Rehabilitation method and system for children with spastic cerebral palsy based on motor imagery brain-computer interface

Abstract

The application provides a rehabilitation method and system for children with spastic cerebral palsy based on motor imagery brain-computer interface. In one aspect, specific motor imagery paradigms are constructed for the severity of children with spastic cerebral palsy, including left hemiplegia paradigm, right hemiplegia paradigm and double hemiplegia paradigm, which can effectively improve the rehabilitation effect of children with spastic cerebral palsy. Moreover, the motor imagery brain-computer interface scheme is used to decode electroencephalogram information, effectively evaluate whether the motor imagery needs to be adjusted, form a closed-loop feedback, and ensure the reliability of the motor imagery brain-computer interface therapy. On the other hand, the rehabilitation method and system constructed by combining the motor imagery brain-computer interface with the conventional rehabilitation training can help improve the effect of rehabilitation training and the rehabilitation participation of patients, and thus provide a more effective and scientific rehabilitation solution for children with spastic cerebral palsy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rehabilitation treatment, in particular to a spastic cerebral palsy child rehabilitation method and system based on motor imagery brain-computer interface. BACKGROUND

[0002] Spastic cerebral palsy (SCP) is a movement and posture disorder caused by non-progressive damage to the developing fetus or infant brain, characterized by high muscle tension and severe reflex hyperactivity. Therefore, improving motor function and self-care ability is one of the important goals of spastic cerebral palsy child rehabilitation. Spastic cerebral palsy is usually accompanied by damage to the sensory motor pathway. Because the brain has significant neural plasticity during development, how to enhance the connection of brain neural network to improve the motor function of children has become a hot issue in current research. At present, the clinical rehabilitation strategy for spastic cerebral palsy children mainly includes motor therapy and occupational therapy, supplemented by assistive orthotic devices and physical factor therapy. The structured rehabilitation strategy mainly enhances the plasticity of the affected limb in the cortical motor representation through bottom-up training of the four limbs.

[0003] Motor imagery (MI) is a treatment method that promotes rehabilitation by repeatedly simulating specific limb movements in the mind without actual movement or muscle stimulation. This therapy helps patients recover motor function by activating neural networks in the brain related to actual movement. Unlike traditional rehabilitation training, motor imagery therapy first mobilizes the core area of brain motor control, activates other motor brain areas and body periphery in a top-down manner, thereby improving neural plasticity and promoting motor cortex reorganization. Studies have shown that motor imagery therapy can improve the upper limb motor function of children with hemiplegia and help improve the ability of motor planning. Because the causes of spastic cerebral palsy are complex and the clinical manifestations are diverse, it is not limited to the rehabilitation of hemiplegic hand function, but also involves diplegic gait movement, and the experimental setup for testing and training adult motor imagery is not completely suitable for children. Therefore, different motor imagery paradigms need to be developed for different types of spastic cerebral palsy children.

[0004] A brain computer interface (BCI) system establishes a direct connection path between the human brain and external devices, analyzes brain activity and converts it into control signals for external devices. The motor imagery brain computer interface (MI-BCI) technology can collect electroencephalogram (EEG) signals in real time during motor imagery and provide neural feedback. The motor imagery brain computer interface technology has achieved good effect in improving the motor dysfunction of adult stroke patients. Although there are differences in the injury mechanism between stroke (ischemia / hemorrhage) and spastic cerebral palsy (trauma), rehabilitation methods can successfully transform into defects that rely on neuroplasticity to restore sensory motor function. In related technologies, the motor imagery brain computer interface technology can improve the motor function rehabilitation of children with cerebral palsy, but whether the motor imagery brain computer interface technology combined with conventional rehabilitation training can promote the recovery of motor dysfunction in children with spastic cerebral palsy still needs further research.

[0005] For example, the Chinese patent with application number 202010939356.7 and the invention name of robot-assisted cerebral palsy rehabilitation expression training system based on brain computer interface; and the Chinese patent with application number CN201710999026.5 and the invention name of a rehabilitation system and method based on interactive virtual reality and neuroelectric stimulation. In the above patents, the application of motor imagery and brain computer interface technology in the field of rehabilitation of children with spastic cerebral palsy is limited, and specific limb rehabilitation methods have not been designed for them. At the same time, the lack of specific motor imagery paradigm according to the severity of the disease of children with spastic cerebral palsy makes the applicability and pertinence of rehabilitation training insufficient. Although these two technologies show a positive effect on improving the rehabilitation effect of children with cerebral palsy, they have not been deeply combined with conventional rehabilitation training, and have not formed a systematic and complete rehabilitation system. Therefore, there is an urgent need for a method that integrates motor imagery brain computer interface and conventional rehabilitation training to fully exert their respective advantages and thus provide a more effective rehabilitation solution for children with spastic cerebral palsy. SUMMARY

[0006] In order to achieve the above-mentioned purposes and other advantages of the present application, the first object of the present application is to provide a spastic cerebral palsy child rehabilitation method based on motor imagery brain computer interface, comprising the following steps:

[0007] Conducting limb function training, fine motor training and self-care ability training based on the conventional rehabilitation training scene;

[0008] Conducting motor imagery training based on the motor imagery brain computer interface, specifically comprising the following steps:

[0009] Obtaining the selected motor imagery paradigm;

[0010] presenting a motor imagery task of a selected motor imagery paradigm to make the child with spastic cerebral palsy continuously imagine motion;

[0011] collecting electroencephalogram signals of the child with spastic cerebral palsy continuously imagining motion;

[0012] decoding the electroencephalogram signals to evaluate whether the motor imagery is effective in a closed-loop feedback manner.

[0013] Further, the motor imagery paradigm includes a left hemiplegia paradigm, a right hemiplegia paradigm and a diplegia paradigm.

[0014] Further, the motor imagery training step based on the motor imagery brain-computer interface further includes:

[0015] In each motor imagery training task, a resting time is first set to prompt the child with spastic cerebral palsy to concentrate;

[0016] After the resting time ends, a prompt information is sent to prompt that the motor imagery task is about to be presented;

[0017] After the motor imagery task is presented, a rest phase is entered.

[0018] Further, the motor imagery training includes multiple rounds of motor imagery training, and each round of motor imagery training includes multiple motor imagery tasks, and after a preset time interval, the next round is entered.

[0019] Further, the electroencephalogram signals are configured to be collected by a multi-channel electroencephalogram cap.

[0020] Further, the decoding of the electroencephalogram signals to evaluate whether the motor imagery is effective in a closed-loop feedback manner adopts a deep learning classification method based on a convolutional neural network and multi-dimensional feature fusion to decode the motor imagery task, and specifically includes the following steps:

[0021] preprocessing, data augmentation and feature extraction are performed on the electroencephalogram signals;

[0022] the effect of the motor imagery task is evaluated through a motor imagery classification model;

[0023] If the probability of the evaluation result is lower than a set threshold, the child is reminded to adjust the motor imagery.

[0024] Further, the method further includes the following steps:

[0025] the collected electroencephalogram signals are divided into a training set and a test set;

[0026] the electroencephalogram signals of the subjects in the training set are preprocessed, data augmented and feature extracted:

[0027] Adaptive joint feature matrix is combined, and a convolutional neural network structure is used for training of the motor imagery classification model.

[0028] The features extracted from the EEG signals of the test set subjects are used for motor imagery classification model testing to obtain a classification result and evaluate the model performance.

[0029] Further, the preprocessing of the EEG signals, data augmentation and feature extraction include the following steps:

[0030] A preset time period after the start of motor imagery is selected by band-pass filtering for subsequent classification, the EEG signals before imagination for a preset time are used as a baseline to remove drift, and independent component analysis is used to remove artifacts;

[0031] The preprocessed signals are augmented by sliding window cropping, and the time series of motor imagery are segmented into multiple segments;

[0032] An adaptive frequency-space-time joint feature matrix is constructed by fusing multi-domain information, specifically including the following steps:

[0033] Based on the energy distribution of the sub-band signals obtained by wavelet packet decomposition, active frequency bands are extracted to realize adaptive frequency spectrum sensing;

[0034] Based on the activation value, each lead signal is adaptively weighted and subjected to one pair of other common spatial pattern spatial filtering;

[0035] Based on the rich time domain information expression of multiple features, the nonlinear and non-stationary information of motor imagery EEG signals at different time points is captured and integrated;

[0036] Through the above steps, adaptive frequency-space-time joint feature learning is realized.

[0037] Further, the step of training the motor imagery classification model by combining the adaptive joint feature matrix and using the convolutional neural network structure includes:

[0038] The convolutional neural network structure performs three-layer spatial convolution;

[0039] The convolutional neural network structure performs two-layer time convolution;

[0040] Two fully connected layers are connected to integrate features and map the learned feature representation to the sample label space;

[0041] Exponential normalization is used as the last layer of the network to map the output to the interval (0, 1) for binary classification;

[0042] The classification accuracy and Kappa coefficient are calculated to select the optimal motor imagery classification model;

[0043] The EEG signals of the test set subjects are preprocessed, data augmented and features extracted.

[0044] A second objective of this invention is to provide a rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface, comprising a conventional rehabilitation training module and a motor imagery brain-computer interface module. The conventional rehabilitation training module is used to implement limb function training, fine motor training, and self-care ability training, while the motor imagery brain-computer interface module is used to implement motor imagery training.

[0045] The motor imagery brain-computer interface module includes a motor imagery paradigm selection module, a motor imagery training module, an EEG device, and an EEG signal processing module.

[0046] The motion imagination paradigm selection module is used to implement the selection of motion imagination paradigms;

[0047] The motor imagery training module is used to present motor imagery tasks of the selected motor imagery paradigm, so that children with spastic cerebral palsy can continuously perform motor imagery.

[0048] The EEG device is used to collect EEG signals from children with spastic cerebral palsy who are continuously imagining movements.

[0049] The EEG signal processing module is used to decode the EEG signal to evaluate the effectiveness of the motor imagery through closed-loop feedback.

[0050] Furthermore, the motor imagery paradigms include the left hemiplegic paradigm, the right hemiplegic paradigm, and the diplegia paradigm.

[0051] Furthermore, the motor imagery training module also includes a rest module, a prompting module, and a rest module: wherein,

[0052] The resting module is used to set a resting time before each motor imagery training task to prompt children with spastic cerebral palsy to concentrate.

[0053] The prompting module is used to issue a prompting message after the rest period ends, indicating that a motion visualization task is about to be presented.

[0054] The rest module is used to enter a rest phase after the motion imagery task is presented.

[0055] Furthermore, the motion imagery training includes multiple rounds of motion imagery training, each round of motion imagery training includes multiple motion imagery tasks, and the next round begins after a preset time interval.

[0056] Furthermore, the EEG device employs a multi-lead EEG cap.

[0057] Furthermore, the EEG signal processing module employs a deep learning classification method based on convolutional neural networks and multi-dimensional feature fusion to decode the motor imagery task, specifically including:

[0058] The EEG signals are preprocessed, amplified, and feature extracted.

[0059] The effectiveness of the motion imagery task was evaluated using a motion imagery classification model.

[0060] If the probability of the assessment result is lower than the set threshold, the child is reminded to adjust their motor imagery.

[0061] Furthermore, it also includes a motor imagery classification model training module, which is used to divide the collected EEG signals into a training set and a test set; preprocess, amplify, and extract features from the EEG signals of the subjects in the training set; train the motor imagery classification model using a convolutional neural network structure in combination with an adaptive joint feature matrix; test the motor imagery classification model using the features extracted from the EEG signals of the subjects in the test set, obtain the classification results, and evaluate the model performance.

[0062] Furthermore, the preprocessing, data amplification, and feature extraction of EEG signals include:

[0063] By using bandpass filtering, a preset time period after the start of motor imagery is selected for subsequent classification. The EEG signal at the preset time before imagery is used as a baseline to remove drift, and then independent component analysis is used to remove artifacts.

[0064] The preprocessed signal is pruned and amplified using a sliding window to divide the time series of motion images into multiple segments;

[0065] Constructing an adaptive frequency-space-time joint feature matrix that integrates multi-domain information includes the following steps:

[0066] Adaptive spectrum sensing is achieved by extracting active frequency bands based on the energy distribution of sub-band signals decomposed by wavelet packet decomposition.

[0067] Based on the activation value, each lead signal is adaptively weighted and a pair of other common spatial modes are spatially filtered.

[0068] Based on multiple features to enrich the temporal information expression, the nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated;

[0069] Through the above steps, adaptive frequency-space-time joint feature learning is achieved.

[0070] Furthermore, the step of training the motion imagery classification model using a convolutional neural network structure, incorporating an adaptive joint feature matrix, includes...

[0071] The convolutional neural network structure performs three layers of spatial convolution;

[0072] The convolutional neural network structure performs two layers of temporal convolution;

[0073] Connecting two fully connected layers is used to integrate features, mapping the learned feature representations to the sample label space;

[0074] Exponential normalization is used as the last layer of the network to map the output to the (0,1) interval for binary classification;

[0075] Calculate classification accuracy and Kappa coefficient to select the optimal motion imagery classification model;

[0076] The EEG signals of the subjects in the test set were preprocessed, data amplified, and feature extracted.

[0077] A third objective of the present invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0078] A fourth objective of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0079] Compared with the prior art, the beneficial effects of the present invention are:

[0080] This invention addresses two key aspects. First, it constructs specific motor imagery paradigms tailored to the severity of spastic cerebral palsy in children, including left-sided hemiplegic, right-sided hemiplegic, and diplegia paradigms. These paradigms effectively and specifically improve rehabilitation outcomes for children with spastic cerebral palsy. Furthermore, the invention provides a motor imagery brain-computer interface (BCI) solution to decode electroencephalogram (EEG) information, effectively assessing the need for adjustments to motor imagery and creating a closed-loop feedback loop to ensure the reliability of the BCI therapy. Second, the rehabilitation method and system developed in this invention, which integrates the motor imagery BCI with conventional rehabilitation training, helps improve the effectiveness of rehabilitation training and patient participation, thus providing a more effective and scientific rehabilitation solution for children with spastic cerebral palsy.

[0081] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0082] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0083] Figure 1A schematic diagram of a rehabilitation method and system for children with spastic cerebral palsy based on motor imagery brain-computer interface;

[0084] Figure 2 This is a flowchart of the rehabilitation method for children with spastic cerebral palsy based on motor imagery brain-computer interface in Example 1;

[0085] Figure 3 The process of motor imagery training based on brain-computer interface for example 1. Figure 1 ;

[0086] Figure 4 The process of motor imagery training based on brain-computer interface for example 1. Figure 2 ;

[0087] Figure 5 This is a flowchart of the EEG signal decoding process in Example 1;

[0088] Figure 6 This is a flowchart of the training process for the motion imagery classification model in Example 1;

[0089] Figure 7 The flowchart for training a motion image classification model using a convolutional neural network structure, combined with the adaptive joint feature matrix, is shown in Example 1.

[0090] Figure 8 This is a flowchart of the preprocessing, data amplification, and feature extraction of EEG signals in Example 1;

[0091] Figure 9 The flowchart for constructing the adaptive frequency-space-time joint feature matrix that integrates multi-domain information is shown in Example 1.

[0092] Figure 10 This is a schematic diagram of the rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface, as described in Example 2.

[0093] Figure 11 This is a schematic diagram of the motor imagery brain-computer interface module in Example 2;

[0094] Figure 12 This is a schematic diagram of the computer device in Example 3;

[0095] Figure 13 This is a schematic diagram of a computer-readable storage medium according to Example 4. Detailed Implementation

[0096] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0097] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0098] The drawing numbers in this application are only used to distinguish the steps in the scheme and are not used to limit the execution order of the steps. The specific execution order is as described in the specification.

[0099] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0100] Example 1

[0101] A rehabilitation method for children with spastic cerebral palsy based on motor imagery brain-computer interface can help improve rehabilitation efficacy, such as... Figure 1 , Figure 2 As shown, the method includes the following steps:

[0102] S1. Conduct limb function training, fine motor training and self-care ability training based on conventional rehabilitation training scenarios. The daily training time can be set according to the actual situation; for example, set the daily training time to 30 minutes.

[0103] S2. Motor imagery training based on brain-computer interface for motor imagery, such as Figure 3 As shown, the specific steps include:

[0104] In the actual operation of the imagery training, the first step is to put an EEG device, such as a 64-lead EEG cap, on the child with spastic cerebral palsy. Taking the 64-lead EEG cap as an example, conductive gel is injected and the needle-type movable electrodes are inserted into the corresponding electrode holes to check the accurate placement of the electrodes and the stability of the signal.

[0105] Then, appropriate motor imagery paradigms are selected for children with spastic cerebral palsy; among them, motor imagery paradigms include left hemiplegic paradigm, right hemiplegic paradigm, and diplegia paradigm.

[0106] S21. Obtain the selected paradigm of motion visualization;

[0107] S22. Present the selected motor imagery paradigm with a motor imagery task to enable children with spastic cerebral palsy to continuously engage in motor imagery.

[0108] In some embodiments, such as Figure 4 As shown, the steps for motor imagery training based on the motor imagery brain-computer interface further include:

[0109] S220. In each motor imagery training task, first set a resting time, such as a 3-second resting time, to prompt children with spastic cerebral palsy to concentrate.

[0110] S221. After the rest period ends, a prompt message is given to indicate that a motor imagery task is about to be presented; for example, after a "beep" prompt, a diagram of the motor imagery task is presented, and the child continues to imagine the movement for 6 seconds after the diagram appears.

[0111] S222. After the motor imagery task is presented, a rest phase begins. For example, after the motor imagery task ends, a 3-second rest phase begins.

[0112] The motor imagery training includes multiple rounds of training, each round containing multiple motor imagery tasks, with a preset time interval between rounds. For example, each round may contain 40 motor imagery tasks, with a 5-minute interval between rounds, for a total of 80 rounds.

[0113] S23. Collecting EEG signals from children with spastic cerebral palsy during sustained motor imagery; in some embodiments, the EEG signals are configured to be collected via a multi-lead EEG cap.

[0114] S24. Decode the EEG signal and use closed-loop feedback to evaluate whether the motor imagery is effective.

[0115] In some embodiments, such as Figure 5 As shown, the decoding of the EEG signal to evaluate the effectiveness of motor imagery using closed-loop feedback employs a deep learning classification method based on convolutional neural networks and multi-dimensional feature fusion to decode the motor imagery task, specifically including the following steps:

[0116] S241. Preprocess, amplify, and extract features from the electroencephalogram (EEG) signal. The specific steps are described in the relevant embodiments below and will not be repeated here.

[0117] S242. Evaluate the effectiveness of the motor imagery task using a motor imagery classification model;

[0118] S243. If the probability of the assessment result is lower than the set threshold, remind the child to adjust their motor imagery. For example, if the probability of the assessment result is lower than the set threshold, play a 3-second beeping sound to remind the child to adjust their motor imagery.

[0119] In some embodiments, such as Figure 6 As shown, it also includes the following steps:

[0120] S200. Divide the collected EEG signals into a training set and a test set; for example, divide the collected EEG signals into a training set and a test set in a ratio of 8:2.

[0121] S201. The EEG signals of the subjects in the training set are preprocessed, data amplified, and feature extracted. The specific steps are described in the relevant embodiments below and will not be repeated here:

[0122] S202. Combining the adaptive joint feature matrix, a convolutional neural network structure is used to train the motion image classification model, such as... Figure 7 As shown, the specific steps include:

[0123] S2021. A convolutional neural network structure is used to perform three-layer spatial convolution; for example, three convolutional kernels of [5×1], [10×1], and [15×1] are used to learn spatial dimension information, following the learning rule from low complexity to high complexity, and the number of the three convolutional kernels is set to 8, 16, and 32 respectively.

[0124] S2022, The convolutional neural network structure performs two layers of temporal convolution; for example, [1×2] and [1×4] convolutional kernels are set to learn the features provided by the temporal dimension, and the number of convolutional kernels is set to 32.

[0125] S2023: Connecting two fully connected layers is used to integrate features, mapping the learned feature representations to the sample label space;

[0126] S2024. Using exponential normalization (softmax) as the last layer of the network, the output is mapped to the (0,1) interval for binary classification.

[0127] S2025. Calculate the classification accuracy and Kappa coefficient to select the optimal motion imagery classification model;

[0128] S2026. The EEG signals of the test set subjects are preprocessed, data amplified, and feature extracted. The specific steps are described in the relevant embodiments below and will not be repeated here.

[0129] S203. The features extracted from the EEG signals of the subjects in the test set are used to test the motor imagery classification model, the classification results are obtained, and the model performance is evaluated.

[0130] In some alternative embodiments, such as Figure 8 As shown, the preprocessing, data amplification, and feature extraction of EEG signals include the following steps:

[0131] S210. Select a preset time period after the start of motor imagery using bandpass filtering to participate in subsequent classification. Use the EEG signal at the preset time before imagery as a baseline to remove drift, and then use independent component analysis to remove artifacts. For example, set the bandpass filter to 0.1-32Hz, select the 0-6s time period after the start of motor imagery to participate in subsequent classification, use the signal at 0.5s before imagery as a baseline to remove drift, and then use independent component analysis to remove artifacts to eliminate obvious artifact components such as electrooculography, eye movement, and head movement.

[0132] S211. The preprocessed signal is pruned and amplified using a sliding window to divide the motion sequence into multiple segments; for example, the window width can be set to 3s and the step size to 100 sampling points to divide the 6s time series into 16 segments.

[0133] S212. Construct an adaptive frequency-space-time joint feature matrix that integrates multi-domain information, such as... Figure 9 As shown, the specific steps include:

[0134] S2121. First, active frequency bands are extracted based on the energy distribution of sub-band signals from wavelet packet decomposition (WPD) to achieve adaptive spectrum sensing.

[0135] S2122. Secondly, based on the activation value, the signals of each lead are adaptively weighted and subjected to "One-Versus-Rest Common Spatial Pattern" (OVR-CSP) spatial filtering. Through the above two steps, the frequency-spatial characteristics of the subject are described.

[0136] S2123. Based on multiple features to enrich the temporal information expression, capture and integrate the nonlinear and nonstationary information of motor imagery EEG signals at different time points, and further provide intuitive information supplementation for the above adaptive frequency-space characteristics; for example, based on eight features to enrich the temporal information expression, including maximum value, peak value, peak-to-peak value, mean value, absolute mean value, root square amplitude, variance and standard deviation, nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated.

[0137] Through the above steps (steps S2121 to S2123), adaptive frequency-space-time joint feature learning is achieved.

[0138] This embodiment constructs a rehabilitation method for spastic cerebral palsy based on a combination of motor imagery brain-computer interface and conventional rehabilitation training. Targeting the degree of motor impairment in children with spastic cerebral palsy, this embodiment designs three motor imagery paradigms: a left hemiplegic paradigm, a right hemiplegic paradigm, and a bilateral hemiplegic paradigm, to meet the rehabilitation needs of different patients. In the hemiplegic paradigm, grasping of the left / right upper limbs and dorsiflexion of the lower limbs are emphasized, while the bilateral hemiplegic paradigm focuses on the coordinated training of dorsiflexion of both lower limbs. This embodiment adds motor imagery brain-computer interface training to conventional rehabilitation training, continuously activating neural networks in the brain related to actual movement through repetitive imagery, synergistically working with conventional rehabilitation therapies to promote the recovery of motor function. This rehabilitation system overcomes the shortcomings of conventional rehabilitation training, such as its monotonous and tedious process and low patient participation, thereby improving patient motivation and intervention effectiveness.

[0139] Example 2

[0140] A rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface, applying the method described above. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here. Figure 1 , Figure 10 As shown, the system 300 includes a conventional rehabilitation training module 310 and a motor imagery brain-computer interface module 320. The conventional rehabilitation training module is used to implement limb function training, fine motor training, and self-care ability training. The daily training time can be set according to actual conditions; for example, the daily training time can be set to 30 minutes. The motor imagery brain-computer interface module is used to implement motor imagery training; wherein,

[0141] like Figure 11 As shown, the motor imagery brain-computer interface module 320 includes a motor imagery paradigm selection module 321, a motor imagery training module 322, an EEG device 323, and an EEG signal processing module 324.

[0142] In the actual operation of the imagery training, the first step is to put an EEG device, such as a 64-lead EEG cap, on the child with spastic cerebral palsy. Taking the 64-lead EEG cap as an example, conductive gel is injected and the needle-type movable electrodes are inserted into the corresponding electrode holes to check the accurate placement of the electrodes and the stability of the signal.

[0143] Then, appropriate motor imagery paradigms are selected for children with spastic cerebral palsy; among them, motor imagery paradigms include left hemiplegic paradigm, right hemiplegic paradigm, and diplegia paradigm.

[0144] The motion imagination paradigm selection module is used to implement the selection of motion imagination paradigms;

[0145] The motor imagery training module is used to present motor imagery tasks of the selected motor imagery paradigm, so that children with spastic cerebral palsy can continuously perform motor imagery.

[0146] In some embodiments, the motor imagery training module further includes a rest module, a prompting module, and a rest module: wherein,

[0147] The resting module is used to set a resting time, such as 3 seconds, for each motor imagery training task to prompt children with spastic cerebral palsy to concentrate.

[0148] The prompting module is used to issue a prompting message after the rest period ends to indicate that a motor imagery task is about to be presented; for example, after a "beep" prompt, a diagram of the motor imagery task is presented, and the child continues to imagine the movement for 6 seconds after the diagram appears.

[0149] The rest module is used to enter a rest phase after the motor imagery task is presented. For example, after the motor imagery ends, a 3-second rest phase begins.

[0150] The motor imagery training includes multiple rounds of training, each round containing multiple motor imagery tasks, with a preset time interval between rounds. For example, each round of training contains 40 motor imagery tasks, with a 5-minute interval between rounds, for a total of 80 training sessions.

[0151] The EEG device is used to collect EEG signals from children with spastic cerebral palsy who are continuously imagining movements; optionally, the EEG device is a multi-lead EEG cap.

[0152] The EEG signal processing module is used to decode the EEG signal to evaluate the effectiveness of the motor imagery through closed-loop feedback.

[0153] In some embodiments, the EEG signal processing module employs a deep learning classification method based on convolutional neural networks and multi-dimensional feature fusion to decode the motor imagery task, specifically including:

[0154] The EEG signals are preprocessed, data amplified, and feature extracted. The specific steps are described in the relevant embodiments below and will not be repeated here.

[0155] The effectiveness of the motion imagery task was evaluated using a motion imagery classification model.

[0156] If the probability of the assessment result is lower than a set threshold, the child is prompted to adjust their motor imagery. For example, if the probability of the assessment result is lower than the set threshold, a 3-second beep sound is played to remind the child to adjust their motor imagery.

[0157] In some embodiments, the system further includes a motor imagery classification model training module for dividing the acquired EEG signals into a training set and a test set; for example, dividing the acquired EEG signals into a training set and a test set in an 8:2 ratio.

[0158] The EEG signals of the subjects in the training set were preprocessed, amplified, and feature extracted. The specific steps are described in the relevant embodiments below and will not be repeated here:

[0159] The motion image classification model is trained using a convolutional neural network structure, combining an adaptive joint feature matrix, specifically including:

[0160] The convolutional neural network structure performs three layers of spatial convolution; for example, three convolutional kernels of [5×1], [10×1], and [15×1] are used to learn spatial dimension information, following the learning rule from low complexity to high complexity, and the number of the three convolutional kernels is set to 8, 16, and 32 respectively.

[0161] The convolutional neural network structure performs two layers of temporal convolution; for example, [1×2] and [1×4] convolutional kernels are set to learn the features provided by the temporal dimension, and the number of convolutional kernels is set to 32.

[0162] Connecting two fully connected layers is used to integrate features, mapping the learned feature representations to the sample label space;

[0163] The network uses exponential normalization (softmax) as the last layer to map the output to the (0,1) interval for binary classification;

[0164] Calculate classification accuracy and Kappa coefficient to select the optimal motion imagery classification model;

[0165] The EEG signals of the test set subjects were preprocessed, data amplified, and feature extracted. The specific steps are described in the relevant embodiments below and will not be repeated here.

[0166] The features extracted from the EEG signals of the subjects in the test set were used to test the motor imagery classification model, and the classification results were obtained to evaluate the model performance.

[0167] In some alternative embodiments, the preprocessing, data amplification, and feature extraction of EEG signals include:

[0168] By using bandpass filtering to select a preset time period after the start of motor imagery for subsequent classification, the EEG signal at the preset time before imagery is used as a baseline to remove drift, and then independent component analysis is used to remove artifacts. For example, if the bandpass filter is set to 0.1-32Hz, the time period 0-6s after the start of motor imagery is selected for subsequent classification, the signal 0.5s before imagery is used as a baseline to remove drift, and then independent component analysis is used to remove artifacts, so as to eliminate obvious artifact components such as electrooculography, eye movement, and head movement.

[0169] The preprocessed signal is pruned and amplified using a sliding window to divide the motion sequence into multiple segments; for example, the window width can be set to 3s and the step size to 100 sampling points to divide a 6s time series into 16 segments.

[0170] Constructing an adaptive frequency-space-time joint feature matrix that integrates multi-domain information includes the following steps:

[0171] First, active frequency bands are extracted based on the energy distribution of sub-band signals from wavelet packet decomposition (WPD) to achieve adaptive spectrum sensing.

[0172] Secondly, based on the activation value, the signals of each lead are adaptively weighted and spatially filtered using a "one-versus-Rest Common Spatial Pattern" (OVR-CSP). Through these two steps, the frequency-spatial characteristics of the subject are described.

[0173] Based on a variety of rich time-domain information representations, nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated, further providing intuitive information supplementation to the above adaptive frequency-space characteristics; for example, based on eight rich time-domain information representations, including maximum value, peak value, peak-to-peak value, mean value, absolute mean value, root square amplitude, variance, and standard deviation, nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated.

[0174] Through the above steps, adaptive frequency-space-time joint feature learning is achieved.

[0175] This embodiment constructs a rehabilitation system for spastic cerebral palsy based on a combination of motor imagery brain-computer interface and conventional rehabilitation training. Targeting the degree of motor impairment in children with spastic cerebral palsy, this embodiment designs three motor imagery paradigms: a left hemiplegic paradigm, a right hemiplegic paradigm, and a bilateral paraplegic paradigm, to meet the rehabilitation needs of different patients. In the hemiplegic paradigm, grasping of the left / right upper limbs and dorsiflexion of the lower limbs are emphasized in training, while the bilateral paraplegic paradigm focuses on the coordinated training of dorsiflexion of both lower limbs. This embodiment adds motor imagery brain-computer interface training to conventional rehabilitation training, continuously activating neural networks in the brain related to actual movement through repetitive imagery, working synergistically with conventional rehabilitation therapies to promote the recovery of motor function. This rehabilitation system overcomes the shortcomings of conventional rehabilitation training, such as its monotonous and tedious process and low patient participation, thereby improving patient motivation and intervention effectiveness.

[0176] Example 3

[0177] A computer device 400, such as Figure 12As shown, the system includes a memory 410, a processor 420, and a computer program 430 stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here.

[0178] Example 4

[0179] A computer-readable storage medium, such as Figure 13 As shown, a computer program is stored thereon. When executed by a processor, the computer program implements the steps of a rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0180] Example 5

[0181] A computer program product includes a computer program that, when executed by a processor, implements the steps of a rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0182] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0183] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0184] The apparatus, computer device, and non-volatile computer storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, computer device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, computer device, and non-volatile computer storage medium will not be repeated here.

[0185] Those skilled in the art will also know that, besides implementing the controller in the form of purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller take the form of logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices included within it for implementing various functions can also be considered structures within that hardware component. Alternatively, the devices for implementing various functions can be considered as both software units implementing the method and structures within a hardware component.

[0186] The systems, apparatuses, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above apparatuses are described separately as various units based on their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0187] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0189] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0190] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0191] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0192] This specification may be described in the general context of computer-executable instructions, such as program units, that are executed by a computer. Generally, program units include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program units may reside in local and remote computer storage media, including storage devices.

[0193] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0194] The above description is merely an embodiment of this specification and is not intended to limit the scope of one or more embodiments of this specification. Various modifications and variations can be made to one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of one or more embodiments of this specification.

Claims

1. A rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface, characterized in that, Includes the following steps: Based on conventional rehabilitation training scenarios, conduct limb function training, fine motor training, and self-care ability training. Motor imagery training based on a brain-computer interface for motor imagery includes the following steps: Acquire a chosen paradigm of motion imagination; Present motor imagery tasks using the selected motor imagery paradigm to enable children with spastic cerebral palsy to continuously engage in motor imagery. Collect electroencephalogram (EEG) signals from children with spastic cerebral palsy during sustained motor imagery. The EEG signals were decoded to assess the effectiveness of the motor imagery using closed-loop feedback. The motor imagery paradigms include the left hemiplegic paradigm, the right hemiplegic paradigm, and the diplegia paradigm; The decoding of the EEG signals, using closed-loop feedback to evaluate the effectiveness of motor imagery, employs a deep learning classification method based on convolutional neural networks and multi-dimensional feature fusion to decode the motor imagery task, specifically including the following steps: The EEG signals are preprocessed, amplified, and feature extracted. The effectiveness of the motion imagery task was evaluated using a motion imagery classification model. If the probability of the assessment result is lower than the set threshold, the child is reminded to adjust their motor imagery. It also includes the following steps: The collected EEG signals were divided into training and testing sets; The EEG signals of the subjects in the training set were preprocessed, amplified, and feature extracted: The motion image classification model is trained using a convolutional neural network structure, incorporating an adaptive joint feature matrix. The features extracted from the EEG signals of the subjects in the test set are used to test the motor imagery classification model, and the classification results are obtained to evaluate the model performance. The step of training the motion imagery classification model using a convolutional neural network structure, combined with an adaptive joint feature matrix, includes: The convolutional neural network structure performs three layers of spatial convolution; The convolutional neural network structure performs two layers of temporal convolution; Connecting two fully connected layers is used to integrate features, mapping the learned feature representations to the sample label space; Exponential normalization is used as the last layer of the network to map the output to the (0,1) interval for binary classification; Calculate classification accuracy and Kappa coefficient to select the optimal motion imagery classification model; The EEG signals of the subjects in the test set were preprocessed, data amplified, and feature extracted.

2. The rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 1, characterized in that, The steps for motor imagery training based on a brain-computer interface for motor imagery also include: In each motor imagery training task, a rest period is set first to prompt children with spastic cerebral palsy to concentrate. After the rest period ends, a prompt message is issued to indicate that a motion visualization task is about to be presented; After the motion visualization task is presented, the rest phase begins.

3. The rehabilitation method for children with spastic cerebral palsy based on motor imagery brain-computer interface as described in claim 2, characterized in that: The motion imagery training includes multiple rounds of motion imagery training, each round of motion imagery training includes multiple motion imagery tasks, and the next round begins after a preset interval.

4. The rehabilitation method for children with spastic cerebral palsy based on motor imagery brain-computer interface as described in claim 1, characterized in that: The EEG signals were configured to be acquired using a multi-lead EEG cap.

5. A rehabilitation method for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 1, characterized in that, The preprocessing, data amplification, and feature extraction of EEG signals include the following steps: By using bandpass filtering, a preset time period after the start of motor imagery is selected for subsequent classification. The EEG signal at the preset time before imagery is used as a baseline to remove drift, and then independent component analysis is used to remove artifacts. The preprocessed signal is pruned and amplified using a sliding window to divide the time series of motion images into multiple segments; Constructing an adaptive frequency-space-time joint feature matrix that integrates multi-domain information includes the following steps: Adaptive spectrum sensing is achieved by extracting active frequency bands based on the energy distribution of sub-band signals decomposed by wavelet packet decomposition. Based on the activation value, each lead signal is adaptively weighted and a pair of other common spatial modes are spatially filtered. Based on multiple features to enrich the temporal information expression, the nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated; Through the above steps, adaptive frequency-space-time joint feature learning is achieved.

6. A rehabilitation system for children with spastic cerebral palsy based on a brain-computer interface for motor imagery, characterized in that: The system includes a conventional rehabilitation training module and a motor imagery brain-computer interface module. The conventional rehabilitation training module is used to implement limb function training, fine motor training, and self-care ability training. The motor imagery brain-computer interface module is used to implement motor imagery training. The motor imagery brain-computer interface module includes a motor imagery paradigm selection module, a motor imagery training module, an EEG device, and an EEG signal processing module. The motion imagination paradigm selection module is used to implement the selection of motion imagination paradigms; The motor imagery training module is used to present motor imagery tasks of the selected motor imagery paradigm, so that children with spastic cerebral palsy can continuously perform motor imagery. The EEG device is used to collect EEG signals from children with spastic cerebral palsy who are continuously imagining movements. The EEG signal processing module is used to decode the EEG signal and evaluate the effectiveness of the motor imagery using closed-loop feedback. The motor imagery paradigms include the left hemiplegic paradigm, the right hemiplegic paradigm, and the diplegia paradigm; The EEG signal processing module employs a deep learning classification method based on convolutional neural networks and multi-dimensional feature fusion to decode the motor imagery task, specifically including: The EEG signals are preprocessed, amplified, and feature extracted. The effectiveness of the motion imagery task was evaluated using a motion imagery classification model. If the probability of the assessment result is lower than the set threshold, the child is reminded to adjust their motor imagery. It also includes a motor imagery classification model training module, which is used to divide the collected EEG signals into training set and test set; preprocess, amplify and extract features from the EEG signals of the subjects in the training set; train the motor imagery classification model using a convolutional neural network structure in combination with an adaptive joint feature matrix; test the motor imagery classification model using the features extracted from the EEG signals of the subjects in the test set, obtain the classification results and evaluate the model performance. The process of training the motion imagery classification model using an adaptive joint feature matrix and a convolutional neural network structure includes... The convolutional neural network structure performs three layers of spatial convolution; The convolutional neural network structure performs two layers of temporal convolution; Connecting two fully connected layers is used to integrate features, mapping the learned feature representations to the sample label space; Exponential normalization is used as the last layer of the network to map the output to the (0,1) interval for binary classification; Calculate classification accuracy and Kappa coefficient to select the optimal motion imagery classification model; The EEG signals of the subjects in the test set were preprocessed, data amplified, and feature extracted.

7. A rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 6, characterized in that: The motor imagery training module also includes a rest module, a prompting module, and a relaxation module: among which, The resting module is used to set a resting time before each motor imagery training task to prompt children with spastic cerebral palsy to concentrate. The prompting module is used to issue a prompting message after the rest period ends, indicating that a motion visualization task is about to be presented. The rest module is used to enter a rest phase after the motion imagery task is presented.

8. A rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 7, characterized in that: The motion imagery training includes multiple rounds of motion imagery training, each round of motion imagery training includes multiple motion imagery tasks, and the next round begins after a preset interval.

9. A rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 6, characterized in that: The EEG device uses a multi-lead EEG cap.

10. A rehabilitation system for children with spastic cerebral palsy based on a motor imagery brain-computer interface as described in claim 6, characterized in that: Preprocessing, data amplification, and feature extraction of EEG signals include: By using bandpass filtering, a preset time period after the start of motor imagery is selected for subsequent classification. The EEG signal at the preset time before imagery is used as a baseline to remove drift, and then independent component analysis is used to remove artifacts. The preprocessed signal is pruned and amplified using a sliding window to divide the time series of motion images into multiple segments; Constructing an adaptive frequency-space-time joint feature matrix that integrates multi-domain information includes the following steps: Adaptive spectrum sensing is achieved by extracting active frequency bands based on the energy distribution of sub-band signals decomposed by wavelet packet decomposition. Based on the activation value, each lead signal is adaptively weighted and a pair of other common spatial modes are spatially filtered. Based on multiple features to enrich the temporal information expression, the nonlinear and nonstationary information of motor imagery EEG signals at different time points can be captured and integrated; Through the above steps, adaptive frequency-space-time joint feature learning is achieved.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 5.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 5.

Images