Computer assisted planning in cranio-maxillofacial surgery often requires the segmentation and reconstruction of the mandibular bone from CT data. A common imaging modality is cone-beam volumetric tomography, which requires only low doses of radiation yet suffers from small signal to noise ratio and strong artefacts in the presence of metal. This work explores the ability of model-based segmentation using a 3D statistical mandible model for automatic segmentation in such data. Apart from the statistical model, a key ingredient for this method is the deformation strategy for detecting the mandibular bone. Quantitative results support the feasibility of the proposed approach.