Electroencephalography (EEG) coherence provides a quantitative measure of functional brain connectivity which is calculated between pairs of signals as a function of frequency. Without hypotheses, traditional coherence analysis would be cumbersome for high-density EEG which employs a large number of electrodes. One problem is to find the most relevant regions and coherences between those regions in individuals and groups. Therefore, we previously developed a data-driven approach for individual as well as group analyses of high-density EEG coherence. Its data-driven regions of interest (ROIs) are referred to as functional units (FUs) and are defined as spatially connected sets of electrodes that record pairwise significantly coherent signals. Here, we apply our data-driven approach to a case study of mental fatigue. We show that our approach overcomes the severe limitations of conventional hypothesis-driven methods which depend on previous investigations and leads to a selection of coherences of interest taking full advantage of the recordings under investigation. The presented visualization of (group) FU maps provides a very economical data summary of extensive experimental results, which otherwise would be very difficult and time-consuming to assess. Our approach leads to an FU selection which may serve as a basis for subsequent conventional quantitative analysis; thus it complements rather than replaces the hypothesis-driven approach.