The stages of NREM sleep are defined by specific changes in brain oscillatory behaviour, resulting in surface EEG patterns like delta activity. Combining simultaneous EEG and fMRI acquisition allows to investigate the brain structures and networks involved in generating and maintaining these oscillations in human subjects. We investigated the regional hemodynamic patterns linked to sleep-related oscillations in 15 subjects (7m, 8f; mean age 25.5±4.5 years) during the first sleep cycles, starting from 23: 00 hrs onwards after 36 hours of sleep deprivation (1.5T, 20 slices, 33 min acquisition time). After eliminating artefacts induced by the switching gradients from the EEG, the spectral delta, theta, alpha and spindle power were computed using an in-house software and used as regressors for analysis in SPM2. Removing trials containing insufficient sleep or movement artefacts resulted in 15 trials (9 subjects) for analysis. Alpha activity (8–12Hz) correlated with activity in the thalamus and cingulate gyrus; spindle activity (13–15Hz) with activation including frontal and parietal areas. Increases of slow oscillations (delta and theta frequency band) correlated with a decrease of hemodynamic response in the cingulate cortex, thalamus and insula. The results are in accordance with results of PET and EEG studies, and allow for a higher temporal and spatial accuracy of the dynamic patterns. The findings strengthen the influence of dynamic vigilance changes on brain activation.