Organizing RadLex lexicon for efficient retrieval of radiology documents

As more and more medical reports go online there is a move in the medical community to standardize the terminology used in these reports. The expectation is that uniformity of vocabulary can reduce variations and hence significantly aid the organization and retrieval of medical documents with modern search engines. Even medical specialties are beginning to realize the potential benefits and are getting together to develop standards for the individual specialties. In the specialty of Radiology, RadLex is the standardized lexicon developed by the Radiological Society of North America. A standardized term in RadLex is assigned one unique category from among: preferred name, synonym and abbreviation. The size of RadLex (over 12,000 terms) makes it difficult to recall a term and all its synonyms and abbreviations. From a usability perspective it is highly desirable to search with a single RadLex term and let the search engine transparently retrieve not only all the documents in which the term appears but also all the other documents created with related terms (synonyms and abbreviations) Towards that we propose an automata-based organization of RadLex lexicon to facilitate efficient retrieval of radiology documents with RadLex search terms. The automaton is used at index creation time to identify RadLex terms in a document and augment its indexes with additional terms related to those identified by the automaton. A distinguishing aspect of the approach is its computational efficiency and its broad applicability to medical Information Retrieval (IR) applications.