A Multi-Scale Cortex-Muscle Coupling Network Analysis Method Based on Ordinal Patterns

Abstract

This invention proposes a multi-scale cortical-muscle coupling network analysis method based on ordinal patterns. First, the invention uses multivariate variational mode decomposition to decompose brain electromyography (EMG) signals at the same frequency scale. Then, it uses a multivariate binary temporal partitioning transfer network method to calculate the coupling strength and coupling adjacency matrix of the multi-component signals, constructing a cortical-muscle coupling network. Complex network parameters are then used to analyze the characteristics of this coupling network. This method has advantages in representing the characteristics of information transmission between the cortex and muscle and their variations across different grip force patterns. It also demonstrates the hierarchical nature of causal links between brain EMG signals, confirming the method's potential in deconstructing the intrinsic connections between the cortex and muscles during upper limb movement and the mechanisms of neural control of muscle movement.

Description

Technical Field

[0001] This invention belongs to the field of signal processing and relates to a method for analyzing multi-scale cortical-muscle coupled networks based on ordinal patterns. Background Technology

[0002] In motor control, the cerebral cortex is the primary actuator, responsible for generating and coordinating motor commands. Cortical regions are directly or indirectly connected to muscles via neurotransmitters and electrical signals. Studying cortical-muscle coupling helps us understand how the brain controls movement, its coordination mechanisms, and the process of motor learning. The motor areas of the cerebral cortex mainly include the motor cortex and the motor planning area. The motor cortex, comprising the primary motor cortex and the premotor cortex, is responsible for generating and executing motor commands. The motor planning area is involved in planning and regulating complex motor sequences. Cortical regions connect to the spinal cord via neural projection fibers, forming cortical-spinal pathways. These pathways transmit motor commands from the cerebral cortex to muscles through the transmission of nerve impulses to control the execution of movement. Studies have shown that there is a temporal correlation between the electrical activity of the cerebral cortex and muscle movement during cortical-muscle coupling. Especially in low-frequency bands (e.g., 8-30 Hz) neural oscillations, cortical and muscle activity can exhibit synchronization and coupling. By studying cortical-muscle coupling, scholars hope to gain a deeper understanding of the mechanisms by which the brain controls movement, providing a foundation and application value for rehabilitation therapy, brain-computer interface technology, and the treatment of movement disorders.

[0003] Studying the coupling relationship between cortical and muscle muscles from a multi-scale perspective can identify the connections between the control and sensory information of the cerebral cortex and the response information of related muscles under different motor states, and explore the characteristics of the hierarchical effect of coupling between EEG and EMG signals. Research on the control mechanisms of motor function can be used to analyze the characteristics and texture of muscle damage caused by brain diseases, and is of great significance for the diagnosis, treatment, and rehabilitation assessment of motor neuron diseases such as stroke, unilateral limb weakness caused by cerebral ischemia or cerebral infarction, and multi-limb weakness caused by subarachnoid hemorrhage, cerebral hemorrhage, and intracranial injury paralysis. Summary of the Invention

[0004] To address the problem that existing technologies cannot systematically analyze the causal links between multiple electroencephalogram (EEG) and electromyogram (EMG) signals, and to analyze the hierarchical characteristics of the coupling network between EEG and EMG signals, this invention proposes a multi-scale cortical-muscle coupling network analysis method based on ordinal patterns.

[0005] This invention first uses the Multivariate Variational Modal Decomposition (MVMD) method to decompose the electromyography (EMG) signal at the same frequency scale. Then, it uses the Multiple Bipartite Ordinal Partition Transition Networks (m-OPTNs) method to calculate the coupling strength of the multi-component signal and construct a cortex-muscle coupling network. Finally, it uses complex network parameters to analyze the characteristics of the coupling network.

[0006] A multi-scale cortex-muscle coupling network analysis method based on ordinal patterns, comprising the following steps:

[0007] Step 1: Decompose the signals of the EEG and EMG channels of each subject at the same frequency scale using the MVMD method to obtain the Intrinsic Mode Functions (IMFs) of each frequency band of the EEG signal.

[0008] Step 2: For each subject, the m-OPTNs causal inference algorithm is used to calculate the coupling strength between brain-muscle electrochannel signals under different delays, and the delay with the highest coupling strength is selected as the delay of the subject's cortical-muscle coupling network.

[0009] The coupling strength is calculated as follows: For a time series X = x c (t) The phase space trajectory is reconstructed using Takens' delayed embedding theorem to obtain the embedding vector.

[0010] v c (t)={x c (t), x c (t+d), ..., x c [t+(m-1)d]} (1)

[0011] Assume a s = (a0, a1, ..., a m-1 If (v) is an ordering of (0, 1, ..., m-1), then for the embedding vector v c Member x in (t) c (t), x c (t+d), ..., x c The ordering relationship of [t+(m-1)d] can correspond to a unique a. s ,satisfy

[0012] x c (t+a0d)≤x c (t+a1d)≤x c (t+a2d)≤…≤xc [t+a m-1 d] (2)

[0013] And when x c (ta l-1 )=x c (ta l When setting a) l-1 <a l Each type of a s This is called an OP pattern, where m is the embedding dimension; for an embedding dimension m, there exist m! different OP patterns, denoted by the symbols π1, π2, ..., π. M！ This indicates that the conditional entropy of time series X1 paired with X2 at a delay of τ is calculated using the OP (operational) mode.

[0014]

[0015] express and The frequency of simultaneous occurrence, Indicates OP mode in hour The frequency of occurrence of conditions.

[0016] Constructing m-OPTNs matrices using conditional entropy

[0017]

[0018] Setting a hard threshold for the m-OPTNs matrix yields according to The coupling strength between brain-muscle electroacupuncture channels was obtained;

[0019] Step 3: Treat each sampling channel as a node of the network, and use the weights in the multi-layer weighted network obtained by the m-OPTNs algorithm as the network edges. For each IMF component obtained by MVMD decomposition, calculate the adjacency matrix according to the delay of the subject's coupling network, and construct a multi-layer electroencephalogram (EEG) signal coupling network.

[0020] To ensure the accuracy of causal link identification and suppress spurious links when constructing a cortex-muscle coupled network, the following steps are required:

[0021] For each node X in this multi-layer network m Its k can be found by the following formula. m Parent nodes and l m child nodes

[0022]

[0023]

[0024] h mnτ Representing the adjacency matrix The value in the m-th row and n-th column is the coupling strength of the m-sequence to the n-sequence in the multilayer weighted network with a delay of τ.

[0025] For two nodes X m and X n Determining the causal relationship between them requires defining a minimal set of conditions.

[0026]

[0027]

[0028] By setting an upper limit on the number of elements in the minimum condition set, the most important neighbor set can be obtained through filtering.

[0029] Define parameters and

[0030]

[0031]

[0032] Through the Set a threshold to remove some indirect causal links in the network. Then X n Viewed as X m Fake links.

[0033] Step 4: Analyze the coupling characteristics between brain electromyography signals at different scales at the network and node levels using several network parameters, including node degree, node strength, betweenness number, clustering coefficient, and characteristic path length.

[0034] Preferably, the delay setting range between EEG and EMG signals in step 2 is 20ms to 40ms. The delay of the EMG signal coupling network varies among subjects, and the delay of bidirectional transmission of information between the cortex and muscle also varies.

[0035] Preferably, in step 2, a hard threshold is set for the m-OPTNs matrix constructed using conditional entropy, as follows.

[0036]

[0037] H max =log2M! (12)

[0038] Wherein, λ is empirically set to 0.99~1, and H is set to max Then it is assumed that there is no link between the two nodes.

[0039] Preferably, in step 3, the m-OPTNs method selects a dimension m of 3.

[0040] The beneficial effects of this invention are:

[0041] First, the present invention uses the m-OPTNs algorithm to measure the coupling strength between brain electromyography signals, which has advantages in identifying direct and indirect links between time series and can better characterize the differences in causal links between signals between different modes of human upper limb movement.

[0042] Second, this invention uses MVMD to decompose brain electromyography signals at the same frequency scale, analyzes the hierarchical characteristics of causal links between brain electromyography signals, and reflects the scaling characteristics of information exchange between the cortex and muscle.

[0043] Third, by combining the MVMD and m-OPTNs algorithms, a complex network approach is used to analyze the intrinsic causal links between brain electromyography signals, providing a new approach for systematically analyzing the information exchange between the cortex and muscles during human upper limb movement. Attached Figure Description

[0044] Figure 1 The process of constructing a cortex-muscle coupling network;

[0045] Figure 2 The results of preprocessing the electromyography signals collected in the experiment;

[0046] Figure 3(a) shows the spectrum of each component obtained by decomposition using the MVMD algorithm;

[0047] Figure 3(b) shows the IMF component map obtained by decomposition using the MVMD algorithm;

[0048] Figure 4 The coupling adjacency matrix was calculated using the m-OPTNs algorithm. In the figure, 1-6 represent the IMF1 components of C4, CP6, left finger flexor muscles, C3, CP5, and right finger flexor muscles, respectively; 7-12 represent the IMF2 components of the 6 channels; 13-18 represent the IMF3 components; 19-24 represent the IMF4 components; 25-30 represent the IMF5 components; 31-36 represent the IMF6 components; and 37-42 represent the IMF7 components.

[0049] Figure 5(a) shows the bar chart of the node strength of the brain-muscle electrocoupling network at various scales;

[0050] Figure 5(b) shows the side betweenness chord diagram of the brain-muscle electrocoupling network at various scales;

[0051] Figure 5(c) shows the betweenness coefficient histogram of the brain-muscle electrocoupling network nodes at various scales; where, in the left-hand experiment: black: C4 channel; gray: CP6 channel; white: left finger flexor muscle channel; in the right-hand experiment: black: C3 channel; gray: CP5 channel; white: right finger flexor muscle channel;

[0052] Figure 5(d) shows the characteristic path lengths of the brain-muscle electrocoupling network at each scale;

[0053] Figure 5(e) shows the aggregation coefficients of the brain-muscle electrocoupling network at various scales. Detailed Implementation

[0054] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings: This embodiment is implemented using a laboratory-collected dataset of constant grip force output electroencephalogram (EEG) data of the human upper limb, based on the technical solution of the present invention. Detailed implementation methods and specific operating procedures are provided.

[0055] like Figure 1 As shown, this embodiment of the patent includes the following steps:

[0056] Step 1: Preprocess the acquired EEG and EMG signal data using the EEGLAB toolbox to remove interference from artifacts such as those in EEG and ECG, and then use wavelet denoising methods for noise reduction, such as... Figure 2 As shown, this represents a comparison of EEG and sEMG signals before and after noise reduction.

[0057] Step 2: Use the MVMD algorithm to decompose the brain electromyography signal at the same frequency scale. The decomposition results are shown in Figure 3(a) and Figure 3(b).

[0058] Step 3: Use the m-OPTNs method to calculate the coupling strength between the electroencephalogram (EEG) signal components in each frequency band, and construct a cross-coupling adjacency matrix, such as... Figure 4 As shown, channels 1 to 36 are the 6-channel EEG and EMG signals of IMF1 to IMF6, respectively, and channels 1 to 6 are the 6-channel EMG signals of the IMF1 level, and so on.

[0059] Step 4: Analyze the cortical-muscle coupling network of different grip force outputs in the human upper limb using complex network parameters. The results are as follows: Figures 5(a)-5(e) As shown.

[0060] In the brain-muscle electrocoupling network of upper limb movement, the C3 and C4 EEG channels play a crucial role in information transmission. The superficial flexor muscles of the left and right hands simultaneously receive and transmit signals. Information is transmitted bidirectionally between the scalp EEG and muscles, with the descending direction exhibiting greater transmission strength than the ascending direction. This aligns with reality, as human movement is accomplished by the coordinated action of multiple muscles, and there are coupling relationships between muscle signals at different locations. The experimental results also demonstrate the hierarchical characteristics of information transmission between the cortex and muscles. Cross-band experiments show that in the high-frequency band, the coupling peak between brain-muscle electrocoupling signals occurs at the same frequency scale, and the coupling strength in the high-frequency band increases with increasing grip strength. In the low-frequency band, there is a strong coupling between low-frequency electromyographic signals (IMF1) and electroencephalogram (EEG) signals, and the low-frequency IMF channel plays an important role in information transmission in this directed network. As the grip strength level increases, the role of higher frequency band (IMF4, IMF6) EEG components in information transmission in this coupled network increases. These scale components are mainly in the gamma band, which is consistent with the transfer characteristics of the cerebral cortex to the muscular system in terms of neural oscillation when static force is adjusted to dynamic force output. Furthermore, the experimental results of the left and right hands are consistent, which confirms the effectiveness of the method in this invention.

Claims

1. A method for multi-scale cortical-muscular coupling network analysis based on ordinal pattern, characterized in that: The method comprises the following steps: Step 1: signal decomposition of the electroencephalogram and electromyogram channels of each subject to obtain IMF components of the brain-muscle electrical signal in each frequency band; Step 2: For each frequency band IMF component of the brain and muscle electrical signals of each subject, the m-OPTNs matrix is calculated using the multivariate bivariate time series partition transfer network causal inference algorithm under different time delays, and a hard threshold is set for the m-OPTNs matrix to obtain According to The coupling strength between the brain and muscle electrical channel signals is obtained; and the time delay with the highest coupling strength is selected as the time delay of the cortical-muscular coupling network of the subject. Step 3: taking each sampling channel as a node of the network and taking the weight in the multi-layer weighted network obtained by the m-OPTNs algorithm as the edge of the network, for each IMF component obtained by using the MVMD decomposition, calculating the adjacency matrix according to the highest coupling strength of the subject and constructing a multi-layer brain-muscle electrical signal coupling network; When constructing the cortex-muscle coupling network, in order to ensure the accuracy of the causal link identification result and inhibit false links, the following steps are required: For each node X in this multilayer brain-muscle electrosynthesis signal coupling network m Its k can be found by the following formula. m Parent nodes and l m child nodes h mnτ denotes the adjacency matrix the value in the mth row and nth column of the matrix, i.e. the coupling strength value of the mth sequence to the nth sequence on the layer with delay τ For the determination of the causal relationship between two nodes X m and X n , define the minimum condition set By setting an upper limit to the number of elements in the minimum condition set, the most important neighbor set is screened. Definition parameters and Through the Set a threshold to remove some indirect causal links in the network. Then X n Viewed as X m The pseudo-links; based on the constructed multilayer brain-myoelectric signal coupling network, the node degree, node strength, betweenness, clustering coefficient and characteristic path length are obtained; Step 4: using several network parameters in the multi-layer brain-muscle electrical signal coupling network, such as node degree, node strength, betweenness, clustering coefficient and characteristic path length, to analyze the coupling characteristics between the brain-muscle electrical signals at different scales from the network level and the node level.

2. The ordinal pattern based multi-scale cortical-muscular coupling network analysis method according to claim 1, wherein: The signal decomposition of the electroencephalogram and electromyogram channels of each subject in step 1 is performed by using the multivariate variational mode decomposition method for simultaneous frequency-scale decomposition.

3. The ordinal pattern based multi-scale cortical-muscular coupling network analysis method according to claim 1, wherein: The coupling strength is calculated as follows: For a time series X = x c (t) reconstruct the phase space trajectory using Takens delay embedding theorem to get the embedding vector v c (t) = {x c (t), x c (t + d),..., x c [t + (m - 1)d]} (7) Assume a s =(a0,a1,…,a m-1 If (v) is an ordering of (0, 1, ..., m-1), then for the embedding vector v c Member x in (t) c (t),x c (t+d),…,x c The ordering relationship of [t+(m-1)d] corresponds to a unique a. s ,satisfy x c (t + aod) < x c (t + a1d) < x c (t + a2d) <... < x c [t + a m-1 d] (8) and when x c (t-a l-1 ) = x c (t-a l ) is set a l-1 <a l ; each a s is called an OP pattern, where m is the embedding dimension; for embedding dimension m, there are m! different OP patterns, which are denoted by symbols π1, π2, …, π M！ The conditional entropy of time series X1 on X2 at delay τ is calculated using OP patterns representing and the frequency of co-occurrence, representing the OP mode in when the conditional frequency of occurrence; The m-OPTNs matrix is constructed by using conditional entropy Setting hard threshold to m-OPTNs matrix gets According to The coupling strength between the brain and muscle electrical channel signals is obtained.

4. The ordinal pattern based multi-scale cortical-muscular coupling network analysis method according to claim 1, wherein: In step 2, the delay with the highest coupling strength is set to a range of 20ms to 40ms.

5. The ordinal pattern based multi-scale cortical-muscular coupling network analysis method according to claim 1, wherein: In step 2, the m-OPTNs matrix constructed by using conditional entropy is set with a hard threshold value, and the method is as follows: H max = log2M! (12) where λ is set to 0.99-1 empirically, and set to H max It is considered that there is no link between two nodes.

6. The ordinal pattern based multi-scale cortical-muscular coupling network analysis method according to claim 1, wherein: In step 3, the m-OPTNs method selects the dimension m as 3.

Images