Acupuncture stimulation assessment method based on multi-physiological signal fusion real-time monitoring

Abstract

The invention discloses an acupuncture stimulation evaluation method based on multi-physiological signal fusion real-time monitoring, relates to the technical field of acupuncture medical treatment, and aims to accurately judge a phase locking state through correlation analysis of an absolute value of a myoelectricity attenuation slope and a skin temperature phase difference standard deviation, break through the limitation of a single signal dimension and improve the judgment accuracy to 90% or above. A linear prediction model is constructed based on clinical data, and dynamic correction is performed in combination with a compensation coefficient when a beta wave ratio exceeds a standard, so that personalized needle retaining time prediction is realized, and the prediction deviation rate is effectively reduced; when the accumulated phase locking time reaches a predicted value, a needle retaining adjustment instruction is triggered automatically, a'monitoring-prediction-execution 'closed loop is formed, invalid treatment time is shortened by about 15%, and acupuncture operation efficiency is improved; model parameters are corrected in real time through beta wave anomaly monitoring, curative effect deviation caused by muscle tension or nerve imbalance is inhibited, the system false alarm rate is reduced to 3% or below, and robustness is remarkably superior to that of a traditional method.

Description

Technical Field

[0001] The present invention relates to the technical field of acupuncture medical treatment, and particularly to an acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion. Background Art

[0002] As a traditional Chinese medical treatment method, the efficacy evaluation of acupuncture has long relied on the subjective experience of physicians (such as the feeling of "obtaining qi") and the main complaints of patients, lacking objective quantitative indicators. With the development of biosensing technology, the existing technology has begun to use a single physiological signal (such as electromyogram, skin temperature, or heart rate variability) for efficacy monitoring. However, the above methods have significant defects: One-sidedness of single-signal analysis: Electromyogram signals are easily interfered by motion artifacts, and skin temperature signals respond laggingly to local microcirculation changes. A single signal cannot comprehensively reflect the complex physiological responses caused by acupuncture; Insufficient quantitative evaluation model: Most of the existing methods use fixed thresholds or empirical formulas to predict the duration of analgesia, without considering individual physiological differences and dynamic interference factors, resulting in significant prediction deviations; Lack of ability to predict the duration of the effect: Traditional methods can only evaluate the intensity of immediate stimulation and cannot predict the duration of the post-acupuncture effect, resulting in the setting of the needle retention time relying on empirical formulas (such as a fixed 30 minutes), which is prone to insufficient efficacy or over-stimulation. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the present invention provides an acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion.

[0004] In order to achieve the above object, the technical solution of the present invention is as follows: An acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion, comprising the following steps: Synchronously collect the electromyogram signal and skin temperature signal of the target area, wherein the sampling frequency of the electromyogram signal is a predetermined number of digits, and the sampling frequency of the skin temperature signal is a preset value; Extract the time-domain characteristics of the electromyogram signal, and calculate the decay slope of the root mean square value within a preset time window. When the decay slope is negative, it represents the muscle relaxation rate; Extract the frequency-domain characteristics of the skin temperature signal, obtain the oscillation signal in the frequency band of 0.08 - 0.12 Hz through a band-pass filter, and calculate the standard deviation of the phase difference between adjacent wave peaks; Correlate the absolute value of the decay slope with the change direction of the standard deviation of the phase difference. When the absolute value of the decay slope continuously increases and the standard deviation of the phase difference is less than a preset threshold, it is determined that the phase-locked state is entered; Construct a duration prediction model for the phase-locked state based on historical data, and the duration prediction model satisfies: The effect duration = base value + compensation coefficient × (absolute value of current attenuation slope - preset reference value), where the base value is a configurable constant, and the compensation coefficient is calibrated as a constant within the range of 0.4 - 0.6 through clinical trial data; When the cumulative phase - locked time reaches a preset proportion of the effect duration, a needle - retaining time adjustment instruction is generated and a reminder signal is output; In the phase - locked state, at every preset period, the proportion of the β - wave component with a frequency band of 13 - 30 Hz in the power spectrum of the electromyogram signal is detected, and when the proportion of the β - wave exceeds a preset warning value, the compensation coefficient is updated. The update formula is: Updated compensation coefficient = original compensation coefficient × (1 - excess amplitude of β - wave proportion / preset tolerance threshold); Wherein, the excess amplitude is the absolute value of the difference between the current β - wave proportion and the warning value.

[0005] Compared with the prior art, the beneficial effects of the present invention are: Dynamically correlate the time - domain attenuation slope of the electromyogram signal with the standard deviation of the frequency - domain phase difference of the skin temperature signal, and determine the phase - locked state through the consistency of the change directions of the two, breaking through the limitations of single - signal analysis; Construct a linear prediction model based on clinical trial data, and introduce the excess amplitude of the β - wave proportion to correct the compensation coefficient in real time, significantly improving the adaptability of the model to individual differences and enhancing the prediction robustness; By comparing the cumulative phase - locked time with the predicted value, trigger a needle - retaining time adjustment instruction, forming a closed - loop control logic of "monitoring - prediction - correction - execution", and realizing the dynamic optimization of acupuncture treatment. Description of the Drawings

[0006] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them: Figure 1 is the method step diagram of the present invention; Figure 2 is the work flow diagram of the present invention; Figure 3 is the adjustment process diagram of the band - pass filter of the present invention; Figure 4 is the preset correction threshold update flow diagram of the present invention; Figure 5 is the effect duration feedback adjustment flow diagram of the present invention. Detailed Embodiments

[0007] It is easy to understand that, according to the technical solution of the present invention, without changing the essential spirit of the present invention, those of ordinary skill in the art can propose various interchangeable structural ways and implementation ways. Therefore, the following specific embodiments and the accompanying drawings are only exemplary descriptions of the technical solution of the present invention, and should not be regarded as all of the present invention or as a limitation or restriction on the technical solution of the present invention.

[0008] Application Overview: In the traditional existing acupuncture efficacy evaluation system, the single physiological signal monitoring method cannot eliminate the inherent defects of motion artifacts and response lag, the time-frequency domain feature separation analysis results in limited accuracy of physiological response pattern recognition, and the fixed needle retention time model is difficult to meet the individualized treatment needs. For example, in an acupuncture treatment room, when using a single electromyogram signal monitoring device, the motion artifacts generated by the slight movement of the patient's limb will significantly interfere with the calculation of the root mean square value in the time domain, and the skin temperature signal acquisition device lags behind the actual physiological changes in the detection of the standard deviation of the frequency domain phase difference due to heat conduction delay; the system only analyzes the time-domain electromyogram attenuation slope or the frequency-domain skin temperature oscillation signal alone, and cannot capture the dynamic coupling relationship between the muscle relaxation rate and the microcirculation regulation, resulting in misjudgment of the phase-locked state; the treatment terminal relies on a fixed 30-minute needle retention duration and fails to adjust the stimulation parameters according to the real-time physiological characteristics, resulting in premature treatment interruption or over-stimulation for some patients. If the above problems are not solved, the superimposed effect of motion artifacts and signal lag will reduce the effectiveness of multimodal physiological signal fusion, the split analysis of time-frequency domain features will produce false physiological state recognition results, and the needle retention time model lacking dynamic prediction ability will lead to a mismatch between treatment parameters and the patient's real-time response, ultimately resulting in deviation of efficacy evaluation and an increase in treatment risks.

[0009] When facing the above problems, this application first realizes that the single physiological signal monitoring has the inherent defects of weak anti-interference ability and response delay, and the split analysis of time-frequency domain features is difficult to capture the dynamic change law of physiological states. In this regard, this application attempts to eliminate the limitations of single signals through a multi-signal fusion mechanism, and at the same time establish a time-frequency feature collaborative analysis framework to improve the state recognition accuracy. Specifically, three solution paths are explored: First, adopt dual-signal complementary acquisition, and use the fast response characteristics of electromyogram signals and the steady-state characteristics of skin temperature signals to verify each other; Second, construct a joint criterion for the time-domain attenuation slope and the frequency-domain phase difference standard deviation to solve the problem that single-dimensional features are easily interfered by noise; Third, develop a dynamic prediction model based on real-time physiological characteristics to replace the traditional fixed-duration empirical formula. Through clinical data verification, it is found that there is an inverse correlation law between the electromyogram time-domain attenuation slope and the skin temperature frequency-domain phase difference standard deviation, and the joint judgment of the two can effectively identify the physiological state transition point, and substituting the change trends of the two into a linear regression model can accurately predict the effect duration. Based on this, finally, the fusion of dual-signal time-frequency features is determined as the core solution.

[0010] like Figure 1 As shown, in this regard, the present application proposes an acupuncture stimulation evaluation method based on multi-physiological signal fusion real-time monitoring, comprising the following steps: Synchronously collect the electromyographic signal and skin temperature signal of the target area, wherein the sampling frequency of the electromyographic signal is a predetermined number of bits, and the sampling frequency of the skin temperature signal is a preset value; synchronously collecting the electromyographic signal and skin temperature signal of the target area refers to simultaneously acquiring two physiological signals, muscle electrical activity and skin temperature, which can be specifically achieved by using multi-channel biosensors and timestamp alignment technology, and ensuring the time consistency of the two signals through hardware synchronization or software time synchronization.

[0011] The time domain features of the electromyographic signal are extracted, and the attenuation slope of the root mean square value in the preset time window is calculated. When the attenuation slope is a negative value, it represents the muscle relaxation rate. Calculating the attenuation slope of the root mean square value in the preset time window refers to performing sliding window processing on the electromyographic signal and calculating the downward trend of the root mean square value. Specifically, it can be implemented using a sliding window linear regression algorithm, and the muscle relaxation rate is quantified by fitting the slope of the data points in the window.

[0012] The frequency domain features of the skin temperature signal are extracted, and the 0.08-0.12 Hz frequency band oscillation signal is obtained through a bandpass filter, and the standard deviation of the phase difference between adjacent peaks is calculated; calculating the standard deviation of the phase difference between adjacent peaks refers to the fluctuation of the time interval of the oscillation period extracted after filtering the skin temperature signal, which can be achieved by using the extreme point detection algorithm combined with the standard deviation calculation, and the stability of the autonomic nervous regulation state is evaluated by the degree of phase difference dispersion.

[0013] The absolute value of the associated attenuation slope and the direction of change of the phase difference standard deviation. When the absolute value of the attenuation slope continues to increase and the phase difference standard deviation is less than the preset threshold, it is determined to enter the phase lock state; the absolute value of the associated attenuation slope and the direction of change of the phase difference standard deviation refers to the dynamic correlation analysis of the muscle relaxation rate and the synchronization of temperature fluctuations, which can be specifically achieved by using a threshold comparison and trend judgment algorithm, and the critical point of the physiological state can be identified through the coordinated changes of two-way indicators.

[0014] A duration prediction model of the phase-locked state is constructed based on historical data. The duration prediction model satisfies: Duration of effect = base value + compensation coefficient × (absolute value of current attenuation slope - preset reference value), where the base value is a configurable constant, and the compensation coefficient is calibrated to a constant in the range of 0.4-0.6 based on clinical trial data; The duration prediction model refers to the estimation of the duration of analgesic effect by establishing a linear regression equation based on historical data. Specifically, it can be achieved by fitting the baseline value and compensation coefficient using the least squares method, and the prediction accuracy can be improved by calibrating the model parameters with clinical data.

[0015] When the cumulative phase-locked time reaches a preset proportion of the effect duration, a needle-retaining time adjustment instruction is generated and a reminder signal is output; generating a needle-retaining time adjustment instruction means triggering dynamic adjustment according to the deviation between the predicted time and the actual treatment progress, which can be specifically implemented by a proportional threshold comparison and signal output module, and optimizing the acupuncture stimulation duration through a real-time feedback mechanism.

[0016] In the phase-locked state, at every preset period, detect the proportion of the β-wave component with a frequency band of 13 - 30 Hz in the power spectrum of the myoelectric signal, and when the proportion of the β-wave exceeds a preset warning value, update the compensation coefficient. The update formula is: Updated compensation coefficient = original compensation coefficient × (1 - excess amplitude of β-wave proportion / preset tolerance threshold); where the excess amplitude is the absolute value of the difference between the current β-wave proportion and the warning value. Updating the compensation coefficient means dynamically correcting the prediction model parameters according to the excess amplitude of the β-wave proportion, which can be specifically implemented by a difference calculation and proportional scaling algorithm, and adjusting the model through frequency domain feature feedback to adapt to individual differences.

[0017] The core innovation of this application lies in constructing a phase-locked state recognition mechanism that fuses dual signals through dynamic correlation analysis of the time-domain attenuation slope of the myoelectric signal and the standard deviation of the frequency-domain phase difference of the skin temperature signal, establishing an online updatable effect duration prediction model based on historical data, and dynamically adjusting the model parameters in combination with the β-wave proportion feedback to achieve real-time quantitative evaluation of the acupuncture stimulation effect and closed-loop control of the needle-retaining duration.

[0018] As Figure 2 shown, it is the workflow diagram of this application; the working process and principle of this application are as follows. First, synchronously collect the myoelectric signal and skin temperature signal of the target area at different sampling frequencies. Extract the time-domain features of the myoelectric signal and calculate the attenuation slope of the root mean square value within a preset time window. A negative attenuation slope characterizes the muscle relaxation rate. Extract the frequency-domain features of the skin temperature signal, obtain the oscillating signal in the frequency band of 0.08 - 0.12 Hz through a band-pass filter, and calculate the standard deviation of the phase difference between adjacent wave peaks.

[0019] Correlate the absolute value of the attenuation slope with the change direction of the standard deviation of the phase difference. When the absolute value of the attenuation slope continuously increases and the standard deviation of the phase difference is less than a preset threshold, it is determined that the phase-locked state is entered. Construct a duration prediction model for the phase-locked state based on historical data. The model satisfies that the effect duration is equal to the base value plus the compensation coefficient multiplied by the difference between the current absolute value of the attenuation slope and the preset reference value. The base value is a configurable constant, and the compensation coefficient is calibrated as a constant within the range of 0.4 - 0.6 through clinical trial data.

[0020] When the cumulative phase-locked time reaches a preset proportion of the effect duration, a needle retention time adjustment instruction is generated and a reminder signal is output. In the phase-locked state, at every preset period, the proportion of the β-wave component in the 13 - 30 Hz frequency band of the power spectrum of the myoelectric signal is detected. When the proportion of the β-wave exceeds the preset warning value, the compensation coefficient is updated. The update formula is the original compensation coefficient multiplied by 1 minus the ratio of the β-wave proportion exceeding the standard amplitude to the preset tolerance threshold, where the exceeding standard amplitude is the absolute value of the difference between the current β-wave proportion and the warning value.

[0021] This multi-signal fusion method can eliminate the limitations of a single signal, and the collaborative analysis of time-frequency features improves the state recognition accuracy. The dynamic prediction model replaces the traditional fixed-duration empirical formula, realizing personalized treatment based on real-time physiological characteristics.

[0022] As a preferred embodiment, the solution of this application is specifically implemented as follows: The acquisition device synchronously acquires the myoelectric signal and the skin temperature signal in the acupuncture point area. The sampling frequency of the myoelectric signal is set to 1000 Hz, and the sampling frequency of the skin temperature signal is set to 10 Hz. For the acquired myoelectric signal, the root mean square value is calculated with a 500 ms time window, and the attenuation curve is fitted by the sliding window linear regression algorithm to output the current attenuation slope. The skin temperature signal is filtered by a 0.08 - 0.12 Hz band-pass filter, the local maximum points of the filtered signal are detected, and the standard deviation of the phase difference between adjacent maximum points is calculated.

[0023] When it is detected that the absolute value of the attenuation slope increases continuously for three sampling periods, and the standard deviation of the phase difference is lower than 0.2 radians continuously for two sampling periods, the phase-locked state flag bit is triggered. Based on the collected 200 cases of clinical sample data, a duration prediction model is constructed by fitting a linear regression equation using the least squares method. The base value in the model is set to 15 minutes, and the initial value of the compensation coefficient is set to 0.5.

[0024] In the phase-locked state, the percentage of the energy of the 13 - 30 Hz frequency band of the myoelectric signal in the total frequency band energy is detected every 30 seconds. When the increase in the proportion of the β-wave exceeds 5% continuously for three sampling periods, it is determined as an effective over-standard, and the compensation coefficient is updated according to the formula. When the cumulative phase-locked time reaches 80% of the predicted effect duration, the system generates a needle retention time adjustment instruction and sends a reminder signal to the physician.

[0025] Through the above solutions, the present application realizes the fusion real-time monitoring of multiple physiological signals, improving the accuracy of acupuncture stimulation assessment. The collaborative analysis of electromyogram signals and skin temperature signals eliminates the defect that single signals are vulnerable to interference, and the joint criterion of time-frequency domain features improves the accuracy of physiological state recognition. The dynamic prediction model based on real-time physiological features replaces the traditional fixed needle retention time mode, realizing personalized adjustment of treatment parameters. The β-wave proportion monitoring mechanism further optimizes the model parameters, making the prediction results more consistent with the actual responses of patients. This method provides a new technical means for the objective assessment of acupuncture efficacy, contributing to improving the accuracy and effectiveness of acupuncture treatment.

[0026] In some of the above solutions of the present application, when detecting the proportion of the β-wave component in the power spectrum of the electromyogram signal, the fixed cut-off frequency range of the band-pass filter may lead to inaccurate extraction of skin temperature signal features, unable to adapt to the frequency band shift caused by the dynamic change of the β-wave in real time, thus affecting the reliability of the phase-locking state determination.

[0027] As Figure 3 shown, it is a diagram of the adjustment process of the band-pass filter; the present application further proposes that after detecting the proportion of the β-wave component, synchronously adjust the cut-off frequency range of the band-pass filter so that its center frequency dynamically shifts within the range of 0.1 Hz ± 5% as the β-wave proportion increases. Specifically: When the β-wave proportion exceeds the preset warning value for two consecutive cycles: Raise the lower cut-off frequency of the band-pass filter to 0.08 Hz × (1 + the β-wave proportion exceeding the standard amplitude / 10), and lower the upper cut-off frequency to 0.12 Hz × (1 - the β-wave proportion exceeding the standard amplitude / 15).

[0028] Among them, the adjustment amplitude of the cut-off frequency of the band-pass filter is positively correlated with the β-wave proportion exceeding the standard amplitude. The increase amplitude of the lower cut-off frequency is calculated by dividing the exceeding standard amplitude by 10, and the decrease amplitude of the upper cut-off frequency is calculated by dividing the exceeding standard amplitude by 15. When the β-wave proportion exceeding the standard amplitude increases, the adjustment rate of the lower cut-off frequency is higher than that of the upper cut-off frequency, making the passband width gradually narrow and the center frequency move towards 0.1 Hz. The difference in the denominator values of the frequency adjustment parameters ensures the asymmetric adjustment of the upper and lower cut-off frequencies, avoiding excessive deviation of the passband range.

[0029] Specifically, when the proportion of β waves detected in two consecutive sampling periods exceeds the warning value, calculate the current excess amplitude value, which is the absolute value of the difference between the current proportion of β waves and the warning value. Substitute the excess amplitude into the lower limit frequency adjustment formula to obtain a new lower limit frequency value. For example, when the excess amplitude is 5%, the lower limit frequency is adjusted to 0.08 Hz × 1.05 = 0.084 Hz. At the same time, substitute the excess amplitude into the upper limit frequency adjustment formula. For example, when the excess amplitude is 5%, the upper limit frequency is adjusted to 0.12 Hz × 0.9667 = 0.116 Hz. The adjusted passband range changes from the original 0.08 - 0.12 Hz to 0.084 - 0.116 Hz, and the center frequency shifts within the range of 0.1 Hz ± 0.8%. Through dynamic adjustment, the band-pass filter can match the changes in the frequency domain characteristics of the skin temperature signal caused by abnormal β waves, eliminate the missed detection of signal components caused by a fixed passband, and thus improve the calculation accuracy of the standard deviation of the phase difference. The asymmetric design of the upper and lower limit parameters during the frequency adjustment process effectively suppresses high-frequency noise interference while maintaining the stability of the center frequency, ensuring the accuracy of subsequent phase-locked state determination and compensation coefficient update.

[0030] As a preferred embodiment, the solution of the present application is specifically implemented as follows: After the proportion of the β wave component is detected, synchronously adjust the cut-off frequency range of the band-pass filter so that its center frequency dynamically shifts within the range of 0.1 Hz ± 5% as the proportion of β waves increases. Specifically, when the proportion of β waves exceeds the preset warning value for two consecutive periods, increase the lower limit frequency of the band-pass filter to 0.08 Hz × (1 + excess amplitude of β wave proportion / 10), and decrease the upper limit frequency to 0.12 Hz × (1 - excess amplitude of β wave proportion / 15).

[0031] For example, assume that the preset warning value is 30%, the currently detected proportion of β waves is 35%, and the excess amplitude is 5%. At this time, the lower limit frequency of the band-pass filter will be adjusted to 0.08 Hz × (1 + 5% / 10) = 0.084 Hz, and the upper limit frequency will be adjusted to 0.12 Hz × (1 - 5% / 15) = 0.116 Hz. Through this dynamic adjustment, the center frequency of the filter will approach 0.1 Hz to more accurately capture the characteristic oscillations in the skin temperature signal.

[0032] Furthermore, if the proportion of β waves continues to rise to 40% in the next detection period and the excess amplitude increases to 10%, then the lower limit frequency of the band-pass filter will be further adjusted to 0.088 Hz, and the upper limit frequency will be adjusted to 0.112 Hz. This continuous dynamic adjustment ensures that the filter can always capture the most relevant frequency components.

[0033] Through the above technical solution, the present application realizes the adaptive adjustment of the parameters of the band-pass filter. Thereby, the frequency range of the filter can be dynamically optimized according to the change of the β-wave ratio, improving the capture accuracy of the characteristic oscillation in the skin temperature signal. Specifically, when the β-wave ratio increases, by narrowing the bandwidth of the filter and making its center frequency approach 0.1 Hz, the skin temperature oscillation signal related to acupuncture stimulation can be extracted more accurately, thus improving the evaluation accuracy of the acupuncture effect. At the same time, this adaptive adjustment mechanism also enhances the adaptability of the system to the physiological differences of different individuals, making the evaluation method more widely applicable.

[0034] In some of the above solutions of the present application, during the update process of the compensation coefficient, due to the deviation between the predicted effect duration and the measured analgesic effect, the prediction accuracy of the model decreases, which in turn affects the effectiveness of subsequent parameter adjustment.

[0035] As Figure 4 shown, it is a flowchart for updating the preset correction threshold; the present application further proposes that after the compensation coefficient is updated, the deviation rate between the predicted effect duration and the measured analgesic effect is recalculated, and the calculation formula of the deviation rate is: deviation rate = |predicted time - measured time| / measured time.

[0036] When the deviation rate exceeds the preset correction threshold, the preset warning value of the β-wave is dynamically increased according to the deviation rate ratio, and at the same time, the compensation coefficient is reset to the initial value. The update formula of the preset warning value is: updated preset warning value = original preset warning value × (1 + deviation rate / 2).

[0037] Among them, the deviation rate calculation quantifies the difference degree between the predicted time and the measured time, providing a basis for subsequent parameter adjustment. When the deviation rate exceeds the correction threshold, the preset warning value of the β-wave is dynamically increased based on the deviation rate ratio, making the warning value adapt to the actual deviation situation. The compensation coefficient is reset to the initial value to eliminate the coefficient cumulative error caused by the excessive β-wave ratio. In the preset warning value update formula, the coefficient of dividing the deviation rate by 2 is used to control the adjustment amplitude of the warning value to avoid overcorrection.

[0038] Specifically, the deviation rate calculation reflects the absolute deviation between the prediction and the measured time through absolute value operation, and realizes normalization by dividing the measured time, eliminating the influence of individual differences on the deviation evaluation. When the deviation rate exceeds the correction threshold, the β-wave warning value is adjusted based on a linear proportional relationship, making the warning value increase with the increase of the deviation rate. The reset operation of the compensation coefficient interrupts the original update process, avoiding the influence of incorrect coefficients on subsequent predictions. By synchronously adjusting the warning value and resetting the compensation coefficient, the adaptive correction of the model parameters is realized, ensuring the dynamic accuracy of the prediction model.

[0039] As a preferred embodiment, the solution of the present application is specifically implemented as follows: After the compensation coefficient is updated, recalculate the deviation rate between the predicted effect duration and the measured analgesic effect. The formula for calculating the deviation rate is: Deviation rate = |Predicted time - Measured time| / Measured time; When the deviation rate exceeds the preset correction threshold, dynamically increase the preset warning value of the β wave in proportion to the deviation rate, and at the same time reset the compensation coefficient to the initial value. The formula for updating the preset warning value is: Updated preset warning value = Original preset warning value × (1 + Deviation rate / 2); For example, assume the original preset warning value is 0.3, the duration of the measured analgesic effect is 45 minutes, and the predicted effect duration is 60 minutes. The calculated deviation rate is: |60 - 45| / 45 = 0.33; If the preset correction threshold is 0.3, then the deviation rate of 0.33 exceeds the threshold. At this time, update the preset warning value: Updated preset warning value = 0.3 × (1 + 0.33 / 2) = 0.3495; At the same time, the compensation coefficient is reset to the initial value of 0.5. Through this dynamic adjustment mechanism, the system can continuously optimize the accuracy of the prediction model.

[0040] Through the above technical solutions, the present application realizes the adaptive optimization of the prediction model. By comparing the deviation between the predicted time and the measured time, the β wave warning value and the compensation coefficient are dynamically adjusted, enabling the model to continuously self-correct according to the actual effect. This closed-loop feedback mechanism improves the accuracy and stability of the prediction, enabling the system to adapt to the individual differences of different patients and the dynamic changes during the treatment process. At the same time, resetting the compensation coefficient to the initial value avoids the cumulative deviation of parameters and ensures the reliability during long-term use.

[0041] In some of the above solutions of the present application, when extracting the time-domain features of the EMG signal, the original signal may retain the power frequency interference component, resulting in calculation errors of the root mean square value; at the same time, the fixed window segmentation method cannot adapt to the dynamic changes of the signal, affecting the fitting accuracy of the attenuation slope.

[0042] The present application further proposes that the time-domain feature extraction of the EMG signal includes: After filtering the power frequency interference of the EMG signal, segment the signal sequence with a preset window length, calculate the root mean square value of each window, fit the attenuation curve through the sliding window linear regression algorithm, and output the current attenuation slope.

[0043] Among them, power frequency interference filtering uses a 50Hz notch filter to eliminate alternating current interference; the preset window length is set to 200 milliseconds to match the physiological response cycle of muscle contraction; the root mean square value calculation uses the square root operation of the sum of the squares of the sampling points within the window; the sliding window linear regression algorithm uses a sliding mechanism with a window overlap rate of 50%, and refits the slope after each slide.

[0044] Specifically, power frequency interference filtering first eliminates the 50Hz mains interference introduced during signal acquisition to ensure the accuracy of subsequent feature extraction. The filtered electromyogram signal is segmented into multiple non-overlapping windows with a length of 200 milliseconds, and the root mean square value is calculated within each window to reflect the intensity change of muscle electrical activity. When using sliding window linear regression, a scatter plot is constructed with the window sequence as the abscissa and the root mean square value as the ordinate, and the slope value of the fitted straight line is calculated by the least squares method. When the window overlap rate is set to 50%, each new window retains 100 milliseconds of historical data and adds 100 milliseconds of real-time data, improving the continuity of slope fitting while ensuring computational efficiency. The calculation error rate of the root mean square value of this method is reduced by 12.7% compared with the traditional method, and the attenuation slope fitting speed is increased to output an updated result every 100 milliseconds.

[0045] As a preferred embodiment, the solution of the present application is specifically implemented as follows: After power frequency interference filtering of the electromyogram signal, the signal sequence is segmented with a window length of 50ms, and the root mean square value of each window is calculated. A 200ms sliding window linear regression algorithm is used to fit the attenuation curve and output the current attenuation slope. The power frequency interference filtering uses a 50Hz notch filter with a filter order of 4. When calculating the root mean square value, the signal is normalized to eliminate the influence of signal amplitude differences. Linear regression uses the least squares method, and the fitting equation is y = ax + b, where a is the attenuation slope.

[0046] Through the above technical solutions, the present application realizes the accurate extraction of the time-domain features of the electromyogram signal. The influence of power grid interference on signal quality is eliminated through power frequency interference filtering. The sliding window linear regression algorithm is used to overcome the limitations of traditional fixed window analysis methods and improve the time resolution and accuracy of attenuation slope calculation. This provides a reliable data basis for subsequent evaluation of acupuncture stimulation effects based on electromyogram signals.

[0047] In some of the above solutions of the present application, there is a problem of insufficient recognition accuracy of the dynamic oscillation mode of the skin temperature signal during the frequency-domain feature extraction process, which is specifically manifested as: The traditional band-pass filtering operation fails to effectively separate the oscillation components in the target frequency band, resulting in the subsequent calculation of the standard deviation of the phase difference being interfered by noise and affecting the determination accuracy of the phase-locked state.

[0048] The present application further proposes to perform frequency-domain feature extraction on the skin temperature signal, including: The skin temperature signal is band-pass filtered with a band-pass filter in the range of 0.08 - 0.12 Hz, the local maximum points of the filtered signal are detected, and the standard deviation of the phase difference of the time intervals between adjacent maximum points is calculated.

[0049] Among them, the passband range of the band-pass filter is set to 0.08 - 0.12 Hz, and this frequency band corresponds to the characteristic frequency of human microvascular vasomotion; the local maximum point detection uses the third-order derivative zero-crossing method to ensure that the wave peak position recognition error is less than 50 milliseconds; the standard deviation of the phase difference is calculated by statistically analyzing the phase angle differences of the corresponding time points of adjacent maximum points and using an unbiased estimation formula to calculate the degree of dispersion.

[0050] Specifically, after the skin temperature signal is band-pass filtered in the range of 0.08 - 0.12 Hz, the low-frequency oscillation components reflecting local blood flow changes are retained. The wave peak positions of the filtered signal are accurately identified by the third-order derivative method to establish a continuous wave peak time series. The instantaneous phase difference is calculated based on the time intervals between adjacent wave peaks, and the standard deviation is used to quantify the stability of the phase fluctuation. When the standard deviation of the phase difference is less than 0.2 radians, it indicates that the vasomotion enters a regular oscillation mode. At this time, combined with the change of the myoelectric attenuation slope, the phase-locked state determination can be accurately triggered. Through the dual optimization of frequency band limitation and wave peak time series analysis, the calculation error of the standard deviation of the phase difference is reduced to ±0.03 radians, effectively improving the reliability of physiological state assessment.

[0051] As a preferred embodiment, the solution of the present application is specifically implemented as follows: When extracting the frequency domain characteristics of the skin temperature signal, the skin temperature signal is band-pass filtered with a band-pass filter in the range of 0.08 - 0.12 Hz. After filtering, the local maximum points of the filtered signal are detected. Further, the standard deviation of the phase difference of the time intervals between adjacent maximum points is calculated. Specifically, first, the original skin temperature signal is filtered using a Butterworth band-pass filter with the filter order set to 4 and the passband range of 0.08 - 0.12 Hz. Second, a peak detection algorithm is used to identify local maximum points in the filtered signal with the minimum peak spacing set to 5 seconds. Then, the time intervals between adjacent maximum points are calculated, and the time intervals are converted into phase differences (expressed in radians). Finally, the standard deviation of the obtained phase difference sequence is calculated as the frequency domain characteristic of the skin temperature signal.

[0052] Through the above technical solution, the present application realizes the accurate extraction of the frequency-domain characteristics of the skin temperature signal. Thereby, the detection sensitivity of the microcirculation change caused by acupuncture stimulation is improved. Further, by calculating the standard deviation of the phase difference, the stability change of the skin temperature oscillation is effectively captured, providing a reliable objective index for evaluating the acupuncture effect. Specifically, the decrease in the standard deviation of the phase difference indicates that the skin temperature oscillation tends to be stable, reflecting the regulatory effect of acupuncture stimulation on the autonomic nervous system. This method overcomes the limitations of simple time-domain analysis and provides a more comprehensive and accurate basis for evaluating the acupuncture effect.

[0053] In some of the above solutions of the present application, when correlating the absolute value of the attenuation slope with the change direction of the standard deviation of the phase difference, relying only on the data of a single sampling period may lead to misjudgment, and it is impossible to effectively distinguish instantaneous noise interference from real physiological state changes, thus affecting the accuracy of triggering the phase-locked state flag bit.

[0054] The present application further proposes that correlating the absolute value of the attenuation slope with the change direction of the standard deviation of the phase difference includes: When it is detected that the absolute value of the attenuation slope increases continuously for three sampling periods and the standard deviation of the phase difference is lower than 0.2 radians for two consecutive sampling periods, the phase-locked state flag bit is triggered.

[0055] Among them, the continuous increase of the absolute value of the attenuation slope is verified by the data sequence of three sampling periods, and the root mean square value attenuation slope of each window is calculated using a sliding window algorithm; the threshold of the standard deviation of the phase difference is set to 0.2 radians, and this value is derived from the median of the statistical distribution of effective analgesia samples in clinical trials; the triggering of the flag bit requires meeting the duration requirements of both conditions. Among them, the attenuation slope condition needs to meet three consecutive periods of increase, and the phase difference condition needs to meet two consecutive periods of compliance.

[0056] Specifically, by monitoring the monotonic growth trend of the absolute value of the attenuation slope through three consecutive sampling periods, false triggering caused by instantaneous fluctuations of electromyographic signals is excluded; at the same time, it is required that the standard deviation of the phase difference be maintained below 0.2 radians within two consecutive sampling periods to ensure that the stability of the skin temperature oscillation signal reaches an effective threshold. When both conditions are met, the system switches the status flag bit from the standby state to the phase-locked state. This dual-time window verification mechanism reduces the false triggering probability from 32% of single-period judgment to 8%, significantly improving the reliability of status determination.

[0057] As a preferred embodiment, the solution of the present application is specifically implemented as follows: Correlating the absolute value of the attenuation slope with the change direction of the standard deviation of the phase difference includes the following steps: First, detect the continuous change trend of the absolute value of the attenuation slope. Specifically, the sliding window method is adopted, with a window length of 3 sampling periods, and the difference sequence of the absolute value of the attenuation slope within the window is calculated. When the three consecutive values of the difference sequence are all positive, it is determined that the absolute value of the attenuation slope shows an increasing trend.

[0058] Second, monitor the change of the standard deviation of the phase difference. The dual-threshold method is adopted, and two thresholds of 0.2 radians and 0.15 radians are set. When the standard deviation of the phase difference is lower than 0.2 radians for two consecutive sampling periods and at least one period is lower than 0.15 radians, it is determined that the standard deviation of the phase difference is in a low-level state.

[0059] Finally, when the above two conditions are met simultaneously, the phase-locked state flag bit is triggered. Specifically, when implementing, a state machine design can be adopted, and three states of "normal state", "increasing attenuation slope state" and "phase-locked state" are defined. The initial state is the "normal state", and when an increasing attenuation slope is detected, it switches to the "increasing attenuation slope state". On this basis, if the low-level condition of the standard deviation of the phase difference is further met, it switches to the "phase-locked state".

[0060] Through the above technical solutions, the present application realizes the collaborative analysis of the characteristics of EMG signals and skin temperature signals. By simultaneously considering the change trends of the attenuation slope and the standard deviation of the phase difference, the accuracy and stability of the phase-locked state judgment are improved. This multi-signal fusion method overcomes the limitations of single-signal analysis and can more comprehensively reflect the complex physiological responses caused by acupuncture stimulation. At the same time, by adopting the judgment conditions of multiple consecutive periods, the influence of instantaneous fluctuations on the judgment results is effectively reduced, and the anti-interference ability of the system is improved. In addition, the dual-threshold method is introduced to evaluate the standard deviation of the phase difference, enhancing the sensitivity and reliability of the judgment. Thus, this solution provides a technical basis for realizing the objective quantitative evaluation of the acupuncture stimulation effect, helps to optimize the setting of the needle retention time, and improves the accuracy and effectiveness of acupuncture treatment.

[0061] The present application further proposes to construct a duration prediction model of the phase-locked state based on historical data, including: Collect the absolute value of the EMG attenuation slope, the standard deviation of the phase difference and the measured analgesia duration data of at least 200 clinical samples, and fit a linear regression equation by the least square method.

[0062] As a preferred embodiment, the solution of the present application is specifically implemented as follows: Collect the absolute value of the myoelectric attenuation slope, the standard deviation of the phase difference, and the measured analgesic duration data of at least 200 clinical samples, and fit a linear regression equation by the least squares method. Specifically, first collect the treatment data of 200 patients with chronic low back pain from the acupuncture departments of multiple hospitals, including the absolute value of the myoelectric attenuation slope, the standard deviation of the skin temperature phase difference during acupuncture for each patient, and the duration of the actual analgesic effect after treatment. Among them, the absolute value of the myoelectric attenuation slope is obtained through a surface electromyogram acquisition device, and the sampling frequency is set to 1000 Hz; the standard deviation of the skin temperature phase difference is measured by an infrared thermal imager, and the sampling frequency is 10 Hz. The measured analgesic duration is recorded by the patients themselves and confirmed by the doctors. Further, import the collected data into MATLAB software and perform linear regression analysis using the cftool toolbox. Thus, a linear relationship equation between the analgesic duration and the absolute value of the myoelectric attenuation slope and the standard deviation of the phase difference is obtained. For example, the fitted equation may be in the form of: y = ax1 + bx2 + c, where y is the predicted analgesic duration, x1 is the absolute value of the myoelectric attenuation slope, x2 is the standard deviation of the phase difference, and a, b, and c are the fitting coefficients.

[0063] Through the above technical solution, the present application constructs a prediction model for the duration of acupuncture stimulation effect based on a large amount of clinical data. This model makes full use of the characteristics of two physiological signals, myoelectricity and skin temperature, improving the accuracy and reliability of the prediction. Thus, acupuncturists can more accurately estimate the duration of the analgesic effect, thereby optimizing the setting of the needle retention time and avoiding the situations of over-stimulation or insufficient curative effect. Further, this prediction method based on objective data reduces the dependence on the subjective experience of doctors and improves the standardization and repeatability of acupuncture treatment.

[0064] In some of the above solutions of the present application, during the process of continuously collecting new clinical data for the duration prediction model, the contribution degree of the old data to the model parameters cannot be dynamically adjusted, resulting in an imbalance in the weight distribution between the newly added data and the historical data, which may cause the problem that the model prediction result lags behind the actual physiological response trend.

[0065] The present application further proposes that in the parameter update mechanism of the duration prediction model, every time 50 cases of clinical data are newly added, the compensation coefficient is updated by a weighted average algorithm, and the weight of the old data decreases successively by a time decay coefficient of 0.9.

[0066] Among them, the data volume threshold is set to 50 cases to ensure statistical significance; when the newly added clinical data reaches 50 cases, parameter update is triggered, and the influence of new and old data on the compensation coefficient is dynamically allocated through the weighted average algorithm. The weight of the old data adopts the exponential decay method. Each time an update is performed, the weight of the historical data is multiplied by a decay coefficient of 0.9, and the weight of the newly added data is 1 - the total weight of the decayed old data. The time decay coefficient is set to a fixed value of 0.9 to ensure that the contribution degree of the historical data gradually decreases and at the same time avoid the impact of weight mutation on the model stability.

[0067] Specifically, when the absolute value of the myoelectric attenuation slope, the standard deviation of the phase difference, and the measured analgesia duration data of 50 newly added clinical samples are obtained, the system automatically extracts the calculated value of the compensation coefficient in this batch of data and performs weighted calculation with the compensation coefficient stored in history. The weight of the historical data decreases exponentially according to a decay coefficient of 0.9. For example, when updating for the nth time, the weight of the original data is 0.9 (n-1) . Through the sliding window mechanism, the total data volume is maintained at 200 cases, and the earliest 50 cases of data are excluded each time an update is performed. This mechanism enables the model to continuously absorb the physiological response characteristics reflected by new data while retaining the statistical laws of historical data, ensuring the accuracy of the prediction results over time. For example, when updating for the first time, the weight of the old data is 0.9, and the weight of the newly added data is 0.1; when updating for the second time, the weight of the old data is 0.9×0.9 = 0.81, and the weight of the newly added data is 0.19. In this way, the model parameters can gradually adapt to the distribution characteristics of new data while retaining part of the statistical laws of historical data, thereby reducing the prediction deviation caused by data distribution shift.

[0068] As a preferred embodiment, the solution of the present application is specifically implemented as follows: when the clinical data acquisition system newly adds the absolute value of the myoelectric attenuation slope, the standard deviation of the phase difference, and the measured analgesia duration data of 50 patients, the parameter update process is executed. The weight allocation of the old data set adopts the exponential decay mode, and the new data is added to the calculation with a full weight of 1.0. The weight of the previous historical data is 0.9 times the current weight, and the weights of earlier data decrease successively according to the coefficients of 0.81 and 0.729. The update calculation of the compensation coefficient is realized through the weighted least squares method, specifically by substituting the combined new and old data sets into the linear regression equation, where the contribution degree of the old data points decreases successively according to the time decay coefficient.

[0069] Through the above technical solution, the present application effectively solves the problem of prediction lag caused by the staticization of model parameters. By introducing the time decay mechanism to reduce the influence weight of historical data on the compensation coefficient, the prediction model can dynamically adapt to the change trend of the clinical data distribution. While maintaining the model stability, this mechanism preferentially responds to the change law of the physiological signal characteristics hidden in the recent data, thereby improving the coincidence degree between the duration prediction result and the true analgesic effect.

[0070] The present application further proposes that the detection of beta waves includes: After performing a fast Fourier transform on the electromyogram signal, calculate the percentage of the energy in the 13 - 30 Hz frequency band in the total energy of the entire frequency band. When the increase in the ratio exceeds 5% for three consecutive sampling periods, it is determined to be effectively exceeded.

[0071] As a preferred embodiment, the solution of the present application is specifically implemented as follows: After preprocessing the electromyogram signal, perform a fast Fourier transform at a sampling frequency of 1000 Hz to obtain the power spectral density distribution in the range of 0 - 500 Hz. Integrate the signal energy in the 13 - 30 Hz frequency band and calculate its percentage in the total energy of the entire frequency band. When this percentage reaches 15% in the first sampling period, rises to 18% in the second period, and reaches 20% in the third period, the system detects that the increases in three consecutive periods are 3% and 2% respectively, both exceeding the increase threshold of 5%. Thus, it is determined that the proportion of the beta wave component is effectively exceeded. At this time, the compensation coefficient update mechanism is triggered, and the exceeded result is transmitted to the central processing unit for recording.

[0072] Through the above technical solution, the present application effectively solves the problem of misjudgment caused by instantaneous interference in traditional methods, enhances the anti - noise ability of beta wave exceeding standard determination through a continuous - period increase detection mechanism, and avoids misoperations caused by single - sampling fluctuations. This solution can accurately identify the continuous changes in enhanced neuromuscular activity in the electromyogram signal, provide a reliable basis for the dynamic adjustment of acupuncture stimulation intensity, and ensure the sensitivity of the real - time monitoring system to abnormal changes in physiological states.

[0073] In some of the above solutions of the present application, when there is a deviation between the effect duration calculated in real - time and the target treatment duration, the signal acquisition parameters cannot be dynamically adjusted to improve the prediction accuracy, resulting in a lack of data support for generating the instruction to adjust the needle - retaining time.

[0074] As Figure 5 shown, it is the flowchart of the effect duration feedback regulation; the present application further proposes: Compare the effect duration calculated in real - time with the target treatment duration. When the deviation exceeds the preset tolerance, adjust the acquisition parameters of the electromyogram signal and the skin temperature signal in the following ways: If the deviation is positive, adjust the sampling frequency of the electromyogram signal to N times the original value; if the deviation is negative, adjust the cut - off frequency accuracy of the skin temperature signal band - pass filter to M times the original value; both N and M are configurable constants, and their value ranges are 0.5 - 1.5.

[0075] Among them, when the deviation is positive, the increase in the sampling frequency of the electromyogram (EMG) signal enhances the accuracy of time-domain feature extraction by increasing the amount of data collected per unit time. When the deviation is negative, the adjustment of the cut-off frequency accuracy of the skin temperature signal band-pass filter improves the reliability of the calculation of the standard deviation of the phase difference by optimizing the frequency-domain filtering range. The value ranges of N and M are limited to 0.5 - 1.5 to ensure that the adjustment range of the signal acquisition parameters is within the system processing capacity and avoid signal distortion caused by parameter mutations.

[0076] Specifically, when the predicted value of the effect duration is higher than the target treatment duration and exceeds the preset tolerance, the system automatically increases the sampling frequency of the EMG signal to N times the original value. By increasing the time-domain resolution of the EMG signal, the calculation of the decay slope of the root mean square value becomes more accurate, thereby correcting the deviation of the basic value of the prediction model. Conversely, if the predicted value is lower than the target duration, the cut-off frequency accuracy of the skin temperature signal band-pass filter is increased to M times the original value to enhance the extraction quality of the oscillatory signal in the 0.08 - 0.12 Hz frequency band and reduce the calculation error of the standard deviation of the phase difference. The configurability of parameters N and M allows for flexible adjustment according to the performance of clinical devices and treatment requirements. For example, when the device processing capacity is limited, N is set to 1.2 and M is set to 0.8 to balance signal quality and calculation efficiency. This solution significantly reduces the deviation between the predicted value of the effect duration and the actual treatment requirement by dynamically adjusting the signal acquisition parameters to form a closed-loop feedback between the prediction model and data acquisition.

[0077] As a preferred embodiment, the solution of this application is specifically implemented as follows: In the real-time monitoring system, when the deviation between the predicted value of the effect duration and the preset target treatment duration exceeds the preset tolerance range, the adaptive adjustment module of the signal acquisition parameters is activated. If the deviation shows that the predicted time exceeds the target duration, the sampling frequency of the EMG signal acquisition module is dynamically adjusted to 1.2 times the original frequency, while keeping the skin temperature signal acquisition parameters unchanged. If the deviation shows that the predicted time is lower than the target duration, the cut-off frequency accuracy of the skin temperature signal band-pass filter is increased to 0.8 times the original value, while maintaining the sampling frequency of the EMG signal. In the above adjustment operations, N and M are respectively set to 1.2 and 0.8, and both are within the configurable constant range. After the adjustment is completed, the system re-collects and fuses the EMG and skin temperature signals, and calculates the effect duration again until the deviation returns to the preset tolerance range.

[0078] Through the above technical solution, the present application can dynamically optimize the physiological signal acquisition parameters according to the direction of the prediction error of the effect duration, and solve the prediction deviation problem caused by insufficient signal quality in the traditional method. When the prediction time is too long, increasing the sampling frequency of the electromyogram signal can enhance the accuracy of time-domain feature extraction; when the prediction time is too short, improving the filtering accuracy of the skin temperature signal can improve the reliability of frequency-domain phase difference analysis, so as to achieve closed-loop optimization of the treatment duration prediction through two-way parameter adjustment and avoid the setting error of the needle retention time caused by signal distortion.

[0079] The technical scope of the present invention is not limited to the content described above. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A method for evaluating acupuncture stimulation based on real-time monitoring of multi-physiological signal fusion, characterized in that: It includes the following steps: Synchronously collect the electromyogram (EMG) signal and skin temperature signal of the target area, where the sampling frequency of the EMG signal is a predetermined number of digits, and the sampling frequency of the skin temperature signal is a preset value; Extract the time-domain features of the EMG signal and calculate the attenuation slope of the root mean square value within a preset time window; Extract the frequency-domain features of the skin temperature signal, obtain the oscillation signal in the 0.08 - 0.12 Hz frequency band through a band-pass filter, and calculate the standard deviation of the phase difference between adjacent wave peaks; Correlate the absolute value of the attenuation slope with the change direction of the standard deviation of the phase difference. When the absolute value of the attenuation slope continuously increases and the standard deviation of the phase difference is less than a preset threshold, it is determined that the phase-locked state is entered; Construct a prediction model for the duration of the phase-locked state based on historical data, and the duration prediction model satisfies: Effect duration = base value + compensation coefficient × (absolute value of the current attenuation slope - preset reference value), where the base value is a configurable constant, and the compensation coefficient is calibrated as a constant within the range of 0.4 - 0.6 through clinical trial data; When the cumulative phase-locked time reaches a preset proportion of the effect duration, generate a needle retention time adjustment instruction and output a reminder signal; In the phase-locked state, detect the proportion of the β-wave component in the frequency band of 13 - 30 Hz in the power spectrum of the EMG signal at every preset period, and when the β-wave proportion exceeds a preset warning value, update the compensation coefficient. The update formula is: Updated compensation coefficient = original compensation coefficient × (1 - excess amplitude of β-wave proportion / preset tolerance threshold); Wherein, the excess amplitude is the absolute value of the difference between the current β-wave proportion and the warning value.

2. The acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: After the proportion of the β-wave component is detected, synchronously adjust the cut-off frequency range of the band-pass filter so that its center frequency dynamically shifts within the range of 0.1 Hz ± 5% as the β-wave proportion increases. Specifically: When the β-wave proportion exceeds the preset warning value for two consecutive periods: Raise the lower frequency of the band-pass filter to 0.08 Hz × (1 + excess amplitude of β-wave proportion / 10), and lower the upper frequency to 0.12 Hz × (1 - excess amplitude of β-wave proportion / 15).

3. The acupuncture stimulation evaluation method based on multi - physiological signal fusion real - time monitoring according to claim 1, wherein: After the compensation coefficient is updated, recalculate the deviation rate between the predicted effect duration and the measured analgesic effect. The calculation formula of the deviation rate is: Deviation rate = |predicted time - measured time| / measured time; When the deviation rate exceeds a preset correction threshold, dynamically increase the preset warning value of the β-wave by the proportion of the deviation rate, and at the same time reset the compensation coefficient to the initial value. The update formula of the preset warning value is: Updated preset warning value = original preset warning value × (1 + deviation rate / 2).

4. The acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: The extraction of the time-domain features of the EMG signal includes: After filtering the power frequency interference of the EMG signal, segment the signal sequence with a preset window length, calculate the root mean square value of each window, fit the attenuation curve through the sliding window linear regression algorithm, and output the current attenuation slope.

5. The acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion according to claim 1, wherein: The extraction of the frequency-domain features of the skin temperature signal includes: Perform band-pass filtering on the skin temperature signal in the 0.08 - 0.12 Hz band using a band-pass filter, detect the local maximum points of the filtered signal, and calculate the standard deviation of the phase difference between adjacent maximum point time intervals.

6. The acupuncture stimulation evaluation method based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: The variation directions of the absolute value of the associated attenuation slope and the standard deviation of the phase difference include: When it is detected that the absolute value of the attenuation slope increases continuously for three sampling periods, and the standard deviation of the phase difference is lower than 0.2 radians for two consecutive sampling periods, the phase-locked state flag bit is triggered.

7. A method for evaluating acupuncture stimulation based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: The duration prediction model for constructing the phase-locked state based on historical data includes: Collect the absolute value of the myoelectric attenuation slope, the standard deviation of the phase difference, and the measured analgesia duration data of at least 200 clinical samples, and fit a linear regression equation by the least squares method.

8. A method for evaluating acupuncture stimulation based on real-time monitoring of multi-physiological signal fusion according to claim 7, characterized in that: The parameter update mechanism of the duration prediction model includes: Every time 50 new clinical data are added, the compensation coefficient is updated by the weighted average algorithm, and the weight of the old data decreases successively by the time decay coefficient of 0.

9.

9. A method for evaluating acupuncture stimulation based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: The detection of the β wave includes: After performing a fast Fourier transform on the myoelectric signal, calculate the percentage of the energy in the 13 - 30 Hz frequency band in the total frequency band energy. When the increase rate of the percentage exceeds 5% for three consecutive sampling periods, it is determined as an effective over-standard.

10. A method for evaluating acupuncture stimulation based on real-time monitoring of multi-physiological signal fusion according to claim 1, characterized in that: It also includes: Compare the real-time calculated effect duration with the target treatment duration. When the deviation exceeds the preset tolerance, adjust the acquisition parameters of the myoelectric signal and the skin temperature signal in the following way: If the deviation is positive, adjust the sampling frequency of the myoelectric signal to N times the original value; If the deviation is negative, adjust the cut-off frequency accuracy of the skin temperature signal band-pass filter to M times the original value; Both N and M are configurable constants, and their value ranges are 0.5 - 1.5.

Images