Bio-medical structures always are extremely irregular-shaped and nonhomogeneous in material. These characteristics cause troubles not only to CAD systems, but also pre-processors and CAE systems. The simulation for them has becomes a difficult problem. Currently, thanks to the CT scan (Computerized Tomography) and related geometric tools, the geometry models of bio-medical structures can be constructed with STL geometry format and used to generate finite element meshes. However, the STL model can only represent a homogeneous model because it contains only geometric data of the surfaces of objects; This is inadequate to represent the bio-medical objects, e.g., bones, of which the internal bone material is not homogeneous, so does the stiffness. For example, the bone density which affects the stiffness of the bone is randomly nonhomogeneous; even the bone densities of the same person are different at different portions and vary over time. These characteristics make it very difficult to generate the analysis models which have randomly nonhomogeneous material. Without the real material information, the simulation will be unreal and inaccurate. This study proposes a mesh generation method which is based on the scanned image data derived from CT scan, which includes bone density data by the grey scale, to generate the analysis models which can represent not only the irregular shape but also the randomly distributed bone density, which is related to the stiffness of real bones. Even for bones of the same person, the bone stiffness changes over time although the shape could be still the same. With the proposed method and realistic models, this type of problems can be effectively simulated. For example, for a bone of the same person, the bone stiffness changes over time can be effectively analysed to evaluate the effects of the age; or, to evaluate the effects of the osteoporosis or certain treatment.