学会ホームページにおける類似症例検索システムの実運用と評価

摘要

Clinical case reports are very important resources for a clinician to retrieve similar clinical cases when he encounters a difficult one to make a diagnosis or to decide a treatment. Of course in order to retrieve similar cases there has been much progress by using the computer with keywords from the era of looking into Index Medix in a library. However, the keyword retrieval is not always the best way. Even after keyword retrieval, we must sometimes select again from many chosen cases. It is ideal if the system tells us similar cases in order of degree of contents. We introduced morphological analysis and tf-idf method in this field to compare all the words in abstract and find similar ones in order of degree of similarity. We collected case reports from presentations of the local committee meeting of the Japanese Society of Internal Medicine (JSIM). These presentation meetings are well known in this country and more than 24,000 cases in the last decade have already been digitalized and stored. Now we started to provide Japanese physicians the option to retrieve for similar cases through the internet by using the text mining technique in JSIM homepage (http://www.naika.or.jp/) from June 17, 2011.