EEG was recorded using a 128-channel Hydrocel Geodesic Sensor Net system (Electrical Geodesics, Inc., Eugene, Oregon), sampled at 250 Hz (bandwidth 0.1–100 Hz; impedances < 100 kΩ), referenced online to Cz. Data were acquired in eight 1-minute segments (four eyes open, four eyes closed), which were randomized and counterbalanced across participants. Consistent with prior EEG research on depression [30], only eyes closed data were analyzed. Data processing occurred offline using BrainVision Analyzer 2.0 (Brain Products GmbH, Gilching, Germany). First, muscle artifacts were manually removed, then blinks and electrocardiogram were removed using independent components analysis [31]. Due to the influence of ICA correction on coherence measures [32], only ICA components without visible neural activity were removed. Corrupted channels were interpolated using a spline interpolation [33]. The EEG was then visually inspected, remaining artifacts removed, and re-referenced to the average reference. After processing, non-overlapping 2.048 s segments were extracted for connectivity analyses. As recommended by Pascual-Marqui et al. [17], all participants had a minimum of 40 s of artifact-free data available for analysis.
