Brain-muscle coupling and brain network analysis method based on improved time delay maximum information coefficient

Abstract

The application discloses a brain-muscle coupling and brain network analysis method based on improved time delay maximum information coefficient, first, arranging subjects to perform corresponding actions to record synchronous sEMG and EEG. Through wavelet denoising, sEMG is preprocessed, and EEG is subjected to band-pass filtering and ICA independent component analysis. The processed EEG and corresponding sEMG are segmented, and EEG and corresponding sEMG during hand movement are extracted. Then, the BEMD method is used to obtain the IMF component of EEG and corresponding sEMG, and the TDMIC method is used to calculate the coupling value of different IMF components of EEG and corresponding sEMG and EEG and EEG. Finally, the obtained results are subjected to statistical analysis. The application evaluates the difference of FCMC and BFCN changes of the tester in a short time after being subjected to different waveforms of NMES in different local frequency bands. And the experimental results are quantified by normalizing the significant area. These studies will provide a new idea for the stroke rehabilitation method based on NMES and provide a new angle for the motor function evaluation method.

Description

Technical Field

[0001] This invention relates to the field of bioelectric signal processing technology, specifically to a brain-muscle coupling and brain network analysis method based on an improved time-delay maximum information coefficient. Background Technology

[0002] Stroke is a neurological disorder caused by ischemia or hemorrhage in the brain, resulting in acute or focal brain dysfunction. This presents unprecedented challenges to stroke prevention and treatment, necessitating continued efforts to strengthen stroke management strategies. Motor dysfunction, speech and swallowing disorders, and cognitive impairment are among the common sequelae experienced by stroke patients. Motor dysfunction is particularly prevalent after a stroke. Current techniques for treating stroke include thrombolysis, endovascular intervention, carotid angioplasty, and electrical stimulation. Among these, electrical stimulation is a promising treatment method that can promote functional improvement by reconstructing damaged neuromuscular connections.

[0003] Neuromuscular electrical stimulation (NMES) is widely used to improve motor function after stroke. NMES-based rehabilitation has been shown to effectively prevent muscle atrophy and improve muscle strength and coordination. Wei et al. found that combining NMES with conventional rehabilitation was more effective than conventional rehabilitation alone in improving upper limb function in patients with chronic stroke. Furthermore, Bennie et al. found that a sine wave requires the minimum average stimulation current to achieve the desired contractile force and was judged as the most comfortable stimulation waveform. These findings were confirmed by Petrovsky et al., who compared the same waveform and used the same current level. However, several other studies found no difference in subjective comfort when stimulating the quadriceps with sine, sawtooth, and square waves. Nevertheless, the evaluation factors used in these studies were subjective, such as subjective comfort. Therefore, identifying relatively objective evaluation factors to determine the optimal stimulation waveform is crucial.

[0004] Since Conway et al. initially discovered the correlation between electroencephalography (EEG) and surface electromyography (sEMG), functional corticomuscular coupling (FCMC) has become a powerful tool for examining the degree of coupling between EEG and sEMG signals. Recent evidence suggests that non-motorized muscle stimulation (NMES) enhances cortical excitability and promotes motor cortical remodeling in stroke patients. NMES not only increases muscle function and reduces muscle spasticity but also induces sensory input to the spinal cord and cerebral cortex. Bao et al. observed a significant increase in cortical-muscular functional coupling between the ipsilateral cerebral cortex and the affected lower limb after NMES-based rehabilitation training in chronic stroke patients. Furthermore, significant differences were observed in both descending and ascending cortical-muscular pathways after NMES training, as demonstrated by FCMC. Although the impact of NMES-based rehabilitation measures on neuromotor control processes remains unclear, research on the brain functional connectivity network (BFCN) has become widespread and can serve as an indicator for evaluating the effectiveness of NMES. During voluntary human movement, the motor cortex transmits commands through motor neural pathways to control muscle movements, while sensory neural pathways feed back sensory information from the muscles to the cortex, ensuring precise execution of the movement. Although NMES has a direct impact on muscles, the motor cortex may change after feedback, thus affecting the BFCN.

[0005] Currently, many methods are applied to the study of FCMC, such as mutual information (MI), transfer entropy (TE), and time-delayed maximal information coefficient (TDMIC). The MI method measures the shared information between two time series X and Y to examine their interactions. It can detect linear and nonlinear correlations between two time series and is therefore widely used in neuroscience. However, it cannot reflect the direction of internal information flow in the FCMC. The TE method is a model-free method based on information entropy, capable of assessing linear and nonlinear coupling between two signals. Recently, due to the asymmetry and transition probability calculation characteristics of TE and TDMIC, they have been considered effective methods for detecting causal relationships between neurophysiological signals. Since different EEG rhythm components represent different brain functional states, previous studies have shown that the FCMC during sustained contraction and slow upper limb movement is dominated by the α band, during steady-state exertion by the β (15-35Hz) band, and during dynamic tasks and large force outputs by the γ (35-60Hz) band. These findings suggest that FCMCs with different rhythmic oscillations can reflect various functional modes of the sensorimotor system. However, the TE and TDMIC methods are not suitable for characterizing information interaction in local frequency bands within the FCMC.

[0006] Currently, methods for extracting specific frequency band signals from EEG or sEMG include wavelet decomposition, FIR filters, variational mode decomposition (VMD), empirical mode decomposition (EMD), and bivariate empirical mode decomposition (BEMD). FIR filters may disrupt the temporal order of the signal, leading to false information detection. Wavelet decomposition results are often affected by the choice of basis functions. Furthermore, VMD and EMD methods can only handle single-channel signals and suffer from mode mixing and uniqueness issues. To address these problems, a bivariate empirical mode decomposition-time delay maximal information coefficient (BTDMIC) method is proposed to quantify FCMC results. By simultaneously performing bivariate empirical mode decomposition on both the real and imaginary components of the binary signal, the problems associated with mode mixing and uniqueness are effectively mitigated. Summary of the Invention

[0007] This invention addresses the shortcomings of existing technologies by proposing a brain-muscle coupling and brain network analysis method based on an improved time-delay maximal information coefficient (BTDMIC), which quantifies the results of FCMC (Focused Functional Motor Response). The bivariate empirical mode decomposition-time delay maximal information coefficient (BTDMIC) can be used to assess the coupling strength, local frequency bands, and directional characteristics of FCMC between the motor cortex and muscle activity, further revealing the details of the dynamic characteristics of physiological signals at multiple time-frequency scales. Furthermore, the impact of NMES (Natural Motor Emission System) on BFCN (Brain-Focused Functional Network) is investigated using the BTDMIC method.

[0008] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0009] A brain-muscle coupling and brain network analysis method based on an improved time-delay maximum information coefficient includes the following steps:

[0010] Step 1: While the subject performs the predetermined action, collect EEG signals through different EEG electrodes and simultaneously collect EMG signals from the corresponding surfaces. Collect EMG signals from the relevant muscles in the subject's arm and record the start and end points of each action.

[0011] Step 2: Preprocess the EEG and EMG signals collected in Step 1, segment the processed EEG and EMG signals, and extract the EEG and EMG signals during hand movements.

[0012] Step 3: Calculate the coupling value between the denoised EEG signal and the corresponding surface EMG signal, as well as the coupling value between different EEG electrodes using the BTDMIC method.

[0013] Step 4: Use repeated ANOVA to examine the statistical differences in coupling values ​​between EEG and EMG signals in different local frequency bands;

[0014] Step 5: Use repeated ANOVA to examine the statistical differences in brain network parameters under different conditions in different local frequency bands.

[0015] Preferably, in step 1, the electromyographic signal acquisition location is the right radial wrist flexor muscle.

[0016] Preferably, in step 1, EEG signals are acquired using different EEG electrodes, with the acquisition locations being channels Fz, F1, F2, F3, F4, F5, F6, FCz, FC1, FC2, FC3, FC4, FC5, FC6, Cz, C1, C2, C3, C4, C5, C6, CP1, CP2, CP3, CP4, CP5, and CP6.

[0017] Preferably, in step 1, when collecting signals, the prescribed action is to raise the right wrist and hold it for 5 seconds, and to collect EEG and EMG signals before and after the NMES.

[0018] Preferably, step 2 includes the following sub-steps:

[0019] 2.1 The EEG signal was preprocessed using the EEGLAB toolbox. If significant artifacts were found in the 1-60Hz range, the EEG signal and its corresponding electromyography (EMG) signal segments were discarded.

[0020] Furthermore, the entire experimental process was divided into different phases. Each subject performed 24 repetitive wrist-raising tasks before NMES and 24 repetitive wrist-raising tasks after NMES, defined as experimental phases. The number of EEG and corresponding surface electromyography (EMG) signals retained by all participants across all experimental phases ranged from 114 to 137. To avoid difficulties in statistical analysis due to uneven distribution of retained EEG and EMG signals, 110 EEG signals and their corresponding EMG signals were randomly selected from the remaining signals of all participants.

[0021] 2.2 For EEG signal preprocessing, a 60Hz low-pass filter was used in EEGLAB's FIR digital filter to remove baseline drift and overflow; independent component analysis was used to eliminate ECG and eye movement artifacts.

[0022] 2.3 For electromyography (EMG) signals, a low-pass FIR digital filter with a cutoff frequency of 60Hz was used. Subsequently, wavelet thresholding was used to denoise the EMG signals, with Sym8 wavelet as the denoising wavelet basis function in the symmetric wavelet system.

[0023] 2.4 A fixed-hysteresis Kalman smoother was used to filter the 50Hz power signal in the EEG and EMG signals.

[0024] As a preferred embodiment, the specific steps of the BTDMIC method described in step 3 are as follows:

[0025] (1) Define the electroencephalogram (EEG) signal and electromyogram (EMG) signal as x(t) and y(t) respectively, and define the complex-valued vector z(t) = x(t) + jy(t). The complex-valued vector z(t) in γ d The projection in the direction is:

[0026]

[0027] Where γ d = 2dπ / D, and The maximum value is at Obtained from China.

[0028] (2) Pair set Perform spline interpolation to obtain γ d Envelope curve in direction And calculate the average value m(t). Then subtract m(t) from the input signal z(t) to obtain h(t):

[0029] h(t) = z(t) - m(t)

[0030] (3) Then check whether h(t) is an intrinsic mode function (IMF). If it is, repeat the above steps with the residual signal; otherwise, replace z(t) with h(t) and repeat the above steps. The result of the BEMD method can be expressed as:

[0031]

[0032] Where h i (t) represents the i-th IMF component, and r(t) is the residual term of the signal.

[0033] (4) The IMF components obtained from step (3) all have real parts w R (t) and the imaginary part w I (t). Finally, x(t) and y(t) are decomposed into a set of IMFs with specific frequency bands, x = (w R1 ,w R2 ,...,w RD ), y = (w I1 ,w I2 ,...,w ID Therefore, BTDMIC can be calculated using the following formula:

[0034]

[0035] in and Let represent the i-th IMF (Intrinsic Mode Function) component of the EEG signal and the surface electromyography signal, respectively, and τ be the time delay coefficient. P(·) is the joint probability among the variables.

[0036] As a preferred option, the analysis method in step 4 is as follows:

[0037] 4.1 Causal Fourier transform (CFT) analysis was used to verify the statistical significance of the BTDMIC values ​​in the frequency domain. This method disrupts direct causal coupling by randomizing the amplitude and phase of the original sequence while preserving the power spectrum and phase difference. For each time series, 1000 iterations of alternative calculations were performed, and the average value was used as the significance threshold for BTDMIC. sig To determine statistical significance, the area under the normalized BTDMIC curve (A) is used. S This curve quantifies the functional coupling characteristics between each pair of EEG and EMG signals within a specific frequency band, and is used to quantify the given frequency band f. l1 -f l2 The information interaction between the two signals is expressed by the following expression:

[0038]

[0039] Where Δf represents the frequency resolution, if BTDMIC(f) ≤ BTDMIC sig (f) then A S (f)=0, A S The larger the value, the stronger the coupling between the two signals;

[0040] 4.2 For each phase of the action task for each participant, firstly, the bidirectional BTDMIC between the electroencephalogram (EEG) and the electromyography (EMG) of the right radial wrist flexor muscle in the frequency range of 1–60 Hz was calculated.

[0041] 4.3 Based on the average BTDMIC value curve of the action task in each period, calculate the A frequency band for different frequency bands. S value;

[0042] 4.4 Finally, an analysis of variance was performed on the results to output the differences in coupling values ​​between EEG signals and corresponding surface electromyography signals in different local frequency bands.

[0043] Preferably, in step 4.3, A is calculated. S The frequency bands for the values ​​are α: 8-14Hz, β1: 15-25Hz, β2: 25-35Hz, γ1: 35-45Hz, and γ2: 45-60Hz.

[0044] Preferably, the analysis method in step 5 is as follows:

[0045] 5.1 Select network nodes, define a threshold coefficient matrix, set coupling values ​​below the threshold to 0, and vice versa, establish network connections based on the values ​​of the threshold coefficient matrix, and use clustering coefficients, network efficiency, and network density to describe the changes in BFCN after different NMES interventions.

[0046] Clustering coefficients represent the probability of connections between nodes connected to a given node. As an indicator of the aggregation level in a network, their calculation formula is as follows:

[0047]

[0048] In the formula, N is the number of nodes, e i k represents the number of edges between adjacent nodes of node i. i This represents the number of nodes connected to node i;

[0049] Network efficiency is an indicator of network communication capability, representing the average of the reciprocals of the shortest path distances between any two nodes in a network. The formula for calculating network efficiency is as follows:

[0050]

[0051] In the formula, d ij This represents the shortest path between node i and node j.

[0052] Network density is the ratio of the total number of edges in a network to the maximum possible number of edges between nodes. It represents the degree of interconnectivity between nodes in the network, and its calculation method is as follows:

[0053]

[0054] Where k i It is the degree of node i;

[0055] 5.2 We selected the threshold that simultaneously makes statistically significant differences in network parameters such as clustering coefficient, network efficiency, and network density as the optimal threshold to construct the brain functional connectivity network and perform multifactor ANOVA.

[0056] This invention has the following characteristics and beneficial effects:

[0057] Using the above technical solution, this invention applies the data-driven BEMD method to decompose EEG and EMG signals into a finite number of paired IMFs, and performs decomposition simultaneously based on the rotational characteristics of the two variables, thereby overcoming the uniqueness and modal aliasing problems of IMFs. The results are then combined with the traditional TDMIC method. This method is used to evaluate the differences in FCMC and BFCN changes in subjects shortly after exposure to NMES with different waveforms (including rectangular waves, sine waves, and sawtooth waves) at different local frequency bands. The experimental results are quantified by normalizing significant regions. These studies will provide new ideas for NMES-based stroke rehabilitation methods and offer a new perspective on motor function evaluation methods.

[0058] This invention proposes a bivariate empirical mode decomposition-time delay maximal information coefficient (BTDMIC) method to quantify FCMC results. BTDMIC can be used to assess the coupling strength, local frequency bands, and directional characteristics of FCMC between the motor cortex and muscle activity, further revealing details of the dynamic characteristics of physiological signals across multiple time-frequency scales. Furthermore, the influence of non-molecular mode optimization (NMES) on basilar fibroblastic neural network (BFCN) was investigated using the BTDMIC method. The results show that sinusoidal NMES is superior only in promoting FCMC, while rectangular NMES is most significant in promoting both FCMC and BFCN. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating an implementation method for brain-muscle coupling and brain network analysis based on an improved time-delay maximum information coefficient, according to an embodiment of the present invention.

[0061] Figure 2 The diagram shows the experimental setup, where (A) shows the placement of the stimulation electrodes and EMG acquisition device, (B) shows the experimental procedure flow, (C) shows a participant in the experimental process, and (D) shows three waveforms of NMES.

[0062] Figure 3Brain topography for BTDMIC increments. (A) shows the result before NMES, and (B) shows the results after rectangular wave (sq-NMES), sine wave (si-NMES), and sawtooth wave (sa-NMES), respectively. The first column represents the α band, the second, third, and fourth columns represent the β1, β2, and γ1 bands, respectively, and the last column represents the γ2 band.

[0063] Figure 4 The statistical differences in As increments under different conditions are shown. (A) compares results after different NMES procedures, and (B) compares results in different directions. "***" indicates p < 0.001.

[0064] Figure 5 The network features are defined by different thresholds. (A) represents the clustering coefficient, (B) represents the network density, and (C) represents the network efficiency. "*" indicates p < 0.05, "**" indicates p < 0.01, and "***" indicates p < 0.001.

[0065] Figure 6 The brain network diagrams are shown with a threshold of 0.25, where (A) is pre-NMES, (B) is sq-NMES, (C) is si-NMES, and (D) is sa-NMES.

[0066] Figure 7 This is a comparison of BTDMIC in two directions of C3. Note: The thickness of the line indicates the global size.

[0067] Figure 8 The statistical differences in brain network parameters of sq-NMES in different frequency bands are shown, where (A) is the clustering coefficient, (B) is the network density, and (C) is the network efficiency. "*" indicates p < 0.05, "**" indicates p < 0.01, and "***" indicates p < 0.001.

[0068] Figure 9 Statistical differences in brain network parameters between pre-NMES and sq-NMES are given, where (A) is the clustering coefficient, (B) is the network density, and (C) is the network efficiency. "*" indicates p < 0.05, "**" indicates p < 0.01, and "***" indicates p < 0.001. Detailed Implementation

[0069] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0070] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0071] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0072] This invention provides a method for brain-muscle coupling and brain network analysis based on an improved time-delay maximum information coefficient, such as... Figure 1 As shown, it includes the following steps:

[0073] Step 1: Record electromyography (EMG) and electroencephalography (EEG) signals during different periods of wrist raising in the subjects. The EMG signal acquisition location was the right radial flexor carpi radialis (FCR) muscle. The EEG signal channels were Fz, F1, F2, F3, F4, F5, F6, FCz, FC1, FC2, FC3, FC4, FC5, FC6, Cz, C1, C2, C3, C4, C5, C6, CP1, CP2, CP3, CP4, CP5, and CP6. The NMES waveforms selected were rectangular wave, sine wave, and sawtooth wave.

[0074] In this embodiment, healthy males were selected as the experimental subjects. Figure 2(B) represents the experimental paradigm. To mitigate the impact of residual effects on the results, a three-cycle experiment was conducted, with a different NMES waveform used in each cycle and a one-day interval between cycles. Although the stimulus waveforms for each participant were randomly generated within each cycle, the waveforms used in the three cycles were not identical. Each phase consisted of a 240-second pre-stimulation action phase, followed by three stimulus phases and a post-stimulation action phase (300 + 80 seconds), with a 160-second rest period between each phase. The pre-stimulation action phase comprised 24 trials, while each post-stimulation action phase consisted of 8 trials. Each trial included 1 second of wrist flexion, 5 seconds of wrist grip, 1 second of wrist relaxation, and a 3-second rest period.

[0075] This embodiment was conducted in a room with electromagnetic shielding. In each trial, participants followed the instructions displayed on the monitor. The electrical stimulation waveforms used were square, sinusoidal, and sawtooth patterns, all at a frequency of 50 Hz. Figure 2 (D) During the “hold” period, each participant is asked to contract their wrist to the maximum extent possible. The stimulation intensity is determined by the peak current that each participant can tolerate. The peak current starts at 5 mA and increases by 1 mA each time until the participant feels uncomfortable and requests the current to be stopped (maximum current). The peak current is then set to be 1 mA less than the maximum current.

[0076] Step 2: High-frequency noise and eye movement artifacts in the EEG signal are removed by bandpass filtering from 1 to 60 Hz and independent component analysis (ICA); the EMG signal is preprocessed by wavelet packet denoising.

[0077] Specifically, 2.1 The EEG signal is preprocessed using the EEGLAB toolbox. If there are significant artifacts in the 1-60Hz range, the EEG and its corresponding electromyography signal segments are discarded.

[0078] 2.2 For EEG signal preprocessing, a 60Hz low-pass filter was used in EEGLAB's FIR digital filter to remove baseline drift and overflow; independent component analysis was used to eliminate ECG and eye movement artifacts.

[0079] 2.3 For electromyography (EMG) signals, a low-pass FIR digital filter with a cutoff frequency of 60Hz was used. Subsequently, wavelet thresholding was used to denoise the EMG signals, and the Sym8 wavelet in the symmetric wavelet system was used as the denoising wavelet basis function.

[0080] 2.4 A fixed-hysteresis Kalman smoother was used to filter the 50Hz power signal in the EEG and EMG signals.

[0081] Step 3: Calculate the coupling value between the denoised EEG signal and the corresponding surface electromyography signal using the BTDMIC method, such as... Figure 3 As shown, this figure is a brain topography map of BTDMIC increments, such as... Figure 3 As shown in (A), the peak values ​​of brain topography prior to NMES were mainly concentrated in the C1 and C3 channels, while Figure 3 (B) indicates that in the α and β1 bands, the peak values ​​of the incremental brain topography of BTDMIC after all three stimuli relative to NMES were mainly concentrated in the CP3 channel, especially in the β1 band. In the γ1 and γ2 bands, the peak values ​​of the incremental brain topography of BTDMIC after all three stimuli were mainly concentrated in the CP1 channel. In the β2 band, the peak values ​​after sq-NMES were mainly concentrated in FC1, FC3, Cz, C1, C3, C5, and FC6; the peak values ​​after si-NMES were mainly concentrated in FCz; and the peak values ​​after sa-NMES were mainly concentrated in FC6. Furthermore, only in the β2 band were the number of high BTDMIC incremental channels after sq-NMES the highest, while in the remaining bands, the number of high BTDMIC incremental channels after si-NMES was significantly higher than the other two groups. Therefore, overall, the number of high BTDMIC incremental channels after si-NMES was the highest, indicating the closest information interaction between EEG and sEMG.

[0082] Step 4: Based on previous research indicating a significant interaction between the C3 channel and sEMG during right upper limb movement, the C3 channel was selected for further analysis. To compare the differences in the descent and ascent paths of FCMC after different NMES at the electrode of interest, the normalized increments of the significant areas after different NMES within five specific frequency bands were calculated. Statistical analysis was performed with direction (EEG→sEMG and sEMG→EEG), stimulus type (sq-NMES, si-NMES, and sa-NMES), and frequency band (α, β1, β2, γ1, and γ2) as independent variables and the normalized increment of As as the dependent variable. Three-way ANOVA results showed a bidirectional interaction between stimulus type and frequency band, while no interaction existed between direction, stimulus type, and frequency band. Figure 4As shown in (A), for electrode C3, the As increment after si-NMES was significantly higher than the other two groups in both the α and β1 frequency bands, while in the β2 and γ1 frequency bands, the As increment after sq-NMES was significantly higher than the other two groups in all directions. Furthermore, there was no significant difference in the γ2 frequency band. Subsequent pairwise comparisons of the As increments after si-NMES and sq-NMES revealed that in the β1 and β2 frequency bands, the As increments after both si-NMES and sq-NMES were significantly stronger in the EEG→sEMG direction than in the opposite direction. Additionally, in the γ1 frequency band, the As increments after sq-NMES were significantly stronger in the EEG→sEMG direction than in the opposite direction, while there was no significant difference in the As increments after si-NMES. Figure 4 As shown in (B).

[0083] Step 5: Calculate the coupling values ​​between different EEG electrodes after denoising using the BTDMIC method. The BTDMIC matrix was calculated at different thresholds. The network clustering coefficients, efficiency, and density within the threshold range of 0.20-0.36 are shown below. Figure 5 As shown in the figure, one-way ANOVA revealed that for clustering coefficients, the values ​​after sq-NMES were significantly higher than those after pre-NMES at almost all thresholds; conversely, the values ​​of si-NMES and pre-NMES did not differ significantly at any threshold. For network density, when the threshold was between 0.20 and 0.26, the values ​​after sq-NMES were significantly higher than those after pre-NMES, si-NMES, and sa-NMES. For network efficiency, when the threshold was between 0.23 and 0.26, the values ​​after sq-NMES were significantly higher than those after pre-NMES, si-NMES, and sa-NMES. Furthermore, when the threshold was 0.25, the clustering coefficient, density, and efficiency after sq-NMES were significantly higher than those of the other groups.

[0084] Step 6: Based on the findings in Step 5, the optimal threshold was selected to construct the brain network, such as... Figure 6 The brain region (BFCN) after NMES treatment was more complex than that before NMES treatment. Notably, the change in BFCN value was most significant after sq-NMES treatment. It is speculated that sq-NMES can further activate the contralateral brain regions corresponding to upper limb movement.

[0085] Step 7: To investigate the effects of different NMES on information exchange between the C3 electrode and other EEG electrodes, in Figure 7The BTDMIC values ​​of the C3 electrode and the other 26 EEG electrodes were described. Analysis showed that in the others→C3 direction, the BTDMIC values ​​of pre-NMES and sa-NMES were not significantly different, while the BTDMIC values ​​of sq-NMES and si-NMES were slightly higher. In the C3→others direction, the BTDMIC values ​​of sq-NMES, si-NMES, and sa-NMES were all higher than those of pre-NMES, and the BTDMIC value of sq-NMES was higher than that of the other three groups. Furthermore, the results indicated that only sq-NMES and si-NMES had significantly higher BTDMIC values ​​in the C3→others direction than in the opposite direction. Conversely, the BTDMIC values ​​of the other two groups showed no significant difference in either direction.

[0086] Step 8: To further analyze the promoting effect of sq-NMES on information exchange between the motor cortex of the brain at different frequency bands, the clustering coefficients, densities, and efficiency of all participants after pre-NMES and sq-NMES were calculated, and a one-way ANOVA was performed. The results are as follows: Figure 8 and 9 As shown. Figure 8 The differences in clustering coefficient, density, and efficiency after sq-NMES are shown across different frequency bands. For clustering coefficient and efficiency, the values ​​in the α band are significantly higher than those in the β2, γ1, and γ2 bands, and the values ​​in the β1 band are also significantly higher than those in the β2, γ1, and γ2 bands. For density, the values ​​in the α band are significantly higher than those in the β1, β2, γ1, and γ2 bands, and the values ​​in the β1 band are also significantly higher than those in the β2, γ1, and γ2 bands. The differences between pre-NMES and sq-NMES are shown in... Figure 9 The results are shown in the diagram. For clustering coefficients, the values ​​after sq-NMES are significantly higher than those after pre-NMES across all frequency bands, and the values ​​in the β1 band are also significantly higher than those in the β2, γ1, and γ2 bands. For density, the values ​​after sq-NMES are significantly higher than those after pre-NMES in the α, β1, and γ2 frequency bands. For efficiency, the values ​​after sq-NMES are significantly higher than those after pre-NMES in the β2 and γ2 frequency bands.

[0087] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments, including components, without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A method for analyzing brain-muscle coupling and brain networks based on an improved time-delay maximum information coefficient, characterized in that, Includes the following steps: Step 1: While the subject performs a predetermined action, collect electroencephalogram (EEG) signals through different EEG electrodes and simultaneously collect electromyographic (EMG) signals from the corresponding surfaces. Step 2: Preprocess the EEG and EMG signals collected in Step 1, segment the processed EEG and EMG signals, and extract the EEG and EMG signals during hand movements. Step 3: Calculate the coupling value between the denoised EEG signal and the corresponding surface EMG signal, as well as the coupling value between different EEG electrodes using the BTDMIC method. For any two signals, BTDMIC is calculated by the following formula: in and Let represent the i-th intrinsic mode function components of any two signals respectively, τ be the time delay coefficient, and P(·) be the joint probability among the variables; Step 4: Use repeated ANOVA to examine the statistical differences in coupling values ​​between EEG and EMG signals in different local frequency bands; Step 5: Use repeated ANOVA to examine the statistical differences in brain network parameters under different conditions in different local frequency bands.

2. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 1, characterized in that, In step 1, the electromyographic signal acquisition location is the right radial wrist flexor muscle.

3. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 1, characterized in that, In step 1, EEG signals are acquired using different EEG electrodes. The EEG signal acquisition locations are channels Fz, F1, F2, F3, F4, F5, F6, FCz, FC1, FC2, FC3, FC4, FC5, FC6, Cz, C1, C2, C3, C4, C5, C6, CP1, CP2, CP3, CP4, CP5, and CP6.

4. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to any one of claims 1-3, characterized in that, In step 1, when collecting signals, the prescribed action is to raise the right wrist and hold it for 5 seconds, and to collect EEG and EMG signals before and after NMES.

5. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 1, characterized in that, Step 2 includes the following sub-steps: 2.1 The EEG signal was preprocessed using the EEGLAB toolbox. If significant artifacts were found in the 1-60Hz range, the EEG signal and its corresponding electromyography (EMG) signal segments were discarded. 2.2 For EEG signal preprocessing, a 60Hz low-pass filter was used in EEGLAB's FIR digital filter to remove baseline drift and overflow; independent component analysis was used to eliminate ECG and eye movement artifacts. 2.3 For electromyography (EMG) signals, a low-pass FIR digital filter with a cutoff frequency of 60Hz was used. Subsequently, wavelet thresholding was used to denoise the EMG signals, with Sym8 wavelet as the denoising wavelet basis function in the symmetric wavelet system. 2.4 A fixed-hysteresis Kalman smoother was used to filter the 50Hz power frequency signal in the EEG and EMG signals.

6. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 1, characterized in that, In step 4, the analysis method is as follows: 4.1 Using the area of ​​significance (A) under the normalized BTDMIC curve S This curve quantifies the functional coupling characteristics between each pair of EEG and EMG signals within a given frequency band, thus quantifying the given frequency band f. l1 -f l2 Information interaction between two signals, where A S The expression is as follows: Where Δf represents the frequency resolution, if BTDMIC τ (f)≤BTDMIC sig (f) then A S (f)=0, A S The larger the value, the stronger the coupling between the two signals; 4.2 For each phase of the action task for each participant, firstly, the bidirectional BTDMIC between the electroencephalogram (EEG) and the electromyography (EMG) of the right radial wrist flexor muscle in the frequency range of 1–60 Hz was calculated. 4.3 Based on the average BTDMIC value curve of the mission in each period, calculate the A frequency band for different frequency bands. S value; 4.4 Finally, an analysis of variance was performed on the results to output the differences in coupling values ​​between EEG signals and corresponding surface electromyography signals in different local frequency bands.

7. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 6, characterized in that, In step 4.3, A is calculated. S The frequency bands for the values ​​are α: 8-14Hz, β1: 15-25Hz, β2: 25-35Hz, γ1: 35-45Hz, and γ2: 45-60Hz.

8. The brain-muscle coupling and brain network analysis method based on the improved time-delay maximum information coefficient according to claim 1, characterized in that, In step 5, the analysis method is as follows: 5.1 Select network nodes, define a threshold coefficient matrix, set coupling values ​​below the threshold to 0, and vice versa, establish network connections based on the values ​​of the threshold coefficient matrix, and use clustering coefficients, network efficiency, and network density to describe the changes in brain functional connectivity networks after different NMES interventions. Clustering coefficients represent the probability of connections between nodes connected to a given node. As an indicator of the aggregation level in a network, their calculation formula is as follows: In the formula, N is the number of nodes, e i k represents the number of edges between adjacent nodes of node i. i This represents the number of nodes connected to node i; Network efficiency is an indicator of network communication capability, representing the average of the reciprocals of the shortest path distances between any two nodes in a network. The formula for calculating network efficiency is as follows: In the formula, d ij This represents the shortest path between node i and node j. Network density is the ratio of the total number of edges in a network to the maximum possible number of edges between nodes. It represents the degree of interconnectivity between nodes in the network, and its calculation method is as follows: Where k i It is the degree of node i; 5.2 We selected the optimal threshold that simultaneously makes the clustering coefficient, network efficiency, and network density statistically significant differences as the optimal threshold to construct the brain functional connectivity network and perform multifactor ANOVA.

Images