Recent developments of 3D computer technology, as computer graphics, haptic interaction, and 3D rapid prototyping, show promising potential for accurate, patient specific surgical planning. Traumatic surgery may potentially benefit from novel 3D technologies, both in the context of diagnosis and surgical planning. An inevitable prerequisite is accurate morphological segmentation, especially in trauma surgery, where all bone fragments, after a traumatic impact are mostly not clearly distinguishable using common segmentation techniques. A novel method is presented for the separation of these fragment clusters; the concept of adaptive surfaces. A flexible grid of control points is subject to motion, controlled by internal and external attracting forces to assess the outer surface of the bone, applicable for both joints and fractures. Internal forces control the stiffness of the surface and external forces act between the lattice and the landmarks on the bone surface, represented by principal gradient magnitudes. The algorithm yields proper classification of fracture parts and was successfully tested with geometrical phantom data and CT patient data from a heel bone fracture.