> Semantic Web Use Cases and Case Studies
February 2007
The obvious benefits of radiological procedures often come with harmful side effects and high costs. Common misusage of radiology includes investigations that are unnecessary or too frequently, investigations that result in little or no impact on patient management, and wrong investigations. In each case, the patient may be subjected unnecessarily to irradiation and scarce medical resources are wasted.
Clinical guidelines and referral criteria have been established to assist clinicians in determining the appropriate radiological procedures for specific clinical circumstances. For this very purpose, “The Radiation Protection 118 — Referral Guideline for Imaging”1 is published by the European Union and a series of Appropriateness Criteria2 are published by American College of Radiology.
Significant effort has been devoted to representing such guidelines and protocols in a machine-executable and interoperable format3. However, the adaptation of these frameworks has not moved beyond their development community This is because the high cost of translating the text-based guidelines to a machine-executable format and the requirement for a proprietary guideline execution engine have hindered the wider adaptation of existing computerized guideline systems, and the reusability of computerized medical knowledge. Furthermore, the latency of updating guidelines to reflect new modalities and radiology protocols reduces the usability and credibility of the published guidelines and protocols.
Semantic web technology provides an open, standards-based, computer-interpretable and executable framework on which the clinical guidelines and protocols can be published. In such an environment, we do not assume all knowledge needed for clinical decision support systems are contained in one proprietary repository, and in a non-standardized format. Instead, we envisage a web of connected knowledge bases; with each piece representing a type of knowledge that is developed and maintained by appropriate bodies. Figure 1 illustrates the key pieces of knowledge needed for a radiological order entry system. Elements of general clinical knowledge are shown in the upper section of the diagram, and the localization rules are shown in the lower section of the diagram.
We use Resource Description Framework (RDF)4 and Web Ontology Language (OWL)5 to transform the Appropriateness Criteria2 into a machine-readable medical knowledge base using ontologies. The ontology (radGuideline) we use associates clinical problems with preferred investigations, and captures in rules the recommendations and comments that contain the imperative wisdom of the medical community. RadGuideline enables the reuse of existing medical knowledge by linking this specialized information with other medical ontologies such as SNOMED CT6, ACR Codea7 and some OBO ontologies8, which enables a standard description of clinical problems and the patient condition.
Figure 1: Web of Knowledge for Radiology Order Entry System
Semantic web rule engines such as CWM9 and Euler10 are able to generate recommendations for radiology orders and proofs based on the given patient conditions (medical history and physical condition) and a set of rules derived from the Referral Guidelines1 or Appropriateness Criteria2.
The true potential of this approach is its ability to integrate cross-domain knowledge and data seamlessly on explicit and unambiguous terms expressed in ontologies. The explanations generated by proof engines provide evidence to clinicians for a decision. The approach can even provide alternative solution. The final decision is still in the hands of a clinician, but making such key information and evidence readily available is extremely important when there are such large volumes of data to consider. This approach can help to prevent medical errors that are caused by physicians overlooking vital facts.
© Copyright 2007, Agfa Healthcare