Ontological Principles Applied to Biomedical Vocabularies

1
Ontological Principles Applied to Biomedical
Vocabularies
Workshop on Foundations of Clinical
Terminologiesand Classification (FCTC 2006)
Timisoara, Romania, April 8, 2006
Josef Ingenerf Roland Linder Institute for
Medical Informatics University to Lübeck
2
Agenda

• Introduction
• Focus on two ontological principles
• Application to three types of biomedical
  vocabularies
• Terminological systems
• Thesauri
• Statistical classifications
• Conclusions

3
Introduction
Conceptualist ontologies
Real ontologies

• Cimino desiderata (ISO 17117), e.g.
• - Concept orientation
• - Concept permanence
• - Non-semantic identifiers
• - Formal definitions
• - Reject NEC
• - Polyhierarchy

4
Introduction
Conceptualist ontologies
Real ontologies

• Cimino desiderata (ISO 17117), e.g.
• - Concept orientation
• - Concept permanence
• - Non-semantic identifiers
• - Formal definitions
• - Reject NEC
• - Polyhierarchy

• Enabling computer-based deductive reasoning
• Recognising semantic similarity in spite of
  syntactic differences.
• Recognising implicit consequences given
  explicitly stated facts.

5
Focus on two ontological principles

• No ISA-overloadingi.e. the misuse of the generic
  relation between concepts (subsumption).

• No intrusion of epistemology i.e. the
  reflection of the way humans gain knowledge
  about reality.

6
Agenda

• Introduction
• Focus on two ontological principles
• Application to three types of biomedical
  vocabularies
• Terminological systems
• Thesauri
• Statistical classifications
• Conclusions

7
Applicability of Ontological Principles
Ontolog. Principles
No ISA- overloading
No intrusion of epistemology
8
Terminological systems, e.g. SNOMED CT
Concept level
Term level (Descriptions)
51197009 Stomach cramps (finding) Primitive
ISA 21522001 Abdominal pain Finding_site
69695003 Stomach structure
- 36112013 Cramps - 2103135014 Magenkrämpfe - ...
21522001 Abdominal pain (finding) Primitive
ISA 22253000 Pain Finding_site 113345001
Abdominal structure
- 36112013 Abdominal Pain - 1924799018
Bauchschmerzen - ...
29384001 Disorder of stomach (disorder) Fully
defined by ISA 64572001 Disorder Finding_site
69695003 Stomach structure
- 2350923015 Gastropathy - 2349022014 Erkrankung
des Magens - ...
User searching Computer NLP
User navigating, retrieving Computer inferencing
9
Terminological systems ? No ISA-Overloading!
Concept level

• e.g. conceptual modelling decisions like
• Ulcer as morphological abnormality versus
  pathological process?
• Aspirin as substance versus product?
• Helicobacter as an Infectious Agent?
• remission of tumor, prevented pregnancy,

10
Terminological systems ? No Intrusion of
epistemology!
Bodenreider, O, Smith, B, Burgun, A (2004). The
Ontology-Epistemology Divide A Case Study in
Medical Terminology. In Varzi, A, Vieu, L.
(ed.). Proc. FOIS 2004, Turin, 4-6. November
2004. Amsterdam IOS-Press, 185-195.
11
Applicability of Ontological Principles
Biomedical Vocabul.
Terminological System
Statistical Classification
Thesaurus
Ontolog. Principles
Both principles are necessary for guaranteeing
valid inferences.
No ISA- overloading
No intrusion of epistemology
12
Thesauri, e.g. MeSH
Term level (entry terms)
Descriptor level (main heading, subject, )

• D010146 Pain
• BT D015746 Abdominal Pain
• BT D000006 Abdomen, Acute
• BT D003085 Colic
• - D012817 Sign and Symptoms, Digestive
• BT D015746 Abdominal Pain
• BT D000006 Abdomen, Acute
• BT D003085 Colic

• - Colic
• - Colicky Pain
• - Kolik
• - Abdominal Cramps
• Abdominal Cramp
• Abdominelle Krämpfe
• - ...

User searching Computer NLP (?)
User navigating, retrieving Computer
inferencing (?)
13
Thesauri ? No ISA-Overloading?
Descriptor level

• Abdomen BT Stomach
• Diagnosis BT Diagnostic Errors

14
Thesauri ? No Intrusion of epistemology?

• No intrusion of epistemology i.e. the
  reflection of the way humans gain knowledge about
  reality.

Descriptor level
15
Thesauri Intermediate concept level
Descriptor level (main headings, subjects , )
- D003085 Colic
Term level (entry terms)

• T008934 Colic (Preferred Term)
• T520657 Colicky Pain
• - ger0002992 Kolik
• - T008933 Abdominal Cramps
• - T008933 Abdominal Cramp
• - ger0030380 Abdominelle Krämpfe
• - ...

16
Thesauri Intermediate concept level
Descriptor level (main headings, subjects , )
- D003085 Colic
Term level (entry terms)
- T008934 Colic (Preferred Term), ref. to (?)
M0004742 - T520657 Colicky Pain ? M0004742 -
ger0002992 Kolik (PT) ? M0004742 - T008933
Abdominal Cramps (PT) ? M0004742 - T008933
Abdominal Cramp ? M0004742 - ger0030380
Abdominelle Krämpfe (PT) ? M0004742 - ...
17
Applicability of Ontological Principles
Biomedical Vocabul.
Terminological System
Statistical Classification
Thesaurus
Ontolog. Principles
Both principles are necessary for guaranteeing
valid inferences.
No ISA- overloading
No intrusion of epistemology
18
Statistical Classifications, e.g. ICD
Term level (alphabetical index)
Class level

• Magenulkus
• Ulcus pyloricum
• - Infektion durch Helicobacter pylori bei
  Ulcus ventriculi
• - ...

R52 Pain, not elsewhere classified R52.0
Acute pain R52.1 Chronic intractable pain
R52.2 Other chronic pain R52.9 Pain,
unspecified K25 Gastric ulcer K25.0 acute
with haemorrhage K25.1 acute with perforation
K25.2 acute with both haemorrhage and
perforation K25.3 acute without haemorrhage and
perforation K25.4 chronic or unspecified with
haemorrhage ... ... K25.9 unspecified
as acute or chronic, without
haemorrhage or perforation
User searching Computer NLP(?)
User navigating, retrieving Computer
inferencing (?)
19
Statistical Classifications ? No ISA-Overloading?
Class level
20
Statistical Classifications ? No Intrusion of
epistemology?
Class level
21
Statistical Classifications Intermediate concept
level
Class level (tabular list)
K25 Gastric ulcer K25.9 unspecified as acute
or chronic, without
haemorrhage or perforation
Term level (alphabetical index)
- I5057 Magenulkus ? K25.9 - I11881 Ulcus
pyloricum ? K25.9 - I32930 Infektion durch
Helicobacter pylori bei Ulcus ventriculi ? K25.9,
B96.81 - ...
Note B96.81 Helicobacter pylori H. pylori as
the cause of diseases classified to other
chapters is added in ICD-10-GM.
22
Statistical Classifications Intermediate concept
level
Class level
K25 Gastric ulcer K25.9 unspecified as acute
or chronic, without
haemorrhage or perforation
Term level (entry terms)
- I5057 Magenulkus ? K25.9 - I11881 Ulcus
pyloricum ? K25.9 - I32930 Infektion durch
Helicobacter pylori bei Ulcus ventriculi ? K25.9,
B96.81 - ...
Note B96.81 Helicobacter pylori H. pylori as
the cause of diseases classified to other
chapters is added in ICD-10-GM.
23
Applicability of Ontological Principles
Biomedical Vocabul.
Terminological System
Statistical Classification
Thesaurus
Ontolog. Principles
Not applicable! (Descriptors not linked
ontologically by ISA-rel.)
Both principles are necessary for guaranteeing
valid inferences.
No ISA- overloading
Epistemology is reflected indirectly (? basic
level)
No intrusion of epistemology
24
Applicability of Ontological Principles
Biomedical Vocabul.
Terminological System
Statistical Classification
Thesaurus
Ontolog. Principles
Not applicable! (Descriptors not linked
ontologically by ISA-rel.)
Not applicable. (Disjoint classes not linked
ontologically by ISA-relations)
Both principles are necessary for guaranteeing
valid inferences.
No ISA- overloading
Epistemology is reflected indirectly (? basic
level)
Epistemology-loaded terms necessarily included!
No intrusion of epistemology
Is the application of ontological principles
useful?
25
Conclusions
Level of purpose-specific aggregations of
concepts (e.g. descriptor, class)
Making the green ones red by applying ontological
principles?
26
Conclusions
Level of purpose-specific aggregations of
concepts (e.g. descriptor, class)
...
...
...
Level of concepts
27
Conclusions
Level of purpose-specific aggregations of
concepts (e.g. descriptor, class)
purpose-dependent e.g. with Epistemologie
...
...
...
GAP (missmatch)
Level of concepts
purpose-independent e.g. without Epistemologie
Elkin, PL, Brown, SH (2002). Automated
enhancement of description logic-defined
terminologies to facilitate mapping to ICD9-CM.
J Biomed Inform 35 (5-6) 281-8.
28
Conclusions
Level of purpose-specific aggregations of
concepts (e.g. descriptor, class)
Concept-oriented formal terminologies can do the
job of purpose-specific vocabularies!
Level of concepts
29
Medical Terminologies adding and removing
complexity
30
Some links to relevant literature
Nelson, SJ, Johnston, D, Humphreys, BL (2001).
Relationships in Medical Subject Headings
(MeSH), chapter 11. In Bean, CA, Green, R
(eds.). Relationships in the Organization of
Knowledge. New York Kluwer Academic Publishers,
171-184, see http//www.nlm.nih.gov/mesh/meshrels.
html.
For example, the component concepts of the MeSH
descriptor 'Exercise', 'Isometric Exercise' and
'Aerobic Exercise' overlap in meaning. They are
not sufficiently distinct in meaning to be
useful in this thesaurus. In other cases, the
descriptor itself may be clear but its
application would not be. For example, we might
wish to make a distinction between 'DNA
Fingerprints' and 'DNA Fingerprinting.' The
meanings are distinct, but the literature is not.
Discussions do not clearly distinguish between
the process and the product. By way of
contrast, we can note that 'Radiography' and
'Radiographs' are sufficiently distinct in the
literature to warrant making them separate
descriptors.
31
Some links to relevant literature
Soualmia, LF, Golbreich, C, Darmoni, SJ (2004).
Representing the MeSH in OWL Towards a
Semi-Automatic Migration. In Hahn, U (ed.).
Proc. KR-MED 2004, Whistler (BC), Canada.see
http//CEUR-WS.org/Vol-102/, 81-87.
The results of this technique are used to
augment the existing SNOMED-RT relation
ontology, which is a necessary step in
automated concept mapping between systems. The
reference terminology must contain all the
semantics implicit in the classification in
order to map concepts between the two
representations.
32
Some links to relevant literature
Smith, B, Ceusters, W (2006). Ontology as the
Core Discipline of Biomedical Informatics -
Legacies of the Past and Recommendations for the
Future Direction of Research. In Crnkovic, GD,
Stuart, S (eds.). Computing, Philosophy, and
Cognitive Science. Cambridge Scholars Press,
forthcoming
As an example, the ICD-10 class B83.9
Helminthiasis, unspecified does not refer (for
example) to a disease caused by a worm belonging
to the species unspecified (...). Rather, it
refers to a statement (perhaps appearing in some
patient record) made by a physician who for
whatever reason did not specify the actual type
of Helminth which caused the disease the
patient was suffering from.
33
Some links to relevant literature
Elkin, PL, Brown, SH (2002). Automated
enhancement of description logic-defined
terminologies to facilitate mapping to ICD9-CM.
J Biomed Inform 35 (5-6) 281-8.
The results of this technique are used to
augment the existing SNOMED-RT relation ontology,
which is a necessary step in automated concept
mapping between systems. The reference
terminology must contain all the semantics
implicit in the classification in order to map
concepts between the two representations.
34
Library Classifications and Thesauri
35
Library Classifications and Thesauri
36
Ontological principles applied to all biomedical
vocabularies