Patient improvement effect analysis method for controlling spinal cord electrical stimulation through implantable brain-computer interface

Abstract

The invention discloses a patient improvement effect analysis method for controlling spinal cord electrical stimulation through an implantable brain-computer interface, and relates to the technical field of medical rehabilitation, and the method comprises the steps: multi-dimensional collaborative data collection: implanting electrodes in a target brain region and below a spinal cord injury segment, installing a detection element at an exoskeleton key part, and carrying out multi-dimensional collaborative data collection; a sensor is attached to a lower limb preset muscle group, electroencephalogram signals, SCS stimulation parameters, EXS motion data and neuromuscular response data are synchronously collected, and time correlation marks are embedded; according to the method, the reliability of motion intention decoding is remarkably improved by adopting a mode of combining multi-source signal preprocessing and a multi-mode intention recognition model, and in the signal preprocessing stage, the self-adaptive filtering algorithm combining Kalman filtering and wavelet threshold denoising is applied, so that the motion intention decoding efficiency is improved. SCS electrical stimulation interference, EXS motor noise and physiological noise in the electroencephalogram signals are effectively removed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical rehabilitation, in particular to a method for analyzing the improvement effect of a patient controlled by an implanted brain-computer interface on spinal cord electrical stimulation. BACKGROUND

[0002] In the field of medical rehabilitation, spinal cord injury is a disease that seriously threatens the quality of life of patients. As the key channel for neural signal transmission between the brain and various parts of the body, once the spinal cord is damaged, it will directly destroy the neural conduction pathway between the brain and the spinal cord and muscles, leading to varying degrees of motor dysfunction in patients, seriously affecting their self-care ability in daily life and social participation. In recent years, the SCS-EXS system combining spinal cord electrical stimulation and exoskeleton has gradually been applied to clinical rehabilitation treatment. This system provides passive training for patients through the synergistic effect of epidural electrical stimulation of the spinal cord and exoskeleton, which can activate neural muscle activity to a certain extent and promote partial recovery of limb motor function.

[0003] However, the existing SCS-EXS system has obvious limitations in actual application. On the one hand, this system mainly focuses on passive training and cannot respond to the active movement intention of patients. In the process of rehabilitation treatment, the active participation of patients is crucial for the remodeling of neural function and the recovery of motor function. However, the existing system cannot construct an active neural cascade control pathway of brain-spinal cord-muscle, limiting the further improvement of rehabilitation effect. On the other hand, although the brain-computer interface (BCI) electroencephalogram acquisition and intention recognition technology can decode the brain movement intention, it lacks synergistic linkage with spinal cord electrical stimulation and exoskeleton system when applied alone, cannot fully utilize the advantages of each technical means, and is difficult to achieve precise and efficient rehabilitation treatment. In addition, the existing improvement effect analysis method also has shortcomings. It only evaluates passive training or single brain control function and does not cover the multi-dimensional data correlation after the fusion of brain control technology and SCS-EXS system. This one-sided evaluation method cannot comprehensively and accurately reflect the rehabilitation condition of patients and cannot provide sufficient technical basis for the optimization of individualized brain control rehabilitation plan, resulting in the lack of pertinence and effectiveness of rehabilitation treatment. SUMMARY

[0004] The purpose of the present application is to make up for the shortcomings of the prior art, and provide a patient improvement effect analysis method for implantable brain-computer interface control of spinal cord electrical stimulation, which can accurately identify the movement intention of the patient through deep analysis and processing of multi-source synchronous data, and establish a multi-dimensional correlation model to quantitatively analyze the conduction and coordination characteristics of the brain-spinal cord-muscle neural cascade pathway. On this basis, a comprehensive evaluation system is constructed from multiple important dimensions such as neural pathway, motor function and long-term recovery, which can systematically evaluate the reconstruction effect of the brain-spinal cord-muscle neural cascade control pathway. It not only provides solid technical support for the adjustment of rehabilitation programs, but also enables rehabilitation treatment to better meet the individual needs of patients.

[0005] To solve the above technical problems, the present application provides the following technical scheme: a patient improvement effect analysis method for implantable brain-computer interface control of spinal cord electrical stimulation, the method comprising: Multi-dimensional collaborative data acquisition: electrodes are implanted in the target brain area and below the spinal cord injury segment, detection elements are installed at key parts of the exoskeleton, sensors are attached to the pre-set muscle groups of the lower limbs, and brain electrical signals, SCS stimulation parameters, EXS movement data and neuromuscular response data are synchronously collected, and time correlation markers are embedded; Multi-source signal preprocessing and noise suppression: the collected brain electrical signals, SCS, EXS and neuromuscular signals are respectively removed by dynamic filtering, parameter calibration, baseline correction and threshold detection to remove interference and correct deviation, and effective signal features and key moment information are extracted; BCI movement intention recognition and precision verification: multi-dimensional features are extracted from the preprocessed brain electrical signals and spinal cord evoked potential signals, a deep learning model containing feature fusion and mapping layers is constructed and trained, and the model accuracy is verified by accuracy, confusion matrix and robustness test under different stimulation intensities; Brain-spinal cord-muscle cascade pathway correlation analysis: the time interval of brain electrical intention start, spinal cord response and muscle activation is calculated, the causal influence strength of brain electrical signals on other signals is quantified, and the coincidence degree of EXS actual movement trajectory and standard intention trajectory and the matching rate of SCS stimulation and intention are compared; Multi-dimensional improvement effect comprehensive evaluation: a quantitative scoring system is constructed based on the pathway delay and SCEP amplitude change index to evaluate the neural pathway reconstruction, the Fugl-Meyer scale is added with brain control special evaluation and daily life scene test to evaluate the motor function, and the long-term recovery trend is evaluated by monitoring the signal changes at multiple follow-up nodes.

[0006] Further, in the multi-source signal preprocessing and noise suppression step, array metal electrodes are implanted in the motor cortex, supplementary motor area, and parietal sensory cortex to collect electroencephalogram signals containing rhythm signals and event-related synchronization / desynchronization signals in a preset frequency band, and SCS stimulation trigger markers and EXS motion instruction markers are embedded in the electroencephalogram signals; SCS stimulation parameters and SCS stimulation patterns associated with motion intention are recorded, the SCS stimulation parameters including stimulation frequency, intensity, and pulse width; motion parameters of the EXS exoskeleton and the time delay between exoskeleton motion and SCS stimulation are collected, the motion parameters including joint rotation angle, driving torque, and motion speed; a spinal cord evoked potential SCEP is collected through an implanted spinal cord electrode; and an electromyography sensor is attached to the surface of a preset muscle group in the lower limbs to collect motor unit action potentials MUAPs.

[0007] Further, in the multi-source signal preprocessing and noise suppression step, the collected multi-source data is subjected to interference suppression and calibration processing, an adaptive filtering algorithm combining Kalman filtering and wavelet threshold denoising is used to denoise the electroencephalogram signals, SCS electrical stimulation interference, EXS motor noise, and physiological noise are removed, and electroencephalogram signal frequency bands related to motion intention are retained; amplitude drift correction is performed on the SCS stimulation parameters; zero position calibration is performed on the EXS exoskeleton joint rotation angle data, and the zero position calibration is implemented before each training; a linear trend elimination method is used to remove the baseline drift of the SCEP signal; and a band-pass filter is used to process the MUAP signal, a preset threshold is used to detect the peak value of the filtered MUAP signal, and the muscle activation start time is extracted.

[0008] Further, in the BCI motion intention recognition and precision verification step, a motion intention recognition model is constructed based on the preprocessed data and the precision is verified, time-frequency features, spatial features, and ERD / ERS features are extracted from the electroencephalogram signals, peak features of the SCEP signal are extracted, and the extracted features are fused; a multi-modal fusion deep learning model combining EEG-CNN and Transformer is used to train the fused features, baseline period data of a preset number of patients are used as the training set and the verification set, a cross-entropy loss function is used to optimize and train the model; the recognition accuracy and the confusion matrix of multi-class motion intention are calculated through an intention recognition confidence formula, the intention recognition robustness of the model under different SCS stimulation intensities is tested, and the precision stability of the model is verified.

[0009] Further, in the BCI motion intention recognition and precision verification step, the recognition accuracy and the confusion matrix of multi-class motion intention are calculated through an intention recognition confidence formula, the formula being: wherein is the recognition confidence of the i-th class motion intention, is the recognition confidence of the j-th The EEG feature matching value of the class intention, The EEG feature matching value of the class intention, The SCEP feature matching value of the class intention, The cross-modal correlation of the EEG and the SCS, , The maximum feature matching value of all intentions.

[0010] Further, in the brain-spinal cord-muscle cascade pathway correlation analysis step, a multi-dimensional correlation model is established to analyze the conduction and synergy characteristics of the neural cascade pathway, a time series cross-correlation algorithm is used to locate the EEG intention trigger point, the SCEP peak time and the MUAP activation time, the time difference between each time is calculated, and the average value of the time difference of multiple repeated actions is taken; a Granger causality test algorithm is used to analyze the causal influence strength of the EEG intention signal on the SCS stimulation parameters, the EXS movement trajectory and the muscle MUAP activation; based on the BCI intention recognition result, a dynamic time warping algorithm is used to calculate the coincidence degree of the EXS movement trajectory and the intention; and the matching rate of the SCS stimulation segment and parameters and the intention is calculated.

[0011] Further, in the brain-spinal cord-muscle cascade pathway correlation analysis step, the Granger causality test algorithm is used to analyze the causal influence strength of the EEG intention signal on the SCS stimulation parameters, the EXS movement trajectory and the muscle MUAP activation, and the algorithm formula is: wherein, is the Granger causality test of the EEG intention signal on the target signal statistic, used to quantify the causal driving strength of the EEG intention on the target signal, the target signal including the SCS stimulation parameters / EXS movement trajectory / MUAP activation signal, the larger the F value, the more significant the active control correlation of the EEG intention on the target signal, is the residual sum of squares when a prediction model is constructed using the historical data of the target signal itself, reflecting the natural change law of the target signal without external driving, is the residual sum of squares when a prediction model is constructed using the historical data of the target signal itself and the historical data of the EEG intention signal, reflecting the prediction error change of the target signal after the EEG intention is added, is the model lag order, determined according to the signal acquisition frequency of the pathway, is the total number of signal samples used for modeling.

[0012] Further, in the brain-spinal cord-muscle cascade pathway correlation analysis step, a dynamic time warping algorithm is used to calculate the coincidence degree of the EXS movement trajectory and the intention, and the algorithm formula is: wherein, is the coincidence degree of the actual motion trajectory of the EXS exoskeleton and the standard trajectory corresponding to the brain electrical intention, the value is closer to 1, indicating that the matching accuracy of the EXS motion and the brain electrical intention is higher, is the actual trajectory data of the EXS and the intention standard trajectory data The dynamic time warping distance is calculated by constructing the distance matrix of the two trajectories and finding the optimal matching path, and the smaller the distance is, the higher the trajectory similarity is, is the actual motion trajectory data of the key joint of the EXS exoskeleton, is the standard motion trajectory data of the key joint corresponding to the brain electrical intention; is the maximum DTW distance of the EXS trajectory and the standard trajectory under this type of intention, which is calibrated offline by simulating the extreme scenario that the EXS trajectory completely deviates from the intention.

[0013] Further, in the multi-dimension improvement effect comprehensive evaluation step, an evaluation system is constructed from the neural pathway, motor function and long-term recovery dimensions and is quantified, the evaluation indexes of the quantified scoring system include pathway delay, SCEP amplitude change and muscle activation synchronism, the total score of the pathway integrity is calculated according to the scores of each evaluation index; a brain control special score dimension is added based on the Fugl-Meyer scale, including joint active range of motion, gait stability and action completion efficiency; the ability of patients in daily life activities under brain control assistance is tested; a follow-up cycle is set, the rhythm signal energy change of the motor cortex and the SCEP amplitude change are monitored in the follow-up cycle, and the neural plasticity of the brain and the spinal cord is evaluated.

[0014] Compared with the prior art, the implanted brain-computer interface control spinal cord electrical stimulation patient improvement effect analysis method has the following beneficial effects: One, the present application significantly improves the reliability of motion intention decoding by adopting the combination of multi-source signal preprocessing and multi-modal intention recognition model, in the signal preprocessing stage, the adaptive filtering algorithm combining Kalman filtering and wavelet threshold denoising is used to effectively remove the SCS electrical stimulation interference, EXS motor noise and physiological noise in the electroencephalogram, and the frequency band related to the motion intention is retained, and a multi-modal fusion deep learning model is constructed, a plurality of features are fused for training, and the motion intention can be more accurately recognized, the recognition accuracy of multi-class motion intention and the confusion matrix are calculated, and the robustness under different SCS stimulation intensities is tested, ensuring the stability of the model precision, providing a reliable basis for subsequent rehabilitation treatment.

[0015] Secondly, the application realizes quantitative analysis of pathway synergy by applying Granger causality test to neural cascade pathway evaluation, can clearly present the causal influence intensity of the electroencephalogram intention signal on the SCS stimulation parameter, the EXS movement trajectory and the muscle MUAP activation, and constructs a multi-dimensional evaluation system, gives consideration to neural repair and function improvement, and comprehensively evaluates from the neural pathway, movement function and long-term recovery, etc. multi-dimensions, which helps to develop more scientific and effective rehabilitation programs.

[0016] Other advantages, objects, and features of the application will be set forth in part by the description that follows, and in part will become apparent to those skilled in the art upon examination of the following specification or can be learned by practice of the application. The objects and advantages of the application can be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without any creative labor.

[0018] Figure 1 Flow chart of the method for analyzing the improvement effect of the patient controlled by the implantable brain-computer interface for spinal cord electrical stimulation; Figure 2 Flow chart of the BCI movement intention recognition and precision verification step of the method for analyzing the improvement effect of the patient controlled by the implantable brain-computer interface for spinal cord electrical stimulation. DETAILED DESCRIPTION

[0019] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the following will combine the drawings and preferred embodiments to specifically describe the specific embodiments, structures, features and effects of the present application.

[0020] Embodiment one The patient is a 52-year-old male, who was injured by a heavy object and caused complete spinal cord injury at T9 segment, 8 months after injury, the lower limb motor function was completely lost (Fugl-Meyer lower limb motor score 0), and there was no any voluntary muscle contraction sign, and the patient received 4 months of SCS-EXS passive training before, and the lower limb joint activity was not obviously improved.

[0021] Under general anesthesia, 4x4 array platinum-iridium alloy electroencephalogram electrodes were implanted in the motor cortex (M1 area) and supplementary motor area of the patient through stereotactic surgery, with electrode contact diameter of 0.5 mm and spacing of 1 mm; 8-contact SCS stimulation electrodes were implanted in the T11-L5 segment of the spinal cord, with electrode length of 2 cm; according to the size of the patient's lower limbs (thigh length 55 cm, calf length 48 cm), the lower limb EXS exoskeleton was customized, and angle sensors with precision of ±0.5° and torque sensors with precision of ±0.1 N·m were installed at the hip, knee and ankle joints, respectively; the EXS controller was connected with the implanted signal transmission module through Bluetooth 5.0; surface electromyography sensors were attached to the surface of the patient's lower limb quadriceps, hamstrings, tibialis anterior and gastrocnemius muscles; a time synchronizer with precision of 1 μs was used to time calibrate the electroencephalogram acquisition device, SCS stimulator, EXS controller and electromyography sensors, to ensure that the time stamp deviation of each device data was ≤1 ms.

[0022] Three data collection stages were set. The baseline period was 1 week before treatment, data was collected once a day, each collection lasted for 60 minutes, including 30 minutes of passive SCS-EXS training data (SCS stimulation frequency 40 Hz, intensity 3.5 mA, pulse width 250 μs), 20 minutes of BCI alone intention recognition data (patient imagines 8 types of movement intentions such as lifting legs, stepping, standing, sitting, etc.), and 10 minutes of neuroelectrophysiological baseline data (SCEP, electromyographic latency); the treatment period was 12 weeks, training was performed 3 times a week, each training lasted for 60 minutes, 5 minutes before training, static neuroelectrophysiological data was collected, 30 minutes during training, brain-controlled SCS-EXS collaborative data (including electroencephalogram signal, SCS stimulation mode, EXS movement parameters, electromyography data) was collected, and 25 minutes after training, dynamic neuroelectrophysiological test data was collected; the follow-up period was 1 month, 3 months and 6 months after treatment, data collection was repeated according to the baseline period at each follow-up node, and each collection lasted for 40 minutes.

[0023] The 40 Hz power frequency interference and harmonics generated by SCS stimulation are suppressed by Kalman filtering algorithm, and the 120-350 Hz noise generated by EXS motor operation is removed by wavelet threshold denoising algorithm, while physiological artifacts such as ECG and EOG are also removed, and finally the mu rhythm, beta rhythm and ERD / ERS signals related to motor intention in the 8-30 Hz frequency band are retained; the processed EEG signal is baseline corrected, taking the signal mean value during the unintentional imagination period as the reference, and the signal baseline drift is eliminated; the SCS stimulation parameter data is calibrated once a week using a high-precision signal acquisition instrument to calibrate the output end parameters of the stimulator, and the stimulation intensity is corrected by comparing the actual measurement value with the recorded value, ensuring that the intensity recording error is ≤0.1 mA; before each training, the patient's lower limbs are controlled to be in a standing position-the zero position attitude of the completely straightened lower limbs, and the output value of the angle sensor at this time is taken as the reference to calibrate the joint angle data to zero; the SCEP signal is removed by linear trend elimination method, and the electromyographic signal is subjected to 20-500 Hz band-pass filtering, and the baseline noise 3 times the standard deviation is set as the peak detection threshold to extract the muscle activation starting time, ensuring that the signal-to-noise ratio of the electromyographic signal is ≥15 dB.

[0024] The EEG data and SCEP data of the patients in the baseline period and the first 4 weeks of the treatment period are collected as training samples, a total of 15000 samples, which are divided into a training set (10500 samples) and a validation set (4500 samples) in a ratio of 7:3; a multi-modal fusion deep learning model combining EEG-CNN and Transformer is used to input the time-frequency features (extracted by short-time Fourier transform), spatial features (electrode array topology correlation) of EEG signals and peak features of SCEP signals, and the model is optimized and trained using the cross-entropy loss function; during the training process, the intention recognition confidence formula Evaluate the recognition reliability-take the "leg lifting" intention ) as an example, the baseline period , , the cross-modal correlation coefficient , , , the calculation is , after 4 weeks of treatment , , , the confidence is significantly improved, before each training, the patient completes the imagination test of 8 types of motor intention in turn, each type of intention test is performed 3 times, a total of 24 tests; the recognition accuracy of each type of intention is calculated by the model, when the overall accuracy is stable at 88% or above, the training on the same day is started; if the accuracy is less than 85%, the impedance tester is used to detect the EEG electrode impedance, if the impedance is out of range, the electrode position is adjusted or the electrode contact is cleaned through surgery, and the test is performed again after recalibration until the accuracy meets the standard.

[0025] The MATLAB time series analysis tool is used to locate the movement intention trigger point (the starting time of mu rhythm ERD), the peak time of SCEP signal and the starting time of muscle activation signal in the electroencephalogram signal; taking the leg lifting intention as an example, the time difference between intention trigger-spinal cord response and spinal cord response-muscle activation is calculated, and the average value of the time difference of 30 repeated actions is taken; before treatment, the total delay of the channel is 450 ms, in which the intention trigger-spinal cord response delay is 280 ms, and the spinal cord response-muscle activation delay is 170 ms; after 12 weeks of treatment, the total delay of the channel is reduced to 230 ms, the intention trigger-spinal cord response delay is 140 ms, and the spinal cord response-muscle activation delay is 90 ms; Granger causality test is realized through the Python deep learning framework, and the formula The causal influence of the electroencephalogram intention on the target signal is analyzed, the electroencephalogram has a causal influence on the SCS stimulation before treatment , and the causal influence is increased by 76% after 12 weeks of treatment The causal influence of the electroencephalogram on the muscle activation is increased from 1.5 to 3.9, which is increased by 67%, and the BCI intention recognition result is used as a benchmark, and the DTW trajectory coincidence degree formula is used to calculate the coincidence degree of the EXS trajectory and the intention, the “leg lifting” intention has a coincidence degree of 0.5 before treatment, and the coincidence degree is increased to 0.9 after 12 weeks of treatment The matching rate of the SCS stimulation segment and the parameter and the intention is calculated, the matching rate is 48% before treatment, and the matching rate is increased to 90% after 12 weeks of treatment.

[0026] According to the scoring criteria of the channel delay (40 points), the SCEP amplitude change (30 points) and the muscle activation synchronism (30 points), the channel integrity score is completed; the channel integrity score of the patient is 20 points before treatment, and is increased to 72 points after 12 weeks of treatment; based on the Fugl-Meyer scale, a brain control special score (total score 20 points) is added, the score is 0 before treatment, and is increased to 16 points after 12 weeks of treatment; the patient can complete autonomous leg lifting and continuous walking 15 times under the assistance of brain control SCS-EXS, and the walking speed reaches 0.25 m / s; after 6 months of follow-up after treatment, the motor cortex mu rhythm energy is increased by 20% compared with that at the end of treatment, and the SCEP amplitude is increased by 28%; the test of the activity of daily living shows that the patient can independently complete up and down 8 stairs, stand independently for 10 minutes, and independently control the lower limbs to complete simple turning actions, and the self-care ability is significantly improved.

[0027] Embodiment two A 45-year-old male patient was admitted to our hospital 6 months after a high fall caused an incomplete spinal cord injury at the L1 segment. The patient had partial loss of motor function in the lower limbs. The left quadriceps muscle strength was grade 1, the right quadriceps muscle strength was grade 2, the bilateral gastrocnemius and tibialis anterior muscle strength was grade 1, and the Fugl-Meyer lower limb motor score was 18. The patient could not stand independently and had to rely on a walking aid to move slowly. He depended on his family for daily transfer. He had received 3 months of traditional SCS-EXS passive training before this treatment, which only slightly improved the range of motion of the lower limbs and did not significantly improve the active movement control ability. During the training, the patient often experienced numbness in the lower limbs due to the discomfort of the SCS stimulation intensity.

[0028] Under general anesthesia, 3×4 array platinum-iridium brain electrodes were implanted in the motor cortex (M1 area, responsible for lower limb motor control area) and supplementary motor area of the patient. The electrode contact diameter was 0.4 mm, and the spacing was 1.2 mm. Intraoperative neurophysiological monitoring confirmed that the electrode contacts could effectively collect motor-related brain electrical signals (at least 6 contacts could detect significant beta rhythm suppression when the patient imagined lifting his leg). A 10-contact SCS stimulation electrode was implanted in the T12-L4 segment of the spinal cord, with an electrode length of 2.5 cm to ensure that the stimulation contacts covered the key motor nerve pathways below the spinal cord injury segment. The lower limb EXS exoskeleton was customized according to the patient's lower limb size (thigh length 52 cm, calf length 46 cm). Angle sensors with an accuracy of ±0.3° and torque sensors with an accuracy of ±0.08 N・m were installed at the hip, knee, and ankle joints. The EXS knee and ankle joint contact parts used breathable medical sponge padding (thickness 6 mm) to avoid skin pressure injuries caused by long-term wear. Disposable surface electromyography sensors were attached to the surface of the bilateral quadriceps, gastrocnemius, tibialis anterior, and gastrocnemius muscles. A time synchronizer with an accuracy of 0.5 μs was used to time-calibrate the brain electrical signal acquisition equipment, SCS stimulator, EXS controller, and electromyography sensors. After calibration, the time stamp deviation of each device data was controlled within 0.8 ms, ensuring the time consistency of multi-source data.

[0029] Three data collection stages are set. The baseline period is 10 days before treatment, a total of 5 data are collected, each collection is 55 minutes, each collection contains 25 minutes of passive SCS-EXS training data (initial SCS stimulation frequency 35 Hz, intensity 3.2 mA, pulse width 220 μs, according to patient tolerance, avoid obvious numbness discomfort), 20 minutes of BCI alone intention recognition data (patients imagine four basic motion intentions of bilateral leg lifting, knee bending, knee extending and stepping), 10 minutes of neurophysiological baseline data (SCEP latency, electromyographic resting amplitude are collected); the treatment period is 14 weeks, training 4 times a week, each training is 55 minutes, 8 minutes before training to collect static neurophysiological data (SCEP, electromyographic baseline value), 30 minutes in training to collect brain-controlled SCS-EXS coordination data (real-time recording of electroencephalogram signal, SCS stimulation parameter adjustment process, EXS motion trajectory data, electromyographic activation signal), 17 minutes after training to collect dynamic neurophysiological test data (including SCEP response under different SCS stimulation intensity, electromyographic contraction force test); the follow-up period is 2 months, 4 months and 6 months after treatment, each follow-up node repeats data collection according to the baseline period collection dimension, each collection is 45 minutes, and the brain-controlled autonomous motion data (patient active motion related data after passive assistance) are supplemented.

[0030] The collected raw EEG signals are processed by using MATLAB signal processing toolbox. First, the 35 Hz power frequency interference and 2 times and 3 times harmonic interference generated by SCS stimulation are suppressed by Kalman filtering algorithm. Then, the 100-300 Hz wide frequency noise generated by EXS motor operation is removed by wavelet threshold denoising algorithm. At the same time, the electrocardiogram and electrooculogram physiological noises are removed by independent component analysis technology. Finally, the mu rhythm (8-13 Hz), beta rhythm (13-28 Hz) and ERD / ERS signals related to motor intention in the 7-28 Hz frequency band are reserved. After processing, the EEG signal is baseline corrected. The mean value of the EEG signal of the patient in the resting state (5 minutes of closed eyes relaxation) is used as the baseline to eliminate the signal baseline drift and ensure that the signal-to-noise ratio of the processed EEG signal is greater than or equal to 13 dB. For SCS stimulation parameter data, the actual parameters of the stimulator output are calibrated once every 3 days by using a high-precision signal analyzer. The deviation between the recorded value and the actual value is compared. The amplitude drift correction is performed on the subsequent recorded stimulation intensity and frequency data by using a linear correction formula to ensure that the parameter recording error is less than or equal to 0.08 mA (intensity) and less than or equal to 1 Hz (frequency). Before each training, the patient's lower limbs are placed in the supine position-zero position attitude of natural straightening of the lower limbs by controlling the EXS exoskeleton. The output value of the angle sensor at this time is used as the baseline to calibrate the zero position of the hip, knee and ankle joint angle data and eliminate the measurement error caused by the mechanical wear of the exoskeleton. The linear trend elimination method is used to remove the baseline drift of the SCEP signal, so that the signal baseline fluctuation range is controlled within ±2 μV. The electromyogram signal is subjected to 15-550 Hz band-pass filtering. The 2.5 times standard deviation of the electromyogram signal baseline noise is set as the peak detection threshold. The signal peaks exceeding the threshold are automatically identified and marked. The time when the peak value appears is used as the starting time of muscle activation to ensure that the electromyogram activation time identification error is less than or equal to 5 ms.

[0031] The EEG data (total of 12000 samples) and SCEP data (total of 8000 samples) collected from the patient at the baseline stage were collected as initial training data, and were divided into a training set (16000 samples) and a validation set (4000 samples) in a ratio of 8:2; a multi-modal fusion deep learning model combining EEG-CNN and Transformer was used, time-frequency features (frequency distribution of different time periods extracted by short-time Fourier transform), spatial features (signal correlation between different EEG electrode contacts) and ERD / ERS features (β rhythm suppression amplitude, duration) of the EEG signal were input, and the peak amplitude and latency features of the SCEP signal were input, and the above features were fused and input into the model for training; the model was optimized using a cross-entropy loss function, and the training period was set to 60 rounds, and the model performance was evaluated using the validation set after each round of training, and the initial training was stopped when the accuracy of the validation set was stable and exceeded 90%; the latest training data (about 3000 samples) of the patient was collected every 3 weeks during the treatment period, and the model parameters were incrementally fine-tuned to adapt to the dynamic changes of the patient's EEG signal with the progress of the treatment (such as the enhancement of the amplitude of the motor intention related EEG feature).

[0032] Before each training, the patient completed the imagination test of 4 types of motor intention in turn, 12 times for each type of intention, a total of 48 tests, and the recognition accuracy of each type of intention and the overall accuracy were calculated by the model, and the daily training was started when the overall accuracy was ≥88%; if the accuracy was <85%, first check the EEG electrode impedance (normal range 4-14 kΩ), if the impedance is out of range, clean the electrode contact and scalp contact part with physiological saline, reconnect and test again; if the impedance is normal, adjust the SCS stimulation base strength (±0.2 mA) to reduce the interference of stimulation on the EEG signal until the accuracy meets the standard, such as before the training of the 5th week, the patient's walking intention recognition accuracy was 82%, and the detection found that the impedance of 2 EEG electrodes reached 18 kΩ, and after cleaning, the impedance decreased to 10 kΩ, and the retest accuracy increased to 91%, meeting the training start standard.

[0033] Using MATLAB time series analysis tools, the motor intention trigger point (β rhythm ERD start time) in the EEG signal, the SCEP signal peak time (maximum amplitude time of spinal cord neural response), and the muscle activation start time (first time the muscle peak value exceeds the threshold value) were located by cross-correlation algorithm; taking bilateral leg lifting intention as an example, the time difference between intention trigger-spinal response and spinal response-muscle activation was calculated, and the average value of the time difference was taken as the pathway delay index, before treatment, the intention trigger-spinal response delay was 260 ms, the spinal response-muscle activation delay was 150 ms, and the total pathway delay was 410 ms; after 14 weeks of treatment, the intention trigger-spinal response delay was 130 ms, the spinal response-muscle activation delay was 80 ms, and the total pathway delay was reduced to 210 ms, and the conduction efficiency of the pathway was significantly improved.

[0034] The Granger causality test is implemented through a deep learning framework to analyze the causal influence intensity of the EEG intention signal on SCS stimulation parameter adjustment, EXS trajectory change, and EMG activation. Before treatment, the causal influence of the EEG intention on the SCS stimulation parameter is weak, and the causal influence on the EMG activation can only explain 15%; after 14 weeks of treatment, the causal influence of the EEG intention on the SCS stimulation parameter can explain 52% of the parameter change, and the causal influence on the EMG activation can explain 48%, indicating that the EEG intention can effectively drive the downstream pathway response; based on the BCI intention recognition result, the dynamic time warping (DTW) algorithm is used to calculate the coincidence degree of the EXS motion trajectory and the standard trajectory of the intention. Before treatment, the coincidence degree of the knee flexion intention trajectory is only 58%, and after 14 weeks of treatment, it increases to 90%; the matching rate of SCS stimulation segment and parameter with intention is calculated. Before treatment, the matching rate is 65%, and after 14 weeks of treatment, it increases to 93%, ensuring that the SCS stimulation can accurately match the EEG intention demand.

[0035] According to the quantitative scoring system of pathway delay (35 points), SCEP amplitude change (30 points), and muscle activation synchronicity (35 points), the pathway integrity score is completed. Before treatment, the patient's pathway integrity score is 22 points, and after 14 weeks of treatment, it increases to 86 points; based on the Fugl-Meyer scale, a brain control special scoring dimension (total score 25 points) is added, including joint active range of motion 8 points, gait stability 9 points, and action completion efficiency 8 points. Before treatment, the brain control special score is 3 points, and after 14 weeks of treatment, it increases to 21 points. At this time, the patient can independently complete bilateral leg lifting and continuous walking 20 times under the assistance of brain-controlled SCS-EXS, with a walking speed of 0.3 m / s, and the SCS stimulation intensity can be automatically adjusted according to the EEG intention. The patient has no obvious numbness and discomfort.

[0036] Set up three follow-up periods of 2 months, 4 months, and 6 months after treatment, and collect motor cortex rhythm signal and SCEP amplitude data in each period. At the 6-month follow-up after treatment, the patient's motor cortex beta rhythm energy increased by 28% compared to at the end of treatment, and the SCEP amplitude induced by SCS stimulation increased by 32% compared to at the end of treatment, indicating that the brain and spinal cord neural plasticity continues to improve; in the activities of daily living test, the patient can independently complete 12 steps of stairs, walk independently for 80 meters, and independently put on and take off shoes and socks. The Fugl-Meyer lower extremity motor score increases to 42 points, completely breaks away from the dependence on a walking aid, and the self-care ability is greatly improved. The patient's satisfaction with the rehabilitation effect is significantly improved, and actively expresses willingness to participate in community rehabilitation exchange activities.

[0037] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in any form. Although the present application has been disclosed with the preferred embodiments as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes, as long as the changes or modifications do not deviate from the technical solution of the present application. Any modification, change, equivalent change and modification of the above embodiments made according to the technical essence of the present application still belong to the scope of the technical solution of the present application.

Claims

1. A method for analyzing the improvement effect of implanted brain-computer interface-controlled spinal cord electrical stimulation on patients, characterized in that, The method includes: Multi-dimensional collaborative data acquisition: By implanting electrodes in the target brain region and below the spinal cord injury segment, installing detection elements in key parts of the exoskeleton, and attaching sensors to preset muscle groups in the lower limbs, EEG signals, SCS stimulation parameters, EXS motion data and neuromuscular response data are collected simultaneously and time-related markers are embedded. Multi-source signal preprocessing and noise suppression: For the acquired EEG, SCS, EXS and neuromuscular signals, interference and deviations are removed and deviations are corrected by dynamic filtering, parameter calibration, baseline correction and threshold detection, respectively, and effective signal features and key moment information are extracted; BCI Motion Intent Recognition and Accuracy Verification: Multi-dimensional features are extracted from preprocessed EEG and spinal cord evoked potential signals. A deep learning model including feature fusion and mapping layers is constructed and trained. The model accuracy is verified through accuracy, confusion matrix and robustness tests under different stimulus intensities. Brain-spinal cord-muscle cascade pathway association analysis: Calculate the time interval between EEG intention initiation, spinal cord response, and muscle activation; quantify the causal influence of EEG signals on other signals; and compare the overlap between the actual EXS movement trajectory and the standard intention trajectory, as well as the matching rate between SCS stimulation and intention. Comprehensive evaluation of improvement effects across multiple dimensions: A quantitative scoring system based on pathway delay and SCEP amplitude changes is constructed to assess neural pathway reconstruction. The Fugl-Meyer scale is combined with additional brain control-specific assessments and daily life scenario tests to assess motor function. Long-term recovery trends are assessed by monitoring signal changes at multiple follow-up nodes.

2. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 1, characterized in that, In the multi-source signal preprocessing and noise suppression steps, array-type metal electrodes are implanted in the motor cortex, supplementary motor area, and parietal sensory cortex to collect EEG signals containing rhythmic signals and event-related synchronization / desynchronization signals within a preset frequency band. SCS stimulation trigger markers and EXS motor command markers are embedded in the EEG signals. SCS stimulation parameters and SCS stimulation patterns associated with motor intentions are recorded. SCS stimulation parameters include stimulation frequency, intensity, and pulse width. The motion parameters of the EXS exoskeleton and the time delay between exoskeleton movement and SCS stimulation are collected. The motion parameters include joint rotation angle, driving torque, and movement speed. Spinal cord evoked potentials (SCEPs) are collected through implanted spinal cord electrodes. Electromyography (EMG) sensors are attached to the surface of preset muscle groups in the lower limbs to collect motor unit action potentials (MUAPs).

3. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 1, characterized in that, In the multi-source signal preprocessing and noise suppression steps, the collected multi-source data undergoes interference suppression and calibration processing. An adaptive filtering algorithm combining Kalman filtering and wavelet threshold denoising is used to denoise the EEG signals, removing SCS electrical stimulation interference, EXS motor noise, and physiological noise, while retaining the EEG signal frequency bands related to motor intention. Amplitude drift correction is performed on the SCS stimulation parameters. Zero-position calibration is performed on the EXS exoskeleton joint rotation data, which is performed before each training session. Linear trend elimination method is used to remove baseline drift from SCEP signals; The MUAP signal is bandpass filtered, and the peak value of the filtered MUAP signal is detected by a preset threshold to extract the muscle activation initiation time.

4. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 1, characterized in that, In the BCI motion intent recognition and accuracy verification step, a motion intent recognition model is constructed based on the preprocessed data and its accuracy is verified. Time-frequency features, spatial features, and ERD / ERS features are extracted from the EEG signal, and peak features of the SCEP signal are extracted. The extracted features are fused. A multimodal fusion deep learning model combining EEG-CNN and Transformer is used to train the fused features. Baseline data of a preset number of patients are used as the training set and validation set. The cross-entropy loss function is used to optimize the training of the model. The recognition accuracy and confusion matrix of multiple types of motion intentions were calculated using the intention recognition confidence formula. The robustness of the model in intention recognition under different SCS stimulus intensities was tested, and the accuracy and stability of the model were verified.

5. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 4, characterized in that, In the BCI motion intent recognition and accuracy verification step, the recognition accuracy and confusion matrix of multiple types of motion intents are calculated using the intent recognition confidence formula, which is as follows: ,in For the first Confidence level for identifying motion intent For the first EEG feature matching values ​​of intention-like behavior For the first SCEP feature matching value of class intent, The cross-modal correlation coefficient between EEG and SCS. , Match the maximum value for all intent features.

6. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 1, characterized in that, In the brain-spinal cord-muscle cascade pathway association analysis step, a multi-dimensional association model is established to analyze the conduction and synergistic characteristics of the neural cascade pathway. A time series cross-correlation algorithm is used to locate the EEG intention trigger point, SCEP peak time, and MUAP activation time, calculate the time difference between each time point, and take the average of the time differences of multiple repeated actions. The Granger causality test algorithm is used to analyze the causal influence of the EEG intention signal on SCS stimulation parameters, EXS movement trajectory, and muscle MUAP activation. Based on the BCI intention recognition results, the dynamic time warping algorithm is used to calculate the overlap between the EXS movement trajectory and the intention. The matching rate between the SCS stimulation segments and parameters and the intention is statistically analyzed.

7. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 6, characterized in that, In the brain-spinal cord-muscle cascade pathway association analysis step, the Granger causality test algorithm is used to analyze the causal influence of EEG intention signals on SCS stimulation parameters, EXS motor trajectory, and muscle MUAP activation. The algorithm formula is as follows: ,in, It is a Granger causality test of EEG intention signal to target signal. The F-statistic is used to quantify the causal driving strength of EEG intention on the target signal. The target signal includes SCS stimulation parameters, EXS motion trajectory, and MUAP activation signal. The larger the F-value, the more significant the active control association between EEG intention and the target signal. It is the sum of squared residuals when constructing a prediction model using the historical data of the target signal itself, reflecting the natural variation pattern of the target signal without external driving force. It is the sum of squared residuals when a prediction model is constructed using both historical data of the target signal itself and historical data of the EEG intention signal. It reflects the change in the prediction error of the target signal after incorporating EEG intention-driven parameters. It is the model lag order, determined based on the frequency of the path signal acquisition. It is the total number of signal samples used for modeling.

8. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 6, characterized in that, In the brain-spinal cord-muscle cascade pathway association analysis step, the dynamic time warping algorithm is used to calculate the overlap between the EXS motion trajectory and the intention. The algorithm formula is as follows: ,in, This represents the degree of overlap between the actual movement trajectory of the EXS exoskeleton and the standard trajectory corresponding to the EEG intention. The closer the value is to 1, the higher the matching accuracy between the EXS movement and the EEG intention. It is the actual trajectory data of EXS. With intention standard trajectory data The dynamic time-warped distance is calculated by constructing the distance matrix between the two trajectories and finding the optimal matching path. The smaller the distance, the higher the trajectory similarity. These are the actual motion trajectory data of key joints in the EXS exoskeleton. These are the standard motion trajectory data of key joints corresponding to the brainwave intent; It is the maximum DTW distance between the EXS trajectory and the standard trajectory under this type of intent, and is calibrated offline by simulating extreme scenarios when the EXS trajectory deviates completely from the intent.

9. The method for analyzing the improvement effect of implantable brain-computer interface-controlled spinal cord electrical stimulation on patients according to claim 1, characterized in that, In the comprehensive evaluation step of the multi-dimensional improvement effect, an evaluation system is constructed and quantified from the dimensions of neural pathways, motor function and long-term recovery. The evaluation indicators of the quantitative scoring system include pathway delay, SCEP amplitude change and muscle activation synchronicity. The total score of pathway integrity is calculated based on the scores of each evaluation indicator. Based on the Fugl-Meyer scale, a brain-controlled specific scoring dimension was added, including joint active range of motion, gait stability, and movement completion efficiency; the ability of patients to perform daily living activities with brain-controlled assistance was tested; a follow-up period was set to monitor changes in the energy of motor cortex rhythmic signals and changes in SCEP amplitude during the follow-up period, and to assess the neuroplasticity of the brain and spinal cord.

Images