Finite element analysis (FEA) is widely adopted to investigate the mechanical behaviour of bone structures. Computed tomography (CT) data are frequently used to generate FE models of bone. If properly calibrated, CT images are capable of providing accurate information about the bone morphology and tissue density. The aim of this work was to develop a special program able to read a CT data set as well as the FEA mesh generated from it, and to assign to each element of the mesh the material properties derived from the bone tissue density at the element location. The program was tested on phantom data sets and was adopted to evaluate the effects of the discrete description of the bone material properties. A three-dimensional FE model was generated automatically from a 16 bit CT data set of a distal femur acquired in vivo. The strain energy density (SED) was evaluated for each model element for increasing model complexity (number of different material cards assigned to the model). The computed SED were strongly dependent on the material mapping strategy.