Conventional planning of orthopedic injuries often relies on manual measurements obtained from radiographic images or CT slices. Computer assisted (CA) methods can support the surgeon in the preoperative planning to improve the reliability of the surgery. In this talk I will present novel CA methods for the planning of two types of interventions, namely the reduction of complex humerus fractures and the surgical correction of deformed forearm bones. Both tasks rely on patient-specific bone models segmented from CT. The segmentation is performed with a graph cut approach, particularly designed to tackle common problems in bone segmentation. For the reconstruction of humerus fractures, an interactive approach and a semi-automatic method are presented. Interactively, the reconstruction is carried out in an immersive, multi-modal planning system integrating stereoscopic rendering and haptic feedback. In the semi-automatic method, the contralateral (uninjured) bone acts as a template to obtain an initial reconstruction. The contralateral matching is efficiently performed using search-space sampling and GPU-based registration. Based on this initial assembly, the most correct reconstruction is finally computed by applying registration and multi-piece alignment techniques to the fracture surfaces. In the last part of my talk, the CA planning of corrective forearm osteotomies will be described. These methods include several 3-d based techniques for an accurate quantification of forearm deformities. Furthermore, a kinematic model of the forearm motion is presented which provides valuable information to the surgeon.