Recent functional magnetic resonance imaging (FMRI) replication studies show a high variability of active voxels within subjects and across runs - a potentially harmful situation for clinical applications. We tried to reduce these uncertainties inherent in current presurgical FMRI. For this, a new high quality head fixation device was used to detect reliably activated voxels over repeated measurements. In addition high correlation thresholds were applied to define the areas with highest probability of activation. The results show a focussing of such functional high risk areas to only a few voxels which localized close to intraoperative cortical stimulation. The generation of such FMRI risk maps may improve validity of clinical localization and facilitate the development of currently missing standards for maximized but still safe tumor resection.