Image-guided neurosurgery allows surgeons to navigate and localize lesion through the patient’s cranium with a 3D image guidance. The model of the head is reconstructed using preoperative computed tomography or magnetic resonance images and real and virtual spaces are aligned by means of fiducial markers placed on the patient. In this paper, a new method for the optimal placement of the fiducial markers in order to reduce misalignment is presented. Using routine diagnostic images, a customized 3D model of the patient’s cranium is reconstructed. A genetic algorithm calculates optimal positions of the marker in order to minimize the target registration error. The fiducial set is shown to the surgeons on the 3D model to help him/her in placement of them.