A stroke rehabilitation evaluation method based on brain muscle function network features

Abstract

The present application relates to the technical field of pattern recognition, in particular to a stroke rehabilitation evaluation method based on brain muscle function network features, comprising the following steps: S1, collecting electroencephalogram signals and electromyogram signals, collecting electroencephalogram signals and electromyogram signals of a stroke patient under a hand dynamometer test; S2, respectively pre-processing the electroencephalogram signals and electromyogram signals; S3, constructing a variational mode decomposition transfer entropy; S4, constructing a CMFN model based on EEG and EMG; S5, brain muscle function network analysis. The method considers that the research on motor recovery of stroke stays in analyzing the consistency or coupling relationship changes between the cerebral cortex and the muscle, realizes the quantification of stroke evaluation, and greatly improves the evaluation accuracy.

Description

Technical Field

[0001] This invention relates to the field of pattern recognition technology, specifically to a method for stroke rehabilitation assessment based on brain muscle functional network characteristics. Background Technology

[0002] Stroke is the most common disease among acute hospitalized patients in neurology departments in my country. Like other cardiovascular diseases, the incidence and disability rate of stroke are increasing year by year. According to data from the "China Stroke Prevention and Control Report 2019" [1], the overall lifetime risk of stroke in my country is 39.3%, ranking first in the world, and the mortality rate is 5 times that of North America. In 2018, about 1.94 million people died from stroke, of which the mortality rate of urban residents was 129 / 100,000 and the mortality rate of rural residents was 160 / 100,000. It is estimated that there are about 13 million existing stroke patients over the age of 40. Therefore, there will be more demand for improving stroke care capabilities and improving postoperative rehabilitation of stroke through rich neurorehabilitation methods.

[0003] Current techniques for treating stroke include thrombolysis, endovascular recanalization, carotid artery angioplasty, and interventional and surgical clipping of aneurysms. While thrombolysis effectively reduces stroke morbidity and mortality, most patients (more than 50%) who undergo thrombolysis still experience postoperative neurological dysfunction, although significantly worse than untreated patients. To date, breakthroughs have not been achieved in addressing the problems encountered in the treatment of acute stroke. The main reason is that we understand more about the causes of stroke (such as ruptured cerebral blood vessels and thrombosis) than about the causes of functional recovery. Therefore, to promote the rehabilitation of stroke patients, it is necessary to understand the physiological processes of stroke neurological deficits and recovery. Since Broca et al. discovered specific defects caused by focal brain injury, our understanding of the mechanisms of stroke defects has been influenced by a localized perspective. Later, through neuroimaging of lesion structures, it was suggested that neurological defects are caused by the disruption of the localized and functional specificity of brain regions. Similarly, the recovery of neurological deficits is related to the traditional localized perspective, namely, the reorganization of restricted, preserved brain regions near the damaged site. Many scholars have also demonstrated in their research that functional recovery after stroke rehabilitation involves different cortical regions. For example, Nudo et al., in their study of brain plasticity after stroke in non-human primates, showed that recovery from neurological deficits can be achieved through the reorganization of functions previously performed by damaged tissues. Nelles et al., using positron emission tomography (PET) to study the functional reorganization of the motor and sensory systems in hemiplegic stroke patients, observed increased cerebral blood flow in the sensorimotor cortex and both parietal lobes. Cramer et al., using functional magnetic resonance imaging (fMRI) in hemiplegic stroke patients, found that patients showed greater activation of the same motor areas and contralateral hemisphere as normal controls. These findings can enhance the understanding of the role of different cortical regions in the rehabilitation process of stroke patients; therefore, rehabilitation therapy can focus on promoting the repeated and intensive activation of these cortical regions.

[0004] In the context of exploring the causes of stroke rehabilitation, establishing the unique functional coupling between the cortex and muscles is a major goal of neuroscience. Motor deficits following stroke are widely considered to be due to the loss of functional coupling between the motor cortex and muscles, and restoring or supporting alternatives to this coupling can promote the recovery of motor abilities in stroke survivors. Electroencephalography (EEG) and electromyography (EMG) help us explore the coupling between the cortex and muscles by recording weak electrical signals generated by cortical activity and muscle contraction. First, isometric muscle contractions in healthy subjects showed cortical-muscle synchronization, supporting the hypothesis that the motor cortex sends movement commands to muscle groups and that the sensation of muscle contraction is fed back to the cortex via neurons. Second, numerous cortical-muscle coupling peaks were found in the contralateral cortical region in highly impaired stroke patients, supporting the hypothesis of functional recovery. Third, the coherence between the cortex and muscles decreases after stroke, and a recent study suggests that this coherence can assess post-stroke motor recovery. Therefore, the consistency between the cortex and muscles can serve as a biomarker for motor recovery. Current research primarily analyzes changes in the consistency or coupling between the cerebral cortex and muscles, without incorporating functional networks. This invention proposes a method for assessing the rehabilitation of stroke patients by establishing a cortico-muscular functional network (CMFN) using synchronized EEG and EMG data, and analyzing changes in network parameters and cortico-muscular functional coupling (CMFC) at various stages of recovery. Summary of the Invention

[0005] This invention proposes a stroke rehabilitation assessment method based on the characteristics of the brain-muscle functional network. Considering that research on stroke motor recovery is limited to analyzing changes in the consistency or coupling relationship between the cerebral cortex and muscles, this invention achieves the quantification of stroke assessment, greatly improving the accuracy of the assessment.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A stroke rehabilitation assessment method based on brain-muscle functional network characteristics includes the following steps:

[0008] S1. Collect EEG and EMG signals. Collect EEG and EMG signals of stroke patients under grip strength test.

[0009] S2. Preprocess the EEG and EMG signals separately:

[0010] S3. Constructing the variational mode decomposition transfer entropy

[0011] S3-1. The time series of the relevant frequency bands of the EEG signal obtained by bandpass filtering is denoted as y(t):

[0012]

[0013] S3-2. Information at different scales of electromyography signals is extracted using VMD to obtain the time series at relevant time scales as x. k (t):

[0014] {x k (t)}(k=1,2,…,K)

[0015] Where K is the number of VMD decomposition layers;

[0016] S3-3, Calculate y(t) and x k The average TE between (t) is used as the functional coupling value between the cerebral cortex and trunk muscles. The specific calculation process of VMD-TE from EEG signal to EMG signal is as follows:

[0017]

[0018] Where n is the discrete time series; p(*) is the joint probability between variables. The larger the VMD-TE value, the stronger the CMFC between EEG and EMG in this frequency band, and vice versa.

[0019] S4. Construct a CMFN model based on EEG and EMG;

[0020] S5. Brain-muscle functional network analysis.

[0021] Preferably, the acquisition methods for EEG and EMG in step S1 are as follows:

[0022] Phase 1: Pre-experiment preparation. Stroke patients sit quietly in front of the screen, adjust their mood, and stabilize their resting EEG signals.

[0023] Phase 2: During the experiment, stroke patients gripped the hand gripper all the way down within 3 seconds according to the on-screen instructions, then rested for 1 minute, and repeated 20 times.

[0024] Phase 3: The experiment was conducted again. The stroke patients rested for 20 minutes after the previous experiment, and then the Phase 2 experiment was repeated. The experiment was repeated a total of 3 times, and a total of 60 sets of data were collected per person per cycle.

[0025] Preferably, the method for preprocessing the acquired EEG signals in step S2 is as follows:

[0026] Removal of power frequency interference from EEG data using a 50Hz notch filter;

[0027] Spatial constraint ICA is used to reduce the energy loss of the artifact-free portion of ICs. That is, a spatial constraint is defined on A to represent specific prior knowledge or prior assumptions related to the spatial terrain projected by certain source sensors. The A after SC is shown below:

[0028]

[0029] Where A is an unknown and invertible m×n matrix, representing the transfer function between the signal source and the received signal. It is a constrained column, A u It is an unconstrained column;

[0030] The discrete wavelet transform (DWT) is used to denoise the extracted spatially constrained independent components. The DWT can be defined by the following formula:

[0031]

[0032] Where f(x) can be used to describe SC-ICs, k is the time shift, j is the decomposition scale, and φ j,k (x) is the mother wavelet, d j,k The wavelet coefficients are shown below, and the inverse wavelet transform is as follows:

[0033] φ j,k (x)=2 -j / 2 φ(2 -j / 2 xk),j,k∈Z.

[0034] Preferably, the method for preprocessing the acquired electromyographic signals in step S2 is as follows:

[0035] Empirical mode decomposition is used to derive the first-order intrinsic mode function from the EMG signal, and the original signal is then reconstructed.

[0036] The IMF is decomposed into a pair of positive and negative semi-definite functions, denoted as IMF. + and IMF -

[0037]

[0038] Each original IMF can be reconstructed after decomposition:

[0039] IMF (k) (t)=IMF +(k) (t)+IMF -(k) (t)

[0040] The IMFs were upsampled by a factor of 2 using a standard low-channel interpolation filter, resulting in two sets of higher-order IMFs, as follows:

[0041] IMF +(k) ←EMD(↑(IMF +(k-1) )),

[0042] IMF -(k) ←EMD(↑(IMF -(k-1) ))

[0043] Where ↑(*) represents the 2x upsampling process, and EMD(*) represents the EMD process.

[0044] Thus, the original low-order IMF can be easily reconstructed from the high-order IMF, as shown below:

[0045]

[0046] Where ↓(*) represents a 2x downsampling process;

[0047] The signal after artifact removal is reconstructed to obtain the denoised EMG signal obtained by removing low-order IMFs from the original EMG signal.

[0048]

[0049] Preferably, in step S2, empirical mode decomposition is used to derive the first-order intrinsic mode function from the EMG signal, and the original signal is reconstructed. The specific processing method is as follows:

[0050] (1) Identify the local maximum and minimum values ​​of the EMG signal S(t), where t is time.

[0051] (2) Perform cubic spline interpolation between the maximum and minimum values ​​to obtain the envelope curve, where the upper envelope curve is E. max (t), the lower envelope curve is E min (t);

[0052] (3) Calculate the average value M(t) of the upper and lower envelope curves:

[0053]

[0054] (4) Calculate the difference between S(t) and M(t):

[0055] C1(t) = S(t) - M(t)

[0056] (5) If the number of local extrema of C1(t) is equal to or differs from the number of intersections with 0 by 1, and the average value of C1(t) is close to 0, then the component IMF1 of the first stage is C1(t). If the above two conditions are not met, C1(t) will continue to repeat steps (1) to (4) until the conditions of IMF in step (5) are met.

[0057] (6) Calculate the residual amount:

[0058] R1(t) = S(t) - C1(t)

[0059] (7) If R1(t) is not a monotonic function or a constant sequence, repeat the above steps for R1(t) to obtain the next IMF and the new residual.

[0060] (8) Obtain n orthogonal IMFs, and reconstruct the original signal as shown in the following equation:

[0061]

[0062] Preferably, the method for constructing the CMFN model in step S4 is as follows:

[0063] S4-1 Determining the Location and Number of Network Nodes in CMFN

[0064] EEG data from 19 electrodes were selected to represent the concentrated manifestation of physiological activity in the cerebral cortex, and EMG signals from a muscle group related to movement were selected to represent the physiological activity of muscle contraction. Therefore, the CMFN has a total of 20 nodes.

[0065] S4-2. Quantify the coupling between CMFN nodes.

[0066] In CMFN, there is information exchange between any two nodes, representing the information transmission between the brain and neuromuscular system. VMD-TE, which is an improvement on VMD, is used to calculate the temporal coupling value of each node in CMFN. Since the information transmission between the cerebral cortex and neuromuscular system is bidirectional, the coupling value calculated by VMD-TE is directional, with uplink EMG→EEG and downlink EEG→EMG. It also includes the information transmission between EEG nodes, EEG→EEG, resulting in an N×N connection matrix, N=20, with 19 EEG nodes and 1 EMG node.

[0067] S4-3, CMFN Threshold Selection

[0068] The top 15% of the connection strength values ​​in each connection matrix are selected as the threshold for binarization of CMFN.

[0069] As a preferred embodiment, the specific analysis method for step S5 is as follows:

[0070] S5-1, Calculation of brain-muscle functional network characteristics;

[0071] S5-2. Compare the changes in the characteristic values ​​of the brain muscle function network at different stages of rehabilitation for stroke patients, and develop an assessment method.

[0072] Preferably, in step S5-1, the brain muscle functional network feature calculation is specifically as follows:

[0073] (1) Clustering coefficient

[0074] The clustering coefficient C(i) of CMFN nodes is defined as follows:

[0075]

[0076] For the entire CMFN, the clustering coefficient (C) is the mean of each node C(i), as shown below:

[0077]

[0078] C can generally be considered an indicator of the information processing efficiency of the local functional area of ​​CMFN;

[0079] (2) Global efficiency

[0080] The global efficiency is the average of the reciprocals of L across all nodes in CMFN:

[0081]

[0082] Where N is the number of nodes and L is the average shortest path length. When calculating L in CMFN, if node i and node j are not connected, that is, there is no path in CMFN that connects i and j, then l... ij The value of is either nonexistent or infinite, therefore E global This problem was solved by calculating the average of the reciprocals;

[0083] (3) Small World Attributes

[0084] In small-world networks, L and C lie between those of regular and random networks.

[0085] The small-world coefficient σ is an indicator that represents whether a network has the small-world property. σ is defined as:

[0086]

[0087] Where L is the average shortest path length and C is the clustering coefficient.

[0088] Preferably, in step S5-2, the evaluation method is formulated as follows:

[0089] (1) Analyze the EEG and EMG data of stroke patients in the 1st, 5th and 8th weeks after hospitalization. The EEG and EMG data recorded in the 1st week are defined as the early stage of rehabilitation, the EEG and EMG data recorded in the 5th week are defined as the middle stage of rehabilitation, and the EEG and EMG data recorded in the 8th week are defined as the late stage of rehabilitation. The preprocessed EEG signals are divided into 3 frequency bands: theta wave: 4-8Hz, alpha wave: 8-13Hz, and beta wave: 13-30Hz.

[0090] (2) Calculate the network characteristics of CMFN, compare the mean and standard deviation of clustering coefficient, global efficiency and small-world properties of stroke patients in three rehabilitation periods, and formulate a method for evaluating stroke rehabilitation based on brain muscle function network characteristics.

[0091] This invention has the following characteristics and beneficial effects:

[0092] Using the above technical solution, VMD is introduced to perform time-frequency scaling of EMG on the basis of TE, and a VMD-TE model is constructed for the calculation of CMFC. Experiments were conducted using synchronous EEG and EMG data from two healthy subjects, determining the optimal number of decomposition layers to be 5, and verifying that VMD-TE significantly improves CMFC calculation compared to TE. Most scholars primarily use brain functional networks (CMFNs) to study stroke. This invention studies the human brain and trunk muscles as a comprehensive network of functional interactions. By calculating the network characteristics of the CMFN, the mean and standard deviation of clustering coefficients, global efficiency, and small-world properties of stroke patients in three rehabilitation periods are compared. Based on this, a method for assessing stroke rehabilitation based on CMFN characteristics is developed. Since the network characteristics of the CMFN are calculable, the rehabilitation status of stroke patients can be quantitatively assessed based on these characteristics, helping doctors to develop rehabilitation plans according to the patient's condition. Attached Figure Description

[0093] 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.

[0094] Figure 1 This is a flowchart of an embodiment of the present invention.

[0095] Figure 2 The locations of the EEG and EMG electrodes are shown in this embodiment of the invention.

[0096] Figure 3 This describes the EEG noise reduction process in an embodiment of the present invention.

[0097] Figure 4 The waveforms of the EEG signal before and after noise reduction are shown in this embodiment of the invention.

[0098] Figure 5 This describes the EMG noise reduction process in an embodiment of the present invention.

[0099] Figure 6 The waveforms of the EMG signal before and after noise reduction are shown in an embodiment of the present invention.

[0100] Figure 7 (a) in the figure is a graph showing the relationship between the EEG→EMG coupling value and K in an embodiment of the present invention.

[0101] Figure 7 (b) in the figure is a graph showing the relationship between the EMG→EEG coupling value and K in an embodiment of the present invention.

[0102] Figure 8 In the table, (a) represents the mean and standard deviation of the clustering coefficients for stroke patients during the three rehabilitation periods.

[0103] Figure 8 (b) in the figure represents the mean and standard deviation of the global efficiency of stroke patients in three rehabilitation periods.

[0104] Figure 8 In the equation (c), the mean and standard deviation of the small-world properties of the CMFN at different times and frequency bands in stroke patients are given. Detailed Implementation

[0105] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0106] 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.

[0107] 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.

[0108] This invention provides a stroke rehabilitation assessment method based on brain-muscle functional network characteristics, such as... Figure 1 As shown, it includes the following steps:

[0109] S1. Collect EEG and EMG signals. Collect EEG and EMG signals of stroke patients under grip strength test.

[0110] S2. Preprocess the EEG and EMG signals separately:

[0111] S3. Construct the variational mode decomposition transfer entropy;

[0112] S4. Construct a CMFN model based on EEG and EMG;

[0113] S5. Brain-muscle functional network analysis.

[0114] Specifically, in this embodiment, the methods for acquiring EEG and EMG are as follows:

[0115] Phase 1: Pre-experiment preparation. Stroke patients sit quietly in front of the screen, adjust their mood, and stabilize their resting EEG signals.

[0116] Phase 2: During the experiment, stroke patients gripped the hand gripper all the way down within 3 seconds according to the on-screen instructions, then rested for 1 minute, and repeated 20 times.

[0117] Phase 3: The experiment was conducted again. The stroke patients rested for 20 minutes after the previous experiment, and then the Phase 2 experiment was repeated. The experiment was repeated a total of 3 times, and a total of 60 sets of data were collected per person per cycle.

[0118] Furthermore, considering that patients may have undergone relevant rehabilitation training on the same day, which could indirectly affect the experimental data, the experiment was scheduled to begin between 8:00 and 9:30. The laboratory was required to maintain a temperature of 24°C to provide a comfortable experimental environment for the subjects.

[0119] Under the above conditions, electroencephalography (EEG) and electromyography (EMG) signals were acquired from the patient according to the experimental paradigm. The sampling frequency for EEG signal acquisition was 1000 Hz. Fifty-nine scalp electrodes were used, conforming to the international 10-20 lead standard. Electrode positions were as follows: Figure 2 As shown, 19 channels were selected from 59 channels (electrode channels: AF3, AF4, F3, FZ, F4, FC3, FCZ, FC4, C3, CZ, C4, CP3, CP4, P3, PZ, P4, PO3, POZ, PO4). This electrode selection facilitates a comprehensive assessment of the motor cortex and reduces mutual interference between electrodes. During EEG signal acquisition, to minimize noise interference from other components, subjects were required to clean and dry their scalps before the experiment. Secondly, conductive ointment was injected to reduce the electrode impedance to below 20kΩ. Furthermore, two electrode pads were used to acquire electrooculogram (EOG) signals. The EMG signal sampling frequency was 1926Hz. The first dorsal interosseous (FDI) muscle was selected as the location for EMG signal acquisition, as shown below. Figure 2 As shown.

[0120] A further provision of the present invention includes a method for preprocessing the acquired electroencephalogram (EEG) signals, such as... Figure 3 As shown:

[0121] Removal of power frequency interference from EEG data using a 50Hz notch filter;

[0122] Spatial Constraint ICA (SCFastICA) is used to reduce energy loss in the artifact-free portion of ICs. The main idea is to define a spatial constraint (SC) on A to represent specific prior knowledge or assumptions related to the spatial terrain projected by certain source sensors. A after SC is shown below:

[0123]

[0124] Where A is an unknown and invertible m×n matrix, representing the transfer function between the signal source and the received signal. It is a constrained column, A u It is an unconstrained column;

[0125] The extracted Spatial Constraint-Independent Components (SC-ICs) are denoised using Discrete Wavelet Transform (DWT). DWT can be defined by the following formula:

[0126]

[0127] Where f(x) can be used to describe SC-ICs, k is the time shift, j is the decomposition scale, and φ j,k (x) is the mother wavelet, d j,k The wavelet coefficients are shown below, and the inverse wavelet transform is as follows:

[0128] φ j,k (x)=2 -j / 2 φ(2 -j / 2 xk),j,k∈Z.

[0129] The changes in the preprocessed EEG signal obtained after the above preprocessing are as follows: Figure 4 As shown.

[0130] Further, methods for preprocessing the acquired electromyographic signals, such as... Figure 5 As shown:

[0131] Empirical Mode Decomposition (EMD) is used to derive first-order intrinsic mode functions (IMFs) from the EMG signal, and the original signal is then reconstructed. The specific processing method is as follows:

[0132] (1) Identify the local maximum and minimum values ​​of the EMG signal S(t), where t is time.

[0133] (2) Perform cubic spline interpolation between the maximum and minimum values ​​to obtain the envelope curve, where the upper envelope curve is E. max (t), the lower envelope curve is E min (t);

[0134] (3) Calculate the average value M(t) of the upper and lower envelope curves:

[0135]

[0136] (4) Calculate the difference between S(t) and M(t):

[0137] C1(t) = S(t) - M(t)

[0138] (5) If the number of local extrema of C1(t) is equal to or differs from the number of intersections with 0 by 1, and the average value of C1(t) is close to 0, then the component IMF1 of the first stage is C1(t). If the above two conditions are not met, C1(t) will continue to repeat steps (1) to (4) until the conditions of IMF in step (5) are met.

[0139] (6) Calculate the residual amount:

[0140] R1(t) = S(t) - C1(t)

[0141] (8) If R1(t) is not a monotonic function or a constant sequence, repeat the above steps for R1(t) to obtain the next IMF and the new residual.

[0142] (8) Obtain n orthogonal IMFs, and reconstruct the original signal as shown in the following equation:

[0143]

[0144] The IMF is decomposed into a pair of positive and negative semi-definite functions, denoted as IMF. + and IMF -

[0145]

[0146] Each original IMF can be reconstructed after decomposition:

[0147] IMF (k) (t)=IMF +(k) (t)+IMF -(k) (t)

[0148] The IMFs were upsampled by a factor of 2 using a standard low-channel interpolation filter, resulting in two sets of higher-order IMFs, as follows:

[0149] IMF +(k) ←EMD(↑(IMF +(k-1) )),

[0150] IMF -(k) ←EMD(↑(IMF -(k-1) ))

[0151] Where ↑(*) represents the 2x upsampling process, and EMD(*) represents the EMD process.

[0152] Thus, the original low-order IMF can be easily reconstructed from the high-order IMF, as shown below:

[0153]

[0154] Where ↓(*) represents a 2x downsampling process;

[0155] The signal after artifact removal is reconstructed to obtain the denoised EMG signal obtained by removing low-order IMFs from the original EMG signal.

[0156]

[0157] The changes in the preprocessed EMG signal obtained after the above preprocessing are as follows: Figure 6 As shown.

[0158] A further provision of the present invention provides a method for constructing the transfer entropy of variational mode decomposition as follows:

[0159] S3-1. The time series of the relevant frequency bands of the EEG signal obtained by bandpass filtering is denoted as y(t):

[0160]

[0161] S3-2. Information at different scales of electromyography signals is extracted using VMD to obtain the time series at relevant time scales as x. k (t):

[0162] {x k (t)}(k=1,2,…,K)

[0163] Where K is the number of VMD decomposition layers;

[0164] S3-3, Calculate y(t) and x k The average TE between (t) is used as the functional coupling value between the cerebral cortex and trunk muscles. The specific calculation process of VMD-TE from EEG signal to EMG signal is as follows:

[0165]

[0166] Where n is the discrete time series; p(*) is the joint probability between variables. The larger the VMD-TE value, the stronger the CMFC between EEG and EMG in this frequency band, and vice versa.

[0167] Using the above method, in this embodiment, firstly, the time series of the EEG-related frequency band is obtained by bandpass filtering, denoted as y(t). Secondly, information at different scales of EMG is extracted by VMD to obtain the time series of the relevant time scale, denoted as x. k (t), where K is the number of VMD decomposition layers. Finally, calculate y(t) and x. k The average TE value between (t) is used as the functional coupling value between the cerebral cortex and trunk muscles. To better reflect the influence of the selection of K value on the VMD-TE value, this embodiment records synchronous EEG and EMG data (a total of 20 sets) of two healthy subjects (two male subjects, average age 25) performing the same experimental action. Then, the recorded EEG and EMG are denoised in step S2, and then the EMG is bandpass filtered to 20-150Hz. The mean CMFC values ​​of EEG→EMG and EMG→EEG directions are calculated in three EEG frequency bands for EMG without VMD decomposition (K=0), K=2, K=3, K=4, K=5, K=6, K=7 and K=8. Figure 5 As can be seen from .1 and 5.2, the functional coupling values ​​in the EEG→EMG and EMG→EEG directions calculated after EMG decomposition by VMD are significantly improved compared to when K=0, such as Figure 7 (a) and Figure 7 In (b), the specific similarity of CMFC in the EEG→EMG and EMG→EEG directions can be seen. These results indicate that the VMD-TE algorithm proposed in this invention can be used to calculate the CMFC between the cerebral cortex and trunk muscles. Regardless of whether it is the EEG→EMG or EMG→EEG direction, it is very clear that the CMFC increases with the increase of K, but it basically reaches an "extreme value" when K=5. Therefore, K is set to 5.

[0168] Furthermore, in step S4, the method for constructing the CMFN model is as follows:

[0169] S4-1 Determining the Location and Number of Network Nodes in CMFN

[0170] The EEG acquisition device conforms to the international 10 / 20 lead standard. It selects 19 electrodes to collect EEG data to represent the concentrated manifestation of physiological activities in the cerebral cortex. It also selects a muscle group related to movement to collect EMG signals as a manifestation of physiological activities of muscle contraction. Therefore, the CMFN has a total of 20 nodes.

[0171] S4-2. Quantify the coupling between CMFN nodes.

[0172] In CMFN, there is information exchange between any two nodes, representing the information transmission between the brain and neuromuscular system. VMD-TE, which is an improvement on VMD, is used to calculate the temporal coupling value of each node in CMFN. Since the information transmission between the cerebral cortex and neuromuscular system is bidirectional, the coupling value calculated by VMD-TE is directional, with uplink EMG→EEG and downlink EEG→EMG. It also includes the information transmission between EEG nodes, EEG→EEG, resulting in an N×N connection matrix, N=20, with 19 EEG nodes and 1 EMG node.

[0173] S4-3, CMFN Threshold Selection

[0174] In this invention, the top 15% of the connection strength values ​​in each connection matrix are selected as the threshold for binarizing the CMFN, a preference shared by many researchers. This threshold selection ensures that the CMFN is neither a very sparse nor a very dense graph.

[0175] Specifically, the specific analysis method for step S5 is as follows:

[0176] S5-1, Calculation of Brain-Muscle Functional Network Features

[0177] (1) Clustering coefficient

[0178] The clustering coefficient C(i) of CMFN nodes is defined as follows:

[0179]

[0180] For the entire CMFN, the clustering coefficient (C) is the mean of each node C(i), as shown below:

[0181]

[0182] C can generally be considered an indicator of the information processing efficiency of the local functional area of ​​CMFN;

[0183] (2) Global efficiency

[0184] The global efficiency is the average of the reciprocals of L across all nodes in CMFN:

[0185]

[0186] Where N is the number of nodes and L is the average shortest path length. When calculating L in CMFN, if node i and node j are not connected, that is, there is no path in CMFN that connects i and j, then l... ij The value of is either nonexistent or infinite, therefore E global This problem was solved by calculating the average of the reciprocals;

[0187] (3) Small World Attributes

[0188] In small-world networks, L and C lie between those of regular and random networks.

[0189] The small-world coefficient σ is an indicator that represents whether a network has the small-world property. σ is defined as:

[0190]

[0191] Where L is the average shortest path length and C is the clustering coefficient.

[0192] S5-2. Compare the changes in the characteristic values ​​of the brain muscle function network at different stages of rehabilitation for stroke patients, and develop an assessment method.

[0193] (1) Analyze the EEG and EMG data of stroke patients in the 1st, 5th and 8th weeks after hospitalization. The EEG and EMG data recorded in the 1st week are defined as the early stage of rehabilitation, the EEG and EMG data recorded in the 5th week are defined as the middle stage of rehabilitation, and the EEG and EMG data recorded in the 8th week are defined as the late stage of rehabilitation. The preprocessed EEG signals are divided into 3 frequency bands: theta wave: 4-8Hz, alpha wave: 8-13Hz, and beta wave: 13-30Hz.

[0194] (2) Calculate the network characteristics of CMFN, compare the mean and standard deviation of clustering coefficient, global efficiency and small-world properties of stroke patients in three rehabilitation periods, and formulate a method for evaluating stroke rehabilitation based on brain muscle function network characteristics.

[0195] Understandably, the above analytical method utilizes Matlab to calculate its network characteristics. It obtains the clustering coefficients, global efficiency, and the mean and standard deviation of small-world properties for stroke patients across three rehabilitation phases. For example... Figure 8 (a) Figure 8 (b) and Figure 8 As shown in (c) in the figure. The results indicate that the beta wave of EEG is the main carrier of information exchange between the brain and trunk muscles when stroke patients perform experimental movements. The clustering coefficient, global efficiency, and small-world property of IP->MP->LP on the beta wave all increase with the progress of rehabilitation time, indicating that these features can be used to assess the rehabilitation status of stroke patients.

[0196] 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 stroke rehabilitation assessment method based on brain-muscle functional network characteristics, characterized in that, Includes the following steps: S1. Collect EEG and EMG signals. Collect EEG and EMG signals of stroke patients under grip strength test. S2. Preprocess the EEG and EMG signals respectively; S3. Construct the variational mode decomposition transfer entropy to obtain the functional coupling value between EEG and EMG signals; S3-1. The time series of the relevant frequency bands of the EEG signal obtained by bandpass filtering is denoted as y(t): S3-2. Information at different scales of electromyography signals is extracted using VMD to obtain the time series at relevant time scales as x. k (t): {x k (t)}(k=1,2,…,K) Where K is the number of VMD decomposition layers; S3-3, Calculate y(t) and x k The average TE between (t) is used as the functional coupling value between the cerebral cortex and trunk muscles. The specific calculation process of VMD-TE from EEG signal to EMG signal is as follows: Where n is the discrete time series; p(*) is the joint probability between variables. The larger the VMD-TE value, the stronger the CMFC between EEG and EMG in this frequency band, and vice versa. S4. Construct a CMFN model based on the functional coupling value between EEG and EMG signals; S5. Brain-muscle functional network analysis.

2. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 1, characterized in that, In S1, the methods for acquiring EEG and EMG are as follows: Phase 1: Pre-experiment preparation. Stroke patients sit quietly in front of the screen, adjust their mood, and stabilize their resting EEG signals. Phase 2: During the experiment, stroke patients gripped the hand gripper all the way down within 3 seconds according to the on-screen instructions, then rested for 1 minute, and repeated 20 times. Phase 3: The experiment was conducted again. The stroke patients rested for 20 minutes after the previous experiment, and then the Phase 2 experiment was repeated. The experiment was repeated a total of 3 times, and a total of 60 sets of data were collected per person per cycle.

3. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 1, characterized in that, In step S2, the method for preprocessing the acquired electroencephalogram (EEG) signals is as follows: Removal of power frequency interference from EEG data using a 50Hz notch filter; Spatial constraint ICA is used to reduce the energy loss of the artifact-free portion of ICs. That is, a spatial constraint is defined on A to represent prior knowledge or prior assumptions related to the spatial terrain projected by some source sensors. A after SC is shown below: Where A is an unknown and invertible m×n matrix, representing the transfer function between the signal source and the received signal. It is a constrained column, A u It is an unconstrained column; The discrete wavelet transform (DWT) is used to denoise the extracted spatially constrained independent components. The DWT can be defined by the following formula: Where f(x) can be used to describe SC-ICs, k is the time shift, j is the decomposition scale, and φ j,k (x) is the mother wavelet, d j,k The wavelet coefficients are shown below, and the inverse wavelet transform is as follows:

4. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 1, characterized in that, In step S2, the method for preprocessing the acquired electromyographic signals is as follows: Empirical mode decomposition is used to derive the first-order intrinsic mode function from the EMG signal, and the original signal is then reconstructed. The IMF is decomposed into a pair of positive and negative semi-definite functions, denoted as IMF. + and IMF - Each original IMF can be reconstructed after decomposition: IMF (k) (t)=IMF +(k) (t)+IMF -(k) (t) The IMFs were upsampled by a factor of 2 using a standard low-channel interpolation filter, resulting in two sets of higher-order IMFs, as follows: IMF +(k) ←EMD(↑(IMF +(k-1) )), IMF -(k) ←EMD(↑(IMF -(k-1) )) Where ↑(*) represents the 2x upsampling process, and EMD(*) represents the EMD process. Thus, the original low-order IMF can be easily reconstructed from the high-order IMF, as shown below: Where ↓(*) represents a 2x downsampling process; The signal after artifact removal is reconstructed to obtain the denoised EMG signal obtained by removing low-order IMFs from the original EMG signal.

5. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 4, characterized in that, In step S2, empirical mode decomposition is used to derive the first-order intrinsic mode function from the EMG signal, and the original signal is reconstructed. The specific processing method is as follows: (1) Identify the local maximum and minimum values ​​of the EMG signal S(t), where t is time; (2) Perform cubic spline interpolation between the maximum and minimum values ​​to obtain the envelope curve, where the upper envelope curve is E. max (t), the lower envelope curve is E min (t); (3) Calculate the average value M(t) of the upper and lower envelope curves: (4) Calculate the difference between S(t) and M(t): C1(t) = S(t) - M(t) (5) If the number of local extrema of C1(t) is equal to or differs from the number of intersections with 0 by 1, and the average value of C1(t) is close to 0, then the component IMF1 of the first stage is C1(t). If this condition is not met, C1(t) will continue to repeat steps (1) to (4) until the condition of IMF in step (5) is met. (6) Calculate the residual amount: R1(t) = S(t) - C1(t) (7) If R1(t) is not a monotonic function or a constant sequence, repeat the above steps for R1(t) to obtain the next IMF and the new residual. (8) Obtain n orthogonal IMFs, and reconstruct the original signal as shown in the following equation:

6. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 5, characterized in that, The method for constructing the CMFN model in S4 is as follows: S4-1 Determining the Location and Number of Network Nodes in CMFN EEG data from 19 electrodes were selected to represent the concentrated manifestation of physiological activity in the cerebral cortex, and EMG signals from a muscle group related to movement were selected to represent the physiological activity of muscle contraction. Therefore, the CMFN has a total of 20 nodes. S4-2. Quantify the coupling between CMFN nodes. In CMFN, there is information exchange between any two nodes, representing the information transmission between the brain and neuromuscular system. VMD-TE, which is an improvement on VMD, is used to calculate the temporal coupling value of each node in CMFN. Since the information transmission between the cerebral cortex and neuromuscular system is bidirectional, the coupling value calculated by VMD-TE is directional, with uplink EMG→EEG and downlink EEG→EMG. It also includes the information transmission between EEG nodes, EEG→EEG, resulting in an N×N connection matrix, N=20, with 19 EEG nodes and 1 EMG node. S4-3, CMFN Threshold Selection The top 15% of the connection strength values ​​in each connection matrix are selected as the threshold for binarization of CMFN.

7. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 6, characterized in that, The specific analysis method of S5 is as follows: S5-1, Calculation of brain-muscle functional network characteristics; S5-2. Compare the changes in the characteristic values ​​of the brain muscle function network at different stages of rehabilitation for stroke patients, and develop an assessment method.

8. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 7, characterized in that, In S5-1, the specific calculation of the brain muscle functional network features is as follows: (1) Clustering coefficient The clustering coefficient C(i) of CMFN nodes is defined as follows: For the entire CMFN, the clustering coefficient (C) is the mean of each node C(i), as shown below: C is an indicator of the information processing efficiency of the local functional area of ​​CMFN, and N is the number of nodes; (2) Global efficiency The global efficiency is the average of the reciprocals of L across all nodes in CMFN: Where N is the number of nodes and L is the average shortest path length. When calculating L in CMFN, if node i and node j are not connected, that is, there is no path in CMFN that connects i and j, then l... ij The value of is either nonexistent or infinite, therefore E global This problem was solved by calculating the average of the reciprocals; (3) Small World Attributes In small-world networks, L and C lie between those of regular and random networks. The small-world coefficient σ is an indicator that represents whether a network has the small-world property. σ is defined as: Where L is the average shortest path length and C is the clustering coefficient.

9. The stroke rehabilitation assessment method based on brain-muscle functional network characteristics according to claim 8, characterized in that, In S5-2, the evaluation method is formulated as follows: (1) Analyze the EEG and EMG data of stroke patients in the 1st, 5th and 8th weeks after hospitalization. The EEG and EMG experimental data recorded in the 1st week are defined as the early stage of rehabilitation, the EEG and EMG experimental data recorded in the 5th week are defined as the middle stage of rehabilitation, and the EEG and EMG experimental data recorded in the 8th week are defined as the late stage of rehabilitation. The pre-processed EEG signals are divided into 3 frequency bands, namely theta wave, alpha wave and beta wave. (2) Calculate the network characteristics of CMFN, compare the mean and standard deviation of clustering coefficient, global efficiency and small-world properties of stroke patients in three rehabilitation periods, and formulate a method for evaluating stroke rehabilitation based on brain muscle function network characteristics.

Images