Effects of brain network segregation and integration on motor imagery sensorimotor rhythm

Abstract

Diverse cognitive processes place different demands on locally segregated and globally integrated brain activities. Motor imagery (MI) is a complex mental operation characterized by sensorimotor rhythms. However, how the brain network acts on MI rhythms is not well-known. The present work aimed to explore the effects of brain integration and segregation on brain rhythmic oscillations. The power spectrum dynamics, topography distribution, and brain network metrics in the alpha and beta bands were calculated. And the correlations were investigated with the network metrics and sensorimotor rhythm. The results showed that the degree of event-related desynchronization/synchronization (ERD/ERS) was higher in alpha band than in beta band during [−1, 1 s] (p < 0.01). The topography of the alpha band demonstrated a bilateral distribution during MI processing, while the beta band had more diffuse distributions around the centre. Moreover, global efficiency was associated with bilateral ERD, and the transitivity was related to contralateral local power. These results suggested that network functions could facilitate the completion of behavior tasks. The integration was related to bilateral hemisphere coordination and the segregation was related to local activation, which shaped the local neural modulation of individuals in MI.

Keywords:

• Motor imagery
• segregation
• integration
• sensorimotor rhythm

1. Introduction

Brain–computer interface (BCI) can provide direct communication between humans and external devices [1,2]. Motor imagery (MI) is a mentally rehearsed manifestation of subjective motor intention [3,4] and usually involves imagining limbs’ actions without performing the actual movement. MI has been widely used in motor rehabilitation, BCI control, etc., to promote communication and functional mobility [5,6].

The specific feature of MI is event-related desynchronization/synchronization (ERD/ERS) [7], which the time scale and spatial distribution can explain. For the time scale, the ERD represents the power decrease in alpha and beta bands occurring during the MI process, while ERS represents the power increase after MI [8,9]. For the spatial distribution, the ERD manifests as a significant power decrease in the contralateral sensorimotor area compared with the ipsilateral area during the MI process [10,11]. It is shown that the distributed specializations of alpha and beta oscillations are closely associated with the long-range inter-regional organization [12]. In MI, the alpha rhythm desynchronization is concentrated in the lateral sensorimotor areas, while the beta rhythm desynchronization is focused on the parietal region [13]. Meanwhile, related studies have shown that the alpha rhythm describes the local processing of motor information, and the beta rhythm acts on the coordination of the sensorimotor tasks [14,15]. However, the relationship between the different rhythms and the functional network architecture during MI is still unclear. As MI involves an extensive frontal–parietal network, not just the primary local brain, large-scale network analyses would effectively explore MI’s spatiotemporal rhythmic characteristics.

In recent years, many studies have focused on the functional connectivity characteristics among cortical regions [16,17], reflecting the statistical temporal covariation among spatial distant brain regions. Functional connectivity during resting could predict task performance [18,19]. Studies have investigated the network function using coherence [20,21] and phase lag index (PLI) [22] and found that the network properties were directly related to MI–BCI performance [23,24]. And for the brain function, the segregation and integration of information processing across large scale-networks can quantify it. Relevant studies have shown that the segregation and integration functions of the resting state will reach a balance to better complete the corresponding cognitive tasks [25], and different cognitive tasks have additional requirements for segregation and integration [26]. A higher degree of segregation is associated with simple tasks, and a higher degree of integration seems to underlie task performance with a more significant cognitive load [27,28]. In motor-related studies, higher segregation levels were associated with successful motion execution and faster processing [29]. Therefore, the relationship between functional network characteristics and the oscillations of different frequencies may provide a new perspective for understanding individual tasks’ performance.

Therefore, the present work explored the spatiotemporal characteristics during different MI periods and examined the effects of brain integration and segregation functions during MI resting intervals on the rhythm metrics of brain oscillations during the MI process, which may contribute to a deeper understanding of how the primary brain network in resting state regulates the sensorimotor rhythms when performing a motor task.

2. Experimental procedures

2.1. Subjects

29 college students (aged 19–27 years old, average age 22.03 years old, two females) were recruited. All subjects were right-handed, had no history of nervous system dysfunction, took no drugs that affected the central nervous system acutely and chronically, and did not consume or inhale any excitatory beverage, food, or gas before the experiment. This experiment was conducted according to the guidelines of the Declaration of Helsinki and was approved by the Ethics Committee of the University of Electronic Science and Technology of China (UESTC). Written informed consent was obtained from all participants.

2.2. Data collection

EEG datasets were recorded by 32 Ag/AgCl electrodes that complied with the 10/20 international electrode placement system (Biosemi Amplifier). The sampling rate of data recording was 2048 Hz. The electrode Fz served as the reference. For all electrodes, the impedance was kept below 5 KΩ during the EEG recording.

2.3. Experimental paradigm

The experiment used the classic MI paradigm, as shown in Figure 1. The stimulus paradigm was prepared in the MATLAB (v2019b) psychology toolbox (Psychtoolbox, v2021) to present the experimental stimuli accurately and synchronously. During the experiment, the subjects completed two repeated MI tasks.

Figure 1. The procedure of the experimental paradigm. (A) one trial of the paradigm, (B) one session of the paradigm.

Each trial contained rest, preparation, and MI tasks. The experimental flow of each trial is shown in Figure 1. When two grey rectangular bars appeared on the screen, the subjects rested. Then, the rectangular position on the screen would change, with the left or right sides turning yellow randomly, lasting for 1 s, which was to prompt the subjects to prepare the MI of the hand in the corresponding direction (Note: it was not required for them to start imagining during the presentation of the prompt). Finally, when the yellow rectangle turned green, the subjects began to imagine the hand movement. The imagination process lasted for 6 s, and the subjects were required to continue imagining throughout the MI period. The imagined movement actions were extension and flexion of one arm. The subjects performed a total of 80 trial tasks in two sessions, including 40 trial tasks imagining the movement of the left hand and 40 trial tasks of the right hand.

3. Methods

3.1. Signal preprocessing

The offline EEG data were analyzed using EEGLAB toolbox [30] (v2021) in MATLAB 2019 b (MathWorks, Natick, NA) as follows: First, the electrodes were located, and the Reference Electrode Standardization Technique (REST) [31] was used to re-reference all EEG data; we checked the data and interpolated the bad electrodes; filtered the data with a 5–30 Hz bandpass filter to obtain the relevant frequency band information; down-sampled the data to 512 Hz; and segmented the data within the period of [−1, 9 s]. After the above steps, the segmented data were checked and put into the independent component analysis (ICA), weakening the disturbance of the electrooculogram (EOG) and EMG. For EOG, a low frequency-dominated power distribution was consistently observed in the prefrontal electrodes, while EMG was distributed above 20 Hz and could be found in most of the electrodes. Finally, the threshold [−75, 75 µV] was set to reject the bad trials with residual EOG artifacts (according to the EEG amplitude range, a trial with >75 µV was considered a bad trial) [32].

3.2. EEG power calculation

The power spectrum of bilateral areas in the left- and right-hand MI tasks was obtained. The power spectral density (PSD) was calculated using the pwelch method [33]. The data of each trial were segmented through the sliding window. Then, each trial’s power spectrum of each subject was averaged to obtain their mean power spectrum.

In addition, we calculated the power difference in the contralateral compared with the ipsilateral to explore the changes in ERD and ERS of sensorimotor rhythms in different frequency bands (alpha and beta) during the MI process. Therefore the normalized Δpower was estimated as follows [34]: (1) $$\Delta \mathit{power}=\frac{\mathit{OSP}-\mathit{SSP}}{\mathit{OSP}+\mathit{SSP}}$$(1) where OSP is the PSD in the contralateral electrodes during the MI tasks, SSP is the PSD of the ipsilateral electrodes, and Δpower is the normalized difference between the contralateral and ipsilateral areas. All the electrodes were located in the left or right sensorimotor areas (left: FC5, FC1, C3, CP1, CP5, and T7; right: FC2, FC6, C4, CP2, CP6, and T8).

3.3. Time-varying power distribution

Time-frequency representation captures the variation in the spectral content of a signal over time, for which the short-time Fourier transform (STFT) is one of the most basic techniques [35], mainly to solve the two limitations of Fourier in spectral analysis for processing non-stationary signals and time-dependent power values. To present the time-varying power distribution, we calculated the power spectrum of all electrodes over time by STFT.

STFT was estimated as follows: (2) $$Y\left(t,f\right)=\int y\left(\tau \right)h(\tau -t)e^{-j2\pi ft}d\tau$$(2) where $y(\mathrm{t})$ is the given signal and $h(t)$ is the window function. Therefore, the spectrogram was (3) $$S={\left|Y\left(t,f\right)\right|}^2$$(3)

All data were filtered into two frequency bands (8–13 Hz and 14–30 Hz). The spectra of all electrodes were drawn into a topography, which displayed the power distributions of the alpha and beta bands in different time epochs during the MI period.

3.4. Brain network analysis

To obtain the properties of the brain network in resting intervals (the resting state following the MI process and preparation), we constructed a brain network based on imaginary coherence. Then, to quantify brain functional integration and segregation, the functional network connectivity indices: global efficiency and transitivity, were calculated based on graph theory. Global efficiency could estimate brain region communication, and transitivity could reflect the node’s relationship and neighbors [29,36].

For the network analysis, although the volume conduction may affect the connectivity measures obtained from the scalp EEG [37], the imaginary coherence was insensitive to false connectivity arising from volume conduction. Therefore, we used the imaginary to estimate the linear relationship between two EEG channels. The imaginary calculation formulas were as follows:

If $x_i(t)$ and $x_j(t)$ are the time series of channels i and j, their complex Fourier transform can be defined as [38]: (4) $$x_i\left(t\right)\stackrel{F}{\to }{x}_i\left(f\right)$$(4) (5) $$x_j\left(t\right)\stackrel{F}{\to }{x}_j\left(f\right)$$(5)

Therefore, the cross-spectrum was (6) $$S_{ij}\left(f\right)=<x_i\left(f\right)x_j^{\mathrm{*}}\left(f\right)>$$(6) where * represents the complex conjugate, and < > is the expectation, averaging over the trials during the MI process. The coherence was defined as the normalized cross-spectrum: (7) $$C_{ij}\left(f\right)=\frac{S_{ij}\left(f\right)}{{\left(S_{ii}\right(f\left)S_{jj}\right(f\left)\right)}^{1/2}}$$(7)

The absolute of the imaginary coherence was: (8) $${\mathit{Coh}}_{ij}\left(f\right)=\left|\mathit{imag}\left(C_{ij}\right(f\left)\right)\right|$$(8)

From the above steps, the imaginary coherence matrix was obtained, and the transitivity and global efficiency were calculated as follows [36]: (9) $$k_i=\sum _{j\in N}{a}_{ij}$$(9) (10) $$t_i=\frac{1}{2}\sum _{j,h\in N}{a}_{ij}{a}_{ih}{a}_{jh}$$(10) (11) $$\mathrm{T}=\frac{{\sum }_{i\in N}2t_i}{{\sum }_{i\in N}{k}_i(k_i-1)}$$(11) where $a_{ij}$ is the connection status between nodes i and j: $a_{ij}$ = 1 when link (i, j) exists (when i and j are neighbors); $a_{ij}$ = 0 otherwise. $k_i$ is the degree of node i; $t_i$ is the number of triangles around node i; and T is the transitivity of the network. (12) $$d_{ij}=\sum _{a_{uv}\in gi\leftrightarrow j}{a}_{uv}$$(12) (13) $$E=\frac{1}{n}{E}_i=\frac{1}{n}\sum _{i\in N}\frac{{\sum }_{j\in N,j\ne i}{d}_{ij}^{-1}}{n-1}$$(13) where $d_{ij}$ is the shortest path length (distance), between nodes i and j, $gi\leftrightarrow j$ represents the shortest path (geodesic) between i and j, and E is the global efficiency of the network.

3.5. Statistical analysis

Group-level statistical tests were conducted for the bilateral power change and the network metrics in the alpha and beta bands. Before the statistical test, the data distribution was examined using the Kolmogorov–Smirnov test.

A paired sample t-test was performed for each paired data. Then, a repeated measurement variance analysis [39] was performed to test the significance of the differences in transitivity and global efficiency among the groups with task and frequency conditions (task: left-MI, right-MI; frequency: alpha, beta). All statistical thresholds were set to p < 0.05, and the p value was subjected to Bonferroni correction [40]. In addition, aiming to reveal the relationship between the brain network and the sensorimotor rhythms of MI, the Pearson correlation coefficient was calculated between brain network transitivity and contralateral sensorimotor rhythm and between the global efficiency and the ERD of the MI task.

4. Results

4.1. Analysis of topography results

We analyzed the alpha and beta rhythms’ topography distribution and spectrum changes. Figure 2 shows the time-varying topographies of all subjects in the alpha and beta bands (8–13 Hz and 14–30 Hz, respectively). The topography’s color bar on the right side was the power value. The distributions of topography between alpha and beta bands are different. The left- or right-hand movement was related to the bilateral desynchronization of the alpha band, and contralateral desynchronization was more evident during the MI process. However, the beta band had more diffuse distributions around the center, and the beta distributions in the central area remained almost unchanged over the whole MI process.

Figure 2. Topography of the alpha and beta bands during the left- and right-hand MI. (A) topography of the alpha band during the left MI task; (B) topography of the beta band during the left MI task; (C) topography of the alpha band during the right MI task; (D) topography of the beta band during the right MI task. The color bar on the right side represents the power value of all the topographies (the redder the color is, the greater the power value). The cue preparation was in [−1, 0 s], the MI ERD was in [0, 6 s], and the rest was in [6, 9 s].

In addition, the power distribution of other frequencies, including delta (1–4 Hz), theta (4–7 Hz), and gamma (30–45 Hz), were also taken into consideration during the study of MI. We found that the distributions of the delta and theta bands were similar to the topography of the alpha band, reflecting the specific power distribution in the sensorimotor area. Meanwhile, the distribution of the gamma band showed diffusion characteristics similar to those of the beta band, but it shifted to the parieto-occipital location (Supplementary material Figures S1–S3).

4.2. Alpha and beta ERD

To calculate the power change during the MI period and the resting intervals of alpha and beta bands, the data in [−1, 1 s] were computed using STFT. Each second was divided into eight segments (0.125 s/segment) through the window with 64-width. The Δpower of the left and right sensorimotor areas (left: FC5, FC1, C3, CP1, CP5, and T7; right: FC2, FC6, C4, CP2, CP6, and T8) in the two frequency bands are displayed in the following histograms (Figure 3). For the preparation stages of all trials in [−1, 0 s], Δpower has changed, which manifested as one of the segments during this period with Δpower < 0, indicating that contralateral ERD existed in the alpha and beta frequency bands during MI preparation. In addition, in [0, 1 s], all Δpower < 0 proved that the power of the contralateral was lower than that of the ipsilateral, that is, the ERD during the MI process. However, for the quantification results of Δpower, the other frequency bands (delta, theta, and gamma) did not show significant dynamic characteristics of the change from ERS to ERD during the MI process (Supplementary material Figure S4).

From the above results, the ERD of the alpha and beta bands appeared in the motor preparation stage. A paired sample t-test was performed for each data pair to compare the power change statistically. The statistical results in Figure 3 show a significant difference in the Δpower between the alpha and beta bands. During the preparation stage [−1, 0 s], compared with the beta band, the value of Δpower in the alpha band was larger (left: alpha–beta, p = 0.001, t = 3.543, df = 28; right: alpha–beta, p = 0.002, t = 3.326, df = 28). However, during the MI process [0, 1 s], the alpha band Δpower was lower than the value of the beta band (left: alpha–beta, p = 0.28 > 0.05, t = −1.068, df = 28; right: alpha–beta, p = 0.03, t = −2.225, df = 28), indicating that the degree of ERD in the alpha band was more obvious.

Figure 3. The Δpower of the alpha and beta bands in [−1, 1 s]. (A) The time-varying Δpower of alpha and beta bands in the left and right MI task (the Δpower of the alpha band in the left MI task; the Δpower of the alpha band in the right MI task; the Δpower of the beta band in the left MI task; the Δpower of the beta band in the right MI task), in which the x-axis represented the time scale, and the y-axis represented the Δpower value (note: Δpower was the relative value between the ipsilateral and contralateral areas). (B) The statistical contrasts between alpha and beta bands in [−1, 0 s] and [0, 1 s]. Where the black columns are the alpha Δpower, and the grey columns are the beta Δpower. (* represents significance p < 0.05, and ** represents significance p < 0.01).

4.3. Correlation analysis between brain network functions and the sensorimotor rhythms in the alpha and beta bands

We compared alpha and beta bands’ global efficiency and transitivity and explored the interaction between brain network functions and sensorimotor rhythms. Before the statistical test, the data distribution was examined used by Kolmogorov–Smirnov. The Pearson correlation coefficient was calculated for the network functions and the sensorimotor rhythms, and the results are shown in Figure 4.

Figure 4. The correlation between brain network functions and the sensorimotor rhythms of the alpha and beta bands (r is the correlation coefficient, and p is the statistical value). (A) The correlation between the transitivity in resting intervals and the contralateral power spectra during the MI tasks in the alpha and beta bands. (B) The correlation between the global efficiency in resting intervals and the Δpower during the MI tasks in the alpha and beta bands. The transitivity was higher in the alpha band (left: alpha–beta, p < 0.001; right: alpha–beta, p < 0.001). Similarly, the global efficiency was also higher in the alpha band (left: alpha–beta, p < 0.001; right: alpha–beta, p = 0.001). *** represents significance.

A repeated measurement variance analysis was performed on the global efficiency and the transitivity among the alpha and beta bands in the left/right MI tasks (2 × 2). The results showed that both the local transitivity and the global efficiency in the alpha band were higher than those in the beta band (transitivity: left_alpha–beta, p < 0.001, t = 5.070, df = 28; right_alpha–beta, p < 0.001, t = 5.084, df = 28; global efficiency: left_ alpha–beta, p < 0.001, t = 5.738, df = 28; right_alpha–beta, p = 0.001, t = 5.788, df = 28). *** represents significance at p ≤ 0.001.

For the interaction between the transitivity and the power of the two rhythms, we found that the transitivity was correlated with the power of the contralateral sensorimotor rhythms of different frequencies during the MI task. The correlation results are depicted in Figure 4. The transitivity of the resting intervals in the alpha band was positively correlated with the power of the contralateral sensorimotor area during the MI tasks, and the result was significant (L_alpha_contra-transitivity, r = 0.53, p = 0.003; R_alpha_contra-transitivity, r = 0.54, p = 0.002). For the beta band, the transitivity was not correlated with the power of the contralateral sensorimotor area. We also calculated the clustering coefficient, which was consistent with the transitivity result (Supplementary material Figure S5).

Then, to explore the effect of brain integration function on ERD, the correlation between the global efficiency during resting intervals and the Δpower of brain bilateral sensorimotor rhythm at different frequencies during the MI task was analysed (Figure 4). The global efficiency in the alpha band negatively correlated with the Δpower between bilateral sensorimotor areas during the MI task, and the result was significant (L_alpha_Δpower-efficiency, r = −0.61, p = 0.0004; R_alpha_Δpower-efficiency, r = −0.5, p = 0.007). However, the global efficiency was not correlated with the Δpower in the beta band. Among them, the Δpower between bilateral sensorimotor areas could reflect the intensity of ERD. Therefore, ERD in the alpha and beta band tasks was related to the global efficiency of the resting intervals.

5. Discussion

The primary purpose of this study was to explore the effect of the segregation and integration of brain function on the sensorimotor rhythm of MI tasks. First, we analyzed the rhythm metric differences between the two frequency bands from the distribution of time-varying topography and the power difference between the bilateral time scales and the contralateral power spectrum. Second, we explored the interaction between brain network properties during the resting intervals and the sensorimotor rhythm during MI.

ERD and ERS are the primary sensory motor rhythms caused by MI [41]. Alpha and beta rhythms are related to motor ability and motor control. However, there are differences in the power spectrum and topography distribution between the two frequency bands [42]. Under the condition that all subjects were right-handed, the present results showed that the left- and right-handed movement was related to the desynchronization of bilateral sensorimotor areas in the alpha band, and the contralateral activity was more substantial than the ipsilateral activity. However, a more diffuse distribution was related to centralized desynchronization in the beta band, and the desynchronization on the contralateral side was slightly potent. This finding is consistent with the previous conclusion that alpha rhythm desynchronization was significantly concentrated in lateral motor-related areas. In contrast, beta rhythm desynchronization is focused in the parietal region and is more diffuse [13]. The mechanism of desynchronization rhythm activity of alpha and beta was different. In addition, the present work found that the topography distributions of the alpha and beta bands were on different time scales. For the alpha band, lateralization appeared when MI was executed, and after the MI process, lateralization gradually disappeared. However, the beta distributions in the central area remained almost unchanged during the MI process. Related research mentioned that the alpha and beta band rhythms had different anatomical distributions and travelled in opposite directions across the sensorimotor cortex, which had other effects on cortical excitability [43]. In addition, alpha rhythm seemed to carry and modulate detailed information on motor execution and assumed the role of local neuronal processing. However, beta rhythm takes and disperses co-ordinative information among different clusters of nodes and modulates activity over more significant spatial regions [15]. Therefore, the alpha and beta rhythms may play different roles in brain oscillation regulation during MI.

MI is a complex cognitive operation that requires memory retrieval and spatial attention and may also require the brain to perform calculations on imagined body movements [44]. ERD and ERS represent the different neuronal activities and information processing strategies during the process of MI. Meanwhile, the efficient information processing of MI reflected the brain’s regional activities and interactions [45]. We found that the different brain regions and the brain states in different periods affected MI performance. Therefore, we researched the effects of functional segregation and integration on the rhythmic activities of the alpha and beta frequency bands during the process of MI.

The results showed that the transitivity in the alpha band positively correlated with the power of the contralateral sensorimotor area. Specifically, when imagining the hand movement, the power spectrum of the alpha and beta bands in the sensorimotor area of the contralateral hemisphere decreased [46]. In the present work, the local power values of the contralateral side were related to the transitivity. Previous research has mentioned that transitivity is one of the essential measures of brain segregation, representing the presence of clusters in local functional networks [47]. The segregation of the brain network may shape individuals’ contralateral local neural modulation in MI, thus facilitating the completion of behavior tasks.

In addition, the global efficiency in the alpha band negatively correlated with the Δpower between bilateral sensorimotor areas during the MI task, and the results were significant. Among them, the power spectrum difference between bilateral sensorimotor areas could reflect the intensity of ERD. Therefore, the increased intensity of the ERD during alpha tasks was accompanied by enhanced global efficiency of the resting intervals. Efficient motor tasks are mainly due to brain lateralization that uses one hemisphere and inhibits the other hemisphere [48–50]. In fact, during movement, the functional contribution of both hemispheres is flexibly driven. This flexibility underlies skilled and adaptive motor behavior [51]. Specific tasks require extensive interactions between related neural regions and functional integration [52]. As movement control is closely associated with inter-hemispheric interactions [51], the bilateral oscillation patterns of MI and inter-hemispheric co-ordination may result from the individuals’ brain network characteristics.

The brain function metrics of the two frequency bands showed differences in the integration and segregation functions. Both the global efficiency and transitivity of the alpha band were significantly higher than those of the beta band. This research showed that the alpha band had higher intra-network and inter-network connectivity than the other frequency bands [53]. This phenomenon may be related to the different network functions in the two frequency bands [54] because alpha is the rhythm-dominating brain activity during MI resting intervals. In addition, the modular structure of brain functional connectivity networks provides a basis for segregated information processing within modules and the integration between them, which requires an optimal balance between local and global processing [29,55]. Therefore, the balance between segregation and integration would be the basis for shaping the patterns of sensorimotor rhythm during the MI process.

The present work also had some limitations. For the analysis, volume conduction might affect the connectivity measures obtained from the scalp EEG [37], but the imaginary coherence was insensitive to false connectivity arising from the volume conduction. Because of the individual differences, the individualized frequency and the network calculation will need to be considered further. For single-modality data collected from EEG, the network computation was imprecise, and it is necessary to integrate fMRI or other acquisition methods to verify the contribution of the resting-state network of the brain to cognitive tasks. For the results, the resting interval time scale was short, the MI recovery periods superimposed with the preparation, and the causality between the brain network and the imaging characteristics was unclear. Therefore, innovative methods or experimental verification designs will need in future studies.

6. Conclusion

The present work mainly explored the effects of resting intervals’ brain network characteristics on brain rhythmic oscillations during the MI process. The results verified that the brain segregation and integration functions were correlated with the metrics of rhythmic oscillation in the alpha band, implying the integration was related to bilateral hemisphere co-ordination, and the segregation was associated with the contralateral local neural activation of individuals in MI tasks. These findings could contribute to a deeper understanding of how the brain network organization regulates the oscillation activity in the motor system.

Ethical approval

The studies involving human participants were reviewed and approved by the Ethics Committee of the University of Electronic Science and Technology of China. The patients/participants provided written informed consent to participate in this study.

Author contributions

M.P., Y.Q., D.L., and T.L. conceived and designed the work. M.P., D.L., and S.L. acquired the data. M.P. and D.L. analysed the data. M.P., Y.Q., D.G., and S.L. wrote the article. All authors revised the work for important intellectual content. All authors contributed to the article and approved the submitted version.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

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

Additional information

Funding

This work was supported by grants from the National Natural Science Foundation of China [grant number: 82102175] and the Department of Science and Technology of Sichuan Province [grant numbers: 2019YJ0411 and 2022NSFSC1410].