Large-scale EEG reveals shared and distinct neural oscillatory features across major psychiatric disorders

Major psychiatric disorders are leading contributors to global disability, yet robust biomarkers to inform diagnosis and treatment remain elusive. Electroencephalography (EEG) offers a scalable, noninvasive approach to capture neural dysfunction, but prior findings have been inconsistent due to small, single-disorder cohorts. Here, we report a large-scale, two-site comparative study of resting-state EEG in schizophrenia (SCZ), bipolar disorder (BD), and major depressive disorder (MDD) compared with healthy controls (HC). Across 9751 participants’ EEG recordings, standardized preprocessing and feature generation yielded spectral power and functional connectivity (FC) measures tested with both statistical and machine-learning approaches. We identified convergent transdiagnostic alterations—reduced alpha power and increased gamma-band FC, implicating shared disruptions in inhibitory control and large-scale network coordination. Additional spectral differences included more pronounced delta-theta elevation in SCZ and a graded delta-theta continuum (SCZ > BD > MDD), suggesting a dimensional progression of neural dysregulation. Machine-learning models robustly distinguished individuals with psychiatric diagnoses from healthy controls within sites, but cross-site generalization and inter-disorder discrimination remained limited, underscoring challenges of heterogeneity. These findings identify reproducible EEG features that capture both shared and disorder-distinct neural alterations, advancing efforts towards early detection and individualized treatment in major psychiatric disorders.