Sympathetic cervical spondylosis autonomic nerve regulation enhancing method based on brain-computer interface

Abstract

The invention discloses a sympathetic cervical spondylosis autonomic nerve regulation enhancing method based on a brain-computer interface, and relates to the technical field of nerve regulation and repair. According to the method, by synchronously collecting electroencephalogram and skin conductance signals and conducting time-frequency domain conjoint analysis and Lempel-Ziv complexity evaluation, accurate quantitative evaluation on sympathetic nerve activity and nerve disorder degree is achieved, double-model collaborative decision-making of machine learning and deep reinforcement learning is utilized, personalized nerve regulation parameters are intelligently generated, and the method has the advantages of being high in accuracy and high in accuracy. High-targeting and safe intervention is implemented through double-target hierarchical regulation and control aiming at neck sympathetic nerves and vagus nerves, and stimulation parameters are dynamically optimized in a closed-loop manner by adopting a genetic algorithm on the basis of real-time feedback of heart rate variability, skin conductance and autonomic nerve balance indexes, so that the stimulation accuracy is improved. Therefore, the sympathetic cervical spondylosis autonomic nerve regulation efficiency and effect are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of neural regulation and repair, and specifically to a method for enhancing autonomic nerve regulation of sympathetic cervical spondylosis based on a brain-computer interface. Background Art

[0002] A brain-computer interface refers to a direct connection created between the human or animal brain and an external device to enable information exchange between the brain and the device. It can be used to restore damaged hearing, vision, and limb movement abilities, and can be used for neural repair, that is, using artificial devices to replace some nerves or sensory organs whose original functions have been weakened.

[0003] The defects of the existing autonomic nerve regulation method for sympathetic cervical spondylosis are:

[0004] 1. Patent document US20140194726A1 discloses ultrasonic neuromodulation for cognitive enhancement. However, the neuromodulation device described in the aforementioned document has difficulty synchronously acquiring high signal-to-noise ratio EEG and skin conductance signals, resulting in large errors in the assessment of sympathetic nerve activity.

[0005] 2. Patent document JP2014074055A discloses a composition for regulating autonomic nervous activity and a method for regulating autonomic nerves. However, the method for regulating autonomic nerves in the above document has a technical problem of low individualized accuracy;

[0006] 3. Patent document JP2011125674A discloses a brain-computer interface, but the brain-computer interface in the above document has a technical problem of being unable to quickly intervene when abnormal physiological responses occur;

[0007] 4. Patent document CN118797560A discloses an autonomic nervous system control method and related equipment based on a brain-computer interface. However, the nervous system control method in the above document has the technical problem that long-term attachment can easily cause skin irritation. Summary of the Invention

[0008] The purpose of the present invention is to provide a method for enhancing autonomic nervous system regulation of sympathetic cervical spondylosis based on a brain-computer interface to solve the technical problems raised in the above-mentioned background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for enhancing autonomic nervous system regulation for sympathetic cervical spondylosis based on a brain-computer interface. The steps for enhancing autonomic nervous system regulation for sympathetic cervical spondylosis based on a brain-computer interface are as follows:

[0010] S1. EEG signal acquisition: EEG and skin conductance (GSR) are collected synchronously using a flexible electrode array, with electrodes covering the frontal (FP1 / FP2), parietal (C3 / C4), and occipital (O1 / O2) brain regions.

[0011] S2. Joint analysis of autonomic nervous system features: Joint analysis of EEG in the time and frequency domains was performed, and wavelet transform was used to extract characteristic frequency band signals related to autonomic nervous system activity, including the instantaneous energy of theta waves (4-8 Hz), alpha waves (8-13 Hz), and high-frequency beta waves (18-30 Hz).

[0012] The power spectral density was calculated to determine the dominant frequency band, and the disorder of neural activity was evaluated based on the Lempel-Ziv complexity (LZc);

[0013] S3, dual-model collaborative decision-making: the theta wave to alpha wave power ratio, beta wave peak frequency, and prefrontal EEG lateralization index (Asym) are input into the machine learning model, which outputs the sympathetic nerve activity score (SNS Index). When the SNS Index exceeds the preset activation threshold, the deep reinforcement learning model is activated to generate targeted neuromodulation parameters.

[0014] S4, dual-target neuromodulation: low-frequency biphasic pulses of 1-10 Hz and 100-200 μs are applied to the cervical sympathetic nerves near the C5-T1 cervical spine;

[0015] For the vagus nerve: transcutaneous stimulation (taVNS) is applied to the concha area. The stimulation pattern is graded according to the SNS Index:

[0016] Mild SNS Index = 1.2-1.5, 1-5 Hz continuous wave;

[0017] Moderate SNS Index = 1.5-2.0, 10-15 Hz pulse train;

[0018] Severe SNS Index>2.0, 25Hz+variable frequency pulse train;

[0019] S5. Closed-loop dynamic optimization: Real-time monitoring of the LF / HF ratio of heart rate variability (HRV), GSR, and blood pressure fluctuations. If the autonomic balance index (ANB = HF / LF × GSR stability coefficient) does not reach 0.8-1.2 within 10 minutes, a genetic algorithm is activated to optimize the stimulation duty cycle. The stimulation intensity is based on 0.5-1 mA and is adjusted by ±0.2 mA every 30 seconds according to the decrease in LF / HF.

[0020] Preferably, the flexible electrode array in S1 includes:

[0021] Contact layer: conductive hydrogel doped with nanodiamonds, conductivity ≥ 100S / m;

[0022] Impedance matching layer: silver fiber mesh, density 80-120 threads / cm 2 , thickness ≤ 50 μm;

[0023] Base layer: Polyimide flexible substrate, Young's modulus <2GPa, its surface impedance is stable at 5~10kΩ in the frequency band of 10~100Hz.

[0024] Preferably, the calculation formula of the SNS Index in S3 is:

[0025]

[0026] The coefficients k1, k2, and k3 are the dynamic weights of the physiological baseline determined during the initial calibration phase of the user, P θ is the integral value of the power spectrum density in the θ frequency band of 4 to 8 Hz, P α is the integrated value of the power spectrum density in the α band from 8 to 13 Hz, f β is the integral value of the power spectrum density in the β band from 18 to 30 Hz, Asym left Asym is the power value of α wave 8~13Hz at the left frontal electrode FP1. right is the power value of α wave 8-13 Hz at the right frontal electrode FP2;

[0027] The weight update formula of coefficients k1, k2 and k3 is:

[0028]

[0029] in is the weight coefficient of the current treatment cycle (i=1, 2, 3 corresponds to k1, k2 and k3 in the formula), is the weight coefficient of the previous treatment cycle, η is the learning rate, is the CSSS drop rate with respect to weight k i , CSSS is the cervical sympathetic nerve symptom score, and the CSSS decrease rate is the rate of change of the score before and after treatment.

[0030] Preferably, the machine learning model in S3 adopts an integrated learning architecture, including:

[0031] First-level classifier: XGBoost model processes theta / alpha power ratio and beta peak frequency;

[0032] Second-level regressor: Support Vector Regression (SVR) processes the Asym index;

[0033] Dynamic weighting module: A three-layer fully connected neural network generates weight coefficients k1, k2, and k3, and the input is physiological baseline parameters.

[0034] Preferably, the deep reinforcement learning model in S3 adopts a dual-stream network architecture, including:

[0035] State stream input: real-time SNS Index, historical treatment data, and individualized tolerance threshold;

[0036] Action flow output: Q matrix in three-dimensional parameter space.

[0037] Preferably, the Q matrix of the three-dimensional parameter space includes:

[0038] Dimension 1: Neck stimulation intensity 0.5-2.0 mA, resolution 0.1 mA;

[0039] Dimension 2: taVNS carrier frequency 1 to 30 Hz, resolution 1 Hz;

[0040] Dimension 3: dual-target phase difference 0 to π radians, resolution π / 8;

[0041] The double-delayed deep deterministic policy gradient (TD3) algorithm is used to update the Q matrix, and the target network is updated every 200 steps.

[0042] Preferably, the dual-target stimulation of S4 adopts a spatiotemporal synergistic protocol:

[0043] Neck stimulation was initiated first, and taVNS was superimposed if the LF / HF ratio did not decrease by 10% within 5 minutes.

[0044] The phase difference between the two target stimulations was fixed at π / 2 radians;

[0045] The pulse width of neck stimulation is 1.5 to 2 times that of taVNS.

[0046] Preferably, the spatiotemporal coordination protocol includes a conflict resolution mechanism:

[0047] When the LF / HF ratio increases by ≥5% within 30 seconds after superimposition of taVNS, a three-level response is initiated immediately:

[0048] First-level response: cut off neck stimulation and maintain taVNS at 5 Hz continuous wave;

[0049] Secondary response: Apply potassium chloride gel to reduce the electrode-skin interface impedance by ≥40%;

[0050] Level 3 response: If the RMSSD of HRV is continuously less than 20ms, switch to the backup sine wave stimulation mode.

[0051] Preferably, the genetic algorithm optimization in step S5 adopts an elite retention strategy, including:

[0052] Chromosome encoding: 8-bit binary string represents 4 bits of stimulus intensity and 4 bits of duty cycle;

[0053] Fitness function:

[0054]

[0055] Where Fitness is the fitness score of the chromosome in the genetic algorithm, △ANB is the absolute deviation between the measured ANB value and the target ANB value, and ANB is the autonomic nervous balance index. It is the baseline mean of the LF / HF ratio within 5 minutes before treatment;

[0056] Selection strategy: The elite retains the top 10% of individuals and directly enters the next generation.

[0057] Preferably, the method further comprises an efficacy verification system, comprising:

[0058] Short-term verification indicators: LF / HF ratio decrease ≥ 15% within 10 minutes after the end of treatment, GSR rising slope ≤ 0.05μS / s;

[0059] Long-term validation indicator: Cervical sympathetic symptom score decreases by >30% after 5 consecutive days of treatment;

[0060] Retraining trigger mechanism: When long-term indicators are not met, expand user data and fine-tune SVR parameters:

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. This invention achieves precise quantitative assessment of sympathetic nerve activity and neurological disorder by synchronously collecting EEG and skin conductance signals and performing joint time-frequency domain analysis and Lempel-Ziv complexity assessment. It utilizes a dual-model collaborative decision-making system of machine learning and deep reinforcement learning to intelligently generate personalized neuromodulation parameters. Through hierarchical regulation of the dual targets of the cervical sympathetic and vagus nerves, highly targeted and safe intervention is implemented. Based on real-time feedback from heart rate variability, skin conductance, and the autonomic balance index, a genetic algorithm is used to dynamically optimize stimulation parameters in a closed loop, thereby enhancing the efficiency and effectiveness of autonomic nerve regulation in sympathetic cervical spondylosis.

[0063] 2. The present invention utilizes a nanodiamond-doped conductive hydrogel in the contact layer, significantly improving signal conduction efficiency while ensuring high biocompatibility and eliminating the risk of skin irritation from long-term application. The impedance matching layer utilizes an ultrathin silver fiber mesh to construct a microcurrent voltage-sharing network, effectively suppressing motion artifacts and stabilizing the electrode-skin interface impedance, ensuring signal integrity in hair-bearing areas such as the frontal lobe. The base layer, based on a flexible polyimide substrate, conforms to the physiological curvature of the neck and reduces signal drift caused by changes in body position. These three elements work together to overcome the challenge of signal fidelity in high-noise environments, providing a reliable hardware foundation for the simultaneous acquisition of high-precision EEG and GSR data, significantly improving the accuracy of sympathetic nerve activity assessment.

[0064] 3. This invention constructs a personalized sympathetic nerve activity assessment model. Its SNS Index calculation formula is based on a three-layer neural network that dynamically updates weight coefficients. A weight learning formula is used to integrate the user's physiological baseline and the rate of change of cervical sympathetic nerve symptom scores in real time, significantly improving individualized quantification accuracy. Secondly, a multimodal integration architecture is adopted. The XGBoost classifier analyzes the θ / α power ratio and β peak frequency, and the SVR regressor processes the prefrontal lateralization index (Asym). A dynamic weighting module collaboratively generates weight coefficients, significantly enhancing the decoding robustness of complex neural states. Finally, a reinforcement learning optimization engine is deployed. A dual-stream network integrates real-time treatment data and individual tolerance thresholds. In the three-dimensional Q matrix parameter space, the TD3 algorithm is used to update the target network strategy every 200 steps, achieving milliampere-level precision control and avoiding the risk of overfitting.

[0065] 4. The present invention uses a stepped stimulation sequence to prioritize neck stimulation and superimpose taVNS only when the LF / HF ratio has not dropped by 10% within 5 minutes, thus avoiding the risk of simultaneous activation of dual targets. A fixed π / 2 radian phase difference is used to enhance neural rhythm coordination, and the neck stimulation pulse width is set to 1.5 to 2 times that of taVNS to match tissue conductivity characteristics. A 30-second three-level response chain is established: if the LF / HF ratio rises by ≥5%, neck stimulation is immediately cut off and taVNS is maintained at a continuous wave of 5 Hz. Potassium chloride gel is applied to ensure that the electrode-skin impedance drops by ≥40%. When the RMSSD lasts for less than 20 ms, the sine wave mode is switched, thereby significantly improving the treatment efficiency and safety of intractable hypersympathetic syndrome. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic diagram of the overall process structure of the present invention;

[0067] Figure 2 This is a schematic diagram of the autonomic nervous system feature analysis process structure of the present invention;

[0068] Figure 3 This is a schematic diagram of the SNS Index calculation process structure of the present invention;

[0069] Figure 4 Schematic diagram of the spatiotemporal co-stimulation protocol structure of the present invention;

[0070] Figure 5 This is a schematic diagram of the genetic algorithm optimization process structure of the present invention;

[0071] Figure 6 Schematic diagram of the process structure of the efficacy verification system of the present invention. DETAILED DESCRIPTION

[0072] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0073] Example 1: Please refer to Figure 1 and Figure 2 The present invention provides an embodiment of a method for enhancing autonomic nervous system regulation of sympathetic cervical spondylosis based on a brain-computer interface. The steps for enhancing autonomic nervous system regulation of sympathetic cervical spondylosis based on a brain-computer interface are as follows:

[0074] S1. EEG signal acquisition: EEG and skin conductance (GSR) are collected synchronously using a flexible electrode array, with electrodes covering the frontal (FP1 / FP2), parietal (C3 / C4), and occipital (O1 / O2) brain regions.

[0075] S2. Joint analysis of autonomic nervous system features: Joint analysis of EEG in the time and frequency domains was performed, and wavelet transform was used to extract characteristic frequency band signals related to autonomic nervous system activity, including the instantaneous energy of theta waves (4-8 Hz), alpha waves (8-13 Hz), and high-frequency beta waves (18-30 Hz).

[0076] The power spectral density was calculated to determine the dominant frequency band, and the disorder of neural activity was evaluated based on the Lempel-Ziv complexity (LZc);

[0077] S3, dual-model collaborative decision-making: the theta wave to alpha wave power ratio, beta wave peak frequency, and prefrontal EEG lateralization index (Asym) are input into the machine learning model, which outputs the sympathetic nerve activity score (SNS Index). When the SNS Index exceeds the preset activation threshold, the deep reinforcement learning model is activated to generate targeted neuromodulation parameters.

[0078] S4, dual-target neuromodulation: low-frequency biphasic pulses of 1-10 Hz and 100-200 μs are applied to the cervical sympathetic nerves near the C5-T1 cervical vertebrae;

[0079] For the vagus nerve: transcutaneous stimulation (taVNS) is applied to the concha area. The stimulation pattern is graded according to the SNS Index:

[0080] Mild SNS Index = 1.2-1.5, 1-5 Hz continuous wave;

[0081] Moderate SNS Index = 1.5-2.0, 10-15 Hz pulse train;

[0082] Severe SNS Index>2.0, 25Hz+variable frequency pulse train;

[0083] S5. Closed-loop dynamic optimization: Real-time monitoring of the LF / HF ratio of heart rate variability (HRV), GSR, and blood pressure fluctuations. If the autonomic balance index (ANB = HF / LF × GSR stability coefficient) does not reach 0.8-1.2 within 10 minutes, the genetic algorithm is activated to optimize the stimulation duty cycle. The stimulation intensity is based on 0.5-1 mA and is adjusted by ±0.2 mA every 30 seconds according to the decrease in LF / HF.

[0084] Furthermore, by synchronously collecting EEG and skin conductance signals and conducting joint time-frequency domain analysis and Lempel-Ziv complexity evaluation, accurate quantitative assessment of sympathetic nerve activity and neurological disorder was achieved. By utilizing the dual-model collaborative decision-making of machine learning and deep reinforcement learning, personalized neuroregulation parameters were intelligently generated. Through dual-target graded regulation of the cervical sympathetic nerves and vagus nerves, highly targeted and safe intervention was implemented. Based on real-time feedback of heart rate variability, skin conductance and autonomic balance index, a genetic algorithm was used to dynamically optimize the stimulation parameters in a closed loop, thus constructing a precise, intelligent, adaptive, closed-loop autonomic nervous function regulation system, thereby significantly enhancing the regulation efficiency and effect of autonomic nervous function disorder in sympathetic cervical spondylosis.

[0085] Example 2: Please refer to Figure 1 , an embodiment provided by the present invention: the flexible electrode array in S1 includes:

[0086] Contact layer: conductive hydrogel doped with nanodiamonds, conductivity ≥ 100S / m;

[0087] Impedance matching layer: silver fiber mesh, density 80-120 threads / cm 2 , thickness ≤ 50 μm;

[0088] Base layer: Polyimide flexible substrate, Young's modulus <2GPa, its surface impedance is stable at 5~10kΩ in the frequency band of 10~100Hz;

[0089] Furthermore, by using a conductive hydrogel doped with nanodiamonds in the contact layer, the signal conduction efficiency is significantly improved while ensuring high biocompatibility, eliminating the risk of skin irritation caused by long-term attachment. The impedance matching layer constructs a microcurrent voltage-equalizing network with an ultra-thin silver fiber grid, which effectively suppresses motion artifacts and stabilizes the electrode-skin interface impedance, ensuring signal integrity in hair areas such as the frontal lobe. The base layer is based on a polyimide flexible substrate, which fits the physiological curvature of the neck and reduces signal drift caused by changes in body position. The three work together to overcome the problem of signal fidelity in high-noise environments, providing a reliable hardware foundation for the simultaneous acquisition of high-precision EEG and GSR data, and significantly improving the accuracy of sympathetic nerve activity assessment.

[0090] Example 3: Please refer to Figure 3 , an embodiment provided by the present invention: the calculation formula of SNS Index in S3 is:

[0091]

[0092] The coefficients k1, k2, and k3 are the dynamic weights of the physiological baseline determined during the initial calibration phase of the user, P θ is the integral value of the power spectrum density in the θ frequency band of 4 to 8 Hz, P α is the integrated value of the power spectrum density in the α band from 8 to 13 Hz, f β is the integral value of the power spectrum density in the β band from 18 to 30 Hz, Asym left Asym is the power value of α wave 8~13Hz at the left frontal electrode FP1. right is the power value of α wave 8-13 Hz at the right frontal electrode FP2;

[0093] The weight update formula of coefficients k1, k2 and k3 is:

[0094]

[0095] in is the weight coefficient of the current treatment cycle (i=1, 2, 3 corresponds to k1, k2 and k3 in the formula), is the weight coefficient of the previous treatment cycle, η is the learning rate, is the CSSS drop rate with respect to weight k i The partial derivative of , CSSS is the cervical sympathetic nerve symptom score, and the CSSS decrease rate is the rate of change of the score before and after treatment;

[0096] The machine learning model in S3 uses an ensemble learning architecture, including:

[0097] First-level classifier: XGBoost model processes theta / alpha power ratio and beta peak frequency;

[0098] Second-level regressor: Support Vector Regression (SVR) processes the Asym index;

[0099] Dynamic weighting module: A three-layer fully connected neural network generates weight coefficients k1, k2, and k3, with physiological baseline parameters as input;

[0100] The deep reinforcement learning model in S3 uses a two-stream network architecture, including:

[0101] State stream input: real-time SNS Index, historical treatment data, and individualized tolerance threshold;

[0102] Action flow output: Q matrix in three-dimensional parameter space;

[0103] The Q matrix of the three-dimensional parameter space includes:

[0104] Dimension 1: Neck stimulation intensity 0.5-2.0 mA, resolution 0.1 mA;

[0105] Dimension 2: taVNS carrier frequency 1 to 30 Hz, resolution 1 Hz;

[0106] Dimension 3: dual-target phase difference 0 to π radians, resolution π / 8;

[0107] The double-delayed deep deterministic policy gradient (TD3) algorithm is used to update the Q matrix, and the target network is updated every 200 steps;

[0108] Furthermore, a personalized sympathetic nerve activity assessment model was first constructed. Its SNS Index calculation formula is based on a three-layer neural network that dynamically updates weight coefficients. The user's physiological baseline and the rate of change of cervical sympathetic nerve symptom scores are integrated in real time through the weight learning formula, significantly improving the individualized quantification accuracy. Secondly, a multimodal integration architecture is adopted. The θ / α power ratio and β peak frequency are analyzed by the XGBoost classifier, and the prefrontal lobe lateralization index (Asym) is processed by the SVR regressor. The dynamic weighting module collaboratively generates weight coefficients, greatly enhancing the decoding robustness of complex neural states. Finally, a reinforcement learning optimization engine is deployed, and a dual-stream network is used to integrate real-time treatment data and individual tolerance thresholds. In the three-dimensional Q matrix parameter space, the TD3 algorithm is used to update the target network strategy every 200 steps, achieving milliampere-level precise control and avoiding the risk of overfitting.

[0109] Example 4: Please refer to Figure 4 , an embodiment provided by the present invention: dual-target stimulation of S4 adopts a spatiotemporal synergistic protocol:

[0110] Neck stimulation was initiated first, and taVNS was superimposed if the LF / HF ratio did not decrease by 10% within 5 minutes.

[0111] The phase difference between the two target stimulations was fixed at π / 2 radians;

[0112] The pulse width of neck stimulation is 1.5 to 2 times that of taVNS;

[0113] The spatiotemporal collaboration protocol includes a conflict resolution mechanism:

[0114] When the LF / HF ratio increases by ≥5% within 30 seconds after superimposition of taVNS, a three-level response is initiated immediately:

[0115] First-level response: cut off neck stimulation and maintain taVNS at 5 Hz continuous wave;

[0116] Secondary response: Apply potassium chloride gel to reduce the electrode-skin interface impedance by ≥40%;

[0117] Level 3 response: If the RMSSD of HRV is continuously less than 20ms, switch to the backup sine wave stimulation mode;

[0118] Furthermore, a stepped stimulation sequence prioritizes neck stimulation and superimposes taVNS only when the LF / HF ratio has not dropped by 10% within 5 minutes, avoiding neural conflicts caused by the simultaneous activation of the two targets. The phase difference between the two targets is fixed at π / 2 radians to enhance neural rhythm coordination, and the neck stimulation pulse width is set to 1.5 to 2 times that of taVNS to match the differences in tissue conductivity properties. A 30-second conflict resolution response chain is established. When the LF / HF ratio rises by ≥5% after superimposing taVNS, the first-level response immediately cuts off neck stimulation and maintains taVNS at a 5Hz continuous wave to block the sympathetic storm. The second-level response uses potassium chloride gel to reduce the electrode-skin impedance by 40% to eliminate abnormal current distribution. The third-level response switches to a sinusoidal wave mode to restore neural balance when the RMSSD of HRV remains <20ms. This protocol deeply integrates timing control, parameter optimization, and emergency protection to build a clinical-grade safety closed loop, significantly improving the efficiency of solving the problem and the safety of treatment for intractable hypersympathetic syndrome.

[0119] Example 5: Please refer to Figure 5 and Figure 6 In one embodiment provided by the present invention, the genetic algorithm optimization in step S5 adopts an elite retention strategy, including:

[0120] Chromosome encoding: 8-bit binary string represents 4 bits of stimulus intensity and 4 bits of duty cycle;

[0121] Fitness function:

[0122]

[0123] Where Fitness is the fitness score of the chromosome in the genetic algorithm, △ANB is the absolute deviation between the measured ANB value and the target ANB value, and ANB is the autonomic nervous balance index. It is the baseline mean of the LF / HF ratio within 5 minutes before treatment;

[0124] Selection strategy: The elite retains the top 10% of individuals and directly enters the next generation;

[0125] The method also includes an efficacy verification system, including:

[0126] Short-term verification indicators: LF / HF ratio decrease ≥ 15% within 10 minutes after the end of treatment, GSR rising slope ≤ 0.05μS / s;

[0127] Long-term validation indicator: Cervical sympathetic symptom score decreases by >30% after 5 consecutive days of treatment;

[0128] Retraining trigger mechanism: When long-term indicators fail to meet standards, expand user data and fine-tune SVR parameters;

[0129] Furthermore, an innovative genetic algorithm adopts an elite retention strategy to accelerate convergence, discretizes the parameter space through 8-bit binary chromosome encoding, and designs a fitness function to simultaneously optimize the ANB target deviation and the autonomic nervous system baseline state, establishing a two-dimensional efficacy verification system. The short-term indicator requires a LF / HF decrease of ≥15% and a GSR rising slope of ≤0.05μS / s within 10 minutes after treatment to confirm the immediate physiological response. The long-term indicator verifies the sustainability of clinical efficacy by a decrease of >30% in the cervical sympathetic symptom score after 5 consecutive days of treatment. A self-evolutionary retraining mechanism is deployed. When the long-term indicators fail to meet the standards, user data is automatically expanded to fine-tune the SVR parameters, driving the model to continuously adapt to the evolution of neural function, and providing a clinically verifiable individualized treatment plan for sympathetic cervical spondylosis.

[0130] Working principle: By synchronously collecting EEG and skin conductance signals and performing joint analysis in the time-frequency domain and Lempel-Ziv complexity evaluation, accurate quantitative evaluation of sympathetic nerve activity and nerve disorder is achieved. The dual-model collaborative decision-making of machine learning and deep reinforcement learning is used to intelligently generate personalized nerve regulation parameters. Through the graded regulation of the dual targets of the cervical sympathetic nerve and vagus nerve, highly targeted and safe intervention is implemented. Based on the real-time feedback of heart rate variability, skin conductance and autonomic balance index, the genetic algorithm is used to dynamically close the loop to optimize the stimulation parameters, thus constructing a precise, intelligent, adaptive, closed-loop autonomic nerve function regulation system, thereby significantly enhancing the regulation efficiency of autonomic nerve dysfunction in sympathetic cervical spondylosis. The contact layer adopts conductive hydrogel doped with nano-diamonds, which significantly improves the signal conduction efficiency while ensuring high biocompatibility and eliminates the risk of skin irritation caused by long-term attachment. The impedance matching layer constructs a micro-current voltage-equalizing network with an ultra-thin silver fiber grid, which effectively suppresses motion artifacts and stabilizes the electrode-skin interface impedance, ensuring the signal integrity of hair areas such as the frontal lobe. The base layer is based on a polyimide flexible substrate, which fits the physiological curvature of the neck and reduces signal drift caused by changes in body position. The three work together to overcome the problem of signal fidelity in high-noise environments, providing a reliable hardware foundation for the simultaneous acquisition of high-precision EEG and GSR data, and significantly improving the accuracy of sympathetic nerve activity assessment. First, a personalized sympathetic nerve activity assessment model was constructed, and its SNS The Asymmetric Index calculation formula is based on a three-layer neural network that dynamically updates weight coefficients. This weighted learning formula integrates the user's physiological baseline and the rate of change in cervical sympathetic symptom scores in real time, significantly improving individualized quantitative accuracy. A multimodal integration architecture is then employed. An XGBoost classifier analyzes the theta / alpha power ratio and beta peak frequency, and an SVR regressor processes the prefrontal lateralization index (Asym). A dynamic weighting module collaboratively generates weight coefficients, significantly enhancing the robustness of decoding complex neural states. Finally, a reinforcement learning optimization engine is deployed, integrating real-time treatment data and individual tolerance thresholds using a two-stream network. Within the three-dimensional Q-matrix parameter space, the TD3 algorithm updates the target network strategy every 200 steps, achieving milliampere-level precision and avoiding overfitting risks. A stepped stimulation sequence prioritizes neck stimulation and only superimposes taVNS if the LF / HF ratio has not decreased by 10% within 5 minutes, avoiding neural conflict caused by simultaneous activation of the two targets. A fixed phase difference of π / 2 radians is used to enhance neural rhythm coordination, and the neck stimulation pulse width is set to 1 / 2 of the taVNS pulse width.5 to 2 times to match the differences in tissue conductivity characteristics, establish a 30-second conflict resolution response chain, when LF / HF rises ≥5% after superimposing taVNS, the first-level response immediately cuts off neck stimulation and maintains taVNS at 5Hz continuous wave to block the sympathetic storm, the second-level response uses potassium chloride gel to reduce the electrode-skin impedance by 40% to eliminate abnormal current distribution, and the third-level response switches to a sine wave mode to restore neural balance when the RMSSD of HRV lasts <20ms. This protocol deeply integrates timing control, parameter optimization and emergency protection to build a clinical-grade safety closed loop, significantly improving the efficiency of cracking and the safety of treatment for intractable hypersympathetic syndrome, and accelerating convergence through an innovative genetic algorithm using an elite retention strategy. By discretizing the parameter space through 8-bit binary chromosome encoding and designing a fitness function, the system simultaneously optimizes the ANB target deviation and the autonomic nervous system baseline state, establishing a two-dimensional efficacy verification system. Short-term indicators require a 15% or greater decrease in LF / HF and a 0.05 μS / s increase in GSR within 10 minutes of treatment to confirm immediate physiological response. Long-term indicators verify the sustainability of clinical efficacy by requiring a >30% decrease in cervical sympathetic symptom score after five consecutive days of treatment. A self-evolutionary retraining mechanism is deployed to automatically expand user data and fine-tune SVR parameters when long-term indicators fall short of target. This drives the model to continuously adapt to the evolution of neural function, providing a clinically verifiable, personalized treatment plan for sympathetic cervical spondylosis.

[0131] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations that come within the meaning and range of equivalents of the claims be embraced therein.

Claims

1. A method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on a brain-computer interface, characterized by: The steps for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface are as follows: S1. EEG signal acquisition: EEG and skin conductance (GSR) are collected synchronously using a flexible electrode array, with electrodes covering the frontal (FP1 / FP2), parietal (C3 / C4), and occipital (O1 / O2) brain regions. S2. Joint analysis of autonomic nervous system features: Joint analysis of EEG in the time and frequency domains was performed, and wavelet transform was used to extract characteristic frequency band signals related to autonomic nervous system activity, including the instantaneous energy of theta waves (4-8 Hz), alpha waves (8-13 Hz), and high-frequency beta waves (18-30 Hz). The power spectral density was calculated to determine the dominant frequency band, and the disorder of neural activity was evaluated based on the Lempel-Ziv complexity (LZc); S3, dual-model collaborative decision-making: the theta wave to alpha wave power ratio, beta wave peak frequency, and prefrontal EEG lateralization index (Asym) are input into the machine learning model, which outputs the sympathetic nerve activity score (SNS Index). When the SNS Index exceeds the preset activation threshold, the deep reinforcement learning model is activated to generate targeted neuromodulation parameters. S4, dual-target neuromodulation; For the cervical sympathetic nerves, low-frequency biphasic pulses of 1 to 10 Hz and a pulse width of 100 to 200 μs were applied near the C5 to T1 cervical vertebrae. For the vagus nerve: transcutaneous stimulation (taVNS) is applied to the concha area. The stimulation pattern is graded according to the SNS Index: Mild SNS Index = 1.2-1.5, 1-5 Hz continuous wave; Moderate SNS Index = 1.5-2.0, 10-15 Hz pulse train; Severe SNS Index>2.0, 25Hz+variable frequency pulse train; S5. Closed-loop dynamic optimization: Real-time monitoring of the LF / HF ratio of heart rate variability (HRV), GSR, and blood pressure fluctuations. If the autonomic balance index (ANB = HF / LF × GSR stability coefficient) does not reach 0.8-1.2 within 10 minutes, the genetic algorithm is activated to optimize the stimulation duty cycle. The stimulation intensity is based on 0.5-1 mA and is adjusted by ±0.2 mA every 30 seconds according to the decrease in LF / HF.

2. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The flexible electrode array in S1 includes: Contact layer: conductive hydrogel doped with nanodiamonds, conductivity ≥ 100S / m; Impedance matching layer: silver fiber mesh, density 80-120 threads / cm 2 , thickness ≤ 50 μm; Base layer: Polyimide flexible substrate, Young's modulus <2GPa, its surface impedance is stable at 5~10kΩ in the frequency band of 10~100Hz.

3. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The calculation formula for the SNS Index in S3 is: The coefficients k1, k2, and k3 are the dynamic weights of the physiological baseline determined during the initial calibration phase of the user, P θ is the integral value of the power spectrum density in the θ frequency band of 4 to 8 Hz, P α is the integrated value of the power spectrum density in the α band from 8 to 13 Hz, f β is the integral value of the power spectrum density in the β band from 18 to 30 Hz, Asym left Asym is the power value of α wave 8~13Hz at the left frontal electrode FP1. right is the power value of α wave 8-13 Hz at the right frontal electrode FP2; The weight update formula of coefficients k1, k2 and k3 is: in is the weight coefficient of the current treatment cycle (i=1, 2, 3 corresponds to k1, k2 and k3 in the formula), is the weight coefficient of the previous treatment cycle, η is the learning rate, is the CSSS drop rate with respect to weight k i , CSSS is the cervical sympathetic nerve symptom score, and the CSSS decrease rate is the rate of change of the score before and after treatment.

4. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The machine learning model in S3 uses an integrated learning architecture, including: First-level classifier: XGBoost model processes theta / alpha power ratio and beta peak frequency; Second-level regressor: Support Vector Regression (SVR) processes the Asym index; Dynamic weighting module: A three-layer fully connected neural network generates weight coefficients k1, k2, and k3, and the input is physiological baseline parameters.

5. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The deep reinforcement learning model in S3 adopts a two-stream network architecture, including: State stream input: real-time SNS Index, historical treatment data, and individualized tolerance threshold; Action flow output: Q matrix in three-dimensional parameter space.

6. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 5, characterized in that: The Q matrix of the three-dimensional parameter space includes: Dimension 1: Neck stimulation intensity 0.5-2.0 mA, resolution 0.1 mA; Dimension 2: taVNS carrier frequency 1 to 30 Hz, resolution 1 Hz; Dimension 3: dual-target phase difference 0 to π radians, resolution π / 8; The double-delayed deep deterministic policy gradient (TD3) algorithm is used to update the Q matrix, and the target network is updated every 200 steps.

7. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The dual-target stimulation of S4 adopts a spatiotemporal coordination protocol: Neck stimulation was initiated first, and taVNS was superimposed if the LF / HF ratio did not decrease by 10% within 5 minutes. The phase difference between the two target stimulations was fixed at π / 2 radians; The pulse width of neck stimulation is 1.5 to 2 times that of taVNS.

8. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 7, characterized in that: The spatiotemporal collaboration protocol includes a conflict resolution mechanism: When the LF / HF ratio increases by ≥5% within 30 seconds after superimposition of taVNS, a three-level response is initiated immediately: First-level response: cut off neck stimulation and maintain taVNS at 5 Hz continuous wave; Secondary response: Apply potassium chloride gel to reduce the electrode-skin interface impedance by ≥40%; Level 3 response: If the RMSSD of HRV is continuously less than 20ms, switch to the backup sine wave stimulation mode.

9. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The genetic algorithm optimization in step S5 adopts an elite retention strategy, including: Chromosome encoding: 8-bit binary string represents 4 bits of stimulus intensity and 4 bits of duty cycle; Fitness function: Where Fitness is the fitness score of the chromosome in the genetic algorithm, △ANB is the absolute deviation between the measured ANB value and the target ANB value, and ANB is the autonomic nervous balance index. It is the baseline mean of the LF / HF ratio within 5 minutes before treatment; Selection strategy: The elite retains the top 10% of individuals and directly enters the next generation.

10. The method for enhancing autonomic nervous system regulation in sympathetic cervical spondylosis based on brain-computer interface according to claim 1, characterized in that: The method further includes an efficacy verification system, comprising: Short-term verification indicators: LF / HF ratio decrease ≥ 15% within 10 minutes after the end of treatment, GSR rising slope ≤ 0.05μS / s; Long-term validation indicator: Cervical sympathetic symptom score decreases by >30% after 5 consecutive days of treatment; Retraining trigger mechanism: When long-term indicators are not met, expand user data and fine-tune SVR parameters.

Images