Disrupted interhemispheric coordination of sensory-motor networks and insula in major depressive disorder

Objective Prior researches have identified distinct differences in neuroimaging characteristics between healthy controls (HCs) and patients with major depressive disorder (MDD). However, the correlations between homotopic connectivity and clinical characteristics in patients with MDD have yet to be fully understood. The present study aimed to investigate common and unique patterns of homotopic connectivity and their relationships with clinical characteristics in patients with MDD. Methods We recruited 42 patients diagnosed with MDD and 42 HCs. We collected a range of clinical variables, as well as exploratory eye movement (EEM), event-related potentials (ERPs) and resting-state functional magnetic resonance imaging (rs-fMRI) data. The data were analyzed using correlation analysis, support vector machine (SVM), and voxel-mirrored homotopic connectivity (VMHC). Results Compared with HCs, patients with MDD showed decreased VMHC in the insula, and increased VMHC in the cerebellum 8/vermis 8/vermis 9 and superior/middle occipital gyrus. SVM analysis using VMHC values in the cerebellum 8/vermis 8/vermis 9 and insula, or VMHC values in the superior/middle occipital gyrus and insula as inputs can distinguish HCs and patients with MDD with high accuracy, sensitivity, and specificity. Conclusion The study demonstrated that decreased VMHC in the insula and increased VMHC values in the sensory-motor networks may be a distinctive neurobiological feature for patients with MDD, which could potentially serve as imaging markers to discriminate HCs and patients with MDD.


Introduction
Major depressive disorder (MDD), a prevalent mental illness, is the foremost cause of disability worldwide (CONVERGE consortium, 2015;Yang et al., 2020). Furthermore, it is a complicated emotional disorder characterized by abnormal clinical symptoms, including cognitive disorders (such as feelings of excessive guilt or worthlessness), autonomic dysfunction (such as changes in appetite or sleep patterns), abnormal psychomotor activities (such as excitement or retardation), and elevated risk of suicide (Lou et al., 2019). Approximately 12% of the global population is affected by MDD, while the lifetime occurrence of MDD in the United States surpasses 20% (Hasin et al., 2018;Hayley et al., 2021;Fries et al., 2022). The incidence of MDD has risen since the outbreak of COVID-19 (COVID-19 Mental Disorders Collaborators, 2021). Approximately 25-50% of those with MDD exhibit deficiencies in one or more cognitive domains, which are considered key features of the condition .
Researches have demonstrated that MDD is a heritable and heterogeneous disorder (Andersson et al., 2019;Kendall et al., 2021;Nguyen et al., 2022), and its clinical features are associated with brain regions that exhibit anatomical differences. Multiple neuroimaging studies have attempted to identify the neurobiology of MDD and distinguished several neuroimaging characteristics between healthy controls (HCs) and patients with MDD. Zhang et al. (2022) found that, in comparison to HCs, patients with MDD exhibited decreased functional connectivity (FC) between the emotional subregion of the anterior cingulate cortex (ACC) and the hippocampus, thalamus, insula, angular gyrus, and posterior cingulate cortex. On the other hand, other studies reported increased connectivity between the hippocampus and the ACC and reduced connectivity between the insula and the ACC in patients with MDD (Krug et al., 2022). Functional brain imaging studies revealed significant changes in the FC in the cerebellum-neocortex and cerebellum-basal ganglia circuits in patients with MDD (Dai et al., 2022). Additionally, an increased long-range positive FC (lpFC) in the left inferior parietal lobule was found to distinguish HCs and patients with MDD (Dziedzic et al., 2022). Given that the architecture of the connectome affects the effectiveness and velocity of information transport, including regions such as anterior and middle cingulate cortex, medial occipital areas, superior frontal areas, post-and precentral gyrus, parahippocampal gyrus, and precuneus, the balance of intra-hemispheric and interhemispheric connectivity plays a crucial role in brain function (Krupnik et al., 2021).
Studies have demonstrated that communication between the left and right hemispheres of the human brain is a crucial aspect of both cognitive and emotional processing (Compton et al., 2005;Toro et al., 2008). Ran et al. (2020) showed that anatomical abnormalities in the corpus callosum in patients with MDD may result from caused by imbalanced inter-hemispheric communication. Concurrently, a magnetic resonance imaging (MRI) study indicated that the deterioration of inter-hemispheric FC is related to the severity of clinical depression and treatment outcomes of patients with MDD (Kozel et al., 2011;Zheng et al., 2022). Additionally, patients with MDD had notably higher overall average (static) functional connectivity (sFC), but lower variability of functional connectivity (vFC) within networks (Demirtas et al., 2016). As per the research by Liu et al. (2021) inter-hemispheric homotopic connection in particular regions could be serve as a potential biomarker to distinguish patients with MDD from HCs. Reduced inter-hemispheric coordination in the posterior defaultmode network and visual regions was also revealed between HCs and patients with MDD (Guo et al., 2018).
Voxel-mirrored homotopic connectivity (VMHC) is a technique used to calculate resting-state FC between voxels in a hemisphere and their corresponding mirror regions in the opposite hemisphere (Jia et al., 2020;Fan et al., 2022). VMHC can be utilized to determine the intensity of functional connections between brain regions in both hemispheres that are located at the same position (Zuo et al., 2010;Jin et al., 2021). Several studies have concluded that the mechanism behind VMHC deficiency may be linked to extensive abnormalities in white matter integrity, dysfunction in local gray matter structure, and the mode of pathway reorganization (Yuan et al., 2012;Ding et al., 2015). Moreover, this method can be used to evaluate the relationship between time series dependent on blood oxygen levels, and to demonstrate how information is exchanged between the two hemispheres of the brain . Some clinical studies have demonstrated altered homotopic FC by measuring VMHC in patients with MDD (Lai and Wu, 2014;Hou et al., 2016). Support vector machine (SVM) is a specialized form of supervised machine learning that multivariable pattern recognition technology that is applied to predict psychosis based on neuroanatomical indicators (Shan et al., 2021). SVM can effectively identify a set of information and features from different brain regions, which can be used to classify patients and HC using neuroimaging data [such as functional magnetic resonance imaging (fMRI) data] (Steardo et al., 2020), making the classification results more convincing. Numerous studies have employed VMHC and SVM methods to study brain disorders, providing possible evidence for the discovery of biological markers in neuroimaging (Chen et al., 2021;Chu et al., 2022;Wu et al., 2022). However, it is still unclear whether abnormal VMHC can be used as an underlying brain imaging symbol to discriminate HCs from MDD using the SVM analysis.
In this study, our objective was to apply the VMHC method to identify the inter-hemispheric functional interaction during resting state in individuals with MDD. This study investigated whether abnormal VMHC might be utilized as a potential marker in patients with MDD by combining VMHC values with cognitive tests, exploratory eye movement (EEM), event-related potentials (ERPs), and other markers. We hypothesized that patients with MDD would exhibit reduced VMHC, which could serve as a potential imaging marker to differentiate between HCs and patients with MDD. In addition, we proposed that aberrant VMHC would be associated with clinical variables in MDD.

Participants
Patients with MDD were recruited from Foshan Third People's Hospital. This study included 32 first-episode patients and 10 recurrent patients with MDD. Forty-two HCs matched for sex and years of education were recruited from the community. The Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5) patient version is used to determine the diagnosis of MDD. For recurrent patients, the use of antidepressants was suspended for at least 2 weeks by themselves. The anxiety and depressive symptoms, social function, personality characteristics, coping style, social support, and psychological cognitive function of the subjects were measured by Hamilton Anxiety Scale (HAMA), Hamilton Depression Scale (HAMD), Social Disability Screening Schedule (SDSS), Eysenck Personality Questionnaire (EPQ), Simplified Coping Style Questionnaire (SCSQ), Social Support Scale (SSS), Wisconsin Card Sorting Test (WCST), and Repeatable Battery for the Assessment of Neuropsychological Status (RBANS). All participants were right-handed and ranged in age from 18 to 60 years old. The exclusion standards for all subjects were as follows: (1) a history of severe physical sickness or alcohol or drug abuse; (2) significant physical impairment, preventing completion of the follow-up study; and (3) the presence of mental retardation, dementia, and severe cognitive impairment in addition to other serious psychiatric problems.
The study was approved by the Research Ethics Committee of the Third People's Hospital of Foshan, and a written informed consent was obtained from all participants.

Event-related potentials (ERP) data acquisition
Event-related potential data were collected using a Japanese Kohden MEB-9402C myoelectric evoked potentiometer. Participants were instructed to sit comfortably and make an effort to concentrate. The electrode placement followed the 10/20 standard by the International Electroencephalogram Association. The ground was put in the center of the hand FPz. The right ear M2 point served as the recording electrode. The central Cz point served as the reference electrode. The electrode impedance was set to 5K, and the filter was 0.2-20 Hz. The analysis duration was 1000 ms. The stimulation is carried out in the traditional "Oddball" auditory stimulation mode, with a frequency of 1 times/s, a duration of 10 ms, and a sensitivity of 5 µV. The detection was accomplished by triggering and activating two systems, which were band-pass filtered 200 times with the low-frequency filter and the high-frequency filters. The settings for the non-target stimulus were set to 80% probability, 70 dB intensity, and 1000 Hz frequency, while the parameters for the target stimulus were set to 20% probability, 90 dB intensity, and 2000 Hz frequency. The two frequencies were randomly interspersed, and each scenario was repeated twice before the average value was calculated. The subject was presented with a target stimulus and the non-target stimulus was not utilized as a response. If a test subject had less than 80% hits, the test was considered invalid. The latencies of the N100, P200, N200, and P300 waves were recorded separately.

Exploratory eye movement (EEM) data acquisition
The EEM data was collected using a Shanghai-made Dekang DEM-2000 eye movement detector. The participants were instructed to sit comfortably in a chair and focus on a small screen in front of them. The angle of eyes movement from the left to the right side of the screen was 33 • , and the distance between their eyes and the screen was 25 cm. The participants were instructed to pay close attention as the initial S-shaped pattern (S) was displayed on the screen for 15 s. Within this time, the device automatically captured the gaze positions and recorded the number of eye fixations (NEF). After that, the second and third S-shaped patterns (S2, S3), which were slightly different from the first image, were displayed on the screen. Every pattern lasted 15 s. When prompted with the question, "What is the difference between the two patterns, " the participants were instructed to pay close attention before responding with, "There is no difference." The responsive seeking score was calculated from the gaze points in seven locations (only one point was considered in each region) for a period of 5 s (RSS). The device can automatically record the eye movement's trajectory, the data are automatically processed by the computer, and the entire procedure may be replayed for later use. A gaze point in EEM analysis is defined as an eye's gaze time that exceeds 200 ms on specific spot (the eyeball is moving within 2 • ). This is defined as the total number of gaze points in 15 s when the eye fixes on the S-pattern. The RSS score is broken down into seven S2 or S3 regions, and the instrument counts the NEF areas for a total of 5 s. No matter how often, the subject's attention on a certain place is worth one point. The maximum RSS score for each picture is thus 7, and for S2 and S3, the maximum RSS overall score is 14. NEF 30 and/or RSS 4 were used as the criterion for abnormality.

Measures
Hamilton Depression Scale (17 items) was used to assess depressive severity. A higher score indicates a more severe level of depressive symptoms (Bagby et al., 2004). The level of anxiety symptoms was evaluated using the HAMA. This scale consists of 14 items, each with a score ranging from 0 to 4. A higher score indicates a greater severity of anxiety symptoms. The SDSS is a 10-item scale that was clinicianadministered to assess the level of functional impairment. A threepoint Likert scale (0-2) was used to rate each item. A higher score indicates a greater level of functional impairment . Factor analysis of the EPQ scale revealed three orthogonal dimensions, leading to the proposed of four basic factors to determine personality: Extraversion (E), Neuroticism (N), Psychoticism (P), and Lie (L). These three dimensions reflect individuals' varying tendencies and levels of performance, resulting in different personality characteristics (Smith and Ellingson, 2002). The SCSQ is developed based on the characteristics of Chinese population. The questionnaire consists of 20 items, each rated on a four-point Likert scale with scores ranging from 0 to 3, and pertains to different coping strategies for handling everyday events . The SSS was used to evaluate social support. A total of 10 components and 3 dimensions made up the scale (i.e., subjective support, objective support, and support utilization). Scores under 20 indicate poor social support, between 20 and 30, moderate social support, and over 30, satisfied social support (Li and Shou, 2021). The WCST is a card-matching task commonly used to evaluate cognitive flexibility and broader aspects of executive function in both research and clinical settings (Miles et al., 2021). The RBANS is a brief neurocognitive assessment tool that provides standardized measures of attention, language, memory, and visuospatial/constructional abilities (Faust et al., 2017).

Imaging data acquisition
Images were collected using a GE 3.0 T scanner (GE 3.0 T Signa Pioneer). The subjects were instructed to remain still, close their eyes, and stay awake. Soft earplugs and foam pads were employed to mitigate the effects of scanner noise and head movements during the imaging process. Scan parameters were repetition time/echo time = 2000/30 ms, 36 slices, 64 × 64 matrix, flip angle 90 • , field of view 24 cm, slice thickness 4 mm, no gap, 250 volumes (500 s).

Data pre-processing
Data Processing Assistant for Resting-State fMRI (DPARSF) software was used to preprocess resting-state data in MATLAB. The slice timing and head motion in the images were corrected.
To ensure the quality of the imaging data, the maximum x, y, or z displacement and angular motion for all subjects should have no more than 2 and 2 • mm, respectively. The functional images were then resampled to 3 mm × 3 mm × 3 mm and normalized (Chao-Gan and Yu-Feng, 2010). An isotropic Gaussian kernel was used to smooth the processed images (fullwidth at half-maximum: 8 mm). Additionally, a linear trend removal and band-pass filtering (0.01-0.08 Hz) were processed (Song et al., 2011). Next, linear regression was used to remove a number of spurious covariates and their temporal derivatives from the data. These variables comprised the signal from a ventricular region of interest (ROI), the signal from a region centered in the white matter, and 24 head motion parameters determined via rigid body correction (Fox et al., 2005).

VMHC analysis
The REST software is used to examine VMHC (Song et al., 2011). In a brief, VMHC maps were created by computing Pearson correlations (Fisher z-transformed) between a given voxel and a mirrored voxel in the opposing hemisphere. The specifics of the VMHC analysis have been previously described in the literature (Zuo et al., 2010).

Statistical analysis
Data analysis in this study was conducted by using SPSS version 25.0. The chi-square test was used to examine the gender differences between the two groups. Two sample t-tests were employed to compare continuous variables such as age, years of education, and clinical scales. The significance level was set at p < 0.05.
The image data was analyzed using the DPARSF software. Two sample t-tests were conducted on each VMHC map to compare the groups. Then multiple comparison correction was performed based on the Gaussian random field (GRF) theory (p < 0.001 for voxellevel significance and p < 0.05 for cluster-level significance). Age, gender, education, and mean framewise displacement were used as covariates to minimize the potential influence of these variables.
For the correlation analyses, the mean VMHC values in aberrant brain areas with substantial differences between depressed patients and HCs were retrieved. The correlations between VMHC levels and clinical factors in patients with MDD were investigated using Pearson or Spearman correlation analyses. The significance level was set at p < 0.05 (corrected according to the Bonferroni's correction).

SVM analysis
The SVM analysis was used to determine the ability of VMHC values extracted from abnormal brain regions to differentiate between HCs and patients with MDD by utilizing the LIBSVM software 1 in MATLAB. The "leave one-out" method was used in the study.

Demographic and clinical data
A total of 46 patients with MDD and 44 HCs were included in this study. However, due to significant head movement, the data of two HCs and four patients were excluded. As a result, the final imaging analysis included 42 patients with MDD and 42 HCs. Detailed demographic and medical information about the individuals can be found in Supplementary Table 1. The age difference between patients and HCs was significant (p = 0.01), and there was no significant difference between gender and years of education. There were significant differences in HAMA (p < 0.001), HAMD (p < 0.001), Neuroticism (N) (p < 0.001), Lie (L) (p < 0.001), Extraversion (E) (p = 0.003), SDSS (p < 0.001), SCSQ subscale scores (p < 0.001, p = 0.001), SSS (p < 0.001), NEF (p < 0.001), RSS (p = 0.025), N200 (p = 0.038), and P300 (p = 0.012) between the two groups. There was no significant difference between the two groups in SCSQ total scores, WCST, RBANS, N100, and P200.

VMHC analysis in depressed patients and HCs
Our results showed that in the patients with MDD, the VMHC values of the superior/middle occipital gyrus and cerebellum 8/vermis 8/vermis 9 increased compared to HCs, while the VMHC values of the insula decreased. The specific details can be found in Table 1 and Figure 1.

SVM analysis
According to abnormal VMHC values in various brain areas and the combination of these clusters, Figure 2A illustrates the effectiveness of differentiating patients with MDD from HCs. The sensitivity, specificity, and accuracy of the differentiation using the VMHC values in the cerebellum 8/vermis 8/vermis 9 and insula were 80.95, 83.33, and 82.14%, respectively ( Figure 2B). From VMHC values in the superior/middle occipital gyrus and insula, the sensitivity was 78.57%, the specificity was 85.71%, and the accuracy was 82.14% ( Figure 2C). In addition, the receiver operating curve (ROC) of applying abnormal VMHC to distinguish patients with MDD from HCs shown in Supplementary Figure 1, and the accuracy, sensitivity, and specificity of classification with ROC and SVM shown in Supplementary Table 2. ROC curve can only analyze a single brain region, while VMHC combined with SVM can analyze two brain regions. This is also an advantage of SVM, which takes into account the spatial and temporal distribution of the brain.

Discussion
In this study, decreased VMHC in the insula and increased VMHC values in the sensory-motor networks may be distinctive neurobiological feature for patients with MDD, which could serve as potential imaging markers to differentiate HCs and patients with MDD. In addition, we found correlations between abnormal VMHC values and clinical/cognitive parameters in MDD. The aberrant VMHC values were further demonstrated to be promising imaging indicators of MDD by using the SVM analysis.
Event-related potential is a mature method to understand brain function in cognitive process, which is expected to make greater contribution to the identification, prediction (Tsai and Liang, 2021), treatment and prevention of mental disorders (Hajcak et al., 2019), and P300 is a typical indicator of neurophysiological ERP (Wada et al., 2019). Our research results show that there are obvious differences between patients with MDD and HCs on the cognitive level, which is consistent with a previous study (Wang et al., 2023). This may be caused by the depression of the patients and the decrease of attention to external events. The number of fixations (NEF) and response search score (RSS) in EEM are related to mental state, active attention and cognitive function (Ross et al., 2000). The neural basis of EEM has been clearly explained by a research (Lencer and Trillenberg, 2008). Studies have proved that compared with the HC group, the eye movements of patients with MDD are abnormal . At the same time, the combination of P300 and eye movement data can improve the accuracy of auxiliary diagnosis of depression (Diao et al., 2022). ERP and EEM combined with other psychological scale data, we can see that there are obvious differences between patients with MDD and HCs.
The insula, surrounded by cortical gyrus, white matter, and basal ganglia structures (Dziedzic et al., 2022), is a center of integration for emotional, visceromotor, autonomic, and interoceptive information. Its diverse functional roles may be due to its strong connections to a broad network of cortical and subcortical regions (Wang R. et al., 2022). Studies have shown that removal of the insular lobe may disrup autonomic nerve function and alter an individual's experience of emotions (Lacuey et al., 2019;Motomura et al., 2019). Processing input from many functioning systems involves an essential integration role for the insular cortex (Starr et al., 2009;Morita et al., 2014;Allen et al., 2016;Berret et al., 2019;Choi et al., 2022). Guo et al. (2015) found that compared with HCs, the insula of the front-limbic circuit, hate circuit, and visual regions of the patients with MDD showed decreased FC. At the same time, FC between the insula and ACC was also reduced (Krug et al., 2022). Patients with MDD showed abnormal FC between the insular subdivisions to the superior temporal sulcus, inferior prefrontal gyrus, amygdala, and posterior parietal cortex (Peng et al., 2018). Multiple studies on patients with MDD have revealed that, in comparison to HCs, their connectivity in insular lobes was reduced (Veer et al., 2010;Sliz and Hayley, 2012;Avery et al., 2014;Penner et al., 2016;Yin et al., 2018). The insula plays a key role in emotional regulation, conscious arousal, and consciousness. The decrease of insular connectivity may reflect the important regulation of negative or arousal stimuli. The decreased connectivity in the insula of patients with MDD may indicate an alteration in their perception of visceral responses and subjective sensory states (Critchley et al., 2004). Consistent with these results, we discovered that VMHC in the insula was significantly lower in patients with MDD compared to HCs, further supporting the insular crucial function in the neurological process underlying MDD. We speculate that the decrease of VMHC in the insula may be an indicator related to the emotional state of patients with MDD.
The cerebellum, located in the posterior fossa, is commonly known to control movement. However, recent studies show that it also plays a significant role in cognition and emotion (Rabellino et al., 2018;Habas, 2021;Su et al., 2021). In this study, compared with HCs, VMHC of the cerebellum 8/vermis 8/vermis 9 in patients with MDD was significantly increased. This increase in FC is often interpreted as a sign of dedifferentiation or compensatory Voxel-mirrored homotopic connectivity (VMHC) values of abnormal regions in patients with MDD. MDD, major depressive disorder; VMHC, voxel-mirrored homotopic connectivity.
redistribution (Cabeza et al., 2002;Grady et al., 2005;Pagen et al., 2020). Inflammatory cytokines (i.e., tumor necrosis factor and interleukin 6) can activate astrocytes and show hyperfunction (increased blood flow and metabolism) (Guo et al., 2013a). Regional hyperfunction may encourage an increase in FC and regional activity Wang et al., 2016). Studies have shown that patients with MDD with MDD experience a significant increase in communication between the cerebellum and the cerebellar-anterior default mode network (Ding et al., 2022). In comparison to HCs, our previous study found a marked decrease in the FC between the cerebellum and cerebral cortex in individuals with both treatmentsensitive and treatment-resistant depression (Guo et al., 2013b). Another study showed that the cerebellar-cerebral dynamic FC of patients with MDD was lower than that of HCs (Zhu et al., 2020). This is because MDD is a multidimensional disease related to emotion, cognition, memory, etc., some special default normal functions of HCs may become dysfunctional in MDD patients, such as the VMHC of the cerebellum 8/vermis 8/vermis 9 in this study.
The findings of this study showed a significant increase in the VMHC value of the superior/medium occipital gyrus. The occipital lobe, located at the back of the cerebral hemisphere, is primarily utilized for processing visual information, and communicating with the cerebral cortex. It plays a crucial role in how facial emotions are perceived and processed (Teng et al., 2018;Li and Wang, 2021). Patients with MDD may experience aberrant cognitive processes, such as attention deficit disorder and motor delay (Yu et al., 2017). The results of this study provide additional evidence to clarify the judgment of MDD. Support vector machine has been widely used for classify mental illnesses. The FC signal as a potential diagnostic index requires that the sensitivity or specificity of SVM be higher than 0.6 (Guo et al., 2011;Shan et al., 2021). The previous SVM analysis showed that using SVM to classify the neuroimaging biomarkers of MDD resulted in a diagnostic accuracy is 98.96% . Compared with a study by Song et al. (2022), our study found that patients with MDD had VMHC abnormalities in extensive brain regions. At this point, the two studies are consistent. Different Support vector machine (SVM) results. (A) The accuracy of classification of six SVM analyses. One represents the VMHC values in the cerebellum 8/vermis 8/vermis 9; 2 represents the VMHC values in the superior/middle occipital gyrus; 3 represents the VMHC values in the insula; 4 represents the combination of the VMHC values in the cerebellum 8/vermis 8/vermis 9 and superior/middle occipital gyrus; 5 represents the combination of the VMHC values in the cerebellum 8/vermis 8/vermis 9 and insula; and 6 represents the combination of the VMHC values in the superior/middle occipital gyrus and insula. (B) SVM analysis applied the combination of the VMHC values in the cerebellum 8/vermis 8/vermis 9 and insula. Sensitivity = 80.95%, specificity = 83.33%, and accuracy = 82.14%. In the left part, the red cross represents patient with MDD, and the green asterisk represents healthy controls. The blue circle represents support vector. (C) SVM analysis applied the combination of the VMHC values in the superior/middle occipital gyrus and insula. Sensitivity = 78.57%, specificity = 85.71%, and accuracy = 82.14%. In the left part, the red cross represents patient with MDD, and the green asterisk represents healthy controls. The blue circle represents support vector. SVM, support vector machines; VMHC, voxel-mirrored homotopic connectivity; MDD, major depressive disorder.
Frontiers in Neuroscience 07 frontiersin.org from Song et al. (2022) research, our research found abnormal VMHC in the insula and the sensory-motor networks. At the same time, we verify the difference between patients with MDD and HCs by measuring clinical data (such as ERP, EEM, and so on), which makes the study more convincing. The current SVM results show that the VMHC values of the insula, cerebellum 8/vermis 8/vermis 9 and superior/medium organic gyrus are greater than 0.78 in the sensitivity, accuracy, and specificity of distinguishing patients with MDD from HCs. Therefore, the VMHC values of the insula, cerebellum 8/vermis 8/vermis 9 and superior/medium organic gyrus may be used as potential imaging markers for MDD.
There are still some limitations in this study. First, the area under curve (AUC) of this study was impressive. But clearly, this is a small preliminary report that needs to be replicated. Second, we recruited patients who did not take medicine at the time of the first episode, and patients who had relapsed and did not take medicine for at least 2 weeks. The dissemination of research findings may be constrained for patients who experience relapses since the impact of antidepressant medications and number of episodes on brain spontaneous activity cannot be entirely ruled out. Finally, we only scanned the patients at baseline, so we did not know the alterations of spontaneous neuronal activity after treatment.

Conclusion
In conclusion, the study's comparison of VMHC alterations in HCs and patients with MDD was groundbreaking. Our research results show that decreased VMHC in the insula and increased VMHC values in the sensory-motor networks may be a distinctive neurobiological feature for patients with MDD, which might be served as potential imaging markers to discriminate HCs and Patients with MDD.

Data availability statement
The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

Ethics statement
This study was approved by the Ethics Committee of the Third People's Hospital of Foshan (FSSY-LS202001).
The patients/participants provided their written informed consent to participate in this study.

Author contributions
CZ, HJ, and HY: writing-original draft, writing-review and editing, methodology, and software. XL, JL, QZ, WL, YO, CP, YY, and WW: validation, investigation, and resources. GX and WG: supervision, project administration, and funding acquisition. All authors contributed to the article and approved the submitted version.

Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher's note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.