阅读说明：本技术 一种多尺度补偿传递熵的皮层肌肉功能耦合方法 (Cortical muscle function coupling method for multi-scale compensation transfer entropy ) 是由 佘青山 张敏 孟明 张启忠 于 2020-12-01 设计创作，主要内容包括：本发明涉及一种多尺度补偿传递熵的皮层肌肉功能耦合方法。本发明首先采用APIT-MEMD方法对同步采集的脑电和肌电信号进行时-频尺度化,然后计算不同尺度上的多尺度补偿传递熵值,分析各个尺度不同耦合方向上的耦合特征。结果表明在恒定握力下,beta频段间的耦合强度最为显著。不同耦合方向的耦合强度随握力的增加会呈现出一定的变化规律。本发明可以定量描述脑肌电信号在不同的耦合方向、多尺度间的皮层功能耦合强度及信息传递量,为探究手部运动控制及患者评价提供一定的理论依据。(The invention relates to a cortical muscle function coupling method for multi-scale compensation transfer entropy. The method comprises the steps of firstly carrying out time-frequency scaling on synchronously acquired electroencephalogram and electromyogram signals by adopting an APIT-MEMD method, then calculating multi-scale compensation transmission entropy values on different scales, and analyzing coupling characteristics of each scale in different coupling directions. The results show that the coupling strength between beta frequency bands is most significant under constant grip strength. The coupling strengths in different coupling directions show a certain change rule with the increase of the grip strength. The invention can quantitatively describe the cortical function coupling strength and the information transmission quantity of the brain electromyographic signals in different coupling directions and among multiple scales, and provides a certain theoretical basis for exploring hand motion control and patient evaluation.)

1. A cortical muscle function coupling method for multi-scale compensation transfer entropy is characterized in that: the method comprises the following main steps:

step (1), synchronously acquiring multi-channel brain and muscle electrical signals;

the method specifically comprises the following steps: simultaneously acquiring multi-channel brain electromyographic signals of the left hand and the right hand of a testee under different grasping actions;

step (2), carrying out time-frequency scaling on the synchronously acquired brain and muscle electrical signals by adopting an APIT-MEMD method;

the method specifically comprises the following steps: firstly, two groups of electroencephalogram signals X ═ X are constructed₁,x₂,…x_i,…,x_MY and electromyographic signal Y ═ Y₁,y₂,…y_i,…,y_MThe time sequence M is the length of the time sequence, then an APIT-MEMD method is adopted to decompose two groups of time sequences to respectively obtain K narrow-band IMF components, and two groups of time sequences of the brain-muscle electricity related to the time-frequency scale are respectivelyAnd

step (3), calculating McTE values of the brain electromyographic signals on different scales;

the method specifically comprises the following steps: two random processes X, Y representing the time variation of the physical system x, y, let x_tAnd y_tIs a random description of the access states of X and Y at time tVariable, and x_n:tIs all x samples of the constituent vector from time n to time t, the compensation delivery entropy cTE is defined as:

cTE_X→Y＝H(y_t|y_1:t-1,x_t)-H(y_t|y_1:t-1,x_1:t) (1)

construction signalTo signalMulti-scale compensation transfer entropy McTE_EEG→EMG：

Wherein McTE_EEG→EMGK representing EEG₁Component to k of EMG₂Multi-scale compensation of components conveys entropy values

The same can be obtained:

wherein McTE_EMG→EEGK representing EMG₂Component to EEG k₁Multi-scale compensation of the components conveys entropy values;

and (4) carrying out cortical muscle coupling analysis in different coupling directions and different frequency bands by adopting the calculation results of the steps (2) and (3).

Technical Field

The invention belongs to the field of research of nervous system motion control mechanisms, and designs a novel multi-scale compensation transfer entropy method.

Background

The interaction between the motor cortex and effector muscles is considered a key feature for effective control of movement. It is reported that cortical muscular functional coupling (FCMC) occurs mainly in the alpha band (8-14 Hz) during continuous contraction, slow finger movement and rapid transition between two stressed targets; during control and maintenance of steady state force output, FCMC occurs in the beta band (15-35 Hz); under the action of strong muscle strength generation and power, the gamma frequency band (35-60 Hz) has a remarkable influence. These studies indicate that FCMC in different frequency bands play different roles in the sensory and motor systems of the subject.

At present, cortical muscle coupling analysis is mainly based on Coherence (Coherence) method to obtain functional connection characteristics between brain motor consciousness and muscle motor response, but traditional Coherence can only describe linear coupling relationship, and in order to better understand functional coupling effect and information transmission characteristics between cerebral cortex and corresponding muscle, the Transfer Entropy (TE) method proposed by Schreiber is widely applied in the field of biological signal processing. Although TE does not employ any model assumptions, causal relationships are defined only in the time domain. However, in actual time series analysis there are transient effects, i.e. effects that occur between two time series within the same time lag, which may reflect a rapid physiologically meaningful interaction, or no physiological significance. In either case, transient effects can affect the calculation of any causal metric.

Recent studies suggest to incorporate the variance terms in the model residual covariance matrix into a so-called partial Granger causal metric or to represent the residual correlation with model coefficients and to define an extended Granger causal metric with the resulting new model structure. However, transient effects are generally not considered in TE calculations of experimental data, since a similar approach cannot be followed in a modeless environment of TE analysis. Thus, Faes proposed cTE a method that combines Conditional Entropy (CE) estimation, non-uniform embedding, and takes into account transient causal relationships to solve the above problem. The conditional entropy estimation must consider the past state of the process, and the problems of overfitting, detection of false causal relationship and the like can be caused by the randomness and redundancy caused by the current embedding scheme which follows uniform multivariable; in addition, although several estimators designed for multidimensional space can be used for conditional entropy estimation, as the embedded vector dimension is increased, the bias increasingly affects the conditional entropy estimation. The method recognizes that TE can be interpreted as CE difference and builds on an efficient CE estimator. The estimator compensates for deviations of the high dimensional condition vectors and follows a sequential embedding process, wherein the condition vectors are formed step by step according to a criterion of CE minimization; and the deviation influencing the conditional entropy estimation is compensated when the embedding dimension is increased, so that the existence of the minimum entropy rate in the termination process is ensured. Although recent studies have begun to elucidate transient causal relationships in the linear parameter framework of multiple autoregressive models, there is little work to address the consequences of eliminating transient effects in model-free causal relationship metric calculations.

However, the physiological system needs to be controlled across multiple space-time scales, so that the brain electromyographic signals have multi-scale characteristics, and therefore, many scholars combine TE and a multi-scale frequency band decomposition algorithm to analyze the cortical muscle coupling relationship in a multi-level manner on a time-frequency domain. For example, a Variable Mode Decomposition (VMD) and a TE method are combined to form a VMD-TE model; the MEMD-TE analysis model is formed by combining Multivariate Empirical Mode Decomposition (MEMD) and a TE method, and is applied to brain-muscle coupling analysis, and the result shows that functional coupling of cortical muscles has bidirectionality, and the coupling strength in the downlink direction of a high-frequency band (40Hz to 75Hz) is greater than that in the uplink direction. EMD is an adaptive data-driven algorithm for analyzing nonlinear and non-stationary time series, but the signal has boundary effect and modal aliasing after decomposition by the method. The MEMD not only can simultaneously process multi-scale analysis of a multi-element time sequence, but also can well eliminate the modal aliasing phenomenon. Because the multivariate signals acquired by the multi-sensor acquisition system cannot avoid power imbalance among different channels and correlation exists among different channels, the vector of the uniform sampling method used by the MEMD is not optimal for capturing the dynamics of multivariate quantity, and the calculation complexity is greatly increased by using projection vectors with the quantity larger than 127; for low (8-31) to medium (32-63) numbers of prediction and imbalance-related data channels, the algorithm provides a sub-optimal estimate of the local mean. Non-uniformly sampled BEMD (NS-BEMD) generates a set of projection vectors representing the highest curvature direction, alleviating this problem. For a three-variable signal, a non-uniform sampled TEMD (NS-TEMD) globally identifies a set of non-uniformly sampled projections by eigen-decomposition of the covariance of the signal and then constructing a non-uniform directional ellipse that matches the signal dynamics. The adaptive projection multivariate empirical mode decomposition (APIT-MEMD) expands an NS-TEMPD method, can better solve the problems of power imbalance and correlation in multi-channel brain electromyographic data, is similar to a standard MEMD, can accommodate nonlinear and non-stationary signals, can relieve mode mixing and generates less IMF components for unbalanced data.

In order to solve the above problems, the invention provides a novel model-free causal relationship analysis method, namely a Multiscale compensated Transfer Entropy (McTE) analysis method, on the basis of cTE. The neuromuscular synchronous coupling between different scales in the process of constant-force output upper limb movement is researched, the coupling characteristics between the brain electromyographic signals and the information transmission method are quantitatively described, and a basis is provided for the generation mechanism of further movement dysfunction and the movement function evaluation method in the rehabilitation process.

Disclosure of Invention

The invention aims to provide an analysis method for obtaining the electroencephalogram and electromyogram coupling strength characteristics of multiple channels of electroencephalogram and electromyogram signals under different frequency band scales.

In order to achieve the purpose, the method mainly comprises the following steps:

step (1), synchronously acquiring multi-channel brain and muscle electrical signals;

the method specifically comprises the following steps: and simultaneously, multi-channel brain and muscle electrical signals of the left hand and the right hand of the testee under different grasping actions are collected.

Step (2), carrying out time-frequency scaling on the synchronously acquired brain and muscle electrical signals by adopting an APIT-MEMD method;

the method specifically comprises the following steps: firstly, two groups of electroencephalogram signals X ═ X are constructed₁,x₂,…x_i,…,x_MY and electromyographic signal Y ═ Y₁,y₂,…y_i,…,y_MThe time sequence M is the length of the time sequence, then two groups of time sequences are decomposed by adopting an APIT-MEMD method to respectively obtain K narrow-band components IMF, and two groups of time sequences of the electroencephalogram and electromyogram related to time-frequency scale are respectivelyAnd

step (3), calculating McTE values of the brain electromyographic signals on different scales;

the method specifically comprises the following steps: two random processes X, Y representing the time variation of the physical system x, y, let x_tAnd y_tIs a random variable describing the access state of X and Y at time t, and X_n:tIs all x samples of the constituent vector from time n to time t, the compensation transfer entropy (cTE) is defined as:

cTE_X→Y＝H(y_t|y_1:t-1,x_t)-H(y_t|y_1:t-1,x_1:t) (1)

where H (-) denotes shannon entropy, which denotes the uncertainty associated with any measure a of the vector random variable a, H (a) sigma_ap(a)logp(a) H (· | ·) represents Conditional Entropy (CE), which represents the uncertainty that still exists when b is known, H (a | b) ═ H (a, b) -H (b). TE, as defined in equation (1), quantifies the transfer of information from X and Y to a target process Y (i.e., Y)_t) Is not included in the past of the source process x (i.e. x)_1:t-1(ii) a The start of time here is usually set to t ═ 1).

Construction signalTo signalMulti-scale compensation transfer entropy McTE_EEG→EMGThe formula is as follows:

wherein McTE_EEG→EMGK representing EEG₁Component to k of EMG₂Multi-scale compensation of components conveys entropy values

The same can be obtained:

wherein McTE_EMG→EEGK representing EMG₂Component to EEG k₁The multi-scale compensation of the components conveys entropy values.

And (4) carrying out cortical muscle coupling analysis in different coupling directions and different frequency bands by adopting the calculation results of the steps (2) and (3).

Compared with the traditional cortical muscle coupling analysis method, the method has the following advantages:

the multi-scale compensation transfer entropy provided by the invention can solve the problems of power imbalance and correlation in multi-channel data, can also relieve the problem of mode mixing, generates less IMF components, effectively describes the energy coupling characteristics between cortical muscles of a subject in different frequency bands and different transfer directions, and provides a certain theoretical basis for researching motion control and feedback information decoding and patient rehabilitation evaluation.

Drawings

FIG. 1 is an experimental diagram of the method of the present invention;

FIG. 2 is a time domain decomposition result graph of the brain and muscle electricity of the subject S3;

FIG. 3 is a time-frequency decomposition result diagram of the electromyographic signal of the subject S1;

FIG. 4 is the McTE and MEMD-cTE values at different scales of subject S3;

FIG. 5 is a graph of McTE values between different scales for EEG and EMG of subject S2;

FIG. 6 is a graph of McTE values between different scales for EEG and EMG of subject S3;

fig. 7 shows the McTE values for different force levels for different subjects in the high frequency range.

Detailed Description

The embodiments of the present invention will be described in detail below with reference to the accompanying drawings: the embodiment is implemented on the premise of the technical scheme of the invention, and a detailed implementation scheme and a specific operation process are given.

The electroencephalogram signals of the cerebral motor cortex and the corresponding electromyogram signals of the opposite sides of the limbs can respectively reflect control information of the brain on the limb movement and feedback information of the muscles on the brain, and functional coupling connection between the cerebral cortex and the corresponding muscles in the limb movement process is disclosed. Therefore, analyzing the coupling characteristics of the multi-channel brain electromyographic signals in different coupling directions of various scales is particularly important for researching the motion control mechanism of the central nervous system. The invention provides a novel multi-scale compensation transfer entropy analysis method. The implementation of the invention mainly comprises the following steps: (1) synchronously acquiring multi-channel brain and muscle electrical signals; (2) carrying out time-frequency scaling on the synchronously acquired brain and muscle electrical signals by adopting an APITMEMD method; (3) and (4) calculating McTE values of the brain electromyographic signals on different scales, and performing cortical muscle coupling analysis in different coupling directions and different frequency bands by adopting the calculation results of the steps (2) and (3).

The respective steps are explained in detail one by one below.

The method comprises the following steps: synchronously acquiring multi-channel brain and muscle electrical signals;

two healthy subjects (S1, S2) and one stroke subject (S3) were recruited for this experiment, both with right-handed, no neurological history, and both were informed of the experimental details, signed with informed consent, and did not exercise vigorously within 24 hours before the test to prevent the effects of sports fatigue. The sampling time is 5 seconds, the sampling frequency is 1000Hz, the patient has a rest for 15s after each grasping action, the patient needs to complete the grasping of 5kg, 10kg and 20kg of left and right hands for 5 times respectively, and the stroke patient can only complete the grasping of 5kg and 10kg of left and right hands due to physical reasons. In the experiment, a 128-lead BrainAmp DC electroencephalogram acquisition system is adopted to synchronously acquire 32 paths of electroencephalogram signals and myoelectric signals of left and right Biceps Brachii (BB) and ulnar wrist Flexor (FCU), the detected part is wiped by alcohol before acquisition, and grease and dandruff on the surface of the skin are removed, which is shown in figure 1.

Step two: carrying out time-frequency scaling on the brain electromyographic signals which are synchronously acquired by adopting an APIT-MEMD method;

consider two random processes X, Y representing the time variation of the physical systems x, y, let x_tAnd y_tIs a random variable describing the access state of X and Y at time t, and X_n:tIs all x samples of the constituent vector from time n to time t, the compensation transfer entropy (cTE) is defined as:

cTE_X→Y＝H(y_t|y_1:t-1,x_t)-H(y_t|y_1:t-1,x_1:t) (1)

aiming at the nonlinear and multi-scale characteristics of electroencephalogram and electromyographic signals, adaptive projection multivariate empirical mode decomposition (APIT-MEMD) can solve the problems of power imbalance and correlation in multichannel electroencephalogram and electromyographic signals, is similar to standard MEMD, can accommodate nonlinear and non-stationary signals, can relieve mode mixing and generates less IMF components for unbalanced data.

Firstly, two groups of electroencephalogram signals X ═ X are constructed₁,x₂,…x_i,…,x_MY and electromyographic signal Y ═ Y₁,y₂,…y_i,…,y_MTime series, and then decomposing the two time series by adopting an API-TMEMD method. Assuming that a multivariate signal x (t) exists, its covariance matrix is C ═ E { x ═^T(t) x (t), determining the direction of the first principal component by eigen decomposition of a covariance matrix, as follows:

C＝VΛV^T (2)

wherein V ═ V₁,v₂,...v_n]Is a matrix of eigenvectors, Λ ═ diag { λ₁,λ₂,...λ_nIs a matrix of eigenvalues, the maximum eigenvalue λ₁Corresponding to the feature vector v₁Feature vector v₁Is the first principal component direction, i.e., the point pointing in the direction of maximum power imbalance.

Subsequently, a building edge v₁Another direction vector v in the opposite direction_o1＝-v₁V is to be₁And v_o1For repositioning all direction vectors previously generated by the uniform projection scheme, the formula is as follows:

wherein the content of the first and second substances,a set of k direction vectors is generated after (n-1) spheres are uniformly sampled for a Hammerseley sequence. Alpha (alpha epsilon [0, 1)]) The degree of power imbalance of the multivariate signal is determined, alpha is 1 to represent high power imbalance among channels, and APIT-MEMD can well process importance sampling of signal space. In an iterative process, x (t) is iterated along these adaptive projection vectors and local means are obtained by MEMD.

Step three: and calculating the McTE value of the brain electromyogram signal on different scales.

After the EEG and EMG signals are decomposed by an API-TMEMD method, K narrow-band components IMF are respectively obtained, and then two groups of time-frequency scale-related electroencephalogram and electromyogram time sequences are respectivelyAnd

next, the coupling strength and information transfer characteristics on various scales of EEG and EMG signals were studied according to the definition of the compensation transfer entropy in equation (1).

Construction signalTo signalMulti-scale compensation transfer entropy McTE_EEG→EMGThe formula is as follows:

McTE_EEG→EMGk representing EEG₁Component to k of EMG₂The multi-scale compensation of the components delivers entropy values, which has the physiological meaning of the information delivery and coupling strength between the cerebral cortex and the muscular tissue in a certain scale up and down (EEG → EMG) brain-to-muscle direction. Similarly, the intensity of feedback of the muscle tissue in the ascending (EMG → EEG) muscle to brain direction to the cortical control commands and to the external information may be compensated for by multi-scale delivery of entropy McTE_EMG→EEGThe formula is as follows:

in the same way, McTE_EMG→EEGThen k represents EMG₂Component to EEG k₁Of a componentThe multi-scale compensation conveys entropy values.

Step four: and (4) carrying out cortical muscle coupling analysis in different coupling directions and different frequency bands by adopting the calculation results of the steps (2) and (3).

The EEG and EMG synchronously acquired by the subjects (S1-S3) are first subjected to APIT-MEMD and MEMD decomposition, respectively. FIG. 2(a) shows an exploded MEMD view of a three-channel brain myoelectric signal C4, BB and FCU of a subject S3. As can be seen from the figure, the MEMD decomposition can simultaneously analyze multi-channel signals, and 13 IMF components are generated after the three-channel brain-muscle electrical signal decomposition, as shown in FIG. 3. FIG. 2(b) shows an exploded view of the subject S3 three-channel brain myoelectric signal APIT-MEMD, which, like MEMD, can accommodate non-linear and non-stationary signals and also mitigate mode mixing, but the APIT-MEMD decomposition produces less IMF component than the MEMD decomposition.

In order to research the coupling characteristics of EEG and EMG in different coupling directions (EEG → EMG, EMG → EEG) and different scales, the invention calculates the McTE values of EEG and EMG signals of 3 subjects in different scales according to formulas (5) and (6), taking a healthy subject S1 as an example, carries out APIT-MEMD decomposition on the EEG and EMG signals (C4 and BB) acquired when the right hand is held for 5kg, and obtains the bandwidth of IMF components and the corresponding frequency band,

to further compare the differences in coupling between cortical muscle bands of the subjects, the McTE and MEMD-cTE values of the electromyographic coupling of the brain and muscle during left and right hand movements of subject S3 (patient) were extracted (similar to other subjects), as shown in fig. 4, in which the black dashed curve represents the McTE value and the black solid curve represents the MEMD-cTE value.

Although the number of IMF components of EEG and EMG signals of the subject S3 after APITMEMD decomposition is different under different grip conditions, the characteristics of coupling and information transmission between cortical muscles are similar. First, the cortical muscle functional coupling of the subject is bi-directional, in the up and down directions, while the black solid line box is the value of McTE for coupling of EEG and EMG between the high gamma bands (50-75Hz) where the EEG → EMG coupling strength is always greater than the EMG → EEG for either left or right hand motion. This may be due toIn the process of maintaining constant grip output, the motor cortex needs to integrate information and transmit the information to muscles, and the difference of directional synchronous oscillation between sensory feedback and a motion control mechanism is reflected. Whether left or right handed of subject S3, McTE_EEG→EMGThe value is always greater than McTE_EMG→EEGThe value is obtained. The black solid line curve in fig. 4 is the MEMD decomposition result, and it can be seen that the result of the method is similar to the result decomposed by the apitmmd in the high gamma frequency band, but the MEMD in the beta frequency band cannot correctly represent the coupling characteristics of the electroencephalogram and the myoelectricity, as shown by the black dashed line box in the figure.

Next, McTE values of coupling among various scales of the subject S2 in the EEG → EMG, EMG → EEG directions are extracted, fig. 5 and 6 show McTE values of coupling of the brain and muscle electricity among various scales of the subject S2 and S3 under different grip strengths of the left hand and the right hand, it can be seen that the coupling strength is most obvious in 6 th to 8 th scales, the result is consistent with the study of EEG-EMG coupling characteristics in the existing static grip strength output experimental mode, the maintenance function of the coupled oscillation of the beta band of the motor cortex on stable motion output is proved, the oscillation in the beta band of the muscle is considered to be the common drive of the corticospinal cone tract and the cortex, and the coupled oscillation of the beta band of the muscle embodies the information transmission between the muscle and the sensory motor cortex.

In order to further explore the coupling strength changes of the EEG and the EMG in different grip strength conditions and different coupling directions, the IMF components corresponding to beta and gamma frequency bands are selected from the signals subjected to APITMEMD decomposition and are reconstructed into new signals, and then the McTE of the subject is calculated according to the formulas (6) and (7)_EEG→EMGValue sum McTE_EMG→EEGThe value is obtained. Fig. 7 shows McTE values for two sets of brain myoelectrical signals (C4 BB and C4 FCU) in different coupling directions for subject S2 (healthy) and subject S3 (patient).

It can be seen that the right hand coupling strength is greater than the left hand coupling strength, which may be due to the right handedness, and thus the brain has greater control over the right hand than the left hand. In fig. 7, the comprehensive analysis of the left and right hand brain myoelectricity shows: whether left-handed or right-handed, McTE_EEG→EMGAll values are greater than McTE_EMG→EEGThe value is obtained. And with increasing grip strength, McTE_EEG→EMGAnd McTE_EMG→EEGThe value of (c) increases.

The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention may be made by those skilled in the art without departing from the spirit of the present invention, which is defined by the claims.