Ontology-based Archetype Interoperability and Management Catalina Martínez-Costa, Marcos Menárguez-Tortosa, J. T. Fernández-Breis Departamento de Informática y Sistemas, Facultad de Informática Universidad de Murcia, CP 30100, Murcia {cmartinezcosta, marcos, jfernand}@um.es Abstract. Semantic interoperability of clinical standards is a major challenge in the ehealth across Europe, because this would allow healthcare professionals to manage the complete EHR of patient. Archetypes are considered a cornerstone to deliver fully interoperable EHRs. Our work is focused on the development of ontology-based methods and techniques for providing semantic interoperability between dierent EHR standards at archetype level. Hence, solutions for the semantic representation, transformation and management of clinical archetypes are described in this work. 1 Introduction The lifelong clinical information of any person supported by electronic means congures his Electronic Healthcare Record (EHR). Nowadays there are dierent advanced standards and architectures [1] for representing and communicating EHRs, such as HL7 [2], OpenEHR [3] and UNE-EN 13606 [4]. Some of these advanced EHR standards, such as OpenEHR and UNE-EN 13606 make use of the dual model architecture approach [5]. This architecture is based on two modelling levels: information and knowledge. The information level is provided by the reference model and the knowledge level by the archetype model. Archetypes dene clinical concepts and are usually built by domain experts. They are a tool for building clinical consensus in a consistent way. The semantic interoperability of clinical standards is a major challenge in the ehealth across Europe, because this would allow healthcare professionals to manage the complete EHR of patient. Clinical archetypes are fundametal for the consecution of semantic interoperability and they are built for particular EHR standards. Our recent work has focused on the development of methods and techniques for providing semantic interoperability between dierent EHR standards at archetype level. First, a methodology for obtaining a semantic representation of archetypes will be presented, describing how syntactic archetypes can automatically be transformed in semantic ones. Next, the semantic interoperability of two dual-model based standards UNE-EN 13606 and OpenEHR will be addressed. Finally, the development of a prototype for the semantic management of clinical archetypes will be described.
2 An Ontological Representation of Clinical Archetypes Clinical archetypes are dened using the Archetype Denition Language (ADL). This is a generic language that does not allow to perform any semantic activity over archetypes. Nevertheless, these activities require the exploitation of information and knowledge: comparisons, classication, integration of information and knowledge coming from dierent, heterogeneous systems. In this way, it can be stated that such activities are knowledge intensive, they require for the semantic management of knowledge and information, and for semantic interoperability between such heterogeneous systems. The advances in the Semantic Web [6] community make it a candidate technology for supporting such knowledge-intensive tasks related to archetypes and EHR systems. This section aims at providing a mechanism for representing archetypes in a Semantic Web-manageable manner. 2.1 An ontological infrastructure for representing clinical archetypes The use of ontologies to represent biomedical knowledge is not new, since ontologies have been widely used in biomedical domains for the last years with dierent purposes [7,8]. For representing clinical information semantics, we have used ontologies modeled by the Web Ontology Language (OWL) [9]. The representation of archetypes in OWL requires the semantic interpretation of clinical archetypes, so both the UNE-EN 13606 and OpenEHR reference and archetype models were analyzed [10]. As result of this process, two main ontologies were built for each standard (see Table 1 for details). The CEN-SP and OpenEHR-SP ontologies represent the clinical data structures and datatypes dened in the corresponding standards, and CEN-AR and OpenEHR-AR that are archetype model ontologies. These ones are available online at http://klt.inf.um.es/~cati/ontologies.htm Ontology Classes DP OP Restrictions CEN-SP 68 16 92 227 CEN-AR 122 76 142 462 OpenEHR-SP 87 14 156 302 OpenEHR-AR 144 75 210 524 Table 1. Details of the OWL ontologies, in terms of classes, dataproperties (DP), objectproperties (OP) and restrictions. 2.2 Transforming ADL archetypes into OWL Once the ontological representation has been provided, to obtain archetypes conforming to this representation from the ADL ones a transformation process is needed. Our technical solution involves three technical spaces (TS) [11]. The Grammar TS to which ADL belongs, the Semantic Web TS to which ontologies belong, and the Model Driven Engineering (MDE) TS is the pivotal TS
in which the transformation takes place. Thus, the transformation process is divided in the following three phases described below and shown in Figure 1. A web tool for transforming ADL archetypes into OWL is available online at http://klt.inf.um.es/~cati/ [12]. Phase One: ADL archetypes are transformed into models according to the Archetype Object Model (AOM). The AOM representation is common to any dual-model EHR standard. At this phase (left side of Figure 1) a change of TS is carried out, from Grammar TS to MDE TS. Archetypes in ADL are processed and serialized in XML by an ADL parser [13]. The AOM metamodel is obtained from its XML Schema and XML archetypes are transformed into models by using EMF [14]. Phase Two: AOM models are transformed into models according to the ontological structure modeled before. The ODM standard [15] denes the representation of OWL ontologies in MDE TS. Protégé [16] implements the transformation from OWL to MDE TS and was used to get the archetypes metamodels (OWL-AR) from the OWL Archetype ontologies. Next, models that conforms the AOM metamodel are transformed into OWL-AR models. This transformation is implemented in RubyTL [17], a model transformation language, and is standard specic. Phase Three: Finally, OWL les are obtained from the ontological models. Its is shown in the right side of the gure. In this phase we move from MDE TS to the Semantic Web TS. To get OWL instances from ontology models, a model-to-text transformation language, MOFScript [18] has been used. Fig. 1. The ADL to OWL transformation process 3 Towards UNE-EN 13606 and OpenEHR archetype-based semantic interoperability Both clinical standards, UNE-EN 13606 and OpenEHR follow the dual model approach. However, they dier in how they structure the EHR domain, that is, they dene dierent reference models. OpenEHR oers a full specication for
the creation, storage, maintenance and querying of EHRs. On the other hand, the UNE-EN 13606 standard was developed to act as an EHR Extract exchange standard, so it does not provide proper version management, workow management, interfaces to other systems, etc. It provides the necessary requirements related to moving pieces of the EHR from one system to another. In summary, OpenEHR allows for dening data more precisely due to its richer data structures and data types. To transform OpenEHR archetypes into UNE-EN 13606 and vice-versa, a similar technological solution to the adopted in the ADL to OWL trasnformation has been used, this allowing us to reuse some previous work. The transformation mechanism starts with the ontological representation of the archetype obtained as explained in Section 2. An integrated ontology has been developed to be used in this process and correspondences with the specic standards and the integrated one have been established. Figure 2 describes the schema of our solution. There, solid arrows represent the correspondences between the three ontologies and dashed arrows shows the possible archetype transformations. The transformations carried out between the three ontologies, take place at the MDE TS, but this has been omitted in the gure for the sake of simplicity. At present, the methodology proposed has been applied to the OpenEHR to UNE-EN 13606 transformation, but in the near future it will be also applied to the opposite transformation. Fig. 2. Ontology-based archetype transformation process 4 Clinical archetypes management Concerning archetype management, a semantic system called Archetype Management System (ArchMS) [19] has been designed. The objective of the system is to support the execution of clinical, semantics activities over archetypes. ArchMS is built on the idea of a virtual archetype repository for dual-model based EHR standards, whose basic unit is the archetype. It is capable of working with UNE- EN 13606 and OpenEHR archetypes, providing the same functionality for both standards. These ones can be semantically annotated in the system, so that these metadata can be used to support semantic searches and comparisons.
Annotating in ArchMS allows for adding semantic metadata to archetypes. This semantic metadata can be associated to a complete archetype or a term of it, allowing them for annotations with dierent granularity. To dene an annotation a classier resource is needed. This has to be an OWL ontology and can be a domain ontology, a terminology and so on. Any OWL resource can be used for this purpose (see Figure 3). Fig. 3. Annotating in ArchMS Semantic annotations allow for searching archetypes holding some specic properties, that is, exploiting the repository in the dierent existing dimensions. The system also allows for suggesting annotations for new archetypes, since searches for similar archetypes can be found. Search mechanisms make use of semantic similarity functions for this purpose. In general, two main searches can be performed: for similar archetypes, and for archetypes holding some properties. The global similarity looks for archetypes similar to a given one by doing semantic comparisons in the context of the archetype ontology available for the particular standard. Archetypes are instances of that ontology, so that instances comparison mechanisms are used. These mechanisms would take into account the following categories: conceptual proximity, property similarity, annotations similarity and linguistic proximity. On the other hand, user can search archetypes holding some properties. These can be either denitional or annotations properties. On the one hand, we might be looking for archetypes written in English, or archetypes including an element measured in a certain unit. On the other hand, we might be looking for archetypes related to a particular disease, being such associations being established through a classier resource. 5 Conclusions Providing an OWL representation for archetypes allows semantic activities such as comparison, classication, selection or consistency checking to be carried out
more eciently. Here, an overview of our work towards semantic interoperability between archetypes has been presented. The rst step was the design and implementation of a methodology for the transformation of ADL archetypes into semantic archetypes expressed in OWL. This methodology has been applied to two dual-model based EHR clinical standards: UNE-EN 13606 and OpenEHR. After that, a similar technological solution has been applied for transforming OpenEHR archetypes into UNE-EN 13606 archetypes and vice-versa. Finally, the ArchMS system for annotating archetypes and to perform dierent types of semantic searches has been presented. References 1. Blobel, B.: Advanced ehr architecturespromises or reality. Methods of Information in Medicine 45(1) (2006) 95101 2. HL7: http://www.hl7.org 3. OpenEHR: http://www.openehr.org 4. UNE-EN13606: http://www.centc251.org 5. Beale, T.: Archetypes and the ehr. Stud Health Technol Inform 96 (2003) 238244 6. Berners-Lee, T., Hendler, J., Lassila, O.: The semantic web. Scientic American May (2001) 2937 7. Schulz, S., Hahn, U.: Part-whole representation and reasoning in formal biomedical ontologies. Articial Intelligence in Medicine 34(3) (2005) 179200 8. Smith, B.: From concepts to clinical reality: An essay on the benchmarking of biomedical terminologies. Journal of Biomedical Informatics 39(3) (2006) 288298 9. OWL-REF: http://www.w3.org/tr/owl-ref/ 10. Fernandez-Breis, J., Vivancos-Vicente, P., Menarguez-Tortosa, M., Moner, D., Maldonado, J., Valencia-Garcia, R., Miranda-Mena, T.: Using semantic technologies to promote interoperability between electronic healthcare records' information models. In: Proc. 28th Annual International Conference of the IEEE Engineering in Medicine and Biology Society EMBS '06. (Aug. 30 2006Sept. 3 2006) 26142617 11. Kurtev, I., Bezivin, J., Aksit, M.: Technological spaces: an initial appraisal. In: CoopIS, DOA'2002 Federated Conferences, Industrial track,. (2003) 12. Martinez-Costa, C., Menarguez-Tortosa, M., Fernandez-Breis, J., Maldonado, J.: A model-driven approach for representing clinical archetypes for semantic web environments. J. of Biomedical Informatics 42(1) (2009) 150164 13. ADL-Parser: http://www.openehr.org/svn/ref_impl_java/trunk/adlparser/ 14. EMF: http://www.eclipse.org/emf/ 15. OMG: Ontology metamodel denition specication. http://www.omg.org/cgibin/doc?ad/2006-05-01.pdf (2006) 16. Protege: http://protege.stanford.edu/ 17. Sanchez-Cuadrado, J., Garcia-Molina, J., Menarguez-Tortosa, M.: Rubytl: A practical, extensible transformation language. In: ECMDA-FA. (2006) 158172 18. MOFScript: http://www.eclipse.org/gmt/mofscript/ 19. Fernandez-Breis, J.T., Menarguez-Tortosa, M., Martinez-Costa, C., Fernandez- Breis, E., Herrero-Sempere, J., Moner, D., Sanchez, J., Valencia-Garcia, R., Robles, M.: A semantic web-based system for managing clinical archetypes. Conf Proc IEEE Eng Med Biol Soc 2008 (2008) 14821485