Medical imaging report indexing: enrichment of index through an algorithm of spreading over a lexico-semantic network

Proceedings of Recent Advances in Natural Language Processing, pages 338–345, Hissar, Bulgaria, Sep 7–9 2015. Medical Imaging Report Indexing: Enrichment of Index through an Algorithm of Spreading over a Lexico-semantic Network Mathieu Lafourcade LIRMM University of Montpellier France mathieu.lafourcade@lirmm.fr Lionel Ramadier IMAIOS 34090 Montpellier France lionel.ramadier@lirmm.fr Abstract In medical imaging domain, digitized da- ta is rapidly expanding Therefore it is of major interest for radiologists to be able to do an efficient and accurate extraction of imaging and clinical data (radiology reports) which are essential for a rigorous diagnosis and for a better management of patients. In daily practice, radiology re- ports are written using a non- standardized language which is often ambiguous and noisy. The queries of ra- diological images can be greatly facilitat- ed through textual indexing of associated reports. In order to improve the quality of the analysis of such reports, it is desir- able to specify an index enlargement al- gorithm based on spreading activations over a general lexical-semantic network. In this paper, we present such an algo- rithm along with its qualitative evalua- tion. 1 Introduction Widespread digitalization in the health care sector and the implementation of customized electronic medical record result in a rapid in- crease in the volume of digital medical data. The medical computer systems allow to archive many and varied information (for example medical record, results of medical analyses, X-rays and radiological reports…). Thus, these data are ac- cessible, either to be completed and compared with new results or to adapt the management of patients, or to provide decision support to im- prove the quality of care. The ability to have eas- ily and efficiently access to these medical data has become a primary objective for health pro- fessionals. Thus, a proper indexing of medical reports (surgical, radiological...) optimizes the search for information, not only in a clinical pur- pose, but also educational. Dinh and al., (2010) then realized a semantic indexing of patients' medical records in order to make it support for some information search procedures. Their in- dexing method uses MeSH (Medical Subject Headings), involves disambiguation, the extrac- tion of clinical values, and weighting of con- cepts. Pouliquen, (2002) also performed an au- tomatic indexation through recognition and ex- traction of medical concepts. He took into ac- count the compound words and word associa- tions to convert a sentence in reference words with the help of a medical thesaurus. In the field of the medical imaging, the quanti- ty of images and reports increases so much that being able to find quickly and easily the infor- mation becomes a major stake. But take full ad- vantage of such a collection of radiological im- ages means being able to quickly identify rele- vant information and requires that they are properly indexed from their reports. To be effec- tive and useful for practitioners, indexing must consider their requests. Several authors, includ- ing Hersh et al., (2001) and Huang et al., (2003) automatically indexed radiological reports using the UMLS metathesaurus. To improve …

Content truncated to 500 words