Physicians & Surgeons, NY, USA

^eDepartment of Computer Science and Engineering, University of South Carolina, Columbia, SC, USA

^fMayo Clinic College of Medicine, Rochester, MN, USA

^gRadiation Oncology Department, University of Washington, WA, USA

^hDepartment of Computer Science at University of Manchester, UK

ⁱMuskegon Surgical Associates, MI, USA

^jDepartment of Medical Informatics, Freiburg University Hospital, Germany

^kDepartment of Philosophy, SUNY at Buffalo, NY, USA

^lOregon Health & Science University at Portland, OR, USA

^mDepartment of Health S&T, Harvard-MIT, MA, USA

ⁿSIM, Paris Hospital & Department of Bioinformatics, University of Paris, France

^oNational Library of Medicine, Bethesda, USA

^pAcademic Medical Center, Amsterdam, The Netherlands

^qUniversity of Medicine and Dentistry of New Jersey

^rArtificial Intelligence Research Laboratory, Iowa State University

^sDepartment of Health Management and Informatics, School of Medicine,University of Missouri-Columbia

 

 

Abstract:

 

In recent years there have been frequent discussions about theoretical foundations and best practice recommendations for biomedical ontology/ terminology building. However there are disparate views amongst experts in the field. This workshop aims at continuing prior e-mail and face-to-face discussions on this topic at the level of the AMIA Knowledge Representation Working Group (KR-WG). It will focus on controversial issues such as concepts vs. reality, single vs. multiple inheritance, "natural" vs. "housekeeping" classes, primitive vs. defined classes, partitions, disjointness, exhaustiveness etc.

 

Biomedical Terminology construction and maintenance is one of the most prominent fields in Medical  Informatics and has a major impact on medical documentation and health services management. Best practices in terminology building have been frequently discussed in recent years. However, there is a range of seemingly disparate views represented by experts in the field. Most biomedical terminologies are built for specific purposes and reflect the need for consensus among medical experts in a specific domain. This has proven insufficient for computer-mediated terminology services which cross domains and application contexts, especially if formal reasoning is to be supported.

The claim for foundational terminology systems which convey principled descriptions of the basic entities of biomedical science is commonly associated with the term ontology, i.e. the representation of basic notions of reality (upper ontology) and specific entities of a given domain (domain ontology) in a formally founded way. Here there are also divergences between the AI approach in which a given ontology should, above all, support concrete reasoning tasks, and the philosophers' approach which is committed to represent the unchanging features of nature, unbiased by concrete application needs.

This contrasts, again, which the linguistic approach which characterizes the building of many (informal) domain terminologies. Here, instead of making absolute assertions about the world, emphasis is laid on clarifying, defining and standardizing the meaning of domain specific terms and organizing them in controlled vocabularies or thesauri. This practice from library and information science represents a long-standing tradition in medical informatics.

The clash between different schools of thinking and different methodologies in the way of describing the domain of biomedicine, in tandem with the emergence of a new generation of large biomedical terminologies (GO, SNOMED) has animated and stimulated numerous discussions and research activities and has also contributed to a wide opening of the medical informatics community to other scientific disciplines.

This move towards a true multidisciplinarity conveys specific challenges. A major problem is to find an agreement on common terms in order to avoid communication errors. Another challenge is to cope with the existence of divergent schools of thinking which - in the case of philosophy - are deeply rooted in history. A symptomatic example for this is the difficulty to find not only commonly agreed words but even commonly agreed notions for what constitutes the very essence of biomedical terminologies, viz. those things terms refer to. The assortment of names is impressive: "concept", "class", "entity", "category", "universal", "kind", "type", "sort", "node", "property" etc. Each of these names reflects different traditions and ways of thinking about closely related, albeit not equivalent notions.

As an AMIA working group committed to promote research and development in medical terminologies, ontologies and knowledge representations we are in a mediating position. On the one hand we take very seriously the need to clarify the foundational issues. On the other hand we must not lose sight of what is relevant for concrete tasks of terminology / ontology building.

 

The proposed workshop is a continuation of intensive email and face-to-face discussions among AMIA’s Knowledge Representation Working Group (KR-WG) members regarding the best practices for ontology building in biomedicine. This work has already been documented in a draft paper.

 

In this context the following issues have been explicitly addressed and will be prepared for the workshop:

 

·        Do different fundamental approaches to ontology lead to substantial differences in domain terminologies/ontologies or does this have only marginal implications ?

·        Are ontologies about concepts or entities in reality? Is this really relevant for practical ontology/terminology engineering ?

·        Which are the basic divergences in concurrent schools of philosophical thinking (realism, relativism, nominalism) and their implications on concrete tasks of ontology engineering in the biomedical domain ?

·        Should the design of formal ontologies be guided by concrete recommendations, e.g.,   single inheritance, avoidance of negative classes, pairwise disjoint siblings, jointly exhaustive siblings? 

·        How to distinguish between "natural" and housekeeping classes in biomedical ontologies?

·        How to deal with different partitions in ontologies, for creation,
maintenance, as well as for use for reasoning?

·        How to deal with primitive vs. defined concepts, and the importance of definitions in ontologies.

·        How to come up with best practice recommendations for biomedical ontologies, which is also a challenge for education in medical informatics

 

The workshop will focus on short, topical presentations with enough space for subsequent discussions. Speakers will have the opportunity to present (and, if possible demonstrate the benefits of) their views without the necessity to reach consensus.

The workshop is also expected to fertilize the continuing discussion at the WG's mailing list and to lay the ground for a jointly authored position paper to be submitted to JAMIA in 2006.

 

Address for correspondence:

 

Stefan Schulz
Abteilung Medizinische Informatik
Universtätsklinikum Freiburg
Stefan-Meier-Str. 26
D-79106 Freiburg (Germany)
stschulz@uni-freiburg.de
http://www.imbi.uni-freiburg.de/ medinf/~schulz.htm