Frequency-decomposed EEG microstate analysis reveals altered cross-frequency coordination in adolescent major depressive disorder.

Adolescent major depressive disorder (MDD) involves alterations in large‑scale brain network dynamics. However, conventional EEG microstate studies typically rely on broadband signals and overlook frequency‑band specificity and cross‑frequency coordination. To address this gap, we collected resting‑state EEG data from 108 first‑episode, treatment‑naïve adolescent MDD patients and 96 healthy controls. We performed frequency‑decomposed microstate analysis (delta to gamma), computed pairwise adjusted mutual information (AMI) to assess band independence, and applied cross‑frequency coordination measures (co‑occurrence rate and effective transfer entropy) to broadband‑narrowband pairs. Discriminative performance was evaluated using machine learning with nested cross‑validation. Results showed that the eyes‑closed condition was more sensitive than eyes‑open, revealing widespread microstate abnormalities in the broadband, alpha, beta, and theta bands. AMI values between frequency bands were very low (maximum 0.0565 for broadband‑alpha), indicating that different bands carry largely independent information. Cross‑frequency coordination analyses revealed opposite alterations: broadband‑alpha coordination was reduced, whereas broadband‑delta coordination was enhanced. A fusion model using only the two best cross‑frequency coordination features achieved high discriminative performance (balanced accuracy = 0.835, AUC = 0.894), and these features did not correlate with clinical symptom severity. Together, our findings provide a frequency‑resolved, system‑level characterization of adolescent MDD and suggest that cross‑frequency coordination features may serve as candidate objective neurodynamic descriptors.