In this paper we present a mesh segmentation technique guided by an a priori knowledge. This knowledge is organized in an ontology. The work is done in an anatomical context, the segmentation aims at identifying different parts of an anatomical organ using their geometric shape. The ontology was extended to include an approximation of the geometric shape of some anatomical parts. The ontology then provides the parameters needed to segment the input mesh corresponding to an organ. The segmentation is then repeated following the hierarchy of the ontology to label the different parts and sub-parts of the starting organ.