EEG microstates analysis after TMS in patients with subacute stroke during the resting state

Abstract

To investigate whether intermittent theta burst stimulation over the cerebellum induces changes in resting-state electroencephalography microstates in patients with subacute stroke and its correlation with cognitive and emotional function. Twenty-four stroke patients and 17 healthy controls were included in this study. Patients and healthy controls were assessed at baseline, including resting-state electroencephalography and neuropsychological scales. Fifteen patients received lateral cerebellar intermittent theta burst stimulation as well as routine rehabilitation training (intermittent theta burst stimulation–RRT group), whereas 9 patients received only conventional rehabilitation training (routine rehabilitation training group). After 2 wk, baseline data were recorded again in both groups. Stroke patients exhibited reduced parameters in microstate D and increased parameters in microstate C compared with healthy controls. However, after the administration of intermittent theta burst stimulation over the lateral cerebellum, significant alterations were observed in the majority of metrics for both microstates D and C. Lateral cerebellar intermittent theta burst stimulation combined with conventional rehabilitation has a stronger tendency to improve emotional and cognitive function in patients with subacute stroke than conventional rehabilitation. The improvement of mood and cognitive function was significantly associated with microstates C and D. We identified electroencephalography microstate spatiotemporal dynamics associated with clinical improvement following a course of intermittent theta burst stimulation therapy.

Introduction

Stroke leads to changes in a wide range of structural and functional brain networks (Salvalaggio et al. 2020). Several studies have demonstrated that although stroke is mostly a focal lesion, the reduction in the connectivity of anatomical/functional areas associated with stroke after its onset often leads to altered integration of brain networks and affects whole-brain function (Munsch et al. 2016). (Xu et al. 2019) identified a significant decline in the overall effects of brain networks among stroke patients experiencing depression, and another study showed that cognitive dysfunction in patients with mild stroke was because of extensive disruption of brain cognitive networks (Marsh et al. 2022). The incidence of cognitive dysfunction in poststroke patients is about 6% to 27% within 3 mo (Droś et al. 2023), whereas the probability of depression 5 yr after stroke is about 30% (Ayerbe et al. 2011). Poststroke mood disorders and cognitive impairment decrease the effectiveness of rehabilitation, seriously affecting the recovery process and increasing the recurrence rate and mortality of stroke (Xu et al. 2022).

In recent years, intermittent theta burst stimulation (iTBS) has become a hot research topic. iTBS not only rapidly modulates brain activity to treat depression (Li et al. 2018) but also can enhance cortical excitability of target areas during the stimulation of the brain region of interest (Suppa et al. 2016; Razza et al. 2023). It was found that iTBS over the left dorsal lateral prefrontal cortex (LDLPFC) significantly reduced depressive symptoms in depressed patients, whereas changes in their salience network and central executive network occurred perhaps as a treatment-specific network effect produced by iTBS (Struckmann et al. 2022). iTBS significantly improved associative memory performance and increased resting-state functional connectivity in the hippocampus and other brain regions (Chen et al. 2022). Consequently, iTBS has modulatory effects on both cognition and mood. Additionally, it has been extensively applied in psychiatric and neurological disorders because of its high efficiency, which requires only a few minutes for a session compared with 30 min or more for other modalities. It has been suggested that the use of iTBS as a neuromodulation technique not only modulates neural circuits but also allows for the examination of neural circuit abnormalities in combination with electroencephalography (EEG), which may provide new solutions for clinical practitioners to treat corresponding diseases and serve as an important guide for future development (Ferrarelli and Phillips 2021).

Anatomically, the output fibers of the cerebellum project to different cortical areas of the brain, which are involved in the processing of cognitive, emotional, and other non-motor functions. The cerebellum is regarded as a central node that integrates information from various neural pathways to perform relevant functions. Therefore, any disruptions in the cerebellar neural network can widely affect emotional, cognitive, and behavioral changes in the brain (Gill and Sillitoe 2019). For example, the cerebellar lesion can result in clinically cerebellar cognitive-emotional syndrome, known as Schmahmann’s syndrome, which is characterized by deficits in executive function, spatial cognition, and neuropsychiatric regulation (Hoche et al. 2018). Currently, theoretical models of the cerebellar-cortical loop suggest that efferent signals from the cerebellar nuclei project to the ventral lateral thalamus nucleus, which project sequentially to the various cortical regions, including the frontal, prefrontal, and posterior parietal cortices. Later, a functional connectivity study revealed that the cerebellum is connected to limbic structures (Sang et al. 2012).

Currently, the protocol of iTBS over the cerebellum has been applied to stroke patients clinically. For example, it has been demonstrated that this protocol is effective for stroke patients in not only ameliorating gait and balance (Koch et al. 2019; Liao et al. 2021) but also improving visuomotor learning (Bonnì et al. 2020), visuospatial neglect (Cao et al. 2016), and swallowing function (Rao et al. 2022). iTBS on the ipsilateral M1 area can promote upper limb recovery in subacute stroke patients (Meng et al. 2020). Similarly, the cerebellum could be an effective target for neurological recovery in stroke patients using iTBS therapy. Previous studies selected LDLPFC as the main stimulation target for iTBS to ameliorate emotional and cognitive functions. As we have previously noted, the cerebellum is involved in the processing of cognition and emotion. Consequently, we propose the hypothesis: Could iTBS stimulation of the lateral cerebellum not only improve postural balance but also enhance the brain’s processing of emotions? However, only a few studies focused on this issue (Bonnì et al. 2020). Therefore, the purposes of the present study were 2-fold. First, we need to verify the effect of balance motor function after iTBS over the lateral cerebellum. Second, we sought to investigate whether iTBS over the cerebellum could effectively modulate neural circuits based on the theoretical basis of cerebellar-limbic system-frontal parietal cortical loops to improve emotional and cognitive disorders in patients with subacute stroke.

EEG—as one of the neurophysiological techniques—is popular because of its relatively low cost and flexibility. Accumulating studies have demonstrated that using EEG microstate (MS) analysis may help elucidate the dynamics of large-scale brain networks in poststroke patients (Zappasodi et al. 2017; Hao et al. 2022; Wang et al. 2022) and assess the likelihood of functional recovery. MS, also known as “atoms of thought,” represent neuronal activity in a certain state of the brain and are widely used in the extraction of electrophysiological indicators and prognosis analysis of neurological and psychiatric disorders (Zappasodi et al. 2017; Schumacher et al. 2019; Gold et al. 2022; Hao et al. 2022). EEG MS are semi-stable topographical maps of voltage topography. One state is often stable for 80 to 120 ms and then changes to another state. MS commonly include 4 typical categories, namely A to D, which have their own specific topographic spatial and temporal dimensional characteristics (Michel and Koenig 2018). The temporal characteristics of the 4 MS and the parameters of their inter-transitions can provide important spatial-temporal information, known as MS dynamics (Hao et al. 2022). MS dynamics is an effective technical tool for studying neuronal activity in the changing nodes of brain networks. For instance, the combination of EEG and functional magnetic resonance imaging (MRI) can be used to study brain network activity and explore the processing mechanism of brain information. It has been shown that the changing patterns of brain activation are significantly associated with MS templates (Britz et al. 2010; Rajkumar et al. 2021; Mikutta et al. 2023). Similarly, 4 MS were significantly associated with the activation areas of the brain and the resting-state local brain network activity: MS A (temporal lobe, auditory network), MS B (occipital lobe, visual network), MS C (cingulate cortex and inferior frontal cortex, salience network), and MS D (frontal and parietal lobes, attention network) (Custo et al. 2017; Tarailis et al. 2023). In summary, we initially investigated MS dynamics alterations in patients with subacute stroke compared with healthy controls (HC). Building on the theoretical framework of cerebella-limbic system-parietal cortical loops, we hypothesized that MS parameters corresponding to the frontoparietal attentional network and cingulate cortex, as well as the inferior frontal cortex salience network, may exhibit characteristic changes following iTBS over the lateral cerebellum. Therefore, our study aimed to investigate whether iTBS over the lateral cerebellum could effectively modulate the cerebella-limbic system-parietal circuit, which may affect the neural network of the frontoparietal lobe and limbic system, reorganize the functional network of the brain, and specifically change the MS dynamics associated with them, thereby improving the emotional and cognitive function in subacute stroke patients.

Materials and methods

Participants

Twenty-four patients who met the inclusion and exclusion criteria and 17 healthy volunteers were recruited from January 2022 to July 2023 at the Department of Rehabilitation Medicine, Second Affiliated Hospital of Kunming Medical University. The inclusion criteria were as follows: (i) first unilateral stroke patients with computed tomography or MRI diagnosis, (ii) aged > 18 yr old, (iii) subacute stroke patients (over 2 wk and <6 mo), (iv) stable condition, and (v) signed informed consent. The exclusion criteria were as follows: (i) the presence of other neurological or psychiatric disorders, (ii) stroke involving bilateral hemispheres, (iii) the current use of psychotropic medications during the whole trial procedure, and (iv) individuals with contraindications to TMS, such as the presence of metals in the body, epilepsy. The patient enrollment flowchart is presented in Fig. 1. Volunteers were >18 yr old, in good health, and without underlying diseases. The study was approved by the Medical Ethics Committee of the Second Affiliated Hospital of Kunming Medical University (ethics number: shen-PJ-ke-2022-22). This trial was registered in the Chinese Clinical Trial Registry (No. ChiCTR2200058553).

Fig. 1.

Patient enrollment flowchart. TMS, transcranial magnetic stimulation.

Experimental settings

Healthy volunteers were defined as HC group. All patients were divided into 2 groups depending on the type of intervention: the iTBS combined with routine rehabilitation training group (iTBS–RRT group) and the routine rehabilitation training group (RRT group). All patients were defined as patients’ group (PT group) at T0. All patients were assessed twice to collect basic demographic information, clinical assessment scales, and resting-state EEG, at admission (T0) and after 2 wk of rehabilitation training (T1). For HC, basic demographic information and resting EEG were collected after recruitment. The schematic diagram of the entire experimental design is shown in Fig. 2.

Fig. 2.

Schematic presentation of the experimental design. Clinical scales and resting state EEG were assessed only at baseline (T0).

E‌EG acquisition

A Neuroscan EEG recorder, Greal V2EEG amplifier, and Curry8.0 software were used for EEG data acquisition and recording. According to the international 10 to 20 electrode arrangement system, 32 conductive poles were placed, and EEG was acquired from subjects in the resting state with eyes closed, requiring a sampling rate of 512 Hz and a mean scalp resistance of <5 kΩ. All participants were instructed to clean their scalp beforehand, assume a relaxed seated position in a sound-attenuated room, and keep their eyes closed during the 6-min recording of resting EEG signals. Participants were advised to minimize blinking, as well as head and body movements.

Clinical assessments

Basic demographic information was recorded, including age, sex, etc. Clinical assessment scales involved in several aspects of balance, emotional, and cognitive function. The balance aspect was assessed by the Berg Balance Scale (BBS). The emotional aspect was assessed uniformly using the Hamilton Depression Scale (HMDS), the Hamilton Anxiety Scale (HAMA), the Montgomery Asberg Depression Self-Rating Scale (MADRS), and the Depressive Symptom Self-Rating Scale 30 (IDSSR). For cognitive assessment, the Mini-mental State Examination Scale (MMSE), the Montreal Cognitive Assessment Scale (MoCA), the Working Memory Test (WMT), and the Boston Naming Test (BNT) were used.

iTBS stimulation protocol

Patients in the iTBS–RRT group underwent iTBS therapy for 4 to 5 min once a day, 5 times a week for 2 wk. Each session started by stimulating the lateral aspect of the contralesional cerebellum (the occipital carina was 1 cm laterally). The parameters of iTBS were as follows: the intensity of iTBS was set at 80% of the active motor threshold; 1,200 pulses per stimulation were designed (Huang et al. 2005). A Magnetic Stimulator (NS5000, YIRUID Medical, Wuhan, China) and a double-cone and butterfly coil (B9076, with 92 mm diameter) were used, with the coil tangential to the scalp and the handle pointing upward.

Data preprocessing

EEG data were preprocessed in MATLAB (R2020b, MathWorks, Natick, MA, USA) using the EEGLAB toolbox (version 2019.0, Swartz Center for Computational Neuroscience, San Diego, CA, USA), and its extensions with some custom MATLAB scripts were combined to perform offline preprocessing. First, the resting-state EEG data were low-pass filtered at 20 Hz, high-pass filtered at 2 Hz, and concave filtered at 48 to 52 Hz. The data were divided into 2-s segments for baseline correction; then, “bad channels” as well as EEG segments were manually removed and replaced with spherical spline interpolation. Next, the EEG data were referenced using whole-brain averaging. Finally, the EEG data were decomposed using independent component analysis in the EEGLAB toolbox to detect any artifactual components (e.g. blink, movement, or muscle artifacts) and manually rejected using visual inspection. The epochs with extreme values (amplitudes exceeding ± 100 μV) were removed.

MS analysis

The Microstate EEGlab toolbox—an EEGLAB-based plug-in—was used for MS analysis to extract MS feature’s from the preprocessed resting-state EEG data and statistical analysis of the feature parameters was performed (Fig. 3C).

Fig. 3.

MS topographies, representation, and study pipeline. A) Four clustered prototype microstate voltage maps. B) The EEG microstate segmentation of one subject was transformed into a representative sample of Global Field Potential (GFP). C) Analysis flowchart. T0, baseline or the time of admission; T1, after 2 weeks of training or the time of discharged.

First, the global field power (GFP) was calculated for each subject to obtain raw EEG topography of the GFP peak, which corresponded to the most stable MS at this time because of its high signal-to-noise ratio. Then the original topographic maps were randomly selected from each subject and stitched together and clustered using modified k-means. N original topographic maps were randomly selected as template maps. The original topographic maps were assigned to a certain class of template maps by calculating spatial correlation/similarity, and then the original maps of the same class were averaged to obtain new N template maps. This was iterated continuously until the preset maximum number of iterations reached 1,000 times. Resting EEG MS analysis generally uses the cross-validation criterion (CV principle) to determine the optimal number of MS. The optimal number of MS was set to 3 to 8, and the optimal number obtained by the final clustering was 4. After the above steps, 5 group-level MS were obtained; 4 MS were fitted to individual EEG data using spatial correlation/similarity metrics, temporal smoothing was performed, and MS characteristics were finally calculated. Occurrence was defined as the average number of occurrences of a MS per second. Duration was defined as the average duration, in milliseconds, before a MS switched to the next MS. The coverage time was the percentage of the total recording time an MS remained active. The transition probability was a 4 × 4 2D matrix that represented the probability of a transition between 2 MS. Global explained variance (GEV) was defined as the percentage of variance of the EEG interpreted at all time points for each MS template measured. The comparison of 4 MS classes (A to D) between groups was conducted through the Topographic Analysis of Variance approach, employing the Ragu program for implementation.

Statistical analyses

Formal analysis was conducted in MATLAB 2020b, IBM SPSS Statistics (v.26.0), and R Statistical Package (v2023.09.0-463). Chi-square test was used to compare categorical variables in demographic characteristics between groups. Two-way ANOVA was performed (group * time). The type of intervention/condition was treated as a between-group factor (iTBS–RRT vs. RRT), whereas time factor was defined as a within-factor (T0 vs. T1). The interaction between the 2 factors on the MS parameters was analyzed. If there was an interaction, simple effect analysis was used to determine how the within-subjects factors affected the MS parameters. Afterwards, paired t-test was used to compare differences between paired groups (T0 vs. T1). To compare MS characteristics between HC group and PT group, 1-way ANOVA was employed. Spearman correlation analyses were used to assess the correlation between clinical scale scores and resting-state EEG MS features in patient groups. A P-value of <0.05 was considered statistically significant, and all statistical tests were 2-tailed.

Results

Baseline characteristics of subjects

Twenty-four patients (15 in the iTBS–RRT group and 9 in the RRT group) and 17 healthy volunteers were enrolled. The demographic and clinical information of subjects are summarized in Table 1. The 3 groups did not significantly differ in terms of basic demographic information, except that subjects in the HC group were significantly younger than those in the iTBS–RRT and RRT groups (P = 0.013). The stroke type was significantly different between the iTBS–RRT group and the RRT group (P = 0.027).

Table 1.

Demographic and clinical information.

	iTBS–RRT group (N = 15)	RRT group (N = 9)	Healthy control (N = 17)	F/x²	P
Age(M ± SD)	57.200 ± 6.405	56.778 ± 8.012	48.059 ± 11.344	4.854	0.013a
Sex(male/female)	13/2	5/4	8/9	5.695	0.058
Handedness(left/right)	1/14	0/9	0/17	1.777	0.411
Lesion side(left/right)	6/9	3/6		0.107	0.744
Stroke type(hemorrhagic/ischemic)	1/14	4/5		4.867	0.027a
Days poststroke at T0(M ± SD)	48.07 ± 54.185	56.44 ± 4.661		0.400	0.693

^a P < 0.05. M ± SD, mean ± standard deviation.

Clinical assessments

Compared with those at T0, MMSE, MoCA, WMT, BNT, and BBS scores increased significantly in the iTBS–RRT group at T1 (P = 0.004, 0.002, 0.016, 0.000, and 0.017, respectively), whereas HMDS and MADRS scores decreased significantly (P = 0.002 and 0.001, respectively) (Fig. 4A to C, E to G, and I). Additionally, BNT and BBS scores increased significantly at T0 compared with those at T1 in the RRT group (P = 0.006 and 0.002, respectively) (Fig. 4G and I), whereas HAMA and IDSSR scores decreased significantly (P = 0.035 and 0.023, respectively) (Fig. 4D and H).

Fig. 4.

Results of paired t-test for clinical scales. A to I) a to i plots represented separately the comparisons of MMSE, MoCA, HMDS, HAMA, MADRS, WMT, BNT, IDSSR, and BBS scores between the iTBS–RRT or RRT group at T0 and T1. Results with significant differences had been emphasized by a box that wrote sig.

MS template

The optimal number of clusters derived using the K-means clustering algorithm in our study was 4, similar to the 4 most classic MS topographies determined in several independent studies, which is consistent with most previous studies on resting-state MS (Koenig et al. 1999; Van de Ville et al. 2010; von Wegner et al. 2021). We manually labeled each set of MS clustering maps as A to D. The 5 sets of MS categories are shown in Fig. 5. On average, these clustering maps accounted for 52.10%, 52.40%, 55.50%, 52.70%, and 55.20% of the GEV in the HC, iTBS–RRT (T0), iTBS–RRT (T1), RRT (T0), and RRT (T1) groups, respectively. No significant variations were observed in the individual MS topographies of each MS class. The topographies generated in HC closely resembled those reported in the literature.

Fig. 5.

MS templates. T0, baseline or the time of admission; T1, after 2 weeks of training or the time of discharged.

MS parameters

Comparison of MS parameters between HC group and all patients at T0 (PT group)

The coverage, duration, GEV, and occurrence of MS D were significantly higher in HC group than that in PT group (Fig. 6). However, the GEV of MS C was significantly higher in PT group compared with that in HC group.

Fig. 6.

Results of between-group ANOVA of MS parameters in HC group and all patients at T0.

Compared with the HC group, the PT group exhibited increased transition probabilities from MS B to MS C and MS C to MS B, as well as decreased transition probabilities from MS A to MS D and MS B to MS D (Fig. 7).

Fig. 7.

Results of the between-group ANOVA for HC and all patients at T0 with respect to each MS transition probability.

Comparison of MS parameters at T0 and T1 between iTBS–RRT and RRT groups

The coverage of both MS C and MS D demonstrated a significant group effect (F_1,44 = 10.182, 5.412; P = 0.003, 0.025; ηp² = 0.188, 0.110, respectively). Moreover, a significant group-by-time interaction effect in the coverage of MS C was observed (F_1,44 = 4.311, P = 0.044, ηp² = 0.089). Simple effect analysis indicated that MS C coverage in the iTBS–RRT group was higher than that in the RRT group at T0 (Fig. 8).

Fig. 8.

Results of 2-way ANOVA with respect to group (iTBS–RRT vs. RRT) and time (T0 vs. T1) for each MS.

The mean duration of MS C exhibited a significant group effect (F_1,44 = 4.674, P = 0.036, ηp² = 0.096), suggesting that the MS C duration in the iTBS–RRT group was higher than in the RRT group. However, group and time interaction effects for all MS classes duration were not significant (Fig. 8).

The GEV of MS A, MS C, and MS D revealed a significant group effect (F_1,44 = 5.042, 11.897, 4.753; P = 0.030, 0.001, 0.035; ηp² = 0.103, 0.213, 0.097, respectively). The GEV of MS A and MS D in the iTBS–RRT group were lower compared with that in the RRT group, whereas the GEV of MS C in the iTBS–RRT group was lower higher compared with that of the RRT group (Fig. 8). There was no significant time by group interaction effect for all MS classes GEV.

The occurrence of MS C and MS D showed significant group effects (F_1,44 = 10.196, 6.126; P = 0.003, 0.017; ηp² = 0.188, 0.122, respectively). The time effect for MS D occurrence was significant (F_1,44 = 4.267, P = 0.045, ηp² = 0.090). In addition, the time by group interaction effect for MS C occurrence rate was significant (F_1,44 = 10.642, P = 0.002, ηp² = 0.195). Simple effect analysis indicated that MS C occurrence in the iTBS–RRT group was higher than that in the RRT group at T0 (Fig. 8).

The group effects revealed that the expected transition probabilities from MS A to MS C (F_1,44 = 7.632, P = 0.008, ηp² = 0.148), from MS B to MS C (F_1,44 = 11.365, P = 0.002, ηp² = 0.205), and from MS D to MS C (F_1,44 = 15.874, P = 0.000, ηp² = 0.265) were greater in the iTBS–RRT group. However, the expected transition probabilities from MS A to MS D (F_1,44 = 9.632, P = 0.003, ηp² = 0.180), from MS B to MS D (F_1,44 = 16.366, P = 0.000, ηp² = 271), from MS D to MS A (F_1,44 = 5.287, P = 0.026, ηp² = 0.107), and from MS D to MS B (F_1,44 = 6.125, P = 0.017, ηp² = 0.122) were lower in the iTBS–RRT group (Fig. 8). The time effect for transition probability from MS A to MS C was significant (F_1,44 = 4.708, P = 0.035, ηp² = 0.097). Furthermore, the time by group interaction effects in expected transition probabilities from MS A to MS C (F_1,44 = 11.540, P = 0.001, ηp² = 0.208) and from MS B to MS C (F_1,44 = 5.834, P = 0.020, ηp² = 0.117) were significant. Simple effect analysis indicated that the transition probabilities from MS A to MS C and from MS B to MS C in the iTBS–RRT group were higher compared with those of the RRT group at T0 (Fig. 9).

Fig. 9.

Results of 2-way ANOVA for each MS transition probability with respect to group (iTBS–RRT vs. RRT) and time (T0 vs. T1).

Correlation between MS parameters and clinical scales at T0 and T1 in iTBS–RRT and the RRT groups

The correlations between MS parameters and clinical scales at T0 and T1 in iTBS–RRT and the RRT groups are displayed in Figs. 10 and 11. In the iTBS–RRT group, HMDS was significantly negatively correlated with the GEV and occurrence of MS C, and transition probability from MS D to MS C at T0 (Fig. 10A to C). Likewise, HMDS was significantly negatively correlated with MS C occurrence at T1 (Fig. 10I). At T1, there were significant positive correlations between BBS and the coverage as well as GEV of MS A. Conversely, BBS showed significant negative correlation with the duration, coverage, and GEV of MS D (Fig. 10D to H). In the RRT group, at T0, MS C coverage exhibited a positive correlation with IDSSR, whereas the opposite was observed for MS D (Fig. 11A and B). At T1, the occurrence of MS D showed significant positive correlation with MMSE and MoCA, whereas a significant negative correlation was noted between MS D occurrence and IDSSR (Fig. 11C to E).

Fig. 10.

Correlation results between MS parameters in the iTBS–RRT group and clinical scales. A-C) A, B and C plots represented separately the correlation analysis between GEV, occurence of MS C and from MS D to MS C and HMDS in the iTBS-RRT group at T0. D-E) D and E plots represented separately the correlation analysis between coverage, GEV of MS A and BBS in the iTBS-RRT group at T1. F-H) F, G and H plots represented separately the correlation analysis between duration, coverage and GEV of MS D and BBS in the iTBS-RRT group at T1. I) I plot represented the correlation analysis between occurence of MS C and HMDS in the iTBS-RRT group at T1.

Fig. 11.

Correlation results between MS parameters in the RRT group and clinical. scales. A-B) A and B plots represented separately the correlation analysis betweencoverage of MS C and MS D and IDSSR in the RRT group at T0. C-E) C, D and E plots represented separately the correlation analysis between MMSE, MoCA, and IDSSR and occurence of MS D in the RRT group at T1.

Discussion

Stroke-induced neuronal death can lead to changes in neural pathways, which can alter the integration of brain networks, affecting their overall function and the physiological electrical signals they generate. Therefore, the present study applied that the differences in MS parameters between HC and stroke patients were concentrated in MS C and MS D. We also observed that cognitive and emotional scores were significantly improved after iTBS over the cerebellum. And then we explored whether iTBS over the cerebellum caused alterations in brain spatiotemporal dynamics and the relationship between MS parameters and clinically relevant emotional and cognitive scales before and after interventions.

Although no overall treatment-associated differences were noted in the clinical scales assessment, the paired t-test before and after treatment in each group was statistically significant. Compared with the pretreatment group, iTBS–RRT group had a powerfuller trend in improving emotional and cognitive function compared with RRT group in stroke patients, which provided the underlying evidence for our study hypothesis. Zhao et al. (2023) found that adolescents with major depressive disorder (MDD) with non-suicidal self-injurious behaviors had lower Depression Scale scores and Ottawa Self-Harm Scale scores when treated with selective serotonin reuptake inhibitors (SSRIs) combined with rTMS compared with medication (SSRIs) alone. Impaired behavioral inhibition was improved to a greater extent after iTBS intervention in heroin-addicted patients (Kang et al. 2022). Another study showed that cerebellar iTBS reduced error adaptation interference, thereby improving adaptive motor learning in healthy subjects (Koch et al. 2020). Cerebellar stimulation with iTBS in subacute stroke patients improved the visuomotor learning and sustained improvements in recently learned skills if retested (Bonnì et al. 2020).

This study revealed that the declining trend in HMDS and IDSSR scores was linked to changes in MS D and MS C parameters, such as the GEV, coverage of MS C, the transition probability from MS D to MS C, and the coverage and occurrence of MS D. Lei et al. reported that following 8 wk of pharmacological treatment in depressed patients, the occurrence and coverage of MS C were significantly elevated, indicating a positive correlation with improved Hamilton Depression Rating Scale (HAMD) scores. Furthermore, the transition probability from MS A to MS C was positively correlated with HAMD scores at 8 wk. These findings suggest that MS D represents an underlying electrophysiological characteristic of MDD (Lei et al. 2022). Gold et al. utilized TMS to address patients with treatment-resistant depression. In this study, 6 MS categories were grouped, revealing a notable connection. An elevation in the occurrence and coverage of MS 2 (associated with MS C) was markedly linked to ameliorated depressive symptoms and heightened activation in the anterior cingulate cortex, inferior frontal cortex, and right insular areas. Conversely, a reduction in the occurrence and coverage of MS 3 (corresponding to MS D) was associated with symptom improvement and activation in frontal and parietal cortical areas (Gold et al. 2022). MS C had been demonstrated to be linked to be the salience network, playing a role in the integration of interoceptive information with emotional salience (Britz et al. 2010). MS D was associated with the frontoparietal cortex attention network, and the frontoparietal cortex played a crucial role in the regulation of emotions (Li et al. 2021a). These positive findings suggest the crucial role of MS C and MS D in emotional impairment.

Our findings indicated a significant correlation between improvements in cognitive function scores and occurrence of MS D. Musaeus et al. (2019) proposed an association between MS A and cognitive dysfunction. Our results appear to be a novel discovery, and there may also be connection between MS D and cognitive function. The most significant difference in duration was found between patients with MCI who progressed and those who remained stable for MS D. MS D may correspond to underlying pathological changes in the progression from MCI to AD (Musaeus et al. 2019). In conclusion, MS D occurrence is closely linked to cognitive function and is applicable to stroke patients. Nevertheless, our current study suggests that the metrics of MS D and MS C may serve as valuable predictors of stroke-specific cognitive and emotion involvement.

We also observed an interesting phenomenon: there was no correlation between MS metrics and BBS scores at T0. However, after iTBS intervention, the coverage and GEV of MS A showed a significant positive correlation with BBS scores, whereas the coverage, duration, and GEV of MS D exhibited a significant negative correlation. A study suggested that cerebellar iTBS promoted gait and balance recovery in stroke patients by influencing cerebello-cortical plasticity, possibly through the activation of cerebello-thalamo-cortical pathways that targeting parietal-frontal networks (Koch et al. 2019). MS D is recognized for its significant involvement in activating the attention network in the frontal and parietal lobes. Our findings seem to provide supporting evidence for the conclusions drawn in the study by Koch et al.

Our observations revealed that the differences in MS parameters between PT and HC groups mainly implicated MS C and MS D. These findings suggest that stroke and related cognitive and emotional disorders may influence the temporal dynamics of EEG MS, possibly through a distinct neurophysiological mechanism. Hao et al. partially supported our findings, indicating that stroke patients without conventional rehabilitation for 2 wk exhibited higher GEV in MS A and lower GEV in MS C compared with HC. Meanwhile, the GEV in MS D showed no significant different between the 2 groups (Hao et al. 2022). There were also distinct alterations in MS parameters across different hemispheres. In cases where the stroke affected the left hemisphere, the duration of MS C was shorter compared with that of MS D. Conversely, when the stroke occurred in the right hemisphere, the results were reversed (Zappasodi et al. 2017). A prior study revealed that individuals with Parkinson’s disease exhibit prolonged MS durations and lower occurrences compared with healthy individuals (Li et al. 2021b). Another study indicated that epilepsy induced a reduction in the occurrence of MS C (Jiang et al. 2021). Therefore, various brain disorders can lead to specific alterations in MS. However, there was heterogeneity in the results of studies focusing on strokes. In summary, our findings will provide valuable insights for future research on MS within the realm of stroke.

Finally, we aimed to validate the hypothesis at the electrophysiological level. Our results suggested that the differences in microstate parameters before and after treatment between the iTBS–RRT and RRT groups were mainly evident in MS C and MS D. This suggests to us that cerebellar iTBS influences brain activation regions specifically related to MS C and MS D. Evidence from previous investigations have indicated that iTBS can improve patients’ mood and cognitive function. For example, iTBS on the LDLPFC significantly ameliorated depressive symptoms and suicidal ideation in patients with refractory depression without side effects (Cole et al. 2020), and also enhanced overall cognitive and memory function in stroke patients (Tsai et al. 2020). MDD was characterized by decreased metabolism and activity of neurons in the left prefrontal region (Bench et al. 1995), and the application of iTBS on the LDLPFC represented targeted site-specific regulation. However, other studies have indicated that the effects of TMS on brain networks are not limited to the target cortical area, but also to the areas connected to this function, and that disturbances to angular gyrus (core area of the default mode network—DMN) and intraparietal sulcus (core area of the dorsal attention network) can propagate to the corresponding networks to which these areas belong (i.e. DMN and DAN) (Croce et al. 2018). In clinical studies, it was found that intervention at the lateral, vermis, or medial cerebellum in healthy individuals using TMS modulated processing of fearful stimuli (Ferrari et al. 2023). In addition, executive control in attentional functions with the frontoparietal network was also associated with cerebellar activation (Xuan et al. 2016). Liu et al. discovered that alterations in functional connectivity among brain regions, such as the posterior cingulate cortex, left posterior cerebellum, right prefrontal cortex, and the limbic system, were associated with emotion regulation and higher cognitive functions in individuals with MDD. Additionally, alterations were observed in the fronto-limbic system in these patients (Liu et al. 2020). In this study, the application of iTBS over the cerebellum led to a significant increase in cerebellar fibers transmitting signals to the limbic system, frontal lobe, and parietal lobe, a phenomenon referred to as remote modulation of neural circuits. A previous review also highlighted that prevailing evidence indicates that TMS is especially well-suited for the examination and modulation of neural circuits (Ferrarelli and Phillips 2021). Moreover, iTBS over the cerebellum has shown promise in ameliorating negative symptoms in patients with schizophrenia by modifying network pathways in the cerebellar-prefrontal cortex (Brady et al. 2019). When applying single-pulse TMS to the cerebellum in healthy individuals, distinct EEG responses emerge in the contralateral prefrontal and parietal cortex. These responses likely signify distinctive activation patterns within the cerebello-dentato-thalamo-cortical pathway (Gassmann et al. 2022). Furthermore, certain thalamic nuclei and the cingulate cortex are part of the limbic system (Arrigo et al. 2014). Ferrari et al. (2018) suggested that the cerebellum may play a pivotal role in the “limbic” network, affecting not only emotion regulation but also emotion perception and recognition (Ferrari et al. 2018). TMS treatment of depressed patients for 6 wk resulted in a significant improvement of depressive symptoms and a decrease in the coverage and occurrence of MS 3 corresponding to the typical MS D (Gold et al. 2022). With the improvement of cognitive and emotional functions in stroke patients, the parameters associated with MS D in the iTBS–RRT group were lower than those in the RRT group after treatment, whereas the parameters for MS C showed the opposite trend. In a study, it was found that all 3 parameters associated with MS D (duration, occurrence, and coverage) were significantly higher during a task compared with wakeful rest with eyes open, and MS C parameters were lower (Seitzman et al. 2017). However, Milz et al. (2016) observed that, compared with rest, the occurrence and coverage of MS D and the duration of MS C were somewhat reduced during the task. It is also possible that cognitive manipulation of a target MS is not straightforward. Consequently, future work should investigate the impact of tasks on MS C and MS D in individuals with brain injuries. This perspective finds supported in our current study, where iTBS exhibited a tendency to more effectively enhance emotion-cognitive function in stroke patients compared with conventional rehabilitation. This suggests that iTBS over the cerebellum potentially regulates the cerebellar-limbic system-frontoparietal cortex circuit. This regulation impacts the neuronal activity of the salience network in the cingulate cortex and the frontoparietal attention network, and influencing the brain network’s ability to dynamically configure. Thus, the mechanisms underlying the improvement of this symptom may involve the modulation of EEG MS.

There are several limitations in our study. First, we did not visualize the proposed hypothesis of neural pathway modulation using MS analysis combined with functional connectivity or sLORETA source localization. Although our study may not provide direct evidence, we note that numerous previous research studies have validated the findings that MS C indicates activation of the cingulate cortex and inferior frontal cortex’s salience network, whereas MS D represents activation of the frontoparietal attention network (Britz et al. 2010; Coquelet et al. 2022; Mikutta et al. 2023). Second, our observation period of 2 wk may have been too short to observe meaningful changes in certain clinical indicators. Third, some MS studies reported that both intraindividual factors and interindividual factors, such as age and gender, can influence the findings (Zanesco et al. 2020; Jabès et al. 2021). In our study, the relatively young age of the HC significantly differed from that of the patient group, potentially impacting the extrapolation of our results. Of note, we only concentrated on patients with subacute stroke, including those with mild symptoms but not those with severe symptoms. However, the variations in ages, gender, stroke sites, stroke sides, stroke types, and time since strokes impacted our findings (Zappasodi et al. 2017). The distribution of stroke types between the 2 patient groups is significantly asymmetrical, indicating that the results of this study may not fully generalize to the entire stroke population. Consequently, future research with large sample size, longer duration of observation, and follow-up times is advocated to obtain a more in-depth and comprehensive understanding of the relationship between MS characteristics and brain network dynamics in stroke patients.

Conclusion

Significant differences exist in the metrics of MS C and MS D when comparing stroke patients with HC. The application of iTBS over the lateral cerebellum predominantly affects the parameters of MS C and MS D in stroke patients. Moreover, significant correlations were observed between certain parameters of MS C and MS D and mood and cognitive scales. Millisecond MS dynamics seem to encode information related to emotional-cognitive function. In summary, our findings enhance our understanding of brain network dynamics and unveil the impact of iTBS over the cerebellum on emotional and cognitive function in stroke patients through resting-state EEG MS analysis. This study introduces novel perspectives for developing therapeutic approaches targeting neuromodulation in subacute stroke patients.

CRediT statement

Hongmei Zhang (Conceptualization, Formal analysis, Funding acquisition, Methodology, Validation, Writing—original draft, Writing—review & editing), Xue Yang (Conceptualization, Formal analysis, Funding acquisition, Methodology, Validation, Writing—original draft, Writing—review & editing), Liqing Yao (Conceptualization, Formal Analysis, Supervision, Writing—review & editing), Qian Liu (Data curation, Funding acquisition), Yihuan Lu (Data curation, Funding acquisition), Xueting Chen (Writing—review & editing), and Tianling Wang (Writing—review & editing)

Funding

The National Key Research and development Program (2022YFC2009700); the Science and Technology Talent and Platform Program (Academician Expert Workstation) (202305AF150032); Yunnan Provincial Science and Technology Department Plan Project (202203AC100007-6); and Kunming Medical University Innovation Fund (2023S074).

Conflict of interest statement: All authors declare no conflicts of interest. There are no conflicts of interest present in our work and articles.