Charles E. Kahn MD, MS, Presenter: Shareholder, Hotlight Inc Officer, Hotlight Inc

Gerald Cameron MS, Abstract Co-Author: Nothing to Disclose

A semi-automated system has been created to translate radiology reporting templates into non-English languages. The availability of templates in a variety of languages has helped encourage interest in structured reporting, and provides an approach to offer reporting templates that allow unified encoding of report data without the barrier of language.

RSNA's radiology report template library (www.radreport.org) offers the option of browsing and selecting reporting templates by language. We developed a semi-automated system to help convert reporting templates from English into any of 58 supported languages.

Authorized editors select a reporting template and target language. Our system extracts reporting elements and text strings from the template and passes them through the Google Translate application programming interface (API) to render provisional translations. The editor then can approve or correct the provided translations. For select languages, the Google Virtual Keyboard helps editors include non-English characters during the translation process. When the editor has completed review, the translated template is saved and can be available immediately for download as an XML document. The original English-language terms are included contextually as annotations within the translated template file. The translated template preserves all of the mappings to ontologies (knowledge models) such as RadLex®, Systematized Nomenclature or Medicine Clinical Terms (SNOMED CT®), and Logical Observation Identifiers Names and Codes (LOINC®). In this way, data can be extracted from the coded data within a structured report regardless of the language in which it is created.

Template editors have experienced variable results with the automated translations, largely due to the specialized terminology of medical practice and radiology. In Turkish, more than 70% of the machine-translated terms have been correct. With other languages, fewer than 50% of provisional translations have been correct. Despite its limited accuracy, editors have found the provisional translations useful. To date, Chinese, Turkish, and Arabic templates have been created using the translation system.

Kahn, C, Cameron, G, A Semi-automated Translation System for Radiology Reporting Templates.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12022655.html