Liqin Wang MS, Abstract Co-Author: Nothing to Disclose

Scott L. Du Vall PhD, Presenter: Partner, Clinical Methods LLC

Our study shows the importance of using RadLex in concert with other UMLS source terminologies. Continued work will inform of the domains and concept types needing further development in RadLex.

RadLex is a terminology developed to unify and supplement the many existing terminologies that are used in imaging. RadLex fills the gaps for concepts used only in imaging: certain medical devices, image acquisition techniques, and some fine anatomy. This presentation represents the initial results in a project to compare the language commonly used by clinicians in imaging reports to RadLex terms across imaging domains at institutions across the nation.

The Clinical Text Analysis and Knowledge Extraction System (cTAKES), a natural language processing (NLP) tool developed at Mayo Clinic, was used to identify 176,091 noun phrases from 47,555 de-identified radiology reports obtained from the University of Pittsburgh NLP Repository. Two methods were used to map these phrases to RadLex: 1) exact mapping, where the words matched before or after removing articles and prepositions; 2) partial mapping, where a phrase was mapped if the majority of words in it were contained in a RadLex term. MetaMap was used to map phrases to UMLS. Only 2.32% of phrases mapped exactly and 50.53% of phrases mapped partially to RadLex. 10.40% and 85.95% of phrases were exactly and partially mapped to UMLS, respectively.

Imaging studies, though they require an extended vocabulary to other clinical domains, are not absent from gross anatomy, diagnoses, and other clinical concepts that can be found in SNOMED CT or CPT. Terms not mapping to RadLex but mapping to UMLS included qualitative/quantitative/spatial/functional/temporal concepts, anatomy, findings, therapeutic/preventative/diagnostic procedures, and medical device. Terms mapping to RadLex but not mapping to UMLS mostly pertained to the types of modifier, imaging observation and characteristic, imaging procedure and attributes, and anatomy. As the study continues to incorporate reports from other medical centers, mapping methods will be updated to use synonyms and acronyms and to better exclude phrases used in reports outside the medical domain.

Wang, L, Du Vall, S, Assessing the Ability of RadLex to Represent the Common Clinical Language in Imaging Reports.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034454.html