A practical approach to create ontology networks in e-health: The NeOn take Tomás Pariente Lobo 1, *, Germán Herrero Cárcel 1, 1 A TOS Research and Innovation, ATOS Origin SAE, 28037 Madrid, Spain. Abstract. Ontological representation of the health domain is widespread. In particular the semantic description of drugs is being tackled in several ongoing initiatives. However, the resultant ontologies tend to be large and unmanageable. In recent years recommendations go towards the use of smaller, dynamic and interlinked ontologies to ease the ontology life-cycle. So far there have been attempts to build ontologies using this approach, but with very few methodological and tooling support. In this paper we propose to apply the notion of networked ontologies, methodology and tools developed within the NeOn project. Keywords: Ontologies, networked, mappings, tools, knowledge management, nomenclature, interoperability 1 Introduction In recent years, there has been an increasing interest in semantic interoperability in e- Health. Semantic interoperability is about sharing and combining data and health records among different systems and actors. It is also related to foster a consistent usage of the terminology (drugs and bio-medical knowledge bases), and the adoption of shared and standard models of clinical data. In short, semantic interoperability goes to the underlying objective of formalizing the health science using shared or linkable models. One of the key aspects to tackle in order to achieve semantic interoperability is the usage of common or interoperable terminologies about drugs, diseases, treatments and so on. Different actors (governmental bodies, hospitals, labs, key industries, etc.) should be able to understand the terminology used by others. To complicate matters, it is quite common that different systems in the same organization do not use the same terminology. In order to overcome this problem, over the past years numerous initiatives, roadmaps and emerging standards have seen an increasingly rapid development. SNOMED-CT [1] is emerging as a de-facto terminological standard for many international initiatives. Examples of this are the information models adopted in Australia (NEHTA) [2], UK (NHS dm+d) [3] or USA. The W3C created in 2008 the Semantic Web Health Care and Life Sciences (HCLS) Interest Group [4]. The EU * Corresponding author: Tomás Pariente Lobo, ATOS Research and Innovation, ATOS Origin SAE, 28037 Madrid, Spain. E-mail: tomas.parientelobo@atosresearch.eu
financed the SemanticHEALTH FP6 project [5] with the objective of delivering a Semantic Interoperability roadmap for Europe. In particular, SemanticHEALTH issues recommendations such as interlinking health models and terminologies by means of modular, multilingual, dynamic (just-intime), collaboratively-designed networks of ontologies [6] [7]. These recommendations also stress the methodological support needed to specify highquality, consistent and scalable ontologies. It also recommends the use of the W3C standard ontology language, OWL [8], because a large and growing community is developing tools and software (in many occasions freely available) that will benefit the integration and maintenance of ontologies based in this language. The tooling and methodological support needed to foster the adoption of interoperable solutions, especially when talking about bridging the gap between huge terminologies, have not followed such a rapid evolution. There are partial solutions that tackle one or several of the issues raised by SemanticHEALTH. However, far too little attention has been paid to the delivery of an overall framework that covers most of the recommendations cited above. 2 The NeOn approach NeOn [9] is a FP6 EU ICT funded project which aim is to create an open infrastructure, and associated methodology, to support the overall development lifecycle of large scale, complex, semantic applications. This infrastructure is based on the notion of networked ontologies. A network of ontologies is a collection of ontologies related together via a variety of different relationships such as mapping, modularization, version, and dependency relationships [10]. NeOn define four main ontology assumptions: Dynamic (ontologies will evolve), Networking (ontologies are interconnected via mappings, alignments or by means of reuse), Shared (ontologies are shared by people and applications), and Contextualized (ontologies are dependent of the context in which are built or are used) [11]. NeOn has defined a service-based reference architecture that covers design and run time aspects of ontology engineering, plus the usage and integration of the networked ontologies into semantic-enabled applications.
Fig. 1. NeOn Architecure [12] As part of the reference implementation of the NeOn architecture, NeOn delivers an open software suite called the NeOn Toolkit [13]. The NeOn Toolkit is an extensible Eclipse-based Ontology Engineering Environment containing plugins for ontology management and visualization. NeOn Toolkit includes some core features such as editing the ontology schema, visualization an browsing of ontology entities, and it allows the usage of OWL, F-Logic and (subsets of) RDF(S) ontologies. There is a number of commercial plugins that extend the functionality of the NeOn Toolkit. Currently, there are more than thirty plugins available to download, providing several funtionalities such as rule support, mapping editors, database integration, dealing with ontology dynamics (modularization, inconsistency checking), collaboration, localization (multilingualism), etc. The NeOn Toolkit offers a simple Eclipse plugin extensibility that allows an easy deployment of new plugins.
NeOn also offers methodological support to dealing with networks of ontologies. The methodological approach is twofold: on the one hand it provides ontology engineering support, and on the other hand, the methodology provides also guidance to develop applications using networked ontologies. The usage of publically available Ontology Design Patterns [14], in order to improve the quality of the ontology design in a variety of scenarios and needs, is also one of the most relevant outcomes of the project. It is clear the alignment of the NeOn objectives and the recommendations issued by SemanticHEALTH. However, NeOn is not targeting specifically the e-health domain, but taking a horizontal approach valid for multiple domains. However, one of the NeOn case studies is focused on the pharmaceutical domain. In particular, the pilot targeting the interoperability between different drugs terminologies is the socalled Semantic Nomenclature case study. The Semantic Nomenclature case study is trying to pave the way towards the use of a network of ontologies to relate different drug terminologies. It defines an ontology network where each actor potentially plugs in its own model as ontology. All ontologies are interconnected and mapped between them to share information. Fig. 2. Semantic Nomenclature ontology network [15] The main ontology in the network is the Nomenclature Reference Ontology. This OWL ontology acts as a bridge between the different application ontologies modeling drugs and domain ontologies, that support some other aspects of the model. The approach followed in this case if to map all the ontologies to the reference ontology. This design decision allows an easy access to the whole ontology network. The ontology is based on the study of several product definitions in other ontologies. Specifically, it follows partially the semantic model of SNOMED-CT as background knowledge, mainly from the Pharmaceutical/Biological product term used in that terminology, allowing the distinction between clinical and commercial drugs.
Fig. 3. Part of the semantic nomenclature reference ontology The main hierarchy is Pharmaceutical_Product, and the underlying Clinical_Drug, Prescription_Drug concepts are mapped to the equivalent concepts of the domain ontologies in the ontology network. On the other hand, Marketed_Drug concept is mapped to the equivalent concepts of the application ontologies which provide access to relevant product information of the pharmaceutical products marketed in Spain. Based on the hierarchy provided by SNOMED, the reference ontology defines a hierarchy that distinguishes between clinical drugs and branded drugs. The main generic concept is Pharmaceutical_Product. The Categorized_Product concept serves to classify the products according to their therapeutic use. In the next level, we define the Clinical_Drug, which can be mapped for instance to active ingredients. The Prescription_Drug concept could be described by the pharmaceutical form and dosage of the Pharmaceutical_Product, and as a matter of example is useful for prescription in hospitals. Finally, the Marketed_Product concept is used to describe branded or commercial drug, as they are dispensed in pharmacies. This last concept is defined by its national code, price, etc. The next figure depicts the hierarchy relation for pharmaceutical products in the reference ontology Pharmaceutical_ Product Categorized_ Product Clinical_Drug Prescription_ Drug Marketed_ Product Fig. 4. Drug-related concepts in the reference ontology
The rest of the ontologies are linked to the different entities of the reference ontology by means of mappings and axioms. 3 Discussion The NeOn approach is promising because it offers a complete and coherent set of tools, APIs and methodological support for ontology engineering and usage. The outcomes of the Semantic Nomenclature case study can be seen as a proof of concept of the NeOn approach applied to the semantic interoperability between drug terminologies. In this sense, the network of ontologies covered by the case study does not intend to be exhaustive or cover the whole set of information coming from external data sources. 3 Future work NeOn is now entering in its final year, which means that some of the results of the project are still not finalized. The true potential of NeOn is expected by summer 2009, when a new Manchester OWL Syntax version of the NeOn Toolkit will be released. In the scope of the Semantic Nomenclature case study, several experiments have been carried out in order to generate semi-automatically mappings using one of the NeOn plugins (the Alignment plugin). During the last year of the project we expect to verify the quality of these alignments and perform some more experiments with other NeOn Toolkit plugins. Besides, by the end of the first semester of 2009, the Semantic Nomenclature case study will release a Web application based on the underlying knowledge base (networked ontologies). This application will show the availability of developing semantic applications based on NeOn technology. 4 Conclusion This paper has given an account of the results of the NeOn project in respect to its usage in the e-health domain. The purpose of this paper was to show the different tools and methodology that NeOn puts at the disposal of the e-health community to model the domain. It was also shown that the NeOn approach towards the use of networked ontologies is clearly in line with some of the ongoing initiatives and roadmaps regarding semantic interoperability in e-health. In summary, the use of NeOn, specially the NeOn Toolkit, the methodology provided by the project, and the approach towards the use of a network of ontologies in order to bridge the gap between different drug terminologies could prove to be beneficial to the health informatics in general and the semantic interoperability in particular.
References [1] SNOMED-CT, http://www.ihtsdo.org/snomed-ct/ [2] NEHTA - National E-Health Transition Authority, http://www.nehta.gov.au/ [3] Dictionary of Medicines and Devices (dm + d), http://www.dmd.nhs.uk/ [4]Semantic Web Health Care and Life Sciences (HCLS) Interest Group, http://www.w3.org/2001/sw/hcls/ [5] SemanticHEALTH project, http://www.semantichealth.org/ [6] SemanticHEALTH partners. Semantic Interoperability Deployment and Research Roadmap. SemanticHEALTH SSA project Deliverable D7.1, 2008 [7] Rector A. Barriers, approaches and research priorities for integrating biomedical ontologies. SemanticHEALTH SSA project Deliverable D6.1, 2008. [8] Web Ontology Language (OWL), http://www.w3.org/2004/owl/ [9] NeOn Project, http://www.neon-project.org/ [10] Haase P, Rudolph S, Wang Y, Brockmans S, 2006. Networked Ontology Model. NeOn Deliverable D1.1.1 [11] Sabou M. et al, 2006. NeOn Requirements and Vision Deliverable [12] Waterfeld W, Erdmann M, Schweitzer T, Haase P. Specification of NeOn architecture and API V2. NeOn Deliverable D6.9.1, 2008 [13] NeOn Toolkit website http://www.neon-toolkit.org/ [14] Ontology Design Patterns, http://ontologydesignpatterns.org [15] Herrero G, Pariente T. Revision of ontologies for Semantic Nomenclature: pharmaceutical networked ontologies. NeOn Deliverable D8.3.2, 2008