Integrating Healthcare Knowledge Artifacts for Clinical Decision Support: Towards Semantic Web Based Healthcare Knowledge Morphing
ABSTRACT Healthcare decision making demands the systematic integration of knowledge from multiple sources, such as clinical guidelines,
clinical pathways, knowledge of practitioners and so on. We present a semantic web based approach for synthesizing health
knowledge through the semantic modeling of healthcare knowledge as ontologies and reasoning over the ontologies to derive
a morphed knowledge object. We demonstrate the application of our approach by generating morphed knowledge about prostate
cancer clinical pathways.
- SourceAvailable from: Marta Sabou[show abstract] [hide abstract]
ABSTRACT: We present a manifesto for a new form of knowledge sharing that is based not on direct sharing of "true" statements about the world but, instead, is based on sharing descriptions of interactions. By making interaction specifications the currency of knowledge sharing we gain a context to interpreting knowledge that can be transmitted between peers. The narrower notion of semantic commitment we thus obtain requires peers only to commit to meanings of terms for the purposes and duration of the interactions in which they appear. This lightweight semantics allows networks of interaction to be formed between peers using comparatively simple means of tackling the perennial issues of query routing, service composition and ontology matching. Although the entire system described in this manifesto has not been built, all its components use established methods; many of these have been deployed in substantial applications; and we summarize a simple means of integration.
- [show abstract] [hide abstract]
ABSTRACT: Ontologies and contexts are complementary disciplines for modeling views. In the area of information integration, ontologies may be viewed as the outcome of a manual effort to model a domain, while contexts are system gener- ated models. In this work, we provide a formal mathematical framework that de- lineates the relationship between contexts and ontologies. We then use the model to handle the uncertainty associated with automatic context extraction from ex- isting documents by providing a ranking method, which ranks ontology concepts according to their suitability to a given context. Throughout this work we moti- vate our research using QUALEG, a European IST project that aims at providing local governments with an effective tool for bi-directional communication with citizens. We empirically evaluated our model using two real-world data sets, com- ing from Reuters and news RSS. Our empirical analysis shows that the proposed model can be adopted in practice. The input needed to accurately define a concept by a context is small, and the classification of documents to concepts is accurate.J. Data Semantics. 01/2007; 9:113-140.
Integrating Healthcare Knowledge Artifacts for
Clinical Decision Support: Towards Semantic
Web Based Healthcare Knowledge Morphing
Sajjad Hussain and Syed Sibte Raza Abidi
NICHE Research Group, Faculty of Computer Science, Dahousie University, Canada
Abstract. Healthcare decision making demands the systematic inte-
gration of knowledge from multiple sources, such as clinical guidelines,
clinical pathways, knowledge of practitioners and so on. We present a
semantic web based approach for synthesizing health knowledge through
the semantic modeling of healthcare knowledge as ontologies and reason-
ing over the ontologies to derive a morphed knowledge object. We demon-
strate the application of our approach by generating morphed knowledge
about prostate cancer clinical pathways.
Healthcare decision making during the diagnostic-treatment cycle is a com-
plex activity. Health professionals make clinical decisions by applying healthcare
knowledge that includes their experiential knowledge and explicit knowledge ‘ar-
tifacts’, such as clinical practice guidelines, best evidence, clinical pathways and
so on . One may note that each healthcare knowledge artifact captures specific
conceptual, contextual and operational aspects of a disease and corresponding
diagnostic/therapeutic procedures. Health professionals, guided by the patient’s
context, are able to select the relevant ‘knowledge objects’ from these different
artifacts and then inter-relate these specific knowledge objects whilst satisfying
clinical relevance and pragmatics constraints. For instance, a health professional
generating a treatment plan for a patient with hypertension and diabetes will
refer to the relevant sections of (a) clinical guidelines for recommendations; (b)
clinical pathways for procedural protocols to exercise these recommendations;
and (c) medical literature to determine the best evidence and outcomes of treat-
ment options. In our work, we attempt to pursue a context-sensitive selection and
integration of medical knowledge from multiple knowledge artifacts to generate
a comprehensive knowledge object for clinical decision support.
We are developing the concept of healthcare knowledge morphing that entails
“the intelligent and autonomous fusion/integration of contextually, conceptually
and functionally related knowledge objects that may exist in different represen-
tation modalities and formalisms, in order to establish a comprehensive, multi-
faceted and networked view of all knowledge pertaining to a domain-specific
problem”–Abidi 2005 . In this paper, we present our healthcare knowledge
C. Combi, Y. Shahar, and A. Abu-Hanna (Eds.): AIME 2009, LNAI 5651, pp. 171–175, 2009.
c ? Springer-Verlag Berlin Heidelberg 2009
172S. Hussain and S.S.R. Abidi
Fig.1. Healthcare Knowledge Morphing
morphing framework K-MORPH that is based on the semantic web approach
that entails: (a) Developing Knowledge Artifact Ontologies (KAOs) to represent
knowledge within CPG, CP and CM ; (b) Specifying the clinical context of
the knowledge morphing activity through a rich morphing construct; (c) Gen-
erating the morphed knowledge by (i) selecting a contextualized sub-ontology,
corresponding to the clinical context, from the KAO; and (ii) merging the se-
lected contextualized sub-ontologies, using reasoning algorithms applied to a set
of domain-specific context-specific axioms, to generate a new sub-ontology that
represents the ‘morphed’ knowledge artifact. We demonstrate the working of
our knowledge morphing framework K-MORPH by morphing three different
location-specific clinical pathways to generate a comprehensive knowledge about
treatments and follow-ups for a clinical context therapeutic decision support (see
Figure 1) .
The literature suggests other approaches for knowledge integration problem from
different perspectives. For instance, the ECOIN framework performs semantic
reconciliation of independent data sources, under a defined context, by defining
conversion functions between contexts as a network. ECOIN takes the single
ontology, multiple views approach , and introduces the notion of modifiers
to explicitly describe the multiple specializations/views of the concepts used in
different data sources. It exploits the modifiers and conversion functions, to en-
able context mediation between data sources, and reconcile and integrate source
schemas with respect to their conceptual specializations. Another recent initia-
tive towards knowledge integration is the OpenKnowledge project  that sup-
ports the knowledge sharing among different knowledge artifacts, not by sharing
their asserted statements, instead by sharing their interaction models. An inter-
action model provides a context in which knowledge can be transmitted between
two (or more) knowledge sources (peers).
Integrating Healthcare Knowledge Artifacts for Clinical Decision Support173
The K-MORPH approach is shown in Figure 1, and its main elements are
described in the following subsections. For further details see .
3.1 Knowledge Representation and Annotation via Ontologies
In K-MORPH, knowledge artifacts are represented using two different (but
inter-related) ontologies, namely: (i) Domain Ontology; and (ii) Knowledge Ar-
tifact Ontology (KAO). A domain ontology serves two purposes: (i) standard-
ization of the domain-specific concepts and relations defined in the knowledge
artifact ontologies; and (ii) specification of abstract knowledge links between con-
textually and functionally congruent knowledge components in different KAOs.
A knowledge artifact ontology (KAO) serves as a lower-level ontology that cap-
tures both the structure and content of a particular knowledge artifact–such as
CPG, CP , clinical cases etc. As a test-case, we used three location-specific
(Halifax, Calgary and Winnipeg) Prostate Cancer (PC) pathways as medical
knowledge artifacts, and represented them in different KAOs .
3.2 Contextualizing Ontologies
Contextualizing an ontology deals with an adaptation of its ontology model
to support a local view . In K-MORPH, each KAO models the procedural
knowledge of a knowledge artifact. But, by contextualizing a KAO we are able to
provide a specialized view that models (i) a specific interpretation of its ontology
concepts, and (ii) an implementation of its procedural knowledge applied in a
particular context. A contextualized sub-ontology is extracted from a KAO based
on the context-specific concepts, and comprises (i) instances (ii) sub-concepts,
(iii) equivalent-concepts, (iv) properties, (v) property domain and range, and
(vi) assertions for the context-specific concepts.
In order to represent the context under which two or more knowledge artifacts
can be morphed to solve a specific problem, we defined a Morphing Construct.
The morphing construct supervises the knowledge morphing process, and pro-
vides a context for determining when, where and how two or more knowledge
artifacts need to be reconciled. A Morphing construct is a tuple that declares a
context-specific knowledge component and its role under a defined context.
The K-MORPH morphing engine inputs the problem-context, ontology-
encoded knowledge artifacts (OKAs), domain ontology, and morphing constructs.
It first employs the problem-context to determine the context-specific knowledge
components (i.e. contextualized sub-ontologies) from different KAOs. Based on
the declarative knowledge of morphing constructs, it identifies correspondences
174S. Hussain and S.S.R. Abidi
between the ontology-entities (concepts, properties, and individuals) of differ-
ent contextualized sub-ontologies. Based on the identified correspondences, the
morphing engine employs the ontology reconciliation process that (i) aligns and
then merges contextualized sub-ontologies; (ii) identifies and resolves logical in-
consistencies, if present; and (iii) generates a morphed ontology, and unresolved
inconsistencies in it.
K-MORPH in Action: Morphing Clinical Pathways
We tested the above mentioned processes in K-MORPH using our PC Test-
case. The test-case involves (i) three medical knowledge artifacts, describing
Prostate Cancer (PC) clinical pathways for three different locations (Halifax,
Calgary, and Winnipeg); (ii) a problem-context; and (iii) the morphing con-
structs. The morphing process for the PC Test-case follows the following steps:
Step # 1: KnowledgeRepresentationandAnnotationofPCArtifacts:Theknowl-
edge of three PC pathway artifacts are encoded into three different KAOs. Each
pathway deals with four major types of tasks, namely (a) Consultation Task; (b)
Non-consulation Task; (c) Referal Task; and (d) Followup Task, represented as
concepts/classes in each KAO. Such tasks are supported (via properties) by other
concepts such as Clinician, Decision Criteria, Frequency, Interval Duration, In-
vestigation, Patient Condition Severity, Test Result, Followup, and Treatment.
Step # 2: Defining a Problem-context: We defined a problem-context thera-
peutic decision support whereby the user is needs to morph all three PC path-
ways in terms of: (i) the treatments, (ii) their durations, (iii) their follow-ups,
(iv) their care-settings and (v) the practitioners involved for them. The given
problem-context represents context-specific interpretations, such as (a) Calgary
and Halifax both share PC Clinicians; and (b) Treatments in the PC Calgary
pathway can be treated as Followups in the PC Winnipeg pathway.
Step # 3: Identifying Contextualized Sub-ontologies: Given the therapeutic de-
cision support context, morphing constructs and three PC pathway ontologies,
three contextualized sub-ontologies were generated. Each contextualized sub-
ontologywas semanticallyvalidated for conceptual consistencyand completeness.
Step # 4: Context-driven Ontology Reconciliation of Sub-ontologies: The K-
MORPH morphing engine initiated an ontology reconciliation process on the
contextualized sub-ontologies, and as a result alignments were found between
the classes Treatment, Followup, Frequency, Interval Duration, and Clinician.
The morphing engine processes these alignments using context-axioms and PC
domain-axioms to generate potential ‘knowledge-links’ between aligned PC
Results: Figure 2 shows the morphed knowledge for the treatment PC-Halifax:
ActiveSurveillance. The morphed knowledge has determined that PC-Halifax:
ActiveSurveillance can now be treated by a Primary Urologist. Based on the
reconciliation of the concepts Clinician, Treatment, Followup and Interval a
knowledge-base was generated in terms of a contextualized sub-ontology.
Integrating Healthcare Knowledge Artifacts for Clinical Decision Support175
Fig.2. PC Test-case: Morphed knowledge for PC-Halifax:ActiveSurveillance
5 Concluding Remarks
Clinical decision support needs a knowledge-base that can be designed from
scratch or created by synthesizing existing healthcare knowledge existing in dif-
ferent modalities. In this paper, we presented our knowledge morphing approach
that allows the systematic synthesis of multiple knowledge artifacts to develop
a comprehensive knowledge-base that can be used by decision support systems.
Our knowledge morphing approach ensures the semantic correctness of the mor-
phed knowledge to the extent that it is comparable to a knowledge-base created
through a knowledge engineering exercise. We showed how our approach is used
to develop a unified prostate cancer clinical pathway by synthesizing three dif-
ferent clinical pathways.
This research is funded by a grant from Agfa Healthcare (Canada).
1. Abidi, S., Abidi, R., Hussain, S., Butler, L.: Ontology-based modeling and merging
of institution-specific prostate cancer clinical pathways. In: Knowledge Management
for Healthcare Processes Workshop at ECAI 2008, Patras, Greece, July 21-25 (2008)
2. Abidi, S.S.R.: Medical knowledge morphing: Towards the integration of medical
knowledge resources. In: Computer-Based Medical Systems, June 23-24 (2005)
3. Firat, A., Madnick, S., Grosof, B.: Contextual alignment of ontologies in the eCOIN
semantic interoperability framework. Inf. Tech. and Management 8(1), 47–63 (2007)
4. Robertson, D., et al.: Open knowledge: Semantic webs through peer-to peer inter-
action. Technical Report DIT-06-034, University of Trento, Povo, Italy (May 2006)
5. Hussain, S., Abidi, S.S.R.: K-MORPH: A semantic web based knowledge repre-
sentation and context-driven morphing framework. In: Workshop on Context and
Ontologies, at ECAI 2008, Patras, Greece, July 21-25 (2008)
6. Segev, A., Gal, A.: Putting things in context: A topological approach to mapping
contexts to ontologies. In: Spaccapietra, S., Atzeni, P., Fages, F., Hacid, M.-S., Kifer,
M., Mylopoulos, J., Pernici, B., Shvaiko, P., Trujillo, J., Zaihrayeu, I. (eds.) Journal
on Data Semantics IX. LNCS, vol. 4601, pp. 113–140. Springer, Heidelberg (2007)