Biomedical term normalization of EHRs with UMLS

Biomedical term normalization of EHRs with UMLS Naiara Perez-Miguel, Montse Cuadros, German Rigau HSLT group at Vicomtech, IXA group at UPV/EHU Donostia/San Sebasti´an, Spain {nperez,mcuadros}@vicomtech.org, german.rigau@ehu.es Abstract This paper presents a novel prototype for biomedical term normalization of electronic health record excerpts with the Unified Medical Language System (UMLS) Metathesaurus, a large, multi-lingual compendium of biomedical and health-related terminologies. Despite the prototype being multilingual and cross-lingual by design, we first focus on processing clinical text in Spanish because there is no existing tool for this language and for this specific purpose. The tool is based on Apache LuceneTM to index the Metathesaurus and generate mapping candidates from input text. It uses the IXA pipeline for basic language processing and resolves lexical ambiguities with the UKB toolkit. It has been evaluated by measuring its agreement with MetaMap –a mature software to discover UMLS concepts in English texts– in two English-Spanish parallel corpora. In addition, we present a web-based interface for the tool. Keywords: term normalization, UMLS, information extraction, biomedical text 1. Introduction Biomedical text mining technologies are becoming a key tool for the efficient exploitation of information contained in unstructured data repositories, including scientific litera- ture, Electronic Health Records (EHRs), patents, biobank metadata, clinical trials and social media. Natural Lan- guage Processing (NLP) and specifically Information Ex- traction (IE) tools, such as term normalization tools, can fa- cilitate knowledge discovery, exchange, and reuse by find- ing relevant terms and semantic structure in those texts. This paper presents a preliminary application that enriches EHRs with links to the Unified Medical Language System (UMLS)1, a multilingual repository of biomedical termi- nologies. The tool is multilingual and cross-lingual by de- sign, but we first focus on Spanish EHR processing be- cause there is no existing tool for this language and for this specific purpose. We propose a sequential pipeline that retrieves mapping candidates from an indexed UMLS Metathesaurus, uses the IXA pipeline (Agerri et al., 2014) for basic language processing and UKB (Agirre and Soroa, 2009) for word sense disambiguation (WSD). In addition to the pipeline itself, this paper also presents a demonstration interface for the tool that will be available on-line2. 2. Related Work Biomedical term normalization is a long-established re- search field in English-speaking countries where termi- nological resources and basic-processing tools for the biomedical domain and this language have been available for decades. Thus, there already exist several mature ap- plications that are being effectively exploited for different purposes and by different organizations as of today. In what follows, we present some of the better-known applications. MetaMap (Aronson, 2001; Aronson, 2006) enriches biomedical text with links to the UMLS Metathesaurus. It is “knowledge intensive” as it relies heavily on the SPE- CIALIST Lexicon, a large syntactic lexicon of biomedical and general English. Meystre and Haug (2005) evaluated 1https://www.nlm.nih.gov/research/umls/ 2http://demos-v2.vicomtech.org/umlsmapper/, user:vicomtech, password:umlsmapper MetaMap with 160 clinical documents of diverse nature (radiology reports, exam reports, and so on). MetaMap’s results were compared to annotations by 8 physicians; the reported precision and recall for …

Content truncated to 500 words