Characterizing brain oscillatory activity at the single-subject level is a key concern in personalized medicine. A recent publication has presented a normative map of the brain’s natural or typical frequencies of oscillation that could be used for detecting potential deviations in healthcare. However, this map represents brain activity in a voxel-by-voxel basis from high-density recordings of a large sample of healthy volunteers, compared to the low-density recordings usually employed on patients. Thus, accounting for less data to construct the single-subject map, a significant portion of noise is leaked into each individual’s voxel-based results. To reduce the amount of high-quality data required for characterizing individual activity, we aimed to define a number of data-driven regions of interest (ROIs) arising from the normative map and implement them for individual mapping. As in the original computation, the normative map was obtained training a k-means clustering algorithm to classify source-space (1cm3-voxels) power spectra from 128 healthy participants (OMEGA-database). Second, to obtain the single-subject maps, individual power spectra were classified into the previous group’s clusters. The natural frequency of each voxel was defined as the peak frequency of the most prevalent power spectrum, normalized. Then, to identify the ROIs, we classified all voxels into the clusters according to their z-scored frequency value and consistency of neighbouring voxels, obtaining 28 ROIs throughout the whole brain. Our results show an improvement in single-subject mapping, with moderate to high correlations with the normative map for most subjects, which represents an important step for personalized medicine.