BACKGROUND AND PURPOSE The aim of the study was to evaluate registration accuracy in an electromagnetic navigation system applied to image-guided intracranial procedures in children. MATERIAL AND METHODS In a group of 34 children aged from 2 weeks to 17 years, 38 procedures were performed using electromagnetic navigation, including 24 neuroendoscopic procedures, 10 craniotomies, and 4 shunting or drainage procedures. Thirty-three registrations based on 7 to 10 anatomical landmarks were digitised, and three-dimensional models of patients' heads were constructed using magnetic resonance images (MRI) (23 cases) or computed tomography scans (CT) (10 cases) and used for further analysis. Registration error calculated by the system was used as a measure of registration accuracy. RESULTS Registration error in the study group ranged from 0.7 mm to 4.4 mm (median 2.1 mm, mean 2.24 +/- 0.7 mm). It was shown that registration accuracy increased with patients' age. Differences between mean registration errors in procedures based on MRI or CT studies were not significant. There was no correlation between the number of landmarks registered and registration accuracy. CONCLUSIONS Electromagnetic neuronavigation, which does not require firm head fixation, can be used in the youngest age group. Nevertheless, registration accuracy is lower in newborns and infants. Neither imaging modality (MRI or CT) used for creation of a three-dimensional model of the patient's head nor registration of more than 6 landmarks have an impact on registration accuracy.