Identification of hunger and satiety states from EEG data

Hunger and satiety states modulate brain areas to adapt behavior to energy needs. Thus, neurophysiological state of an individual measured by EEG may be indicative of one's hunger/satiety state. In this work, we evaluate applicability of spectrotemporal parameters of EEG for classification hunger and satiety. EEG recordings of resting state brain activity were made in 20 individuals after overnight fasting of at least 8 hours (hunger state), and again after a meal (satiated state). The data were acquired with a 64-channel EGI Hydrocel Geodesic Sensor nets. Spectrotemporal features of EEG likely to index metabolic state (peak alpha frequency, peak alpha power, average spectral power and 1/f exponent) were computed from EEG data. In comparison across machine learning models using Leave-One-Participant-Out Cross-Validation, alpha peak frequency yielded the highest accuracy (≥ 85%). Linking the hunger state to neural decision drivers and behavioral outputs promises rich insights into regulation of energy homeostasis, and might allow to categorize mixed metabolic-cognitive states from spectrotemporal EEG features.