New EEG framework predicts psychopathology dimensions with multi-scale features

A new study published on arXiv introduces a granularity-aware EEG feature framework for predicting dimensional psychopathology, offering a systematic approach to analyzing neurophysiological correlates. Led by Haofan Cheng and colleagues, the work organizes EEG descriptors into global, regional, and channel levels to capture multi-scale brain activity patterns. Using the Healthy Brain Network (HBN) cohort—a large pediatric dataset—the researchers evaluated predictions for four key psychopathology dimensions: the general psychopathology factor (p-factor), internalizing, externalizing, and attention problems. The framework was tested across four different EEG paradigms (resting state, task-based, etc.), providing a broad assessment of its generalizability. Tree-based machine learning models combined with a granularity-balanced feature selection strategy yielded modest but consistent improvements over conventional approaches, though effect sizes remained small—reflecting the inherent difficulty of predicting questionnaire-based scores from EEG signals.

Beyond prediction performance, the study analyzed the selected features to reveal dimension-specific spatial and spectral patterns that align with existing neurophysiological knowledge. For instance, attention problems were associated with frontal theta activity, while internalizing symptoms linked to parietal alpha asymmetry. To test robustness, the authors conducted an exploratory cross-dataset sanity check on the independent PEARL cohort, finding that the selection principle remained technically feasible despite differences in recording protocols—though they caution against claiming cross-dataset generalizability. Overall, the research demonstrates that multi-scale EEG features contain weak but detectable signals related to dimensional psychopathology, and the granularity-aware selection approach offers a useful feature-reduction strategy for future EEG-based phenotyping studies, particularly in pediatric mental health research.