Spatial localization of scalp EEG electrodes is a major step for dipole source localization and must be accurate, reproducible and practical. Several methods have been proposed in the last 15 years. The most widely used method is currently electromagnetic digitization. Nevertheless, this method is difficult to use in a clinical environment and has not been validated with a high number of electrodes. In this paper, we introduce a new automatic method for localizing and labeling EEG sensors using MRI. First, we design a new scalp EEG sensor. Secondly, we validate this new technique on a head phantom and then in a clinical environment with volunteers and patients. For this, we compare the reproducibility, accuracy and performance of our method with electromagnetic digitization. We demonstrate that our method provides better reproducibility with a significant difference (p<0.01). Concerning precision, both methods are equally accurate with no statistical differences. To conclude, our method offers the possibility of using MRI volume for both source localization and spatial localization of EEG sensors. Automation makes this method very reproducible and easy to handle in a routine clinical environment.