Introduction: In the present work, the steps of constructing hybrid phantoms have been studied. Mathematical and voxel phantoms are two various kinds of computational human body models which used in dose evaluations and estimations. In mathematical phantoms, organs contour define with mathematical equations and therefore they are not realistic, unlike voxel phantoms are image-based and more real. In turn, the disadvantage of voxel phantoms is extreme dependence of organs contour on CT and MRI image contrast. Hybrid phantoms are more realistic than mathematical phantoms and more desirable than voxel phantoms due to their flexibility in the shape and size of organs. In this approach, organs surface is defined with non-uniform rational B-spline (NURBS) surface which is a mathematical technique used in 3D graphics and animations extensively. Methods: Three steps are carried out to generate a hybrid phantom. (1) Transforming 2D images of human body to 3D model (2) Producing a 3D polygon mesh model of human body and internal organs (3) Creating NURBS. Initially, CT and MRI images for identifying soft and hard tissues are used. Then, two first steps can be constructing with software codes such as 3D-Doctor. For third step, NURBS modeling software can be used such as Rhinoceros. Results: We constructed hybrid phantoms with real CT and MRI images and the result is the Rhinoceros normal outcome file as *.rhp. It can be used for any size of human body because the size of organs is changeable. This pliability is the effect of NURBS control points which is the most important advantage of hybrid phantoms. Conclusion: We used advantages of both mathematical and voxel phantoms in constructing hybrid phantoms and thus they have the desirable shape and flexibility in organs. We should transform this phantom to voxel for applying in Monte carlo codes (MCNP). This voxelisation could be performing with MATLAB codes. Furthermore heart and respiratory motions can be simulated with this technique in 4D phantoms.