Percutaneous radiofrequency ablation is a minimally invasive therapy for the treatment of liver tumors that consists in a destruction of tumors by heat. A correct insertion and placement of the needle inside the tumor is critical and conditions the success of the operation. We are developing a software that uses patients data to help the physician plan the operation. In this context, we propose a method that computes automatically, quickly and accurately the areas on the skin that provide a safe access to the tumor. The borders of the 3D mesh representing insertion areas are refined for a higher precision. Resulting zones are then used to restrict the research domain of the optimization process, and are visualized on the reconstructed patient as an indication for the physician.