=Paper=
{{Paper
|id=Vol-410/paper-12
|storemode=property
|title=Creation and Usage of a "Micro theory" for Long Bone Fractures: An Experience Report
|pdfUrl=https://ceur-ws.org/Vol-410/Paper12.pdf
|volume=Vol-410
|dblpUrl=https://dblp.org/rec/conf/krmed/GoldbergKS08
}}
==Creation and Usage of a "Micro theory" for Long Bone Fractures: An Experience Report==
https://ceur-ws.org/Vol-410/Paper12.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)
Creation and Usage of a “Micro Theory” for Long Bone Fractures: An
Experience Report
Howard S. Goldberg, MD1, Vipul Kashyap, PhD1, Kent A. Spackman, MD, PhD2
1
Clinical Informatics R&D, Partners Healthcare System, Wellesley, MA, USA
2
Oregon Health & Science University, Portland, OR, USA
{hgoldberg, vkashyap1}@partners.org, ksp@ihtsdo.org
We seek to leverage enhanced expressivity in OWL unanticipated inference, especially when scaling to
1.1 via property chain axioms with right identities in large numbers of concepts [1].
order to organize and constrain anatomic concepts
for use in clinical descriptions. Anatomic knowledge For our initial investigation, we focused on a single
represented in SNOMED CT uses SEP triplets; we use case limited to fractures of long bones. We
anticipate that property chains will allow a more adopted an iterative bottom-up process to developing
parsimonious organization of anatomic concepts. a “micro-theory”—an axiomitization that yields
However, these constructs may lead to unanticipated sensible and logically correct inference in a limited
inference, especially when scaling to large numbers domain. At each stage, we tested the incremental
of concepts [1]. We used a bottom-up approach theory against the use case scenario. We re-used
based on targeted use case questions to iteratively content from the Foundational Model of Anatomy
develop a “micro theory” that both identifies the (FMA) [8]. The various distinctions introduced in the
sensible locations of fractures in long bones and also FMA to model partonomy, i.e., systemic-part-of,
supports logic-based classification of fractures. regional-part-of and constitutional-part-of were
Alternative representations of the statement explored. We attempted to design a theory that was
“fractures occur in bone” were explored with the compact and understandable and also gave us the
aim of creating rich clinical descriptors that support correct intended behavior. The model accounts for
classification for inference and data mining. The both anatomic perspectives and functional clinical
process of creating this micro theory is discussed, perspectives. We tested the model by computing the
where pragmatic decisions were made with an appropriate inferences based on the use cases.
intention of both constraining data entry and
enabling inferences within the scope of the use cases. Locative transfer over pathophysiologic processes is a
fundamental property for ontologies that will be used
INTRODUCTION for clinical decision support or data warehousing
applications. Given the sheer number of anatomic
OWL and other forms of description logics have been concepts present in systems such as SNOMED CT, it
used extensively to model spatial relationships for is critical that modeling idioms yield predictable
anatomical knowledge [1-6]. The focus of these results in order to scale. An important goal of the
efforts has been either to investigate the current work is begin to understand the characteristics
computational properties of the description logic or to of these idioms in a limited domain.
develop a generalized set of axioms or theories to
support classification inferences for a wide variety of
Clinical Scenario and Use Case Questions
clinical decision support use cases. We seek to
Typically, a physician creates a clinical descriptor
leverage the enhanced expressivity of OWL 1.1 [7] to
that is of sufficient granularity to support a
organize anatomic concepts for use in creating
management plan—the clinical descriptor is an index
clinical descriptions. In particular, we explore the use
for the general management plan for a given
of property chain axioms with right identities to
pathology. Within the contemporary electronic health
simplify a knowledge base of anatomy without
record, the clinical descriptor may be reused as data
limiting the inferences that can be computed. It has
to drive point-of-care decision support, or as
previously been demonstrated that for anatomical
warehouse data to support reporting. For instance, if
descriptions, inferences after addition of these axioms
we need to report the number of patients who had a
can remain computationally tractable [3]. In contrast
fracture of the proximal femur, we should include the
with other approaches, such as SEP triplets [2], we
number of patients who had a fracture of the femoral
anticipate that property chains will allow a more
neck. In both cases, the original descriptor should
parsimonious organization of anatomic concepts. The
support detailed classification schemes.
downside to using property chains and transitivity,
however, are that these constructs may lead to
66
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)
With respect to bone fractures, it is desirable to hierarchy, we ideally would have the following
describe fractures in detail with respect to the bone classes defined:
features involved—the clinical detail drives the
management plan. The clinical detail may describe StructureOfFemur
EntireFemur Ն StructureOfFemur
either a fracture involving an anatomic landmark or a
FemurPart Ն StructureOfFemur ՈӃpartOf.EntireFemur
functional region where all fractures act similarly. It BoneStructureOfDistalFemur Ն FemurPart
is equally important that the clinical descriptor not EntireDistalFemur Ն BoneStructureOfDistalFemur
admit any nonsensical description. While fractures DistalFemurPart Ն BoneStructureOfDistalFemur
may involve bony landmarks, we generally do not Ո ӃpartOf.EntireDistalFemur
describe fractures of the periosteum—the bone StructureOfDistalEpiphysisOfFemur Ն DistalFemurPart
lining—or the bone marrow. While these are parts of EntireDistalEpiphysisOfFemur
Ն StructureOfDistalEpiphysisOfFemur
bones, they are not generally parts through which
fractures are described to occur. The GALEN project The E-class is instantiated by entire anatomical
used constraints called sanctions to specify the values objects (such as the entire femur), and the P-class by
that could sensibly be applied to relations such as the proper parts of the referred objects (such as the
has-location [6]. Similarly, we constructed our distal femur). The S-class, finally, is instantiated by
ontology fragment with the intent of logically instances that are either entire objects or their parts.
defining the set of all and only locations for fractures. This definition explains the is-a links from the E-
class and the P-class to the S-class, as well as the
Given a need to document, to classify, and possibly to partOf link from the P-class to the E-class. The main
obtain reference information, useful questions that idea underlying the SEP-triplet approach is to
might be posed include† represent a part-whole relationship between two
entity classes not by a part-of link between the E-
1. What bone regions and features are contained in classes, but rather by an is-a link between the S-class
the Distal Epiphysis of the Femur? of the “part” and the P-class of the “whole”. This is,
2. What parts of the Distal Epiphysis of the however, sufficient to simulate transitivity of part-of
Humerus are covered by Articular Cartilage? through the inherently transitive relation is-a:
3. Is a fracture of the Femoral Neck also a fracture
of the Proximal Femur (i.e., is a fracture through EntireDistalEpiphysisOfFemur
an anatomic feature a fracture of a functional Ն StructureOfDistalEpiphysisOfFemur
region)? Ն DistalFemurPart
4. Is a fracture of the Trochlea a fracture of the Ն BoneStructureOfDistalFemur
Distal Epiphysis of the Humerus? Ն FemurPart
Ն partOf.EntireFemur
5. Is a fracture of the Trochlea an intra-articular
fracture?
6. Is a fracture of the Trochlea an intra-articular This allows us to conclude that every Distal Epiphysis
fracture of the Distal Epiphysis of the Humerus? of the Femur is part of some Femur. Since
characteristics are inherited along the is-a hierarchy,
the SEP-triplet encoding also allows us to simulate
MATERIALS
inheritance of characteristics along the part-of
We looked at the following two sources for creating hierarchy. In our example, by connecting a fracture
the fracture ontology: (a) The SNOMED CT via the findingSite property to the S-class, we can
hierarchies spanning the femur and the humerus; and ensure that a fracture located in the Distal Epiphysis
(b) The FMA hierarchies corresponding to the femur of the Femur is classified as a fracture located in the
and the humerus. A brief description of the portion of Femur. Another advantage of the SEP encoding is
these two knowledge sources is described below. that one can suppress such inheritance along the part-
of hierarchy by connecting via findingSite to the E-
SNOMED CT class.
SEP-triplets are extensively employed in the
anatomical part of SNOMED CT. For each There are, however, several problems with the SEP-
SNOMED anatomical class representing one entire triplet encoding. First, from a formal ontological
entity, called entity (or entire) class (E-class), there point of view, it partially conflates the is-a hierarchy
are two auxiliary classes, the structure class (S-class) with the part-of hierarchy, which may lead to
and the part class (P-class). For example, in the femur unintended consequences since the two relationships
are completely different by nature [9]. In SNOMED,
it has indeed turned out that is-a links can be
†
1) The medial and lateral condyles of the Femur; 2) Trochlea and ambiguous, i.e., it is not always clear whether they are
Capitellum; 3) Yes; 4) Yes; 5) Yes; 6) Yes.
67
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)
introduced as part of the SEP-triplet approach, or are defined not only by genetically regulated
supposed to represent a genuine generalization developmental processes (e.g., lobe of lung, cortex of
relationship. Second, the SEP-triplet approach is error kidney, finger), but also by arbitrary landmarks or
prone since it works correctly only if it is employed coordinates, such as used for demarcating the thoracic
with a very strict modeling discipline. In SNOMED, and abdominal parts of the aorta and the fundus of the
triplets are often modeled in an incomplete way; in stomach from adjacent parts of the corresponding
particular, the P-class and the part-of link to it from wholes. A systemic part is defined as a secondary
the E-class are missing in most cases. For example, partition of an anatomic structure in accord with
the following axioms presented earlier were not functional systems.
actually asserted in SNOMED, but were included for
pedagogical purposes (DistalFemurPart does not The distinction between regional parts determined by
currently exist in SNOMED): well defined genetically regulated processes and
arbitrary landmarks and coordinates, is represented
DistalFemurPart Ն BoneStructureOfDistalFemur by associating the attributes anatomical or arbitrary
Ո ӃpartOf.EntireDistalFemur with regional parts. Furthermore, these attributes
StructureOfDistalEpiphysisOfFemur Ն DistalFemurPart provide the basis for the different views of regional
partitions, as in the case of the liver, where its
In addition, the auxiliary S-class is sometimes traditional partition into lobes based on arbitrary
incorrectly used as if it were an entire entity class. landmarks constitutes an arbitrary kind of regional
Third, the approach introduces for every proper class view, while another partition based on the distribution
in the ontology two auxiliary classes, which results in of the tributaries of the hepatic veins or branches of
a significant increase in the ontology size. Finally, the hepatic artery constitutes an anatomical regional
the SEP approach makes it much more difficult to view.
define and maintain the set of sensible locations for
fractures.
The FMA also supports topologic relationships
Foundational Model of Anatomy supporting connectedness and containment.
The FMA ontology defines a set of partonomic Connectedness describes whether structures are
relationships discussed in [10,11] for guiding the continuous with, attached to, or synapsed with other
representation of anatomical parts. This is a smaller structures. Containment deals exclusively with the
set than that used in GALEN [6], and thus one of the containment of a material anatomic entity within an
questions we seek to answer is whether it is sufficient anatomic space, e.g., Right lung -contained in- Right
for clinical modeling, Refinements of the generic half of thoracic cavity. Connectedness and
part-whole relationships for anatomical structures are containment are orthogonal to regionality and
proposed, as anatomical structures have been constitutionality and do not confer parthood [12].
decomposed based on several different contexts. A
partition is defined as the decomposition of the entire
body or any anatomical structure in a given context or METHODS
viewpoint. We now present our approach to developing the long
bone fracture ontology. We draw on the FMA as a
A constitutional part is defined as a primary partition primary source of anatomic content.
of an anatomical structure into its compositionally
distinct anatomical elements. In the context of the Regional vs. Constitutional Partitions
whole, an element is any relatively simple component As previously discussed, the FMA ontology draws a
of which a larger, more complex anatomical structure distinction between a regional partition and a
is compounded; i.e., the partition is compositional constitutional partition. We reviewed this content to
rather than spatial. For example, a stomach may be determine whether it was suitable for reuse within our
viewed as being partitioned into its wall and cavity. A ontology fragment.
regional part on the other hand is defined as a
primary partition that spatially subdivides an 1. The regional partition of long bones is
anatomical structure into sets of diverse constitutional exemplified by the following regional parts of the
parts that share a given location within the whole; i.e., Femur (regPartOf):
the partition is spatial rather than compositional. For
example, a stomach may be viewed as being ProximalEpiphysisOfFemur ՆӃ regPartOf.Femur
partitioned into its fundus, body and pyloric antrum to DiaphysisOfFemur ՆӃ regPartOf.Femur
name a few of such parts. Constitutional parts are DistalEpiphysisOfFemur ՆӃ regPartOf.Femur
genetically determined, whereas regional parts are FemoralNeckOfFemur ՆӃ
regPartOf.ProximalEndOfFemur
68
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)
Disorder
Fracture Ն Disorder ՈӃILQGingSite.BoneRegion
Regional parts of the femur include true anatomic
parts (epiphyses, diaphysis) as well as functional parts The class Bone is effectively the class BoneOrgan in
defined by fiat boundaries (proximal end of femur), the FMA. Within this initial iteration, we are neutral
illustrating the FMA’s anatomic and arbitrary types. regarding the alignment of Bone with the Upper
Ontology of FMA, i.e., aligning with the is-a
hierarchy consisting of CavitatedOrgan, Organ,
2. The constitutional partition of long bones is AnatomicalStructure, MaterialAnatomicalEntity, and
exemplified by the following constitutional parts AnatomicalEntity, as we did not see an impact of this
of the Femur (constPartOf): in the context of the application at hand. The property
findingSite aligns with the SNOMED CT relationship
BonyPartOfFemur ՆӃFRQVW3DUW2I)HPXU which assigns locations to clinical conditions
BoneOfFemur ՆӃFRQVW3DUW2I%RQ\3DUW2I)HPXU
PeriosterumOfFemur Ն We declare the property regionalPartOf to be
ӃFRQVW3DUW2I%RQ\3DUW2I)HPXU reflexive, thereby inducing Bone to be a BoneRegion.
MedullaryCavityOfFemur Ն
This has the important effect of unifying the treatment
ӃFRQVW3DUW2I%RQ\3DUW2I)HPXU
VasculatureOfBonyPartOfFemur Ն of entire long bones and bony landmarks with respect
ӃFRQVW3DUW2I%RQ\3DUW2I)HPXU to findingSite—fractures may be declared to occur
ArticularCartilageOfDistalEpiphyisOfFemur Ն equally within the entire bone or at the landmark. We
ӃFRQVW3DUW2I)HPXU declare regionalPartOf to be transitive to support the
ArticularCartilageProximalEpiphysisOfFemur Ն interrelationships between discrete landmarks, larger
ӃFRQVW3DUW2I)HPXU regions of bone, and the entire bone.
VasculatureOfFemur ՆӃFRQVW3DUW2I)HPXU
CavityOfFemur ՆӃFRQVW3DUW2I)HPXU
The constitutional parts of the femur include the Anatomical vs. Functional Partition
multiple tissue types that combine to form a long We add the following subclasses of BoneRegion into
bone—the bone proper, the articular cartilage, etc. the model:
Note the bone proper also decomposes to include the
bone material itself, the periosteum, and the AnatomicBoneRegion Ն%RQH5HJLRQ
medullary cavity. FunctionalBoneRegion Ն%RQH5HJLRQ
The regional partition includes the structures where In clinical practice, pathology may be attributed to a
clinicians locate fractures and the relationships true anatomic entity or a functional entity where
between these structures. The constituents of long unique pathologies behave similarly, are responsive
bone such as the periosteum, where fractures are not to similar treatments, are aggregated for
described to occur, are conveniently sequestered in epidemiologic purposes, etc. In orthopedics, for
the constitutional partition. We adopted the relevant example, several unique fractures all aggregate to
portions of the regional partition for use in our model. fractures of the proximal femur. As previously noted,
However, we adopted a simpler representation for the the FMA incorporates true anatomic regions and
incorporation of articular cartilage into the model for functional regions. We partition bone regions into
this initial iteration. either anatomic or functional components to support
the independent enumeration of these features, as
described in the use case.
Modeling Design Choices
We now present some high-level classes and object
properties that characterize the entities in which we Propagation of Locative Relationships
are interested. A key functionality that is required to support the use
case questions discussed earlier is the ability to
Bone propagate the location of a fracture from a given
LongBone Ն Bone region to all the regions to which it has regionalPartOf
Femur Ն LongBone relationships. For instance, if a fracture is located in
Humerus Ն LongBone the femoral neck, it is also located in the proximal
ObjectProperty(regionalPartOf) metaphysis of the femur as the femoral neck is a
reflexive(regionalPartOf)
transitive(regionalPartOf)
regional part of the proximal metaphysis of the femur.
BoneRegion Ն ӃUHJLRQDO3DUW2IBone This is represented using the following axiom:
ObjectProperty(findingSite)
domain(findingSite) = Disorder findingSite ӓregionalPartOf ՆfindingSite
69
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)
(Since regionalPartOf ӓregionalPartOfՆ
It may be noted that the transfer of locative regionalPartOf)
relationships is also propagated transitively due to the Ն)UDFWXUHՈӃfindingSite.Femur
transitive nature of regionalPartOf. (Since findingSite ӓregionalPartOf ՆfindingSite)
Ն)HPRUDO)[
regionalPartOf ӓregionalPartOf ՆregionalPartOf
The proof above indicates that we can represent
direct relationships between bones and bone features
Articular Bone Regions and infer regional partonomy relationships between
In order to explore articular fractures—the fracture of them.
a bone region covered by articular cartilage, we
incorporated the following concepts : Articular Fractures
Extending the model to describe and classify articular
ArticularCartilage fractures is also accommodated by the model and
ObjectProperty(coveredBy) creates no additional complications. The articular
ArticularBoneRegionԘ%RQH5HJLRQՈ parts of the distal epiphysis of the humerus—trochlea
ӃFRYHUHG%\$UWLFXODU&DUWLODJH and capitellum—are created as articular regions,
ArticularFracture Ԙ)UDFWXUH Ո while the non-articular parts—the medial and lateral
ӃILQGLQJ6LWHArticularBoneRegion
epicondyle—are created as regular bone regions. The
This representation provides a simple method to only caveat to this approach is that partially-covered
distinguish between articular and non-articular bone regions are not considered articular regions; the distal
regions. epiphysis of the humerus is not considered an
articular bone region by this criterion.
RESULTS Fractures of the parts are created in the usual fashion
by restricting the fracture finding site. Trochlear and
Using the initial ontology, we were able to create a capitellar fractures classify appropriately as articular
series of detailed clinical descriptions which fractures. General fractures of the distal humeral
classified as expected. Some examples are discussed epiphysis and articular fractures are then created in
next. the same way. Articular fractures of the distal
humeral epiphysis are classified as subclasses of
Locative Transfer over Regional Parts general fractures of the distal humeral epiphysis;
rA facture of the Femoral Neck is classified as a trochlear and capitellar fractures are classified as
fracture of the Proximal Femur. further subclasses. Fractures of the epicondyles
classify correctly as general fractures only. We did
not specifically try to define non-articular fractures
FemoralNeckFx Ԙ)UDFWXUHՈ ӃfindingSite.FemoralNeck
(fractures of parts not covered by articular cartilage).
Ն)Uacture Ո
ӃfindingSite.(ӃregionalPartOf.ProximalEndOfFemur) Breach of the Model
(Since FemoralNeck Ն We note one failure of the model in the subset of
ӃregionalPartOf.ProximalEndOfFemur) bones we examined. The FMA contains a regional
Ն)UDFWXUHՈӃfindingSite.ProximalEndOfFemur part of the humerus, ‘Nutrient Foramen of Humerus’,
(Since findingSite ӓregionalPartOf ՆfindingSite) literally, the hole where the nutrient artery enters the
Ն3UR[LPDO)HPXUFx humerus. Because fractures are usually not described
through this feature, this constitutes a failure of the
Transitive Locative Transfer model to constrain the set of sensible locations of
A fracture of the Femoral Neck is classified as a fractures.
fracture of the Femur. Let’s revisit the earlier
example and begin with the following reformulation In the FMA, the nutrient foramen is a subclass of
of FemoralNeck. Immaterial Anatomic Entity. We can certainly
remediate the model to additionally restrict bone
FemoralNeckFx regions to subclasses of Material Anatomic Entity.
Ն)UDFWXUHՈ
ӃfindingSite.(ӃregionalPartOf.ProximalEndOfFemur)
However, the appearance of the nutrient foramen as
Ն)UDFWXUHՈӃfindingSite. an arbitrary bone region bears further discussion.
(ӃregionalPartOf.(regionalPartOf.Femur))
(Since ProximalEndOfFemur ՆӃregionalPartOf.Femur)
Ն)UDFWXUHՈӃfindingSite.(ӃregionalPartOf.Femur)
70
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)
DISCUSSION References
We have begun to explore the creation of a ‘micro- 1. Seidenberg J and Rector A, Representing
theory’ for long bone fractures—an axiomitization Transitive Propagation in OWL. Proc. 25th
that yields both sensible and logically correct International Conference on Conceptual
inference. Using a set of framing axioms in Modeling (ER 2006), November 2006.
combination with content from the FMA regional 2. Schulz S, Romacker M, and Hahn U, Part-whole
partition, we were able to describe and correctly reasoning in medical ontologies revisited:
classify a rich set of fractures, while maintaining a Introducing SEP triplets into classification-based
fairly parsimonious ontology. The resulting ontology description logics. Proc. 1998 AMIA Annual Fall
is quite constrained as compared to an SEP triple Symposium.
approach. 3. Suntisrivaraporn B, Baader F, Schulz S and
Spackmn K, Replacing SEP-Triplets in
Although this initial model successfully fulfills the SNOMED-CT using Tractable Description Logic
use case, the breach raises significant questions. Operators. Proc. 11th International Conference on
Because the FMA regional model admits arbitrary Artificial Intelligence in Medicine (AIME 07),
regions, there is no principled reason why an arbitrary July 2007.
region of a bone can sensibly be a fracture location. 4. Horrocks I and Sattler U, Decidability of SHIQ
Our success seems to be an empirical finding— with complex role inclusion axioms. Artificial
further analysis is necessary to see whether the model Intelligence 160(1-2), December 2004.
hold across all bones, or can extend to parenchymal 5. Schulz S, Hahn U and Romacker M. Modeling
organs such as the lung or the liver—organs made of Anatomic Spatial Relations with Description
the same ‘stuff’. Logics. Proc. 2000 AMIA Annual Fall
Symposium.
Our model may succeed because bones exemplify a 6. Rector A, Bechhofer S, Goble C, Horrocks I,
‘stuff/whole’ partonomy, where arbitrary regions are Nowlan W, Solomon W. The GRAIL concept
compositionally homogenous. This does suggest that modelling language for medical terminology.
if regional parthood could be compositionally Artif Intell Med. 1997 Feb;9(2):139-71.
restricted, this would offer a more convincing model. 7. OWL Working Group,
One obvious possibility for implementing this http://www.w3.org/2007/OWL/wiki/OWL_Work
restriction is to utilize the GALEN partitive attribute ing_Group
“hasSolidDivision”, or perhaps a similar attribute 8. Foundational Model Explorer.
with even more specialized meaning [13]. Currently, http://fme.biostr.washington.edu:8089/FME/inde
compositional properties are available through the x.html
constitutional partition of the FMA, i.e., if a part is 9. Patrick J. Aggregation and generalization in
composed of bone or a particular organ parenchyma, SNOMED CT. Proc. First Semantic Mining
etc. Further work is necessary to reflect Conf. on SNOMED CT, 2006.
compositionality from constitutionality. We note that 10. Mejino J L V, and Rosse C. Symbolic Modeling
currently, wholes and parts have distinct roots in the of structural relationships in the Foundational
FMA; in our model, it is important to treat parts and Model of Anatomy. Proc. First International
wholes similarly as bone regions. Workshop on Formal Biomedical Knowledge
Representation (KR-MED 2004), 2004.
This work provides initial insight into creating safe 11. Mejino J LV, Agoncillo A V, Rickard K L, and
and effective inference over property chains for the Rosse C, Representing complexity in part whole
purpose of creating and classifying clinical relationships within the Foundational Model of
descriptions. Because of the tremendous change Anatomy. Proc. 2003 AMIA Annual Fall
management implications of incorporating new Symposium.
idioms into terminologies such as SNOMED CT, it is 12. Smith B, Mejino JLV, Schulz S, et al.
important that we demonstrate that such idioms are Anatomical Information Science. Proceedings,
safe, effective, and scale. Continuing work will Seventh International Conference on Spatial
investigate constraining the regional idiom with Information Theory COSIT 2005 Lecture Notes
respect to homogenous compositionality, expanding in Computer Science 3693, pages pp. 149-164.
our analysis to a larger set portion of the skeletal 13. Rector AL, Gangemi A, et al. The GALEN
system, and examining the generalizability of the CORE model schemata for anatomy. Proceedings
idiom to additional organ systems. of Medical Informatics Europe, MIE 94, Lisbon,
pp 186-189.
71