The primary goal of the original digitised collection was to support systematic searching of findings in support of teaching and diagnosis. To this end, it captured natural language representations of findings and their localisation. However, the radiological language used to record findings is complex and esoteric: it does not easily lend itself to deconstruction, and non-radiological clinicians find it difficult to understand. From a machine perspective it is also ambiguous. In its original form, REAMS also suffered from a number of technical restrictions. Primary amongst these was that its valuable knowledge base was available only through a single application which utilised a static, legacy database.

Over the past two years we have been working on a new representation of the knowledge base behind REAMS; one which enhances the radiological interpretation of radiographs. It does this by capturing aspects of the prior knowledge necessary for understanding those interpretations. This has involved developing a rich knowledge base capable of classifying, deconstructing and interrelating findings. One of the strengths of the REAMS collection is that all findings are related to radiographs within the knowledge base. The collection of images can therefore act as a reference collection for exemplifying radiological terms. Equally importantly, the technology used offers substantial advantages. It is designed to free the latent knowledge from any specific application or platform, providing a suitable base for the development of innovative applications and allowing straightforward linking with other similarly constructed knowledge bases. This is the vision for the ‘Semantic Web’: where the internet becomes a true knowledge resource, encouraging the reuse of encoded knowledge.

The radiological phenotype ontology for skeletal dysplasias is designed to remove ambiguity from findings. This is not just between practitioners, but at the level of machine interpretation. This is of paramount importance in enabling future bioinformatic applications which integrate dREAMS information with other resources on the semantic web. The knowledge base will be made web accessible, with a SPARQL interface which facilitates integration with other bioinformatic resources. We are working towards its integration with genotypic and scientific information related to skeletal dysplasias, developing a significant resource for translational research on skeletal dysplasias.