Identifying anatomical concepts associated withICD10 diseases

Terminology and Artificial Intelligence 2009.

计算机科学;

Fleur Mougin1 ; Olivier Bodenreider2 et Anita Burgun3 ; 2National Library of Medicine ; Bethesda ; Maryland ; USA ; 3INSERM U936 ; EA3888 ; School of Medicine ; University of Rennes 1 ; IFR 140 ; France

Others  :  http://ceur-ws.org/Vol-578/paper5.pdf
PID  :  42741

学科分类：计算机科学（综合）

来源: CEUR

【 摘 要 】

Unlike recent biomedical terminologies, the International Classification of Diseases (ICD) does not state any explicit associationsbetween a given disease and the corresponding anatomical structure(s). As aconsequence, clinical repositories coded with ICD cannot be searched byanatomical structure. The objective of this work is to find associations betweendiseases from ICD10 and anatomical structures. Toward this end, weinvestigated three approaches (symbolic, lexical, and statistical) which exploitvarious features of the Unified Medical Language System (UMLS). Weevaluated these approaches according to i) the consistency of resultinganatomical concepts with the high-level anatomical concept(s) identified for thechapter in which the disease is listed; and ii) the validity of resulting anatomicalconcepts assessed manually. We show that the symbolic approach is both themost productive and the most accurate approach.