2005 Fiscal Year Annual Research Report
|Research Institution||Hokkaido University |
櫻井 恒太郎 北海道大学, 大学院・医学研究科, 教授
|Foreign Research Fellow
JIANG G. 北海道大学, 大学院・医学研究科, 外国人特別研究員
|Keywords||Cross-Language Information Retrieval / Patient Records / Formal Concept Analysis / UMLS / Consumer Health Vocabulary (CHV) / Consumer Friendly Display (CFD) / Electronic Nursing Records / Clinical Terminology|
Development of a Cross-Language Information Retrieval System in Clinical Domains Using Formal Concept Analysis
Purpose : To develop and evaluate a Japanese-English cross-language information retrieval (CLIR) system using the formal concept analysis (FCA). Methodology : The system was constructed using Protege-2000, an ontology edit environment developed by Stanford Medical Informatics. A collection of discharge summaries in cardiovascular domain was mapped into Protege-2000 as the virtual patient records. And the PubMed was used as the target online information source. A bilingual ontology was constructed by 1) mapping the MSHJPN source in the UMLS2005AA with their corresponding English MeSH Heading terms, and 2) mapping the diagnostic and procedural terms of MEDIS with the UMLS concepts. A concept-oriented view generation system using formal concept analysis, developed as a Protege-2000 tab plug-in in our previous study, was integrated with the bilingual ontology and used as the user i
nterface. Conclusions : The system could automate the cross-language query formation for PubMed within the context possibly reflecting the clinicians' information needs.
Development of Consumer Health Vocabulary Using Nursing Records : A Pilot Study
Purpose : To develop and evaluate an approach for CHV development through using the electronic nursing records.
Methods : A corpus was constructed using the free texts collected from two inquiry items of 16,717 patients, including "The Hardest Symptom" and "The Understanding about the Disease", in the electronic nursing records in Hokkaido University Hospital. Two controlled medical vocabularies, i.e. the MSHJPN from UMLS 2005AA and the MEDIS diagnostic terms were used for identifying the medical terms from the corpus. The identified medical terms with their frequency in corpus, were linked to their concept identifiers i.e. a CUI for a MeSH term or an Interchange Code (IC) in MEDIS, and subsequently to their synonyms. A subset (n= 100) randomly selected from the identified concepts with the frequency above 10 was reviewed by 4 graduate students in Department of Medical Informatics to identify the "consumer-friendly display"(CFD) names for each concept. Results&Conclusion : The corpus-based approach using the electronic nursing records in the pilot study was demonstrated as an effect way for CHV development. Less
Research Products (4 results)