Neuroprosthetic devices can be an innovative solution to support subjects suffering from a limb loss. The possibility of restoring tactile sensations by using sensory feedback represents a research area whose outcomes could significantly improve the quality of life of prosthesis users. One of the best techniques to restore close-to-natural and selective tactile sensations is the electrical current stimulation of nerves using neural electrodes. Their interaction with nerves is an important aspect to be studied. The aim of this work is to deepen, in the framework of computational modeling, the possibility of approximating a realistic 3D human median nerve model, based on anatomical imaging, to a simplified model using the hybrid FEM-Neuron approach. Often, a high computational simulation time is also related to the complexity of the model geometry; therefore using a geometrical simplified model can be an important aspect to be analyzed. The simplified model is built approximating inner fascicles shape to simple geometrical shapes, i.e. ellipses. The electrical current stimulation is studied in Comsol environment by using a ds-FILE electrode model implanted in the human median nerve. The results obtained from computational simulations using both anatomical and simplified models, allow concluding that the percentage activation ranges at different distances from the active site obtained by the two models are comparable.