Brain Wave Signature Tied to Self-Harm in Depressed Teens, Study Finds

Adolescents who self-harm while struggling with major depressive disorder may carry a measurable electrical signature in their brains, according to a new study published in BMC Psychiatry. Researchers in China found that depressed teenagers with a history of self-harm showed abnormally prolonged activity in a specific resting-state brain network, and that this neural abnormality was statistically coupled with a distinctive psychosocial profile: more severe depression, higher reported social support, and weaker attachment to their fathers. The findings, the authors caution, are candidate markers rather than proven diagnostic tools, but they offer one of the most detailed multivariate pictures to date of how brain dynamics and life circumstances intertwine in this vulnerable group.

Self-harm is one of the most alarming features of adolescent depression. It frequently co-occurs with major depressive disorder and is considered a critical risk factor for later suicidal behavior, which makes identifying teenagers at elevated risk a pressing clinical priority. Yet the tools available to clinicians remain largely subjective, relying on interviews and questionnaires. The research team, led by Yaozhi Liu and Xiaona Wei of Suzhou Guangji Hospital and North Sichuan Medical College, set out to ask whether the brain’s spontaneous electrical activity could distinguish depressed adolescents who self-harm from those who do not, and whether any such neural differences relate systematically to their psychological and social circumstances.

The study enrolled 84 adolescents diagnosed with major depressive disorder, split evenly into two groups of 42: those with a history of self-harm and those without. Each participant completed a battery of standardized psychosocial assessments, including the Self-rating Depression Scale to quantify depressive symptoms, the Teenager Social Support Rating Scale to measure perceived and received support, and the Inventory of Parent and Peer Attachment to evaluate the quality of relationships with mothers, fathers, and peers. The researchers then recorded resting-state electroencephalography, or EEG, while participants sat quietly with no task to perform, capturing the brain’s intrinsic activity patterns.

The analytical centerpiece of the study was EEG microstate analysis, a technique that treats the brain’s moment-to-moment electrical landscape as a sequence of recurring topographic patterns. At any instant, the scalp distribution of electrical potentials forms a characteristic map, and decades of research have shown that these maps fall into four canonical classes, labeled A through D. Each class is thought to reflect a different resting-state network: Class A has been associated with phonological or auditory processing, Class B with visual imagery, Class C with salience and emotional evaluation, and Class D with attentional control. The brain flickers among these states every few hundred milliseconds, and the timing, duration, and transitions among them are considered a kind of fingerprint of large-scale network dynamics.

To extract these microstates, the team preprocessed the raw EEG signals, removing artifacts with independent component analysis, and then grouped the topographic maps using an atomize-and-agglomerate hierarchical clustering approach. Group comparisons were performed with permutation analysis of covariance, a robust statistical method that adjusts for demographic and clinical covariates, and all results were corrected for multiple comparisons using the false discovery rate procedure to guard against spurious findings. The result was strikingly specific: compared with depressed adolescents who did not self-harm, those who did showed a significantly longer mean duration of Microstate Class B, meaning the visual-imagery-linked network stayed active for extended stretches before switching to another state. The overall contribution of Class B to total brain activity, however, did not differ significantly between groups, indicating that the abnormality lay in how long each episode persisted rather than how often the state appeared overall.

The researchers were careful to note an important interpretive limitation. Because visual imagery was not directly assessed in the study, the prolonged Class B activity cannot be taken as direct evidence of sustained visual imagery processing in these adolescents. The association with the visual network remains a hypothesis grounded in the established functional interpretation of Class B microstates, not a demonstrated mechanism. This kind of restraint matters in a field where overinterpreting neural correlates is a persistent hazard, and it signals the authors’ intent to frame their findings as candidates for future validation rather than settled conclusions.

Beyond duration, the team examined transition probabilities, the likelihood that one microstate class gives way to another. Here, too, the self-harm group diverged. They showed increased bidirectional transitions between Class B and Class C, the salience and emotion-related network, and decreased transitions between Class A and Class D, the pair linked to phonological and attentional processing. In other words, the brains of adolescents who self-harm appeared to shuttle more frequently between the visual-imagery system and the emotional salience system, while the normal traffic between auditory-language and attentional networks was dampened. These transition differences survived statistical correction, with corrected probability values ranging from 0.014 to 0.038, suggesting a genuine reorganization of how these networks communicate at rest.

The psychosocial findings added a second, equally important layer. Adolescents in the self-harm group reported higher depression severity on the Self-rating Depression Scale, but also higher subjective and objective social support on the Teenager Social Support Rating Scale, alongside lower father attachment and lower total parental attachment on the Inventory of Parent and Peer Attachment. All of these differences were statistically significant after false discovery rate correction. The pattern is counterintuitive at first glance: these teenagers were not socially isolated in the conventional sense. They reported support from others, yet their bonds with their fathers, and with parents overall, were weaker. Class B duration itself correlated positively with depression severity and social support, and negatively with father attachment, with a correlation coefficient of minus 0.506 that was highly significant after correction.

To test whether the neural and psychological variables moved together as an integrated pattern rather than as isolated correlations, the researchers turned to regularized canonical correlation analysis, a multivariate technique that finds weighted combinations of two sets of variables that are maximally correlated with each other. Regularization, optimized with leave-one-out cross-validation, prevents the method from overfitting the modest sample size. The analysis identified a significant canonical correlation of 0.538, with a probability value of 0.027, linking a psychological profile of high depression, high social support, and low parental attachment to hyperactivity of Microstate Class B. This is the study’s central contribution: not merely that a brain measure differs between groups, but that a coherent mind-and-brain configuration emerges when the two domains are analyzed jointly.

The authors conclude that adolescents with depression and self-harm exhibit a distinct neuropsychosocial profile, one in which prolonged Class B microstate activity is coupled with a specific constellation of emotional and relational characteristics. They are explicit that these EEG features are candidate electrophysiological correlates of self-harm that require longitudinal and multi-center validation before they could inform clinical practice. The study was approved by the Ethics Committee of Suzhou Guangji Hospital and conducted in accordance with the Declaration of Helsinki, with written informed consent from all participants and their legal guardians. If future work confirms the pattern in larger and more diverse samples, microstate analysis, a relatively inexpensive and noninvasive technique, could eventually help clinicians move beyond self-report alone in assessing which depressed adolescents face the greatest risk, and point toward interventions that strengthen the fragile parental attachments at the heart of the profile.

Subject of Research: EEG microstate abnormalities and psychosocial characteristics in adolescents with major depressive disorder and self-harm

Article Title: Multivariate coupling of EEG microstate class B hyperactivity with a distinct psychosocial profile in adolescents with major depressive disorder and self-harm

Article References: Liu, Y., Wei, X., Yang, L., Zhu, F., Zhu, Y., Deng, H., Tian, Q., & Du, X. (2026). Multivariate coupling of EEG microstate class B hyperactivity with a distinct psychosocial profile in adolescents with major depressive disorder and self-harm. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08664-z

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08664-z

Keywords: EEG microstates, major depressive disorder, self-harm, adolescents, Microstate Class B, parental attachment, social support, canonical correlation analysis, resting-state EEG, suicide risk, neuropsychiatry, BMC Psychiatry