Cataloguing Spanish Medical Reports with UMLS Terms

UMLS (Unified Medical Language System) is one of the most comprehensive terminological resources for the medical domain. Thus, the provision of instruments to assist in the cataloguing of medical reports with UMLS is of utmost relevance, especially when these reports are written in unstructured free-text natural language. For this purpose, it is possible to use tools that, like MetaMap, enable the automatic annotation of clinical texts with UMLS terms. However, these tools typically work on reports written in English, which seriously hinders their applicability to other languages. In this paper, we describe an approach to mitigate these shortcomings, which pipelines state-of-the-art language translation services with automatic mapping tools. We demonstrate the feasibility of the approach by combining Google Translate with MetaMap and by using the resulting pipeline to catalog, with UMLS, a representative set of Spanish-written X-Ray Thorax reports corresponding to images taken from the Indiana Chest X-ray radiology corpus. The resulting cataloguing is not significantly different, in quality, from that obtained through the direct application of MetaMap on a similar set of reports written in English and selected from this Indiana Chest X-ray corpus.