PROBLEM TO BE SOLVED: To provide a retrieval device for quickly retrieving a proper similar case, based on plural interest areas.SOLUTION: A similar case retrieval server 17 comprises: an individual similarity calculation part 65； a total similarity calculation part 66； and a similar case retrieval part 67. The individual similarity calculation part 65 associates interest areas and plural case lesions existing in case images one to one, compares feature amounts of the associated interest areas and feature amounts of the case lesions respectively, and respectively calculates individual similarity for every interest area. The total similarity calculation part 66 calculates total similarity based on the plural pieces of individual similarity which were respectively calculated to the plural interest area, and calculates the total similarity using the individual similarity, to only a combination of completely different kinds formed by combining plural case lesions which are in a same case and which have a different kind and the plural interest areas. The similarity case retrieval part 67 retrieves a similar case based on the total similarity corresponding to the combination of completely different kinds.COPYRIGHT: (C)2016,JPO&INPIT【課題】複数の関心領域を基に適切な類似症例を迅速に検索する検索装置を提供する。【解決手段】類似症例検索サーバ１７は、個別類似度算出部６５、総合類似度算出部６６、類似症例検索部６７を備える。個別類似度算出部６５は、関心領域と症例画像内に存在する複数件の症例病変とを１対１で対応させて、対応させた関心領域の特徴量と症例病変の特徴量とを比較して、関心領域毎の個別類似度をそれぞれ算出する。総合類似度算出部６６は、複数の関心領域に対してそれぞれ算出した複数の個別類似度に基づいて、総合類似度を算出するものであり、同一症例内の種類が異なる複数の症例病変と複数の関心領域をそれぞれ組み合わせて構成される完全異種の組み合わせに対してのみ、個別類似度を用いて総合類似度を算出する。類似症例検索部６７は、完全異種の組み合わせに対応する総合類似度に基づいて類似症例を検索する。【選択図】図１２
