Reflections on a medical ontology

https://doi.org/10.1016/j.ijhcs.2007.02.005Get rights and content

Abstract

In this paper we confront the divide between the ontologies developed from the requirement of comprehensive and general domain coverage and those devised to meet application-specific requirements. While the generalists typically attach philosophical sophistication to their approach, in supposed contrast to the narrow remit chosen by the application-bound knowledge engineers, we would like to indicate that the latter practice can often reflect a multi-faceted rationale, nuanced by the requirements of the domain.

We demonstrate how the necessity of placing ontology-based systems with the work-practices of domain experts introduces unique demands on design rationales and enforces, often implicitly, a philosophical assessment of the necessary concepts and relations that balance the generality and specificity. Such demands are not addressed by generic approaches to modelling the reality of a domain. Indeed, we articulate the philosophical and practical considerations that we have taken into account when developing an application-specific ontology. We would certainly hope that our experiences can be of help to the development of ontologies in similar applications.

Introduction

Medical ontologies classify and encode relationships between nomenclature and/or concepts invoked in medical procedures and operations. These foster a mutual understanding between human experts with different backgrounds as well as between human and software agents and between software agents alone. They are, thus, well suited to provide knowledge-level support for describing and putting together information-intensive applications in medicine; hence, a wide range of efforts are directed at designing medical ontologies. Among this wide spectrum of approaches which differ in complexity and scale, we can broadly distinguish two groups of approaches to medical ontology development: the general purpose ones that are application-independent (Hahn and Schulz, 2004, Smith et al., 2005), and those set up to drive a particular application. Examples of the former include the efforts towards controlled vocabulary in medical domain such as UMLS Metathesaurus,1 SNOMED,2 MeSH,3 GALEN,4 etc. The universal reach of these efforts requires substantial resources and extensive collaborative efforts, and often leaves a trail of philosophical commentary regarding the degree to which they can be transparent representations of reality (Smith, 2004). In contrast, many of the applications built in short-term projects in the second category tend to develop medical ontologies which appear to be ad hoc to some degree (Abu-Hanna et al., 2005), e.g. the Brazilian National Health Card Ontology,5 the Drug Ontology Project for Elsevier,6 Domain Ontology of EORCA project,7 etc.

Medical imaging with advanced knowledge technologies (MIAKT) was a 2-year project with the aim of providing a prototype for advanced knowledge services to breast cancer screening procedure and breast cancer triple assessment (TA). Breast cancer is one of the leading causes of cancer death among women in the US (Patlak et al., 2001) and is the most common cancer for women in the UK.8 Diagnosis of breast cancer normally involves multi-disciplinary meetings with experts from different medical backgrounds, e.g. radiologists, surgeons, oncologists, histologists, and other clinical staff. As with other disciplines, we expect considerable variability among experts. This provides the motivation for standardising the vocabulary used in the breast cancer screening and diagnosis process. In the context of MIAKT, we have faced problems and difficulties that, we imagine, might be faced by most other ontology developers working on projects of similar size. In this paper, we try to clarify some of the conceptual underpinnings of our approach to ontology building, outlining our attempts to understand the context of application which determines the modelling choices we have made. We focus on examples from our Breast Cancer Imaging Ontology (BCIO) to illustrate essential questions that often need addressing when developing ontologies. Despite its roots in the specific application we have developed, the design of ontologies is meant to facilitate application independence. Hence, we set out the rationale by which we extricate the “ontological core” from the application context and indicate how we tailor the ontology to meet the application-specific requirements.

Section snippets

Setting the baseline for conceptual reflection

The problem we faced was to build a computer-aided patient data system to support routine breast cancer screening sessions, and subsequent patient management meetings. For this purpose, we needed to provide a framework and communication interface for the mutual understanding of clinicians from different backgrounds. We needed to make information accessible and understandable by computers in order to provide paperless storage and remote access and implement procedures for intelligent retrieval

Scrutinising the domain of discourse

Rooted in philosophy, Ontology is the study of existence (i.e. of what is). The recent blossoming of ontology-related research in computer science, however, takes a more practical perspective focussing on how to capture and codify the essence of beings and how to share the codified knowledge. A good example of such a trend can be witnessed in the artificial intelligence (AI) community wherein researchers tend to deliberately blur the differences among different uses of the word “ontology” and

Breast cancer imaging ontology

Breast cancer diagnosis relies heavily on visual and tactile information, e.g. various medical images, results of physical examination, which cannot be accurately and comprehensively represented using mathematical or logical models. Such information can be abstracted either using an approximate model or leaving a “semantic gap” between the source of information and the representations and requesting human interpreters to fill the gap with their experience and wisdom. We expect that a

Discussions

Nowadays, after the explosion of information available from the internet, we have witnessed an explosion of so-called ontologies made public by all sorts of ontology developers. Their “products” are quite diverse in nature ranging from comprehensive ones, e.g. Cyc14 and foundational ones, e.g. IEEE SUMO,15 DOLCE16 to application-specific ones such as BCIO. Despite the number of available ontologies, people working

Acknowledgements

Research for this paper was funded by the British Engineering and Physical Sciences Research Council (EPSRC) under the MIAKT Grant GR/R85150/01 and under the AKT IRC Grant GR/N15764.

References (33)

  • A. Abu-Hanna et al.

    Protégé as a Vehicle for Developing Medical Terminological Systems

    Human-Computer Studies

    (2005)
  • E. Morris

    Illustrated breast MR lexicon

    Seminars of Roentgenology

    (2001)
  • C. Rosse et al.

    A reference ontology for biomedical informatics: the foundational model of anatomy

    Journal of Biomedical Informatics

    (2003)
  • Baader, F., Calvanese, D., McGuinness, D., Nardi, D., Patel-Schneider, P. (Eds.), 2003. The Description Logic Handbook:...
  • Bernaras, A., Laresgoiti, I., Corera, J., 1996. Building and reusing ontologies for electrical network applications....
  • BI-RADS, 1998. Breast Imaging Reporting and Data System: BI-RADS, third ed., American College of...
  • Bodenreider, O., Smith, B., Burgun, A., 2004. The ontology–epistemology divide: a case study in medical terminology....
  • Brewster, C., Alani, H., Dasmahapatra, S., Wilks, Y., 2004. Data driven ontology evaluation. In: Proceedings of...
  • Dasmahapatra, S., O’Hara, K., 2005. Interpretations of ontologies for breast cancer. In: Proceedings of European...
  • Degen, W., Heller, B., Herre, H., Smith, B., 2001. GOL: towards an axiomatized upper-level ontology. In: Proceedings of...
  • Fernández, M., Gómez-Pérez, A., Juristo, N., 1997. METHONTOLOGY: from ontological art towards ontological engineering....
  • J. Fodor

    A Theory of Content

    (1990)
  • Genesereth, M., Nilsson, N. (Eds.), 1987. Logical Foundations of Artificial Intelligence. Morgan Kaufmann, Los Altos,...
  • Hahn, U., Schulz, S., 2004. Building a very large ontology from medical thesauri. In: Staab, S., Studer, R. (Eds.),...
  • Hu, B., Kalfoglou, Y., Alani, H., Dupplaw, D., Lewis, P., Shadbolt, N., 2006. Semantic metrics. In: Proceedings of the...
  • D. Ikeda et al.

    Development, standardization, and testing of a lexicon for reporting contrast-enhanced breast magnetic resonance imaging studies

    Journal of Magnetic Resonance Imaging

    (2001)
  • Cited by (22)

    View full text