Lower limb rehabilitation evaluation method, system and equipment based on brain-computer interface and medium

Abstract

The invention provides a lower limb rehabilitation evaluation method, system and device based on a brain-computer interface and a medium, and relates to the technical field of rehabilitation medicine and medical electronics. According to a lower limb rehabilitation evaluation technology based on a brain-computer interface, a two-way evaluation system of peripheral muscle form-central nervous activity is constructed through synchronous acquisition of wearable A-type ultrasound and electroencephalogram signals; a-type ultrasound monitors the change of the thickness of the rectus femoris in real time, and electroencephalogram signals analyze the coherence of Alpha / Beta frequency bands, so that functional difference evaluation of a stroke patient in a resting state, a passive motion state and an active motion state with different resistances is realized, the limitation of single subjective scoring of a traditional scale is avoided, and a multi-dimensional quantitative index is provided for rehabilitation evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rehabilitation medicine and medical electronics, in particular to a lower limb rehabilitation evaluation method, system, device and medium based on brain-computer interface. BACKGROUND

[0002] Stroke is one of the leading causes of death and disability in adults worldwide. About 80% of stroke survivors are accompanied by lower limb motor dysfunction, which seriously affects the patient's independent walking ability and quality of life. Therefore, accurate and dynamic rehabilitation evaluation of lower limb motor function is a key link to judge the progress of the disease, develop treatment plans and evaluate the effect of rehabilitation, which is directly related to the improvement of prognosis and functional recovery of patients.

[0003] Currently, clinical evaluation mainly relies on subjective evaluation methods such as Fugl-Meyer Motor Function Scale (FMA), which has the limitations of insufficient quantitative accuracy and difficulty in real-time dynamic monitoring, and cannot meet the needs of precise rehabilitation. In terms of technology, traditional surface electromyography (sEMG) evaluation is limited by the detection range of superficial muscles and is easily affected by electromagnetic interference; imaging techniques such as dual-energy X-ray absorptiometry (DEXA) are limited by the heavy equipment and cannot achieve dynamic measurement, so the application scenarios are limited. Although the coherence analysis of Alpha (8-13 Hz) and Beta (14-30 Hz) bands in quantitative electroencephalogram (QEEG) has been proven to be related to motor cortex function, related research has focused on the upper limbs, and the evaluation of lower limb motor function has not been further studied; at the same time, although A-mode ultrasound can effectively measure muscle thickness, its multi-modal fusion application with electroencephalogram signals is still a blank.

[0004] In summary, the existing technology cannot fully depict the functional status of the "central-peripheral" motor control pathway after stroke, and cannot provide comprehensive evaluation basis for lower limb rehabilitation considering central nervous electrical activity and peripheral muscle morphological changes. Therefore, it is urgent to propose a lower limb rehabilitation evaluation technology that integrates multi-modal information to meet the urgent needs of precise and personalized development of lower limb rehabilitation after stroke. SUMMARY

[0005] (I) Technical problems solved

[0006] In view of the deficiencies of the prior art, the present application provides a lower limb rehabilitation evaluation method, system, device and medium based on brain-computer interface, which at least solves the problem of poor accuracy of existing lower limb rehabilitation evaluation.

[0007] (II) Technical solutions

[0008] To achieve the above purpose, the present application is realized by the following technical solutions:

[0009] In a first aspect, the application first provides a lower limb rehabilitation evaluation method based on a brain-computer interface, the method comprising:

[0010] synchronously acquiring muscle thickness data and electroencephalogram data of the patient in multiple states based on the brain-computer interface device;

[0011] extracting multi-modal features based on the muscle thickness data and the electroencephalogram data;

[0012] performing collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank-sum test model, and a neural network model to generate a quantitative evaluation index.

[0013] In one embodiment, the brain-computer interface device comprises a motion acquisition unit for acquiring muscle thickness data of the patient; a BCI acquisition unit for acquiring electroencephalogram data of the patient; and a motion load unit for providing passive, active, non-resistance, and progressive resistance motion modes for the patient.

[0014] Preferably, the motion acquisition unit comprises a wearable A-mode ultrasonic sensor; the BCI acquisition unit comprises a multi-channel electroencephalogram cap; and the motion load unit comprises a limb motion rehabilitation training device.

[0015] In one embodiment, the multi-modal features include fuzzy entropy of the electroencephalogram, a coherence coefficient, a mutual information coefficient of the electroencephalogram and the muscle thickness signal, an electroencephalogram acquisition quality coefficient, and muscle thickness data.

[0016] In one embodiment, extracting multi-modal features based on the muscle thickness data and the electroencephalogram data comprises:

[0017] preprocessing the ultrasonic signal data representing the muscle thickness by median filtering denoising and calculating the muscle thickness based on echo time difference;

[0018] preprocessing the electroencephalogram data by first analyzing the dynamic amplitude of the head motion using a navigation and attitude solving algorithm to remove low-quality electroencephalogram data collected under large dynamic abnormal shaking, then filtering the retained electroencephalogram data by notch filtering and high-order band-pass filtering, and finally removing non-brain-derived components in the filtered electroencephalogram data using independent component analysis method.

[0019] In a preferred embodiment, when the coherence coefficient is acquired, the similarity of two signals is measured by calculating the power spectral density, and the formula is:

[0020]

[0021] wherein PC Z F Z (f) is the cross-power spectrum, representing the correlation strength and phase difference between the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PCZ (f) PF Z (f) is the power spectrum, used to quantify the above three frequency bands C Z and F Z the energy intensity of each channel; C CF (f) is the coherence, indicating the above three frequency bands in C Z and F Z region after normalizing the cross-power spectrum.

[0022] In one embodiment, the three factors in the three-factor variance analysis model include group, frequency, and action state; the rank sum test model is a Wilcoxon rank sum test algorithm model.

[0023] In one embodiment, the quantitative evaluation indicators include: resting muscle thickness difference, active exercise thickness growth rate, Alpha band coherence difference, high Beta band coherence ratio, and Pearson correlation coefficient of muscle thickness and Beta coherence, coherence coefficient of electroencephalogram signal and muscle thickness signal in Alpha band and Beta band, and coherence difference of patient group and healthy group in different exercise states.

[0024] In one embodiment, the muscle thickness is the rectus femoris muscle thickness.

[0025] In one embodiment, the multiple states include: resting state 1, passive pedaling, active unresisted pedaling, resistance 2Nm active pedaling, resistance 4Nm active pedaling, and resting state 2.

[0026] In a preferred embodiment, the multi-channel electroencephalogram cap includes ND-310A.

[0027] In a preferred embodiment, the wearable A-mode ultrasonic sensor includes a multi-channel A-mode driving system.

[0028] In a second aspect, the application further provides a lower limb rehabilitation evaluation system based on a brain-computer interface, the system comprising:

[0029] A data acquisition module configured to synchronously acquire muscle thickness data and electroencephalogram signal data of a patient in multiple states based on a brain-computer interface device;

[0030] A multi-modal feature extraction module configured to extract multi-modal features based on the muscle thickness data and electroencephalogram signal data.

[0031] A quantitative evaluation indicator generation module configured to perform collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate quantitative evaluation indicators.

[0032] In a third aspect, the present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for lower limb rehabilitation assessment based on brain-computer interface when executing the program.

[0033] In a fourth aspect, the present application finally provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method for lower limb rehabilitation assessment based on brain-computer interface.

[0034] (III) Beneficial Effects

[0035] The present application provides a method, system, device and medium for lower limb rehabilitation assessment based on brain-computer interface. Compared with the prior art, the present application has the following beneficial effects:

[0036] The method for lower limb rehabilitation assessment based on brain-computer interface provided by the present application synchronously acquires muscle thickness data and electroencephalogram signal data of a patient in multiple states based on a brain-computer interface device, extracts multi-modal features based on the muscle thickness data and electroencephalogram signal data, performs collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model, and generates a quantitative evaluation index. Through linear coherence and nonlinear mutual information analysis of multi-modal signals, combined with a neural network model, the present application realizes accurate quantification, dynamic monitoring and personalized evaluation of lower limb motor function, and solves the technical problem of strong subjectivity and insufficient quantification of traditional evaluation methods. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the description of the embodiments or the prior art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0038] Figure 1 Flowchart of the method for lower limb rehabilitation assessment based on brain-computer interface of the present application;

[0039] Figure 2 Electrode distribution diagram for electroencephalogram recording in the embodiment of the present application

[0040] Figure 3 Flowchart of the method for lower limb rehabilitation assessment based on brain-computer interface of the present application;

[0041] Figure 4 Flowchart of the method for lower limb rehabilitation assessment based on brain-computer interface of the present application;

[0042] Figure 5Flow chart for EEG analysis in the embodiment of the application;

[0043] Figure 6 Flow chart for obtaining EEG acquisition quality coefficient in the embodiment of the application;

[0044] Figure 7 Flow chart for generating quantitative evaluation index in the embodiment of the application;

[0045] Figure 8 Principle diagram of the lower limb rehabilitation evaluation system based on brain-computer interface in the application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme in the embodiments of the application is described clearly and completely. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the application.

[0047] Stroke is one of the main causes of death and disability in adults worldwide, and about 80% of the survivors have lower limb motor dysfunction.

[0048] The current clinical evaluation methods such as Fugl-Meyer Motor Function Scale (FMA) have strong subjectivity, insufficient quantitative accuracy, and cannot be monitored in real time, which cannot meet the needs of precise rehabilitation. Traditional surface electromyography (sEMG) evaluation technology is limited by the limitations of superficial muscle detection and electromagnetic interference, and imaging methods such as dual-energy X-ray absorptiometry (DEXA) are also insufficient in terms of equipment and dynamic measurement.

[0049] At present, in the field of neuroelectrophysiology, the coherence analysis of Alpha (8-13 Hz) and Beta (14-30 Hz) frequency bands of quantitative electroencephalogram (QEEG) has been proven to be related to motor cortex function, but existing researches mostly focus on upper limbs, and the evaluation research on lower limb motor function is still insufficient. From the perspective of evolution and development, the focus of upper limbs and lower limbs is completely different. Lower limbs evolve more stable skeletal structure (such as the weight-bearing design of hip joint and knee joint) and coordinated movement patterns to adapt to weight-bearing and movement, and the neural control focuses on "energy saving and balance". After being liberated from the function of bearing weight, the upper limbs gradually evolved the ability of fine operation, and the neural control emphasizes "flexibility and precision". In addition, the control of the brain on the upper limbs and lower limbs has significant differences in neural mechanisms, functional division of labor, and movement characteristics, which are closely related to the evolutionary needs of humans and the positioning of limb function. For example, in terms of neural control pathways, functional division of labor and movement characteristics, and neural regulation complexity, there are great differences between the two.

[0050] At the same time, although A-mode ultrasound technology has been proven to be effective in measuring muscle thickness, its application in multimodal fusion with electroencephalogram signals is still a blank.

[0051] It can be seen that there is a lack of a multimodal evaluation technology in the prior art that can integrate central nervous electrical activity and peripheral muscle morphological changes in real time, making it difficult to fully characterize the functional status of the "central-peripheral" motor control pathway after stroke. In view of this, a lower limb rehabilitation evaluation technology based on a combination of A-mode ultrasound and electroencephalogram non-invasive wearable brain-computer interface is proposed, in order to at least solve one or more of the above technical problems.

[0052] The embodiment of the present application provides a lower limb rehabilitation evaluation method and system based on a brain-computer interface, at least solves the problem of poor accuracy of existing lower limb rehabilitation evaluation, and achieves the purpose of accurately evaluating the lower limb rehabilitation of a patient after stroke.

[0053] In order to better understand the above technical solutions, the above technical solutions will be described in detail in combination with the drawings of the specification and specific embodiments.

[0054] Glossary:

[0055] Linear coherence (also known as coherence or spectral coherence) is a statistical measure of the degree of linear correlation between two signals at a specific frequency.

[0056] Nonlinear mutual information (abbreviated as MI) is a statistical measure of the nonlinear dependence between two random variables. It is not limited by whether the relationship between variables is linear or not, and can capture a wider range of associations (including nonlinear, non-monotonic relationships).

[0057] Attitude and heading determination (AHRS) is a core technology in the field of navigation and motion perception, used to calculate the attitude (Attitude) and heading (Heading) of a carrier (such as an airplane, satellite, robot, drone, etc.) in real time through sensor data, i.e. the orientation state of the carrier in three-dimensional space.

[0058] Three-way ANOVA is a statistical method for analyzing the effects of three independent variables (factors) on a continuous dependent variable, suitable for exploring the effects of multiple factors and their interactions on the results.

[0059] Greenhouse-Geisser correction is a statistical correction method used to deal with the violation of sphericity assumption in analysis of variance (ANOVA), mainly applied to repeated measures ANOVA, and the core is to correct the deviation of statistical test caused by data not meeting the sphericity by adjusting the degrees of freedom.

[0060] Fuzzy Entropy (FuzzyEn) is a nonlinear dynamics index for quantifying the complexity or irregularity of time series. It introduces the concept of fuzzy logic based on Approximate Entropy and Sample Entropy, and measures the similarity of patterns in the sequence through fuzzy membership functions, thus more robustly dealing with data noise and short sequence problems.

[0061] Embodiment 1:

[0062] In a first aspect, the present application first proposes a lower limb rehabilitation evaluation method based on brain-computer interface, see Figure 1 The method comprises the following steps:

[0063] S1, synchronously acquiring muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device;

[0064] S2, extracting multi-modal features based on the muscle thickness data and electroencephalogram data;

[0065] S3, for the multi-modal features, performing collaborative analysis based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate a quantitative evaluation index.

[0066] The implementation process of one or more embodiments of the present application will be described in detail below in combination with the accompanying Figures 1-7 and specific explanations of steps S1-S3.

[0067] The lower limb rehabilitation evaluation method based on brain-computer interface proposed in this embodiment comprises:

[0068] S1, synchronously acquiring muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device.

[0069] In this embodiment, the brain-computer interface device used to acquire the muscle thickness data and electroencephalogram data of the patient at least comprises a motion acquisition unit, a BCI acquisition unit, a motion load unit, and an acquisition control system unit. These data acquisition and control units cooperate to synchronously acquire the muscle thickness data and electroencephalogram data of the patient.

[0070] It should be noted that the muscle thickness refers to the cross-sectional area or vertical thickness of the muscle at a specific position, reflecting the morphological structure characteristics of the muscle, and is closely related to the physiological cross-sectional area (CSA) of the muscle, muscle mass, etc. Unlike the existing electromyography signal. Electromyography signal is the weak electrical activity generated by motor units (composed of a motor neuron and the muscle fibers it innervates) in muscle fibers when the muscle contracts or relaxes, which is recorded by electrodes and converted into an electrical signal that can be analyzed.

[0071] In a preferred embodiment, the motion acquisition unit is preferably a wearable A-mode ultrasonic sensor for acquiring thickness data of the lower limb muscles of the patient, specifically the rectus femoris muscle. Specifically, the wearable A-mode ultrasonic sensor uses a multi-channel A-mode driving system developed by the University of Science and Technology of China, with a myRIO master board as the hardware core, equipped with a flexible ultrasonic transducer with a characteristic frequency of 2MHz (silver nanowire diameter 50-100nm, composite PDMS elastic substrate), the transducer contact surface is coated with a medical ultrasonic coupling agent, and is fixed to the midpoint of the quadriceps femoris muscle of the patient (the midpoint of the line connecting the anterior superior iliac spine and the upper edge of the patella) through an elastic bandage, the sampling rate is set to 128Hz, the spatial resolution is 0.08mm, the temperature compensation circuit is integrated to reduce environmental interference, and the pulse emission and signal reception module is integrated, with a sampling frequency greater than or equal to 1kHz.

[0072] It should be noted that when a person is exercising, the change in muscle thickness is extremely rapid, and if the data acquisition speed of the ultrasonic sensor cannot keep up with the speed of the change in muscle thickness, it will not be able to fully acquire complete data, and thus the entire rehabilitation evaluation cannot be completed. The A-mode ultrasonic sensor in this embodiment has a very fast acquisition speed compared to other ultrasonic sensors, and can track the real-time changes in muscle in real time.

[0073] The BCI acquisition unit is preferably a multi-channel electroencephalogram cap for acquiring the electroencephalogram signal of the patient. The electroencephalogram cap, also known as the electroencephalogram (EEG) cap, is a medical device used to record and monitor the electrical signals produced by the brain. In this embodiment, the multi-channel electroencephalogram cap uses the ND-310A developed by the research technology company, with silver chloride comb electrodes, positioned at Cz, C3, C4, Fz, F3, F4, Pz, P3, P4, a total of 9 channels, as shown in Figure 2 The reference electrode is placed at the earlobe position, and the ground electrode is placed at the forehead FPz (FPz corresponds to the frontal pole area of the median line of the forehead), with an impedance control below 5kΩ, and a sampling rate of 256Hz. The multi-channel electroencephalogram cap is equipped with a MEMS attitude sensor, which synchronously acquires heart rate and blood oxygen signals (sampling rate 100Hz).

[0074] The motion load unit is preferably a limb motion rehabilitation training device (in this embodiment, a lower limb treadmill). The lower limb treadmill is used to provide passive, active non-resistance and progressive resistance motion modes for the patient, i.e., corresponding to six states.

[0075] In terms of acquisition control system software, the synchronous acquisition software developed by LabVIEW is used to realize clock synchronization of the A-mode ultrasound (sampling rate 128 Hz) and the electroencephalogram (256 Hz) through a USB interface, the timestamp accuracy is ≤1 ms, real-time data visualization and state marking are supported, and event markers are automatically generated when each state is switched.

[0076] In a preferred embodiment, in order to synchronize and cyclically collect the muscle thickness data and the electroencephalogram signal data of the patient in different states, as shown in Figure 3 six states are set, i.e., a resting state, passive pedaling, active non-resistance pedaling, resistance 2Nm active pedaling, resistance 4Nm active pedaling, and a second resting state, and the related data are cyclically collected by using the brain-computer interface device. Specifically, the collection time length of each state is set to 2 minutes, and the interval between the data collection of each two adjacent states is 10 minutes. During the execution of the motion state, the lower limb treadmill seat is adjusted to control the knee flexion and extension range at 20-40°, the lower limbs of the patient with knee varus or valgus are fixed with an elastic bandage, and the training speed is uniformly set to 20r / min; the resistance gradient of the resistance mode is set to 0Nm, 2Nm and 4Nm.

[0077] It should be noted that the lower limb treadmill designs a multi-state cyclic test scheme, combines the high-frequency synchronous acquisition of the wearable device, and real-time captures the dynamic changes of the muscle thickness and the electroencephalogram coherence caused by the motion intention, thereby providing a continuous monitoring means for the functional evolution in the rehabilitation training.

[0078] S2, extracting multi-modal features based on the muscle thickness data and the electroencephalogram signal data.

[0079] In an embodiment, the ultrasound signal representing the muscle thickness of the patient, the electroencephalogram signal generated by the brain, the head motion signal and the like obtained in S1 are preprocessed, and then multi-modal features for subsequent evaluation of the lower limb rehabilitation of the patient are extracted. For details, please refer to Figure 4 .

[0080] The preprocessing of the ultrasound signal data includes: denoising by median filtering, and calculating the muscle thickness based on the echo time difference. Specifically, when the A-mode ultrasound signal is preprocessed, the original echo signal is subjected to 50-500 kHz band-pass filtering, the threshold method is used to remove noise spikes, the anterior and posterior boundaries of the rectus femoris muscle are extracted by an echo peak detection algorithm, and then the muscle thickness is calculated, and the formula is as follows:

[0081] Thickness = t x v / 2

[0082] wherein Thickness represents the rectus femoris muscle thickness; t is the echo time difference; v is the ultrasound propagation velocity in the muscle.

[0083] The data of each of the six states is subjected to sliding average filtering, and abnormal values exceeding 3 times the standard deviation are removed, and the average value of each 2 min data is calculated.

[0084] The EEG signal is denoised, and the preprocessing process includes: first, the dynamic amplitude of the head motion is analyzed by using the attitude and position calculation algorithm, the real-time impedance of the 9 measurement channels is combined, the EEG acquisition quality coefficient is evaluated, the low-quality EEG data collected under large dynamic abnormal shaking is removed, then the filtered EEG signal is obtained through 50Hz FIR notch filtering and 0.1-40Hz FIR high-order band-pass filtering, finally, independent component analysis is used to remove non-brain components such as electromyogram and eye movement, and the data is divided into six state time periods based on the marking record. Specifically, the EEG signal is preprocessed using MATLAB R2022b: first, 0.1-40Hz Butterworth band-pass filtering is used, and 50Hz notch filtering is used to remove power frequency interference; independent component analysis is performed using the EEGLAB toolbox, eye and muscle artifact components are selected and removed through visual inspection and kurtosis value screening, and the EEG signal is reconstructed.

[0085] It should be noted that since the relevant parameters of the lower limbs of the patients in the motion state are collected, the data collected under the abnormal shaking of the patients is low-quality data, which has a negative impact on the final lower limb rehabilitation evaluation. In order to solve this problem, when the EEG signal is denoised, the dynamic amplitude of the patient's head motion is first analyzed by using the attitude and position calculation algorithm to remove low-quality EEG data.

[0086] Since the data of the six states is collected in a cycle, the signal is divided into six state segments according to the marking time point, each segment is 2 min, and the sampling rate is reduced to 128Hz to synchronize with the ultrasound signal.

[0087] In a more preferred embodiment, when the dynamic amplitude of the head motion is analyzed by using the attitude and position calculation algorithm, the calculated head motion dynamics are transmitted to the display screen on the lower limb treadmill in synchronization, and real-time feedback is provided to the test patient through animation effects, reminding the patient to keep the posture stable as much as possible during the collection process, improving the test cooperation degree of the patient, improving the overall data collection quality of the rehabilitation evaluation system, and further improving the test efficiency.

[0088] In addition, as Figure 5The fuzzy entropy of the electroencephalogram signal is calculated to reflect the complexity of the signal. The fuzzy entropy is calculated based on the filtered electroencephalogram signal, which is one of the input parameters of the subsequent neural network model. The fuzzy entropy reduces the sensitivity of parameter selection through the tolerance parameter and the fuzzy parameter, and provides a smoother entropy estimate. Compared with other electroencephalogram entropy index algorithms, it can better reflect the changes of the electroencephalogram signal.

[0089] In addition, the electroencephalogram acquisition quality coefficient in the preprocessing process is taken as one of the input parameters of the subsequent neural network model. The electroencephalogram acquisition quality coefficient is evaluated based on the head stability and contact impedance in the acquisition process, and reflects the reliability of the electroencephalogram data acquisition. The specific calculation method is as follows:

[0090] Firstly, the posture sensor data and electrode impedance data of the electroencephalogram cap are extracted, and then the tilt evaluation coefficient, the shaking evaluation coefficient and the impedance evaluation coefficient are obtained through time window accumulation and numerical calculation, and finally the three evaluation coefficients are weighted and averaged to obtain the electroencephalogram acquisition quality coefficient. The specific process can be referred to Figure 6 .

[0091] Further, the mutual information coefficient (mutual information value) of the electroencephalogram signal and the rectus femoris thickness signal is calculated to quantify the nonlinear dependence relationship between the electroencephalogram signal and the rectus femoris thickness signal, and analyze the influence of the active movement intention on the central-peripheral signal coupling.

[0092] Mutual information is a core concept in information theory, which is used to quantify the dependence relationship between two random variables. When there is a complex nonlinear relationship between variables (such as the dynamic coupling between brain neural signals and muscle contraction in this embodiment), mutual information is a more suitable tool. In this embodiment, the change of the rectus femoris thickness may not only be directly driven by the EEG signal, but also be generated through a feedback mechanism or a high-order interaction. These relationships cannot be described by a linear model, therefore, mutual information is used to effectively quantify them.

[0093] In the analysis of the electroencephalogram signal and the rectus femoris thickness signal (continuous data, usually moderate sample size), the K nearest neighbor method (Kraskov method) is the most commonly used choice, because it does not require distribution assumption, is sensitive to nonlinear relationships and has stable performance in small samples. If rapid verification is required, the binning method can be used to calculate the initial result, and then the KNN method can be used to refine the result.

[0094] Further, the electroencephalogram frequency band coherence feature coefficient is extracted, including the coherence values of the Cz and Fz channels in the Alpha (8-13 Hz), Beta (14-30 Hz) and Gamma (30-40 Hz) frequency bands.

[0095] The coherence coefficient is a result obtained after a specific processing of the electroencephalogram signal, reflecting the degree of cooperative correlation between two adjacent regions of the brain. In the present embodiment, the electroencephalogram cap has 9 electrodes, and the regulation of lower limb movement involves the cooperative action of the most core regions such as the primary motor cortex (Cz region), the supplementary motor area (Fz region), and the premotor cortex PMC (between Cz and Fz). Therefore, the two points of Cz and Fz are selected, and the coherence coefficients of the alpha, beta, and gamma frequency bands are calculated respectively, and the three results are combined into an array, collectively referred to as "coherence coefficient", as part of the multi-modal feature.

[0096] In a preferred embodiment, the coherence is measured by calculating the power spectral density of the similarity of two signals, and the formula is:

[0097]

[0098] Wherein, PC Z F Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC CF (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC Z (f) is the cross power spectrum, indicating the correlation strength and phase difference of the Cz and Fz channels in the alpha, beta, and gamma frequency bands; PC

[0099] S3, based on the three-factor variance analysis model, the rank sum test model, and the neural network model, the multi-modal features are analyzed cooperatively to generate quantitative evaluation indexes.

[0100] Based on the three-factor variance analysis model, the rank sum test model, and the neural network model, the multi-modal features obtained in step S2 are analyzed cooperatively to generate quantitative evaluation indexes. For details, please refer to Figure 7 .

[0101] In a preferred embodiment, the three factors in the three-factor variance analysis model include group (such as "patient group", "healthy control group", etc.), frequency (such as "low frequency", "medium frequency", "high frequency", etc.), and action state (such as "still", "active movement", "passive movement", etc.), and the Greenhouse-Geisser correction is used, with a significance level of P<0.05. Based on this, the correlation of the multi-modal features is quantitatively evaluated.

[0102] It needs to be explained that through experimental verification, it is proved that the mean difference between the healthy control group and the stroke patients shows different degrees of statistical significance, which shows that the mean difference between the healthy control group and the stroke patients is a real“systematic difference”, rather than an“accidental difference”caused by random factors (such as sampling error, measurement fluctuation, etc.). Specifically:

[0103] The wearable A-type ultrasound was used to measure the rectus femoris thickness of the stroke patients (EX) and the healthy control group (HC) in the optimized sitting position motor rehabilitation evaluation scheme. Independent sample t-test was used for comparison between the two groups. The results show that in the right resting state (Rest1 and Rest2), the mean difference between the healthy control group and the stroke patient group shows different degrees of statistical significance. Specifically, the mean value of the healthy control group in the resting state one is significantly higher than that of the patient group, and there is a significant inter-group difference between the values of the healthy group and the stroke group (P<0.05).

[0104] However, the difference under the condition of resting state two exists, but does not reach a significant level (P>0.05), showing that the difference between the two groups in the resting state is small. In the right passive state (Event1), the measurement value of the healthy control group is significantly higher than that of the EX group, and the difference is also statistically significant (P<0.05), indicating that there is a more obvious difference between the healthy group and the experimental group in the passive state condition. In the active motion condition (Event2, Event3, Event4), the mean value of the healthy control group is generally higher than that of the stroke patient group, and these differences have reached statistical significance (P<0.05), and the research results show that with the increase of the patient's active motion intention, the difference between the healthy control group and the stroke patient group gradually increases, further supporting the significant difference between the two groups under different evaluation conditions.

[0105] In summary, under the conditions of resting state and dynamic state, there is a significant difference between the healthy control group and the stroke patient group, and this difference is more obvious under the condition of different participants having active motion intention, as shown in the following table:

[0106]

[0107] In addition, the rank sum test model is preferably the Wilcoxon rank sum test algorithm model. The neural network model takes the fuzzy entropy, coherence coefficient, mutual information coefficient, electroencephalogram acquisition quality coefficient and rectus femoris thickness data as input, and outputs the Fugl-Meyer motor function score prediction value.

[0108] In the embodiment, the neural network model is preferably a multi-layer perceptron (MLP, multi-layer fully connected network). The multi-layer perceptron has a simple structure, is suitable for processing the splicing of multi-modal features (directly splicing the electroencephalogram features and the muscle thickness data into a one-dimensional vector input), can learn the complex nonlinear correlation between the features (such as the joint influence of the electroencephalogram fuzzy entropy value and the muscle thickness on the motor function) through multi-layer nonlinear transformation (such as RelU activation), has a lower requirement for the sample size, can be stably trained with a small or medium-sized dataset (such as tens to hundreds of samples), and is suitable for the scene where the sample size is limited in clinical research.

[0109] In a preferred implementation, in order to comprehensively and accurately evaluate the lower limb rehabilitation of the patient, the quantitative evaluation indexes of the lower limb rehabilitation of the patient include: a resting-state muscle thickness difference, an active movement thickness growth rate, an Alpha frequency band coherence difference, a high Beta frequency band coherence ratio, and a Pearson correlation coefficient of the muscle thickness and the Beta coherence, the coherence coefficients of the electroencephalogram signal and the rectus femoris thickness signal in the Alpha frequency band and the Beta frequency band are calculated, the coherence differences of the patient group and the healthy group in different movement states are compared and analyzed, and the like, and the quantitative evaluation indexes are all obtained based on the corresponding values in S2. Specifically, the electroencephalogram fuzzy entropy, the coherence coefficient, the mutual information coefficient (mutual information value), the electroencephalogram acquisition quality coefficient, and the rectus femoris thickness data are input into the neural network model, the Fugl-Meyer motor function score is output through the model, and the lower limb movement function state and the rehabilitation progress are quantitatively evaluated in combination with the multi-modal analysis result.

[0110] It should be noted that the multi-modal features are integrated by the three-factor variance analysis and the neural network model, the individual-based motor function quantitative evaluation result is generated, the muscle atrophy degree, the neural-muscle coupling efficiency and the like are quantified, the traditional evaluation mode is broken through, the data support is provided for the personalized rehabilitation scheme such as the progressive resistance training, and the accurate connection from the evaluation to the intervention is realized.

[0111] At this point, the whole process of the lower limb rehabilitation evaluation method based on the brain-computer interface is completed. The lower limb rehabilitation evaluation method based on the brain-computer interface constructs a two-way evaluation system of “peripheral muscle morphology-central nervous activity” through the synchronous acquisition of the wearable A-mode ultrasound and the electroencephalogram signal; the A-mode ultrasound monitors the rectus femoris thickness change in real time, the electroencephalogram signal analyzes the Alpha / Beta frequency band coherence, the function difference evaluation in the resting state, the passive movement and the different resistance active movement states of the stroke patient is realized, the limitation of the single subjective score of the traditional scale is avoided, and the multi-dimensional quantitative index is provided for the rehabilitation evaluation.

[0112] Embodiment 2:

[0113] In a second aspect, the application further provides a lower limb rehabilitation evaluation system based on a brain-computer interface, referring to Figure 8 The system comprises:

[0114] a data acquisition module configured to synchronously acquire muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device;

[0115] a multi-modal feature extraction module configured to extract multi-modal features based on the muscle thickness data and the electroencephalogram data.

[0116] a quantitative evaluation index generation module configured to collaboratively analyze the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate a quantitative evaluation index.

[0117] It can be understood that the lower limb rehabilitation evaluation system based on a brain-computer interface provided by the embodiments of the present application corresponds to the lower limb rehabilitation evaluation method based on a brain-computer interface described above, and the explanation, examples, beneficial effects, etc. of the related content can refer to the corresponding content in the lower limb rehabilitation evaluation method based on a brain-computer interface, which will not be repeated here.

[0118] Embodiment 3:

[0119] In a third aspect, the present application also provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to realize the steps of the lower limb rehabilitation evaluation method based on a brain-computer interface described in any one of the embodiments and the preferred embodiments, and the method mainly comprises:

[0120] S1, synchronously acquiring muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device;

[0121] S2, extracting multi-modal features based on the muscle thickness data and the electroencephalogram data;

[0122] S3, collaboratively analyzing the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate a quantitative evaluation index.

[0123] It can be understood that the computer device provided by the embodiments of the present application corresponds to the lower limb rehabilitation evaluation system and method based on a brain-computer interface described above, and the explanation, examples, beneficial effects, etc. of the related content can refer to the corresponding content in the lower limb rehabilitation evaluation system and method based on a brain-computer interface, which will not be repeated here.

[0124] Embodiment 4:

[0125] In a fourth aspect, the present application also provides a computer readable storage medium comprising a stored computer program, wherein the computer program, when executed, controls a device on which the computer readable storage medium is located to perform the steps of the method for lower limb rehabilitation assessment based on brain-computer interface according to any one of the preceding embodiments and preferred embodiments, which mainly comprises:

[0126] S1, synchronously acquiring muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device;

[0127] S2, extracting multi-modal features based on the muscle thickness data and the electroencephalogram data;

[0128] S3, performing collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate a quantitative evaluation index.

[0129] It can be understood that the computer device provided by the embodiments of the present application corresponds to the lower limb rehabilitation assessment system, method and device based on brain-computer interface, and the explanation, examples, beneficial effects and the like of the related content can refer to the corresponding content in the lower limb rehabilitation assessment system, method and device based on brain-computer interface. Here, it will not be repeated.

[0130] In summary, compared with the prior art, the present application has the following beneficial effects:

[0131] 1. The lower limb rehabilitation assessment method based on brain-computer interface provided by the present application synchronously acquires muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device, extracts multi-modal features based on the muscle thickness data and the electroencephalogram data, performs collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank sum test model, and a neural network model to generate a quantitative evaluation index. Through linear coherence and nonlinear mutual information analysis of multi-modal signals, combined with a neural network model, the present application realizes precise quantification, dynamic monitoring and personalized evaluation of lower limb motor function, and solves the technical problem of strong subjectivity and insufficient quantification of traditional evaluation methods.

[0132] 2. The technology of the present application constructs a two-way evaluation system of "peripheral muscle morphology-central nervous activity" through synchronous acquisition of wearable A-mode ultrasound and electroencephalogram signals. A-mode ultrasound monitors the thickness change of rectus femoris in real time, and electroencephalogram signal analysis of Alpha / Beta band coherence realizes the evaluation of functional differences of stroke patients in resting state, passive movement and different resistance active movement states, avoids the limitations of single subjective score of traditional scales, and provides multi-dimensional quantitative indicators for rehabilitation assessment.

[0133] 3. The technology of the present application designs a multi-state cycle test scheme based on a limb movement rehabilitation training device (lower limbs), combines high-frequency synchronous acquisition of wearable devices, and provides information feedback to the test patients to keep the body posture stable during measurement, improves data acquisition quality, and captures the dynamic changes of muscle thickness and brain electrical coherence caused by movement intention in real time, providing continuous monitoring means for functional evolution in rehabilitation training.

[0134] 4. The technology of the present application integrates multi-modal features through three-factor variance analysis and neural network model, generates individual-based movement function quantitative evaluation results, can quantify muscle atrophy degree, neural-muscle coupling efficiency and other indicators, breaks through the traditional evaluation mode, provides data support for personalized rehabilitation programs such as progressive resistance training, and realizes precise docking from evaluation to intervention.

[0135] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0136] The above embodiments are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A lower limb rehabilitation evaluation method based on a brain-computer interface, characterized in that, The method comprises: Synchronously acquiring muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device; Extracting multi-modal features based on the muscle thickness data and electroencephalogram data; Performing collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank-sum test model, and a neural network model to generate a quantitative evaluation index.

2. The method of claim 1, wherein, The brain-computer interface device comprises: A motion acquisition unit for acquiring muscle thickness data of a patient; A BCI acquisition unit for acquiring electroencephalogram data of a patient; A motion load unit for providing passive, active, non-resistance, and progressive resistance motion modes for a patient.

3. The method of claim 2, wherein, The motion acquisition unit comprises a wearable A-mode ultrasonic sensor; the BCI acquisition unit comprises a multi-channel electroencephalogram cap; and the motion load unit comprises a limb motion rehabilitation training device.

4. The method of claim 1, wherein, The multi-modal features comprise fuzzy entropy of electroencephalogram signals, a coherence coefficient, a mutual information coefficient of electroencephalogram signals and muscle thickness signals, an electroencephalogram acquisition quality coefficient, and muscle thickness data.

5. The method of claim 1, wherein, Extracting multi-modal features based on the muscle thickness data and electroencephalogram data comprises: Pretreating ultrasonic signal data representing muscle thickness by median filtering denoising and calculating muscle thickness based on echo time difference; Pretreating electroencephalogram signal data: first, analyzing dynamic amplitude of head motion by a navigation and attitude solving algorithm to remove low-quality electroencephalogram data collected under large dynamic abnormal shaking; then, filtering the retained electroencephalogram signal data by notch filtering and high-order band-pass filtering; finally, removing non-brain-derived components in the filtered electroencephalogram signal data by independent component analysis.

6. The method of claim 4, wherein, When the coherence coefficient is acquired, the similarity of two signals is measured by calculating power spectral density, and the formula is: where PC Z F Z (f) is the cross-power spectrum, indicating the correlation strength and phase difference between Cz and Fz channels in the alpha, beta, and gamma bands; PC Z (f), PF Z (f) is the auto-power spectrum, used to quantify the energy strength of each of the Cz Z and F Z channels in the above three frequency bands; C CF (f) is the coherence.

7. The method of claim 1, wherein, The quantitative evaluation index comprises resting-state muscle thickness difference, active motion thickness growth rate, Alpha frequency band coherence difference, high Beta frequency band coherence ratio, and Pearson correlation coefficient of muscle thickness and Beta coherence, coherence coefficients of electroencephalogram signals and muscle thickness signals in Alpha and Beta frequency bands, and coherence differences of patient groups and healthy groups in different motion states.

8. A lower limb rehabilitation evaluation system based on brain-computer interface, characterized in that, The system comprises: A data acquisition module configured to synchronously acquire muscle thickness data and electroencephalogram data of a patient in multiple states based on a brain-computer interface device; A multi-modal feature extraction module configured to extract multi-modal features based on the muscle thickness data and electroencephalogram data. A quantitative evaluation index generation module configured to perform collaborative analysis on the multi-modal features based on a three-factor variance analysis model, a rank-sum test model, and a neural network model to generate a quantitative evaluation index.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the lower limb rehabilitation evaluation method based on a brain-computer interface according to any one of claims 1-7. 10.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the lower limb rehabilitation evaluation method based on a brain-computer interface according to any one of claims 1-7.

Images