Major Depressive Disorder (MDD) has become one of the three leading causes for years lived with disability with more than 264 million people affected worldwide (James et al., 2018). For those affected, MDD means personal suffering, reduced functioning and quality of life, social withdrawal, risk for co-morbid medical condition and increased mortality risk (Kessler and Bromet, 2013). Traditional diagnostic practices for MDD, while valuable, rely on subjective assessments that may not adequately reflect the biological foundations of the disorder (Dean and Keshavan, 2017). This highlights the pressing need for objective biomarkers that can enhance diagnostic precision and inform personalized treatment strategies (Ivanets et al., 2021; Carrle et al., 2023).

Neurophysiological methods such as electroencephalography (EEG) have gained traction as potential tools for elucidating the neural correlates of MDD (Lei et al., 2022). EEG’s high temporal resolution enables the detection of subtle changes in brain oscillations that are often associated with MDD, offering a window into the underlying neurophysiology of the disorder. The relevance of EEG in MDD diagnosis is further underscored by its potential to reveal altered spectral power within specific frequency bands linked to the disorder’s emotional dysfunctions (Cao et al., 2022; Han et al., 2022; Teng et al., 2022; Huang et al., 2023; Kavanaugh et al., 2023; Han et al., 2023a). Despite these advances, challenges persist in translating these findings into clinical practice, with issues such as inter-individual variability and the influence of medication status affecting the utility of EEG as a standalone diagnostic tool (Watts et al., 2022). To accurately identify patients with MDD, researchers have explored various analysis methods for EEG signals. Hasanzadeh et al. (2019) integrated multiple nonlinear features to achieve a classification accuracy of up to 91.3%. Bai et al. (2021) employed the k-Nearest Neighbors (KNN) model to analyze the complexity features of the gamma band, reaching an accuracy of 79.63%, while a random forest classifier achieved an accuracy of 65.94% for the fractal dimension of the beta band.

EEG data, which quantifies cortical activity through time series analysis methods like fast Fourier transformation, is invaluable in studying MDD. Recent research has confirmed the significance of EEG frequency band power (Wang et al., 2023; Han et al., 2023b), particularly noting increased alpha and theta power in patients in the early stages of depression (Grin-Yatsenko et al., 2010). This supports the notion that specific EEG patterns, such as alpha activity, play a functional role in MDD and its treatment response. Moreover, the BDNF Val66Met polymorphism has been linked to EEG alpha power, with the MetMet variant associated with low-voltage alpha EEG in MDD patients, suggesting a genetic influence on EEG characteristics (Zoon et al., 2013). Research by Yang et al. (2023) supports the effectiveness of using power spectral density (PSD) features for examining EEG signals across frontal, temporal, and central regions. Their findings suggest that these combined regions yield the highest accuracy in detecting MDD.

Further studies indicate that depression affects the brain’s hemispheres asymmetrically, leading to distinct patterns across various regions (Jiang et al., 2021). This asymmetry in brain activity is a critical aspect of how depression manifests and is detectable through EEG. Chang et al. (2011) reported the expected pattern of decreased alpha power at right frontal sites relative to the left, suggesting a hyperactive right and hypoactive left prefrontal cortex in depression. The assessment of resting-state EEG signals from different brain areas and frequency bands has been consistently highlighted as important, offering insights into the neural mechanisms behind emotional processes. Particularly significant is the practice of combining EEG data from both eyes-open and eyes-closed conditions into a single analytical model (Han et al., 2023b). This integrated approach provides a comprehensive understanding of brain activity, including its asymmetries, which is crucial for developing robust biomarkers for conditions like MDD (Ladeira et al., 2020). Despite its potential, the application of this integrative method in MDD research is still rare, emphasizing the need for further exploration in this area to enhance the diagnosis and understanding of depression.

In this study, we analyzed the resting-state EEG characteristics of individuals with MDD and healthy controls under both eyes-closed (EC) and eyes-open (EO) conditions. We focused on the neural oscillation of different brain regions and the asymmetry of band power between hemispheres. These features, along with clinical characteristics of the disease, were incorporated into a diagnostic regression model for MDD. In contrast to traditional EEG-based methods for MDD classification, which often focus on singular frequency band analysis or simple lateralization indexes, our approach represents a significant advancement. We integrate multi-regional band power assessments with hemispheric asymmetry analysis in both eyes-closed and eyes-open conditions, providing a more detailed picture of the brain’s electrical activity. This method acknowledges the dynamic nature of EEG signals and their variability with different states of arousal, which has been overlooked in previous studies. By doing so, we aim to capture the intricate neural oscillation patterns that are more indicative of MDD, potentially leading to improved diagnostic accuracy.

In the present study, we investigated the alterations in EEG spectral power in individuals with MDD and explored their potential as objective markers for diagnosis and personalized treatment. Our findings revealed significant changes in beta2 activity in the prefrontal cortex, and alterations in beta3 power in the occipital cortex, alpha and beta power in the parietal and temporal regions. Binary logistic regression identified significant EEG predictors of MDD, including beta2 power in PFC with eyes open, beta3 power in MOC with eyes open, beta2 power in RMFC with eyes closed, asymmetry in beta1 power between RT and LT with eyes open, and asymmetry in alpha and theta power with eyes closed. The model’s performance indicating fair discrimination ability. The ROC curve visually represented the model’s performance, demonstrating its superiority over a random classifier. The regression model involving multiple variables demonstrates a better predictive ability for depression than any individual factor. Furthermore, the correlation between EEG spectral features and clinical indicators suggests the potential of EEG as a monitoring tool for the clinical course of MDD (Baskaran et al., 2012; Olbrich and Arns, 2013).

Spectral analysis revealed that individuals with MDD exhibited a marked elevation in beta2 relative band power in the PFC, RMFC, LMFC, and CC when their eyes were open. This augmentation in beta2 band power aligns with neurocognitive models of MDD, which suggest hyperactivity in specific brain circuits, correlating with rumination and negative cognitive biases—hallmarks of depression (Siegle et al., 2007). Such increased beta activity may mirror frontal lobe dysregulation, supporting the frontal lobe hypothesis of depression that associates changes in frontal brain activity with depressive symptoms (Pizzagalli, 2011; Sharpley et al., 2023b). Furthermore, Claverie et al. (2016) provide compelling evidence for the prognostic value of EEG spectral features, particularly beta2 main peak frequency, in identifying vulnerability to depression. This research, conducted on a rat model, demonstrates that individuals exhibiting lower beta2 main peak frequency prior to exposure to stressors were more likely to become vulnerable to depression, as indicated by persistent low serum BDNF levels. The persistence of altered EEG patterns in vulnerable animals across different time points—before stress exposure, immediately after, and one month later—suggests that these electrophysiological markers are stable indicators of susceptibility to depression.

Conversely, in the occipital region, particularly the MOC, the MDD group displayed a significant reduction in beta3 power, potentially indicating anomalies in visual processing or the occipital lobe’s role in mood regulation (Bruder et al., 2001; Liu et al., 2022). Our study’s findings of altered band power in the parietal regions are corroborated by literature that implicates the parietal lobe in the neural circuitry of depression (Fales et al., 2008; Liu et al., 2023). The right parietal cortex, in particular, is associated with attentional control and emotional regulation—capabilities often compromised in MDD (Fales et al., 2008). Additionally, the parietal cortex’s contribution to the default mode network (DMN), known to be disrupted in MDD, further supports our observations (Greicius et al., 2007; Sheline et al., 2009). The DMN is linked to self-referential mental activity, frequently exhibiting a negative bias in depression (Hamilton et al., 2012), which could manifest as altered EEG spectral power in the parietal regions.

Consistent reports of alpha band power alterations in the parietal regions suggest cortical hypoactivation in MDD (Pollock and Schneider, 1990; Bruder et al., 2001). Simultaneously, some studies propose that increased alpha power in the right parietal cortex reflects an internal attentional focus (Benedek et al., 2014) and a causal relationship between parietal alpha activity and spatial auditory attention (Deng et al., 2019). Our findings, consistent with these studies, propose that parietal hypoactivation may serve as a stable neurophysiological marker for MDD. Under the eyes-closed condition, our study identified an increase in alpha power in the RT region and alterations in beta3 power in the RMOC and LMOC among participants with MDD. These results are consistent with existing literature, which reports an increase in alpha band power during eyes-closed rest, potentially more pronounced in MDD (Thatcher, 2012; Chandler et al., 2022). The RT region, known for its role in emotional processing and memory—areas often compromised in MDD—may account for the specific regional increase in alpha power (Drevets, 1998; Bruder et al., 2001).

The power asymmetry between the LT and RT regions in the alpha and beta bands underscores the potential lateralization of brain activity in MDD (Messerotti Benvenuti et al., 2019). The beta3 power asymmetry between the RMOC and LMOC could reflect the lateralized dysfunction in MDD, which is believed to disrupt affective processing and attention (Bruder et al., 2001). Given the occipital cortex’s primary role in visual processing, changes in beta3 power may indicate broader sensory processing issues in MDD (Kaiser et al., 2015). Moreover, these occipital lobe changes in beta3 power might relate to disruptions in the DMN, which includes occipital components and is affected in MDD (Greicius et al., 2007; Sheline et al., 2009). The DMN is associated with self-referential thought and mind-wandering, often negatively biased in MDD, which could be reflected in altered EEG patterns (Mayberg, 1997; Siegle et al., 2007). The observed imbalances may indicate disrupted interhemispheric communication, a factor implicated in the pathophysiology of depression (Uhlhaas and Singer, 2010). This lateralization has been noted in EEG studies, where alpha-band asymmetries correlated with emotional processing and depression severity (Bruder et al., 2001; Xie et al., 2023).

Furthermore, alterations in interhemispheric alpha power have been linked to functional disconnection between the cerebral hemispheres, potentially underlying the cognitive and affective disturbances in MDD (McVoy et al., 2019). Beta-band activity in the LT region is associated with language and executive functions, which are often impaired in MDD, suggesting that beta-band imbalances may correspond to the observed difficulties in cognitive control and verbal communication (Siegle et al., 2007; Pizzagalli, 2011). Additionally, research indicates that the LT region is involved in approach-related emotional processing, while the RT is associated with withdrawal-related emotions (Harmon-Jones et al., 2010). The alpha and beta power imbalance between these regions could reflect the emotional dysregulation and anhedonia commonly reported in MDD. Neuroimaging studies have supported the presence of structural and functional abnormalities in the temporal lobes of individuals with MDD, reinforcing the concept of lateralized dysfunction (Drevets, 1998; Lai, 2014). These abnormalities may be connected to disrupted connectivity within the limbic–cortical networks, which is essential for emotional regulation and stress response (Mayberg, 1997; Sheline et al., 2009).

In our results, binary logistic regression analyses of EEG data have identified certain spectral power features as robust predictors of MDD. Notably, these features include the relative power of the beta2 band in the PFC with eyes open and asymmetries in the power of the beta1, alpha, and theta bands between the RT and LT regions, with eyes open and closed, respectively. The PFC is known for its role in executive functions and emotion regulation, both of which are often impaired in MDD (Siegle et al., 2007; Hiser and Koenigs, 2018). Beta2 activity, in particular, has been associated with active cognitive processes and attention (Engels et al., 2010), and its dysregulation may reflect the cognitive disturbances observed in individuals with MDD (Clark et al., 2017). When eyes are open, asymmetry in beta1 power between the RT and LT regions could indicate the lateralized processing of emotional stimuli and stress response, which are frequently disrupted in MDD. Asymmetries in alpha power, particularly with eyes closed, have been linked to altered arousal and vigilance states, which are characteristic features of MDD (Sharpley et al., 2023a). Furthermore, theta power is associated with memory and emotional processing, and its alteration may correspond to the memory deficits and negative bias in emotional processing characteristic of MDD (H. Jiang et al., 2022). These electrophysiological markers offer a window into the underlying neural mechanisms of MDD and may enhance the accuracy of diagnostic procedures when combined with traditional clinical evaluations (Pizzagalli, 2011; Jaworska and Protzner, 2013). By providing a quantitative measure of brain activity, EEG can offer a more nuanced understanding of the disorder (Baskaran et al., 2012; Olbrich and Brunovsky, 2021).

The analysis of the ROC curve for the predictive utility of single EEG spectral power in MDD resulted in modest Area Under the Curve (AUC) values, such as 0.655 for beta2 in the PFC with eyes open, indicating a fair level of discriminative ability. However, when multiple EEG spectral power and asymmetry were included in a binary logistic regression model, the average AUC improved to 0.7709, indicating a more robust discrimination ability. This variability highlights the complexity of MDD as a disorder and the challenges in identifying a single biomarker with high diagnostic accuracy (Insel et al., 2010). While the modest AUC values obtained from ROC curve analyses of EEG features in predicting MDD do not diminish the potential value of these measures, they emphasize the importance of a multimodal diagnostic approach that integrates EEG, clinical assessments, and other biomarkers. Previous studies have also indicated that composite EEG measurement indices, such as the Antidepressant Treatment Response Index (ATR), exhibit strong predictive accuracy in determining treatment response in MDD. For instance, retrospective analysis of an initial study involving subjects with MDD treated with selective serotonin reuptake inhibitors (SSRIs) or venlafaxine demonstrated that ATR predicted response with an accuracy of 70%, with 82% sensitivity and 54% specificity (Iosifescu et al., 2009). Furthermore, in the Biomarkers for Rapid Identification of Treatment Effectiveness in Major Depression (BRITE-MD) study, ATR showed predictive value by achieving an accuracy of 74% in predicting response and remission, with a sensitivity of 58%, specificity of 91%, positive predictive accuracy of 88%, and negative predictive accuracy of 67% (Leuchter et al., 2009). Our approach aligns with contemporary psychiatric practice, which emphasizes the integration of biological data to inform diagnosis and treatment response, ultimately aims to improve outcomes for individuals with MDD (Insel et al., 2010; Markiewcz, 2017).

Correlation analysis of EEG relative band power with clinical indicators in MDD patients has provided insightful data, revealing a significant positive relationship between HDRS-17 scores and increased beta2 activity in various frontal regions, including the PFC, with coefficients ranging from r = 0.223 to r = 0.240 (Koller-Schlaud et al., 2020). This suggests that as the severity of depression increases, so does the beta2 activity in these areas. The positive correlation between HDRS-17 scores and increased beta2 activity in frontal regions suggests that EEG spectral features could serve as objective indicators for monitoring the clinical course of MDD.

However, this study is not without limitations. Given the relatively small sample size of our study, the sensitivity and specificity reported here may not accurately reflect what would be obtained in a larger, more heterogeneous population. Additionally, the cross-sectional nature of the study design precludes the ability to infer causality or the directionality of the observed relationships. Future research should include longitudinal designs to assess the temporal stability of EEG markers and their predictive value for treatment outcomes. While our findings provide valuable insights into EEG markers for MDD, they should be considered preliminary and warrant validation in larger-scale studies that can offer more definitive evidence of their generalizability. Moreover, the heterogeneity of MDD symptoms and the presence of comorbidities, such as anxiety disorders, may confound the EEG signals. Therefore, subsequent studies should consider stratifying participants based on symptom clusters or comorbid conditions (Frodl and O’Keane, 2013).

The study confirms the potential of multi-region EEG analysis in improving the diagnostic precision for MDD. MDD patients showed increased beta2 band power in the PFC, RMFC, LMFC, and CC under eyes-open conditions, and increased beta2 in RMFC and alpha in RT under eyes-closed conditions. Conversely, beta3 band power was lower in MDD across multiple regions. Notably, asymmetries in alpha and beta bands between right and left temporal cortices emerged as strong predictors of MDD. These EEG markers, along with clinical scores, provide potent diagnostic indicators for MDD.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of Beijing Anding Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

XL: Writing – original draft, Writing – review & editing, Methodology, Supervision, Validation. HZ: Writing – original draft, Writing – review & editing, Data curation, Formal analysis, Investigation, Methodology, Software, Supervision, Validation. YC: Methodology, Software, Visualization, Writing – review & editing, Project administration, Supervision, Validation. TZ: Methodology, Software, Visualization, Writing – review & editing, Supervision, Validation. BW: Data curation, Software, Visualization, Writing – review & editing, Project administration, Resources. XX: Data curation, Methodology, Visualization, Writing – review & editing. SL: Data curation, Software, Validation, Visualization, Writing – review & editing. SS: Project administration, Resources, Validation, Visualization, Writing – review & editing. YY: Project administration, Resources, Writing – review & editing. XZ: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing. LZ: Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing.