Multi-mode nerve regulation electroacupuncture control system for spinal cord injury repair

Abstract

The invention relates to the technical field of spinal cord injury repair, and discloses a multi-mode nerve regulation electric acupuncture control system for spinal cord injury repair, a patient evaluation module is used for establishing a personalized spinal cord injury model of a patient and evaluating the current nerve function state and rehabilitation potential of the patient to obtain a nerve function evaluation result; the parameter setting module is used for setting multi-modal stimulation parameters in electroacupuncture treatment; the signal processing module is used for collecting local field potential of a spinal cord injury area and electromyographic signals around the injury in real time, preprocessing the collected signals and extracting time-frequency characteristics of the signals. The evaluation module is used for inputting the time-frequency characteristics of the signals into an electro-acupuncture evaluation model, evaluating the neural pathway in real time through the electro-acupuncture evaluation model and outputting an evaluation result; the adjusting module is used for adjusting the frequency, waveform and current intensity parameters of the electroacupuncture needle based on the evaluation result and a Q-learning algorithm, searching an optimal electroacupuncture needle parameter combination and feeding back the optimal electroacupuncture needle parameter combination to the electroacupuncture needle; according to the invention, the pertinence and effectiveness of treatment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of spinal cord injury repair, and particularly relates to a multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair. Background Art

[0002] Spinal cord injury is a serious nervous system injury, often leading to limb movement and sensory function disorders in patients, seriously affecting the quality of life of patients. At present, the treatment methods for spinal cord injury mainly include surgical treatment, drug treatment, and rehabilitation treatment, etc. However, the treatment effects of these methods are still very limited. As a product of the combination of traditional Chinese medicine acupuncture and modern electrical stimulation technology, electroacupuncture therapy shows certain potential in spinal cord injury rehabilitation. It promotes nerve function recovery by introducing weak current into acupoints to stimulate nerve tissue. However, the existing electroacupuncture therapy usually adopts a single electrical stimulation mode, which is difficult to accurately simulate the complex characteristics of human nerve signals, unable to fully stimulate the repair potential of nerve cells, and there is still a large room for improvement in the treatment effect. In addition, the single electrical stimulation mode cannot comprehensively consider the differences among spinal cord injury patients in terms of injury degree, nerve recovery status, etc., and it is difficult to achieve personalized and precise treatment. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and design a multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair.

[0004] The present invention provides a multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair, and the system includes: A patient evaluation module, which is used to obtain the basic information of the patient's spinal cord injury and individual physiological characteristic data, establish a personalized spinal cord injury model of the patient, and evaluate the patient's current nerve function status and rehabilitation potential to obtain a nerve function evaluation result; A parameter setting module, which is used to set multi-modal stimulation parameters in electroacupuncture treatment according to the patient's personalized spinal cord injury model and nerve function evaluation result; A signal processing module, which is used to, during the treatment process, collect the local field potential of the spinal cord injury area and the electromyogram signal around the injury in real time through the electronic sensor array on the electroacupuncture, preprocess the collected signals, and extract the time-frequency characteristics of the signals; An evaluation module, which is used to input the time-frequency characteristics of the signals into an electroacupuncture evaluation model, evaluate the nerve pathway in real time through the electroacupuncture evaluation model, and output an evaluation result; An adjustment module, which is used to adjust the frequency, waveform, and current intensity parameters of the electroacupuncture based on the evaluation result and the Q-learning algorithm, find the optimal electroacupuncture parameter combination, and feedback it to the electroacupuncture.

[0005] Optionally, in the first implementation manner of the present invention, the patient evaluation module includes: A building sub-module is used to abstract different regions of the spinal cord, nerve cells, and connected nerve pathways as nodes in a graph, with the connection relationships between the nodes represented by edges, and to build a graph structure reflecting the physiological structure of the spinal cord; An establishing sub-module is used to convert the basic information of the patient's spinal cord injury and individual physiological characteristic data collected into the feature information of nodes and edges, and input it into the constructed graph structure to establish a personalized spinal cord injury model for the patient based on the graph neural network; An output sub-module is used for the nodes in the graph neural network to exchange information with adjacent nodes through edges. Each node receives the feature information from its neighbor nodes, combines its own features, updates its own state, and outputs the predicted nerve function state and rehabilitation potential indicators to obtain the nerve function evaluation result.

[0006] Optionally, in the second implementation manner of the present invention, the multi-modal stimulation parameters include electrical stimulation, magnetic stimulation, and optical stimulation.

[0007] Optionally, in the third implementation manner of the present invention, the signal processing module includes: A decomposition sub-module is used to decompose the locally recorded field potential in the spinal cord injury area and the electromyogram signals around the injury through wavelet transform into different frequencies and time scales to obtain a series of sub-signals with different frequencies and time resolutions; A squeezing sub-module is used to rearrange the decomposed sub-signals according to their frequency components, and perform a synchrosqueezing operation after rearrangement to highlight the time-varying characteristics of different frequency components in the signal; A first extraction sub-module is used to obtain a time-frequency graph after the synchrosqueezing operation, and extract the time-frequency features of the signal from the time-frequency graph.

[0008] Optionally, in the fourth implementation manner of the present invention, the squeezing sub-module includes: For each frequency point, search for its energy distribution at different scales, and squeeze and concentrate the energy originally scattered on adjacent scales to the true frequency position.

[0009] Optionally, in the fifth implementation manner of the present invention, the evaluation module includes: A characterization sub-module is used to perform normalization processing on the time-frequency features of the signal, and add position encoding information to each feature vector to characterize the position relationship of the signal features in the time series; A second extraction sub-module is used to input the processed time-frequency feature vectors into an electroacupuncture evaluation model. The input feature vectors first enter a Transformer network, and the Transformer network extracts the long-range dependent feature relationships and context information in the signal; A judgment sub-module, which is used to input the feature vectors output by the Transformer network into the CapsNet network, obtain a more hierarchical and structured feature representation through the CapsNet network, and judge the states of various parts in the neural pathway; A mapping sub-module, which is used to organize and map after the CapsNet network evaluates the neural pathway, and output the evaluation result.

[0010] Optionally, in the sixth implementation manner of the present invention, the second extraction sub-module includes: Calculate the attention weights between each feature vector and other feature vectors from multiple different perspectives through the multi-head attention mechanism. According to the calculated attention weights, perform weighted summation on the feature vectors, extract the long-range dependent feature relationships and context information in the signal, perform feature transformation and enhancement on the weighted-summed feature vectors through a feed-forward neural network, increase the non-linear expression ability of the features through non-linear activation function processing, output the processed feature vectors, and transmit them to the CapsNet network.

[0011] Optionally, in the seventh implementation manner of the present invention, the judgment sub-module includes: The CapsNet network divides the input feature vectors into multiple different capsules, each capsule represents the features in the neural pathway, calculates the coupling coefficients between the capsules through a dynamic routing algorithm, and based on the coupling coefficients, performs weighted combination on the outputs of the capsules to obtain a more hierarchical and structured feature representation. According to the weighted combination result of the capsule outputs, judge the states of various parts in the neural pathway to evaluate the integrity and functional state of the neural pathway.

[0012] Optionally, in the eighth implementation manner of the present invention, the adjustment module includes: A division sub-module, which is used to divide the evaluation result output by the electroacupuncture evaluation model into several discrete states in the electroacupuncture treatment scenario; A creation sub-module, which is used to create a two-dimensional table as the Q-table. The rows of the table correspond to different states, and the columns correspond to different actions. Initialize the Q-values of all state-action pairs in the Q-table to 0; A selection sub-module, which is used to view the Q-values of all actions corresponding to the current determined state in the Q-table, and adopt -greedy strategy to select actions; An execution sub-module, which is used to correspondingly adjust the frequency, waveform, and current intensity parameters of the electroacupuncture according to the selected actions, and execute the adjusted electroacupuncture treatment.

[0013] Optionally, in the ninth implementation manner of the present invention, the selection sub-module includes: At Randomly select an action with a probability to explore new parameter adjustment combinations, with a probability of 1 - select the action with the largest Q value with a probability of, where is a parameter between 0 and 1.

[0014] In the technical solution provided by the present invention, basic information on the patient's spinal cord injury and individual physiological characteristic data are obtained, a personalized spinal cord injury model of the patient is established, and the patient's current neurological function status and rehabilitation potential are evaluated to obtain a neurological function evaluation result; according to the patient's personalized spinal cord injury model and the neurological function evaluation result, multi-modal stimulation parameters in electroacupuncture treatment are set; during the treatment process, local field potentials in the spinal cord injury area and electromyography signals around the injury are collected in real time through an electronic sensor array on the electroacupuncture needle, the collected signals are preprocessed, and the time-frequency characteristics of the signals are extracted; the time-frequency characteristics of the signals are input into an electroacupuncture evaluation model, and the neural pathway is evaluated in real time through the electroacupuncture evaluation model, and an evaluation result is output; based on the evaluation result and the Q-learning algorithm, the frequency, waveform, and current intensity parameters of the electroacupuncture are adjusted to find the optimal electroacupuncture parameter combination and feedback it to the electroacupuncture; the present invention integrates three modalities of electrical stimulation, magnetic stimulation, and light stimulation, stimulates nerve cells at the spinal cord injury site from different angles, electrical stimulation can directly excite nerve fibers, magnetic stimulation can penetrate deep into tissues and act on nerve cells, and light stimulation can regulate cell metabolism and promote nerve regeneration. The synergistic effect of multiple stimulation modalities can more comprehensively stimulate the repair potential of nerve cells, improve the treatment effect, effectively extract the characteristics of nerve signals, evaluate the integrity of the neural pathway in real time and accurately, provide a reliable basis for adjusting treatment parameters, and through an adaptive adjustment mechanism, it can automatically find the optimal treatment parameter combination according to the real-time changes in the patient's nerve activity, improve the pertinence and effectiveness of the treatment, realize the personalized customization of the treatment plan, fully consider the differences between individual patients, and improve the pertinence and effectiveness of the treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0016] Figure 1 It is a schematic structural diagram of a multi-modal nerve regulation electroacupuncture control system for spinal cord injury repair provided by an embodiment of the present invention; Figure 2 It is a schematic structural diagram of a patient evaluation module provided by an embodiment of the present invention; Figure 3 It is a schematic structural diagram of a signal processing module provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In the description, claims and the above-mentioned drawings of the present invention, terms such as "first", "second", "third", "fourth", etc. (if any) are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or equipment.

[0018] For ease of understanding, the specific process of the embodiments of the present invention will be described below. Please refer to Figure 1 The structural schematic diagram of the multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair provided by the embodiments of the present invention. The system includes: A patient evaluation module, configured to obtain the basic information of the patient's spinal cord injury and individual physiological characteristic data, establish a personalized spinal cord injury model of the patient, and evaluate the patient's current neurological function status and rehabilitation potential to obtain a neurological function evaluation result; A parameter setting module, configured to set multi-modal stimulation parameters in electroacupuncture treatment according to the personalized spinal cord injury model of the patient and the neurological function evaluation result; A signal processing module, configured to, during the treatment process, collect the local field potential in the spinal cord injury area and the electromyogram signal around the injury in real time through the electronic sensor array on the electroacupuncture, preprocess the collected signals, and extract the time-frequency characteristics of the signals; An evaluation module, configured to input the time-frequency characteristics of the signals into the electroacupuncture evaluation model, evaluate the neural pathway in real time through the electroacupuncture evaluation model, and output an evaluation result; An adjustment module, configured to adjust the frequency, waveform and current intensity parameters of the electroacupuncture based on the evaluation result and the Q-learning algorithm, find the optimal combination of electroacupuncture parameters, and feedback to the electroacupuncture.

[0019] In this embodiment, please refer to Figure 2 , the patient evaluation module includes: A construction sub-module, configured to abstract different regions of the spinal cord, nerve cells and connected neural pathways as nodes in a graph, represent the connection relationship between nodes with edges, and construct a graph structure reflecting the physiological structure of the spinal cord; An establishment sub-module, configured to convert the collected basic information of the patient's spinal cord injury and individual physiological characteristic data into the characteristic information of nodes and edges, input it into the constructed graph structure, and establish a personalized spinal cord injury model of the patient based on the graph neural network; An output sub-module is used for nodes in the graph neural network to exchange information with adjacent nodes through edges. Each node receives feature information from neighbor nodes, combines its own features, updates its own state, outputs the predicted neural function state and rehabilitation potential indicators, and obtains the neural function evaluation result.

[0020] In this embodiment, the multi-modal stimulation parameters include electrical stimulation, magnetic stimulation, and optical stimulation.

[0021] In this embodiment, please refer to Figure 3 , the signal processing module includes: A decomposition sub-module is used to decompose the locally recorded field potential in the spinal cord injury area and the electromyogram signals around the injury through wavelet transform into different frequencies and time scales, and obtain a series of sub-signals with different frequency and time resolutions; A squeezing sub-module is used to rearrange the decomposed sub-signals according to their frequency components, and perform a synchrosqueezing operation after rearrangement to highlight the time-varying characteristics of different frequency components in the signal; A first extraction sub-module is used to obtain a time-frequency map after the synchrosqueezing operation, and extract the time-frequency features of the signal from the time-frequency map.

[0022] In this embodiment, the squeezing sub-module includes: for each frequency point, searching for its energy distribution at different scales, and squeezing and concentrating the energy originally scattered on adjacent scales to the true frequency position.

[0023] In this embodiment, the evaluation module includes: A characterization sub-module is used to normalize the time-frequency features of the signal and add position encoding information to each feature vector to characterize the position relationship of the signal features in the time series; A second extraction sub-module is used to input the processed time-frequency feature vectors into the electroacupuncture evaluation model. The input feature vectors first enter the Transformer network, and the Transformer network extracts the long-range dependent feature relationships and context information in the signal; A judgment sub-module is used to enter the feature vectors output from the Transformer network into the CapsNet network, and obtain a more hierarchical and structured feature representation through the CapsNet network to judge the states of various parts in the neural pathway; A mapping sub-module is used to sort and map after the evaluation of the neural pathway by the CapsNet network, and output the evaluation result.

[0024] In this embodiment, the second extraction sub-module includes: calculating the attention weights between each feature vector and other feature vectors from multiple different perspectives through a multi-head attention mechanism, performing weighted summation on the feature vectors according to the calculated attention weights, extracting the long-range dependent feature relationships and context information in the signal, performing feature transformation and enhancement on the weighted-summed feature vectors through a feed-forward neural network, increasing the non-linear expression ability of the features through non-linear activation function processing, and outputting the processed feature vectors and transmitting them to the CapsNet network.

[0025] In this embodiment, the judgment sub-module includes: The CapsNet network divides the input feature vectors into multiple different capsules, each capsule representing the features in the neural pathway, calculates the coupling coefficients between the capsules through a dynamic routing algorithm, performs weighted combination on the outputs of the capsules based on the coupling coefficients to obtain a more hierarchical and structured feature representation, and judges the states of each part in the neural pathway according to the weighted combination result of the capsule outputs to evaluate the integrity and functional state of the neural pathway. In this embodiment, the adjustment module includes: A division sub-module, used to divide the evaluation result output by the electroacupuncture evaluation model into several discrete states in the electroacupuncture treatment scenario; A creation sub-module, used to create a two-dimensional table as the Q-table, where the rows of the table correspond to different states and the columns correspond to different actions, and initialize the Q-values of all state-action pairs in the Q-table to 0; A selection sub-module, used to view the Q-values of all actions corresponding to the current determined state in the Q-table, and adopt -greedy strategy to select an action; An execution sub-module, used to correspondingly adjust the frequency, waveform, and current intensity parameters of the electroacupuncture according to the selected action, and perform the adjusted electroacupuncture treatment.

[0026] In this embodiment, the selection sub-module includes: with probability, randomly select an action to explore new parameter adjustment combinations, and with a probability of 1 - select the action with the largest Q-value, where is a parameter between 0 and 1.

[0027] In some embodiments, the patient's file is retrieved through the hospital electronic medical record system to obtain the patient's individual physiological characteristic data such as age, gender, and contact information, and at the same time understand the basic information such as the cause of injury, injury site, injury degree, and injury course, and combine clinical physical examinations to determine whether it is a complete injury or an incomplete injury.

[0028] In some embodiments, the multi-modal stimulation includes electrical stimulation, magnetic stimulation, and optical stimulation, where: Electrical stimulation parameters: Initially determine that the frequency range of electrical stimulation is 2 - 100 Hz, the waveform switches between square wave and triangular wave, and the current intensity range is 0.1 - 5 mA. For different spinal cord injury stages and symptoms, determine the initial electrical stimulation parameters to promote the recovery of nerve cell activity and enhance the muscle contraction ability; Magnetic stimulation parameters: Set the magnetic field intensity, pulse frequency, and stimulation time of magnetic stimulation. According to the depth of the spinal cord injury site in the patient and the distribution of nerve fibers, adjust the action range and intensity of magnetic stimulation to effectively penetrate tissues and act on the nerve cells at the injury site; Optical stimulation parameters: Determine the optical stimulation wavelength, light intensity, and irradiation time. Red light mainly promotes cell metabolism and blood circulation, and near-infrared light can penetrate deeper tissues to stimulate the regeneration and repair of nerve cells.

[0029] In some embodiments, during the treatment process, a polyimide-based graphene electrode with an impedance <1 kΩ @ 1 kHz is used, integrated on the surface of the acupuncture needle, to collect the local field potential in the spinal cord injury area and the electromyogram signal around the injury in real time. Through short-time Fourier transform and deep learning model, extract the beta oscillation energy characteristics of the corticospinal tract in the local field potential of the spinal cord injury area, correlate the activation patterns of motor units in the electromyogram signal around the injury, and evaluate the integrity of the nerve pathway in real time.

[0030] In some embodiments, a polydopamine coating is micro-nano processed on the electrode surface, loaded with BDNF / NGF nanoparticles, and triggered by electrical stimulation to release through charge inversion.

[0031] In some embodiments, the specific process of the signal processing module using wavelet transform to extract the time-frequency characteristics of the signal includes: Initialize the parameters required for adaptive synchrosqueezing wavelet transform, and input the raw local field potential in the spinal cord injury area and the electromyogram signal around the injury collected in real time into the adaptive synchrosqueezing wavelet transform algorithm; Use wavelet basis functions to perform multi-scale wavelet decomposition on the original signal. During the decomposition process, through convolution operations between wavelet functions of different scales and the original signal, expand the signal in both time and frequency dimensions to obtain a preliminary time-frequency representation; The original signal is decomposed into sub-signals at multiple different scales. For different types of signals and differences between individual patients, an adaptive mechanism is used to adjust the scale of wavelet decomposition Analyze the statistical characteristics of the original signal to capture low-frequency signal characteristics; For each frequency point, search for its energy distribution at different scales, squeeze and concentrate the energy scattered on adjacent scales towards the real frequency position to obtain a clear and energy-concentrated time-frequency diagram, and extract the time-frequency characteristics of the signal from the time-frequency diagram.

[0032] In some embodiments, the synchrosqueezing wavelet transform redistributes the energy according to the magnitude of the modulus of each element in the time-scale plane, and then converts the time-scale domain to the time-frequency domain through a mapping relationship, that is, the wavelet coefficients are converted from the time-scale plane to the time-frequency plane, and then squeezed along the frequency direction to obtain the focused signal time-frequency spectrogram.

[0033] In some embodiments, the local field potential signals in the spinal cord injury area extract the energy distribution characteristics of the corticospinal tract-related frequency bands, the changing trend of frequency over time, etc.; the electromyography signals around the injury extract the frequency components related to the activation of motor units, the time-frequency patterns under different muscle contraction states, and other characteristics.

[0034] In some embodiments, the process of the electroacupuncture evaluation model includes: The Transformer network calculates the attention weights between each feature vector and other feature vectors from multiple different perspectives through the multi-head attention mechanism, and reflects the correlation degree between features at different times and different frequencies through the attention weights; According to the calculated attention weights, weighted summation is performed on the feature vectors to extract the long-range dependent feature relationships and context information in the signal; The weighted-summed feature vectors are further subjected to feature transformation and enhancement through a feedforward neural network, processed by a non-linear activation function in the network to increase the non-linear expression ability of the features, and the processed feature vectors are output and transmitted to the CapsNet network; The CapsNet network divides the output of the Transformer network into multiple different capsules, and each capsule represents a specific type of feature in the neural pathway; Through the dynamic routing algorithm, the coupling coefficients between the capsules are calculated, and the correlation strength between the features represented by different capsules is reflected through the coupling coefficients; Based on the coupling coefficients, weighted combination of the outputs of each capsule is performed to obtain a more hierarchical and structured feature representation; According to the weighted combination result of the capsule output, the states of each part in the neural pathway are judged, and the integrity and functional state of the neural pathway are evaluated After the evaluation of the neural pathway by the CapsNet network, the evaluation results are sorted and mapped, and the evaluation results are output.

[0035] In some embodiments, the CapsNet network uses a set of vector neurons to replace the scalar neuron nodes in traditional neural networks, changing the structure of the traditional neural network where scalars are connected to scalars. In each network layer, the information carried by each capsule increases from one-dimensional to multi-dimensional. Through the dynamic routing mechanism, the calculation results saved by the lower-layer capsules are passed to the upper-layer capsules, thereby reducing information loss while extracting local fault features.

[0036] In some embodiments, the process of the dynamic routing algorithm includes: first multiplying the output vector of the lower-layer capsules by the corresponding weight matrix to obtain a new vector, linearly combining the new vector and the coupling coefficient generated during the dynamic routing process, further correcting the total output vector of the lower-layer capsules for a non-linear mapping, so as to obtain the final output vector. The non-linear mapping function uses the Squashing activation function. Since the norm of the capsule output vector represents the probability value of the category, the Squashing activation function limits the norm of the output vector within [0,1]. The larger the norm of the output vector, the greater the probability. The coupling coefficient represents the connection probability between two layers of capsules during the iteration process.

[0037] In some embodiments, the evaluation results include indicators such as the conduction speed change of the neural pathway, the number of neuron activations, and the synaptic connection strength, which are displayed to the treating doctor in the form of a visualization interface in real time, facilitating the doctor to timely understand the dynamic changes of the patient's nerve function.

[0038] In some embodiments, in the electroacupuncture treatment scenario, based on the evaluation results and the Q-learning algorithm, the frequency, waveform, and current intensity parameters of the electroacupuncture are adjusted to find the optimal combination of electroacupuncture parameters. The specific process includes: In some embodiments, the state of the Q-learning algorithm consists of the evaluation results and the current stimulation parameters. The continuous state space is discretized. For example, the integrity of the neural pathway is divided into five intervals, the frequency is divided into three intervals: low, medium, and high, the waveform is divided into four types: sine wave, square wave, pulse wave, and triangular wave, and the current intensity is divided into three intervals: low, medium, and high, forming a discrete state space; The action is defined as the adjustment method of the stimulation parameters, including frequency adjustment, waveform switching, and current intensity adjustment. Each action corresponds to a specific parameter change; Set the reward function based on the CST coherence index of the local field potential in the spinal cord injury area and the co-contraction rate of the electromyographic signals around the injury; Create a two-dimensional table as the Q-table. The rows of the table correspond to different states, and the columns correspond to different actions; Initialize the Q values of all state-action pairs in the Q-table to 0, where the Q value represents the expected long-term cumulative reward obtained after performing a certain action in a certain state; Update the Q-values in the Q-table to find the optimal action in each state. Using the currently known optimal policy, select the action with the maximum Q-value with a probability of 1 - Select the action with the maximum Q-value with a probability of, select the optimal action according to the Q-value table, adjust the frequency, waveform, and current intensity parameters of the electroacupuncture, and feedback the adjusted parameters to the electroacupuncture device to achieve real-time optimization of the treatment parameters.

[0039] In some embodiments, the Q-learning algorithm selects the optimal action in the current state through -greedy policy, that is, select the action with the maximum Q-value. To explore unknown states and actions, Q-learning selects non-optimal actions with a certain probability for exploration strategy.

[0040] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multimodal neuromodulation electroacupuncture control system for spinal cord injury repair, characterized in that, The system includes: A patient evaluation module, which is used to obtain the basic information of the patient's spinal cord injury and individual physiological characteristic data, establish a personalized spinal cord injury model for the patient, and evaluate the patient's current neurological function status and rehabilitation potential to obtain a neurological function evaluation result; A parameter setting module, which is used to set multi-modal stimulation parameters in electroacupuncture treatment according to the patient's personalized spinal cord injury model and neurological function evaluation result; A signal processing module, which is used to collect the local field potential in the spinal cord injury area and the electromyogram signal around the injury in real time through the electronic sensor array on the electroacupuncture needle during the treatment process, preprocess the collected signals, and extract the time-frequency characteristics of the signals; An evaluation module, which is used to input the time-frequency characteristics of the signals into the electroacupuncture evaluation model, evaluate the neural pathway in real time through the electroacupuncture evaluation model, and output an evaluation result; An adjustment module, which is used to adjust the frequency, waveform and current intensity parameters of the electroacupuncture based on the evaluation result and the Q-learning algorithm, find the optimal combination of electroacupuncture parameters, and feedback to the electroacupuncture.

2. The multimodal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, wherein The patient evaluation module includes: A construction sub-module, which is used to abstract different regions of the spinal cord, nerve cells and connected neural pathways as nodes in a graph, represent the connection relationship between nodes with edges, and construct a graph structure reflecting the physiological structure of the spinal cord; An establishment sub-module, which is used to convert the collected basic information of the patient's spinal cord injury and individual physiological characteristic data into the characteristic information of nodes and edges, input it into the constructed graph structure, and establish a personalized spinal cord injury model for the patient based on the graph neural network; An output sub-module, which is used for the nodes in the graph neural network to exchange information with adjacent nodes through edges, each node receives the characteristic information from neighboring nodes, combines its own characteristics, updates its own state, and outputs the predicted neurological function status and rehabilitation potential indicators to obtain a neurological function evaluation result.

3. The multimodal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, wherein The multi-modal stimulation parameters include electrical stimulation, magnetic stimulation and optical stimulation.

4. The multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, characterized in that, The signal processing module includes: A decomposition sub-module, which is used to decompose the collected local field potential in the spinal cord injury area and the electromyogram signal around the injury to different frequencies and time scales through wavelet transform to obtain a series of sub-signals with different frequencies and time resolutions; A squeezing sub-module, which is used to rearrange the decomposed sub-signals according to their frequency components, and perform a synchrosqueezing operation after rearrangement to highlight the time-varying characteristics of different frequency components in the signal; A first extraction sub-module, which is used to obtain a time-frequency diagram after the synchrosqueezing operation, and extract the time-frequency characteristics of the signal from the time-frequency diagram.

5. The multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 4, wherein, The squeezing sub-module includes: For each frequency point, search for its energy distribution at different scales, and squeeze and concentrate the energy originally scattered at adjacent scales to the true frequency position.

6. The multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, characterized in that, The evaluation module includes: A characterization sub-module, which is used to normalize the time-frequency characteristics of the signal and add position encoding information to each feature vector to characterize the position relationship of the signal features in the time series; The second extraction sub-module is used to input the processed time-frequency feature vector into the electroacupuncture evaluation model. The input feature vector first enters the Transformer network, and the Transformer network extracts the feature relationships and context information of long-distance dependencies in the signal; The judgment sub-module is used to input the feature vector output by the Transformer network into the CapsNet network. Through the CapsNet network, a more hierarchical and structured feature representation is obtained to judge the states of various parts of the neural pathway; The mapping sub-module is used to organize and map after the evaluation of the neural pathway by the CapsNet network, and output the evaluation result.

7. The multi-modal neuromodulation electro-acupuncture control system for spinal cord injury repair according to claim 6, wherein, The second extraction sub-module includes: Calculate the attention weights between each feature vector and other feature vectors from multiple different perspectives through the multi-head attention mechanism. According to the calculated attention weights, perform weighted summation on the feature vectors to extract the feature relationships and context information of long-distance dependencies in the signal. Pass the weighted-summed feature vector through a feed-forward neural network for feature transformation and enhancement, and process it through a non-linear activation function to increase the non-linear expression ability of the features. Output the processed feature vector and transmit it to the CapsNet network.

8. The multi-modal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, characterized in that, The judgment sub-module includes: The CapsNet network divides the input feature vector into multiple different capsules, and each capsule represents the features in the neural pathway. Calculate the coupling coefficients between each capsule through the dynamic routing algorithm, and based on the coupling coefficients, perform weighted combination on the outputs of each capsule to obtain a more hierarchical and structured feature representation. According to the weighted combination result of the capsule output, judge the states of various parts of the neural pathway to evaluate the integrity and functional state of the neural pathway.

9. The multimodal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 1, wherein The adjustment module includes: The division sub-module is used to divide the evaluation result output by the electroacupuncture evaluation model into several discrete states in the electroacupuncture treatment scenario; The creation sub-module is used to create a two-dimensional table as the Q-table. The rows of the table correspond to different states, and the columns correspond to different actions. Initialize the Q-values of all state-action pairs in the Q-table to 0; A selection sub-module, which is used to view the Q-values of all actions corresponding to the current state in the Q-table and select an action using the -greedy strategy; The execution sub-module is used to correspondingly adjust the frequency, waveform, and current intensity parameters of the electroacupuncture according to the selected action, and perform the adjusted electroacupuncture treatment.

10. The multimodal neuromodulation electroacupuncture control system for spinal cord injury repair according to claim 9, wherein, The selection sub-module includes: Randomly select an action with a probability of to explore new parameter adjustment combinations, and select the action with the largest Q value with a probability of 1 - , where is a parameter between 0 and 1.

Images