A medical image processing apparatus which processes a medical image including a plurality of structures, the apparatus including: a memory in which a previous shape model including a set of contours of a specific structure based on an anatomical position is stored as previous information; and a hardware processor that: locates the previous shape model in the medical image; calculates, for the medical image, image feature amounts at positions which overlap the contours in the located previous shape model; and extracts contours of the specific structure by detecting candidate points of the specific structure based on the contours in the previous shape model and the image feature amounts.