PROBLEM TO BE SOLVED: To accurately extract an object from an image in which a plurality of structures are superimposed.SOLUTION: A medical image processing device in which a plurality of structures are included includes storage means 23 for storing a prior shape model consisting of an aggregation of contour lines of a specific structure based on an anatomical position as prior information, reference position extraction means 21 for arranging the prior shape model in a medical image, calculation means for calculating an image feature amount for the medical image in a position overlapped with a plurality of the contour lines of the arranged prior shape model, and extraction means for extracting a contour line of the specific structure by detecting a candidate point of the specific structure based on the contour lines of the arranged prior shape model and the image feature amount.SELECTED DRAWING: Figure 2【課題】複数の構造物が重なっている画像から対象物を精度良く抽出する。【解決手段】複数の構造物が含まれる医用画像処理装置において、事前情報として解剖学的な位置に基づく特定の構造物の輪郭線の集合からなる事前形状モデルを保存する保存手段23と、事前形状モデルを医用画像内に配置する基準位置抽出手段21と、医用画像に対して、配置された事前形状モデルの複数の輪郭線と重合する位置において、画像特徴量を計算する計算手段と、配置された事前形状モデルの輪郭線と画像特徴量とに基づいて特定の構造物の候補点を検出することで特定の構造物の輪郭線を抽出する抽出手段とを備えている。【選択図】図2