=Paper=
{{Paper
|id=Vol-410/paper-3
|storemode=property
|title=Why do it the hard way? The Case for an Expressive Description Logic for SNOMED
|pdfUrl=https://ceur-ws.org/Vol-410/Paper03.pdf
|volume=Vol-410
|dblpUrl=https://dblp.org/rec/conf/krmed/RectorBK08
}}
==Why do it the hard way? The Case for an Expressive Description Logic for SNOMED==
https://ceur-ws.org/Vol-410/Paper03.pdf
Representing and sharing knowledge using SNOMED
Proceedings of the 3rd international conference on Knowledge Representation in Medicine (KR-MED 2008)
R. Cornet, K.A. Spackman (Eds)
Why do it the hard way? The Case for an Expressive
Description Logic for SNOMED
Alan Rector, Sebastian Brandt
School of Computer Science, University of Manchester, Manchester M13 9PL,
(rector | brandt @cs.manchester.ac.uk
Since SNOMED-RT/CT was originally formulated uniformly as fully defined “situations” that in-
in the early to mid 1990s, there have been major clude any context required and that deal with
developments in logic-based formalisms, ontology negation explicitly and formally.
design and associated tools. Combined with the • To represent all sites explicitly as to whether
increase in computing power in the past two dec- they refer to the site in its entirety or to the dis-
ades, these developments mean that many of the junction of the site and its parts.
restrictions that limited SNOMED’s original for-
• To define observables and related findings in
mulation and schemas no longer need apply. We
such a way that the classifier can be used to
contend that future development of SNOMED
recognise the equivalence between a situation
would be made easier if a more expressive formal-
involving an observable with a given value and
ism and more modern tools were adopted.
the corresponding finding of the observable
The difficulties in the existing structure of with that value – e.g., between an observable
SNOMED have been well documented. For exam- of “blood pressure” qualified by “elevated”
ple, Bodenreider (1) examined the specialization and a finding of “elevated blood pressure”.
hierarchy of SNOMED classes. Schulz discussed
• To organise the stated form as a set of modules
‘relationship groups’ (2) and a broad range of other
that can be separated for specific applications.
ontological problems along with potential remedies
(3). Schulz suggested a modest extension of Details of the proposed mechanisms are described
SNOMED’s formalism to one with more clearly in the extended version of this paper and in (4, 5).
defined semantics (EL+) but which still lacks true Although the effort to migrate any large software
negation and disjunction. We argue here that judi- object should not be underestimated, most of the
cious use of a more expressive language, OWL proposed changes would cause few changes to the
1.11, is now practical and would bring great bene- schemas except for “Situations with specific con-
fits including: text,” which are known to be problematic. (How-
• A uniform, clear and understandable schema ever, the proposed analysis would identify many
for all concepts used in clinical records, in- errors to be corrected.) The effort would be more
cluding context and negation. than repaid by providing a more regular and consis-
tent system that would improve usability and sim-
• Elimination of the need for special mecha-
plify software development and query formulation.
nisms to deal with context, partonomy, and
We argue that a feasibility study using a modest
role groups.
subset of around 25K concepts should be an urgent
• More effective leveraging of the underlying priority for the SNOMED community.
logical representation to organise and quality
References
assure the SNOMED hierarchies.
1. Bodenreider O, Smith B, Kumar A, Burgun A. Inves-
• Improved ability to recognise semantic equiva- tigating subsumption in SNOMED CT: An explora-
lence between post-coordinated and pre- tion into large description logic-based biomedical
coordinated expressions and between “observ- terminologies. AI in Medicine. 2007;39:183-195.
ables” with “values” and the corresponding 2. Schulz S, Hanser S, Hahn U, Rogers J. The seman-
“findings.” tics of procedures and diseases in SNOMED CT.
• Improved ability to modularise and segment Meth Inf Med. 2006;45:354-358.
SNOMED for specific purposes 3. Schulz S, Suntisrivaraporn B, Baader F; SNOMED
• Access to the tools and techniques being de- CT's Problem List: Ontologists' and logicians' ther-
apy suggestions.; Medinfo 2007: IOS Press; 802-806.
veloped by the wider Semantic Web and OWL
communities. 4. Rector A, Qamar R, Marley T; Binding ontologies &
coding systems to electronic health records and mes-
In outline, the proposals are: sages. 2006; Formal Biomedical Knowledge Repre-
• To represent all concepts used in clinical re- sentation (KR-MED 2006
cords (findings, observables, and procedures) 5. Rector AL; What's in a code: Towards a formal ac-
count of the relation of ontologies and coding sys-
tems. 2007; Medinfo 2007: Brisbane, Australia: IOS
1
http://www.webont.org/owl/1.1/ Press; 730-734.
15