Electric stimulation method and system based on artificial intelligence

Abstract

The invention provides an artificial intelligence-based electrical stimulation method and system, and relates to the technical field of bioelectronics, and the method comprises the steps: firstly, synchronously collecting an electromyographic signal and a kinematics signal which are triggered by the electrical stimulation acting on a target muscle in a first period; then, analyzing and acquiring a first type of characteristic parameters representing a nerve recruitment dynamic state based on the electromyographic signal, and analyzing and acquiring a second type of characteristic parameters representing a muscle contraction mechanical state based on the kinematics signal; then, carrying out dynamic weight fusion on the first type of characteristic parameters and the second type of characteristic parameters to obtain a comprehensive muscle state index; then, according to the deviation between the comprehensive index and a preset target range, generating an adjusting instruction for adjusting the electrical stimulation parameters; and finally, adjusting the output electrical stimulation waveform parameters in real time according to the adjustment instruction in a second period following the adjustment instruction. Through multi-mode signal fusion and closed-loop feedback, self-adaptive accurate regulation and control of electrical stimulation parameters are realized.

Description

Technical Field

[0001] This application relates to the field of bioelectronics technology, and in particular to an artificial intelligence-based method and system for electrical stimulation. Background Technology

[0002] Functional electrical stimulation (FES) is a therapeutic technique that uses specific electrical currents to stimulate nerves or muscles to restore or improve their function. It is widely used in rehabilitation medicine, such as in the treatment of stroke sequelae, spinal cord injury, sports science, and neuromodulation.

[0003] Existing electrical stimulation devices typically employ preset, fixed stimulation programs or simple closed-loop control based on a single feedback signal, such as electromyography (EMG) amplitude. However, preset programs cannot be adjusted according to the user's real-time, dynamically changing physiological state. Relying solely on macroscopic signals such as EMG amplitude, they cannot distinguish between changes in neural drive and changes in muscle mechanical properties, making it difficult to identify fatigue or efficiency decline early. Furthermore, simple proportional control cannot handle the problem of multi-parameter synergistic optimization, making it difficult to achieve the optimal balance between activation effect and comfort / anti-fatigue performance. Summary of the Invention

[0004] This application provides an artificial intelligence-based electrostimulation method and system to improve the above-mentioned problems.

[0005] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, embodiments of this application propose an artificial intelligence-based electrical stimulation method, characterized in that the method is applicable to an electrical stimulation system, the electrical stimulation system including a controller, and the method is applicable to the controller, including: During the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle were collected. First-class feature parameters for characterizing neural recruitment dynamics are obtained based on electromyographic signals; The second type of feature parameters for characterizing the mechanical state of muscle contraction are obtained based on kinematic signal analysis; Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index; Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions are obtained for adjusting the electrical stimulation parameters; In the second cycle, the parameters of the electrical stimulation waveform output are adjusted based on the adjustment command. The second cycle is a continuous cycle following the first cycle.

[0006] In conjunction with the first aspect, in some implementations, a first type of feature parameters for characterizing neural recruitment dynamics are acquired based on electromyographic signals, including: The electromyographic signal is denoised and decomposed, and the power spectral density ratio of the target frequency band is obtained from the decomposed signal components. Extract the firing sequence of the motor unit of the target muscle and calculate the coefficient of variation corresponding to the firing interval of the motor unit; The power spectral density ratio and the coefficient of variation are combined to form the first type of characteristic parameter.

[0007] In conjunction with the first aspect, in some implementations, kinematic signal analysis obtains a second type of feature parameters for characterizing the mechanical state of muscle contraction, including: Velocity information is obtained based on kinematic signals, and the peak rate of target muscle contraction and the time required to reach the peak rate are obtained based on the velocity information. The peak rate and the time required are used as parameters to characterize the explosive force of contraction. Displacement information is obtained based on kinematic signals, and the effective mechanical work completed by the target muscle contraction is obtained based on a preset load model. The effective mechanical work is used as a parameter to characterize the contraction efficiency. The peak rate, the time required to reach the peak rate, and the effective mechanical work together constitute the second type of characteristic parameters.

[0008] In conjunction with the first aspect, in some implementations, dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index, including: Obtain the signal-to-noise ratio of the first type of feature parameters and the second type of feature parameters respectively; Initial fusion weights are assigned to the two types of feature parameters based on the data signal-to-noise ratio; Based on the consistency of the changing trends of the first type of feature parameters and the second type of feature parameters within adjacent time windows, the initial fusion weights are dynamically corrected. The modified weights are used to perform a weighted summation and normalization of the two types of feature parameters, outputting a comprehensive muscle state index.

[0009] In conjunction with the first aspect, in some implementations, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of comprehensive muscle state indicators from a preset target range, including: The comprehensive muscle status index is compared with the preset upper and lower thresholds. Based on the comparison results, it is determined whether the target muscle is in an under-activated state, an ideal state, or a fatigued state. If the target muscle is underactivated, the first adjustment strategy is determined, which is to primarily increase the stimulation intensity linearly and secondarily adjust the stimulation frequency. If the target muscle is in a state of fatigue, then the second adjustment strategy is determined. The second adjustment strategy is to reduce the stimulation frequency and adjust the ratio of pulse width to stimulation intensity. The first or second adjustment strategy is quantified into specific parameter adjustment amounts.

[0010] In conjunction with the first aspect, in some implementations, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of comprehensive muscle state indicators from a preset target range, including: Establish a three-dimensional parameter lookup table containing stimulus intensity, frequency, and pulse width. The three-dimensional parameter lookup table stores the optimized parameter combinations corresponding to different muscle state index ranges. Based on the current range of the comprehensive muscle status index, the corresponding parameter combination is retrieved from the lookup table as the adjustment target; Adjust the current electrical stimulation parameters to the target level.

[0011] In conjunction with the first aspect, in some embodiments, prior to acquiring electromyographic and kinematic signals induced by electrical stimulation of the target muscle during the first cycle, the following steps are included: Apply a set of low-intensity standard test electrical stimulation sequences; Obtain the basic physiological response signals corresponding to the test electrical stimulation sequence; Based on basic physiological response signals, individualized baseline values ​​of the first type of feature parameters and the second type of feature parameters are determined; Based on individualized baseline values, an initial target range for comprehensive muscle status indicators is set.

[0012] Secondly, embodiments of this application propose an artificial intelligence-based electrical stimulation system, which is configured as follows: During the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle were collected. First-class feature parameters for characterizing neural recruitment dynamics are obtained based on electromyographic signals; The second type of feature parameters for characterizing the mechanical state of muscle contraction are obtained based on kinematic signal analysis; Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index; Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions are obtained for adjusting the electrical stimulation parameters; In the second cycle, the parameters of the electrical stimulation waveform output are adjusted based on the adjustment command. The second cycle is a continuous cycle following the first cycle.

[0013] Optionally, in conjunction with the second aspect, in some embodiments, a first type of feature parameters for characterizing neural recruitment dynamics are acquired based on electromyographic signals, including: The electromyographic signal is denoised and decomposed, and the power spectral density ratio of the target frequency band is obtained from the decomposed signal components. Extract the firing sequence of the motor unit of the target muscle and calculate the coefficient of variation corresponding to the firing interval of the motor unit; The power spectral density ratio and the coefficient of variation are combined to form the first type of characteristic parameter.

[0014] In conjunction with the second aspect, optionally, in some embodiments, kinematic signal analysis obtains a second type of feature parameters for characterizing the mechanical state of muscle contraction, including: Velocity information is obtained based on kinematic signals, and the peak rate of target muscle contraction and the time required to reach the peak rate are obtained based on the velocity information. The peak rate and the time required are used as parameters to characterize the explosive force of contraction. Displacement information is obtained based on kinematic signals, and the effective mechanical work completed by the target muscle contraction is obtained based on a preset load model. The effective mechanical work is used as a parameter to characterize the contraction efficiency. The peak rate, the time required to reach the peak rate, and the effective mechanical work together constitute the second type of characteristic parameters.

[0015] In conjunction with the second aspect, optionally, in some embodiments, dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index, including: Obtain the signal-to-noise ratio of the first type of feature parameters and the second type of feature parameters respectively; Initial fusion weights are assigned to the two types of feature parameters based on the data signal-to-noise ratio; Based on the consistency of the changing trends of the first type of feature parameters and the second type of feature parameters within adjacent time windows, the initial fusion weights are dynamically corrected. The modified weights are used to perform a weighted summation and normalization of the two types of feature parameters, outputting a comprehensive muscle state index.

[0016] In conjunction with the second aspect, optionally, in some embodiments, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of comprehensive muscle state indicators from a preset target range, including: The comprehensive muscle status index is compared with the preset upper and lower thresholds. Based on the comparison results, it is determined whether the target muscle is in an under-activated state, an ideal state, or a fatigued state. If the target muscle is underactivated, the first adjustment strategy is determined, which is to primarily increase the stimulation intensity linearly and secondarily adjust the stimulation frequency. If the target muscle is in a state of fatigue, then the second adjustment strategy is determined. The second adjustment strategy is to reduce the stimulation frequency and adjust the ratio of pulse width to stimulation intensity. The first or second adjustment strategy is quantified into specific parameter adjustment amounts.

[0017] In conjunction with the second aspect, optionally, in some embodiments, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of comprehensive muscle state indicators from a preset target range, including: Establish a three-dimensional parameter lookup table containing stimulus intensity, frequency, and pulse width. The three-dimensional parameter lookup table stores the optimized parameter combinations corresponding to different muscle state index ranges. Based on the current range of the comprehensive muscle status index, the corresponding parameter combination is retrieved from the lookup table as the adjustment target; Adjust the current electrical stimulation parameters to the target level.

[0018] In conjunction with the second aspect, optionally, in some embodiments, prior to acquiring electromyographic and kinematic signals induced by electrical stimulation of the target muscle in the first cycle, the following steps are included: Apply a set of low-intensity standard test electrical stimulation sequences; Obtain the basic physiological response signals corresponding to the test electrical stimulation sequence; Based on basic physiological response signals, individualized baseline values ​​of the first type of feature parameters and the second type of feature parameters are determined; Based on individualized baseline values, an initial target range for comprehensive muscle status indicators is set.

[0019] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method proposed in the first aspect of the present invention.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention.

[0021] In summary, the above method and apparatus have the following technical effects: This application proposes an artificial intelligence-based electrical stimulation method and system. First, in the first cycle, electromyographic (EMG) and kinematic signals induced by electrical stimulation applied to the target muscle are simultaneously acquired. Then, based on EMG signal analysis, a first type of feature parameter characterizing the dynamics of nerve recruitment is obtained, and based on kinematic signal analysis, a second type of feature parameter characterizing the muscle contraction mechanics is obtained. Next, the first and second type feature parameters are dynamically weighted and fused to obtain a comprehensive muscle state index. Then, based on the deviation of this comprehensive index from a preset target range, adjustment instructions for adjusting the electrical stimulation parameters are generated. Finally, in the subsequent second cycle, the output electrical stimulation waveform parameters are adjusted in real time according to the adjustment instructions. This invention achieves adaptive and precise control of electrical stimulation parameters through multimodal signal fusion and closed-loop feedback. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating an artificial intelligence-based electrical stimulation method proposed in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This application proposes an artificial intelligence-based electrostimulation method, characterized in that the method is applicable to an electrostimulation system, the electrostimulation system including a controller, and the method is applicable to the controller. (See also...) Figure 1 The method includes the following steps: S101: In the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle are acquired.

[0025] Understandably, during the current stimulation cycle, i.e., the first cycle, electromyographic signals of the target muscle are synchronously acquired via surface electrodes, while kinematic signals related to muscle contraction, such as displacement, angle, and acceleration, are acquired simultaneously via sensors such as inertial measurement units, accelerometers, and gyroscopes. The acquisition of these two types of signals is strictly synchronized in time to ensure the correspondence in subsequent feature analysis.

[0026] Understandably, within the current, complete electrical stimulation output cycle, two types of physiological response signals generated by the target muscle in response to the electrical stimulation of that cycle are captured synchronously: one type is the electromyographic signal collected using surface electrodes, which is essentially the sum of bioelectrical activities generated by muscle fibers under the drive of nerve electrical impulses, directly reflecting the intensity and pattern of the motor drive commands issued by the central nervous system; the other type is the kinematic signal collected using motion sensors such as inertial measurement units, such as acceleration, angular velocity, or displacement, which objectively records the actual kinematic changes of the external limbs or body parts caused by muscle contraction, quantifying the mechanical output effect of the muscle.

[0027] It is understandable that electrical stimulation therapy varies from person to person; therefore, in a preferred embodiment, an individualized calibration step is included before performing step S101.

[0028] First, a set of low-intensity standard test electrical stimulation sequences is applied, for example, at a frequency of 20 Hz and a pulse width of 200 μs, with the intensity slowly increasing from the sensory threshold to the motor threshold. Understandably, the current intensity is not constant, but rather starts from a low sensory threshold where the user can only perceive a slight tingling sensation, and is slowly and continuously increased in a ramp or stepwise manner until the target muscle exhibits the first visible or palpable slight twitching, at which point the user's individualized motor threshold has been reached.

[0029] Then, the basic physiological response signals generated during this sequence are collected and recorded. These basic signals are then analyzed to determine individualized baseline values ​​for the first and second type of characteristic parameters in subsequent steps, such as resting signal amplitude, noise level, and basic motor unit firing frequency. Based on these baseline values, an initial target range for comprehensive muscle state indicators is set to provide a personalized benchmark for subsequent real-time adjustments.

[0030] S102: First-class feature parameters for characterizing neural recruitment dynamics based on electromyographic signals.

[0031] It is understandable that neural control information is analyzed from both macroscopic and microscopic levels.

[0032] In one specific embodiment, the step further includes: S1021: Bandpass filtering is performed on the acquired raw electromyography signals, such as 20-500Hz, to remove power frequency interference and motion artifacts. Then, signal decomposition is performed using methods such as wavelet transform and empirical mode decomposition.

[0033] For example, the lower limit of 20 Hz is designed to filter out low-frequency motion artifacts caused by slow limb movement or changes in skin electrode contact; the upper limit of 500 Hz is used to filter out 50 Hz or 60 Hz power frequency interference and its harmonics in the environment. Through this operation, the key frequency bands that are actually generated by muscle action potentials and are buried in invalid noise in the original signal are highlighted, thus improving the signal-to-noise ratio of the signal.

[0034] S1022: Calculate the proportion of the power spectral density of the target frequency band, such as the mid-frequency band of 30-150Hz, to the total power spectral density from the decomposed signal components.

[0035] Understandably, the physiological basis for this ratio is related to the muscle fatigue process. During sustained or repetitive muscle contraction, the metabolic environment within muscle fibers changes, leading to a slowdown in the conduction velocity of action potentials along the muscle fiber membrane. This decrease in action potential conduction velocity causes a shift in the distribution of its frequency components towards lower frequencies. Therefore, the decrease in the proportion of energy in the 30 Hz to 150 Hz mid-frequency range has been observed as an electrophysiological phenomenon accompanying muscle fatigue.

[0036] By continuously calculating this ratio, the system obtains a time-series parameter reflecting changes in the electrophysiological state of the muscle. This parameter serves as an electrical signal indicator for assessing whether the muscle is approaching fatigue, providing a basis for subsequent adjustments to electrical stimulation parameters.

[0037] S1023: Using blind source separation algorithms, such as FastICA, or convolutional kernel compensation algorithms, the action potential sequences of individual motor units are separated from high-density surface electromyography (EMG) signals. These motor units are identified and tracked, and their firing sequences, i.e., the time points of each firing, are extracted. The coefficient of variation (standard deviation / mean) of the time difference between adjacent firing intervals for each motor unit is calculated. An increase in this coefficient indicates that the neural firing rhythm has become unstable, which is a sensitive indicator of early neurological fatigue.

[0038] Understandably, blind source separation algorithms or convolution kernel compensation algorithms are used to treat the high-density surface electromyography (EMG) signals acquired from multiple channels as a linear mixture of multiple independent source signals (i.e., action potential sequences generated by different motor units). These source signals are then estimated and separated using mathematical methods to obtain the action potential waveform of a single motor unit and its time sequence.

[0039] After obtaining signals from multiple motion units, the system continuously identifies and tracks each identifiable motion unit. This means that the system can correlate the action potentials generated by repeated discharges of the same motion unit in continuous signal segments, thereby extracting the complete discharge time sequence of that unit.

[0040] Based on the firing time sequence of each motor unit, the system calculates the time interval between adjacent firing intervals. Subsequently, the coefficient of variation (COP) of these firing intervals is calculated, which is the standard deviation of the firing interval divided by its mean. This COP is a dimensionless statistic used to quantify the dispersion of the firing intervals around their mean. An increase in the COP indicates that the firing rhythm of the spinal motor neurons innervating that motor unit has become disordered. This instability of neural firing rhythm has been shown to occur before a decline in muscle mechanical output and is a sensitive electrophysiological marker of adaptive changes or early fatigue in the central or peripheral nervous system. Therefore, this step uses an algorithm to monitor the stability of neural control.

[0041] S103: Obtain the second type of feature parameters for characterizing the mechanical state of muscle contraction based on kinematic signal analysis.

[0042] This step first involves mathematical processing of the kinematic signals. The contraction velocity is obtained by differentiating the displacement signal or integrating the acceleration signal. Based on the obtained velocity information, the system identifies a single contraction cycle and calculates its maximum velocity value, i.e., the peak contraction rate. Simultaneously, the time elapsed from the contraction initiation point to reaching this peak velocity is calculated. These two parameters together describe the muscle's rapid initiation capability and explosive power characteristics.

[0043] Furthermore, step S103 may include the following steps: S1031: Obtain velocity information based on kinematic signals, and obtain the peak rate of target muscle contraction and the time required to reach the peak rate based on the velocity information. The peak rate and the time required are used as parameters to characterize the explosive force of contraction.

[0044] Understandably, the first stage involves acquiring velocity information from the raw signal. The system performs mathematical processing on the acquired kinematic signals. If the raw signal is acceleration, it is integralized over time; if the raw signal is displacement, it is differentially derived over time. This calculation yields a continuous waveform showing the velocity of muscle contraction or limb movement over time.

[0045] The second stage involves extracting two key parameters from the velocity waveform. First, the system identifies and outputs the maximum absolute value that the velocity waveform can reach within a complete contraction cycle, i.e., the peak contraction rate. Second, the system calculates the time interval from the start of the contraction action to the moment the peak rate is reached, i.e., the time required to reach the peak rate.

[0046] In biomechanics, the explosive force of muscle contraction is defined by its ability to produce a change in velocity per unit time. Peak contraction rate directly reflects the highest velocity level that a muscle can achieve during contraction and is a measure of the magnitude of explosive force. The time required to reach peak rate reflects the speed of the muscle's response from initiation to maximum velocity output and is a measure of the efficiency of explosive force generation. Therefore, combining these two parameters allows for the characterization of the explosive force characteristics of muscle contraction.

[0047] S1032: Obtain displacement information based on kinematic signals, and obtain the effective mechanical work completed by the target muscle contraction based on a preset load model. The effective mechanical work is used as a parameter characterizing the contraction efficiency.

[0048] First, the system acquires displacement information based on the original kinematic signals. If the original signal is acceleration, it is converted into displacement through two time integration operations; if it is velocity, it is converted into displacement through one integration. This yields the motion trajectory and displacement of the muscle attachment point or related limb segment during contraction.

[0049] The system then combines this displacement information with a pre-defined load model to calculate the effective mechanical work. This load model includes parameters of the mechanical environment against which the work is done, such as the magnitude and direction of gravity acting on the limb segments, the inertial characteristics of the musculoskeletal system, and any possible elastic or damping loads. The calculation of effective mechanical work essentially estimates the energy transferred by muscle contraction in overcoming these external loads and generating actual displacement.

[0050] S1033: The peak rate, the time required to reach the peak rate, and the effective mechanical work together constitute the second type of characteristic parameter.

[0051] Understandably, the peak rate and the time required to reach that rate together quantify the speed characteristics and rapid force exertion capability of contraction, i.e., the explosive force dimension, while the effective mechanical work quantifies the total amount of contraction that ultimately resists external loads and completes mechanical output, i.e., the efficiency dimension.

[0052] S104: Perform dynamic weight fusion of the first type of feature parameters and the second type of feature parameters, and obtain a comprehensive muscle state index.

[0053] Specifically, the signal-to-noise ratio (SNR) of the first and second type of feature parameters is obtained. Initial fusion weights are assigned to the two types of feature parameters based on the SNR. The initial fusion weights are dynamically adjusted based on the consistency of the changing trends of the first and second type of feature parameters within adjacent time windows. The adjusted weights are then used to perform weighted summation and normalization on the two types of feature parameters to output a comprehensive muscle state index.

[0054] Understandably, the system calculates the signal-to-noise ratio of each of the two types of parameters and analyzes whether their changing trends are consistent within adjacent time windows.

[0055] Based on the above evaluation, the system assigns and continuously adjusts the fusion weights for the two types of parameters. The weight allocation follows a dynamic rule: the parameter category with a higher signal-to-noise ratio will receive a higher initial weight to ensure the reliability of the evaluation. At the same time, if the changing trends of the two types of parameters are highly consistent, their weights will be enhanced. If the trends diverge, the system will adjust the weights according to preset logic, such as increasing the influence of parameter categories that are more relevant to the current external task.

[0056] Using this dynamically determined set of weights, a weighted sum is calculated for the two types of normalized feature parameters. The result is mapped to a scalar value, namely the comprehensive muscle state index, through a preset scaling function. This index integrates both neural drive stability and muscle output efficiency, becoming a single quantitative basis for comprehensively judging muscle activation level, fatigue state, or functional efficiency.

[0057] S105: Based on the deviation between the comprehensive muscle state index and the preset target range, obtain adjustment instructions for adjusting the electrical stimulation parameters.

[0058] Specifically, the comprehensive muscle state index is compared with preset upper and lower thresholds. Based on the comparison results, it is determined whether the target muscle is in an underactivated state, an ideal state, or a fatigued state. If the target muscle is in an underactivated state, the first adjustment strategy is adopted, which mainly involves linearly increasing the stimulation intensity and secondarily adjusting the stimulation frequency. If the target muscle is in a fatigued state, the second adjustment strategy is adopted, which mainly involves reducing the stimulation frequency and secondarily adjusting the ratio of pulse width to stimulation intensity. The first or second adjustment strategy is quantified into specific parameter adjustment amounts.

[0059] The system compares comprehensive muscle condition indicators to a preset target range. This target range is typically defined by a lower threshold and an upper threshold, representing the ideal working interval for muscle maintenance. The system calculates the deviation of the current indicator value from the boundary or midpoint of this range.

[0060] Based on the direction and magnitude of this deviation, the current functional state of the muscle can be determined. Typical state classifications include: under-activation state (indicators are consistently below the lower limit), ideal state (indicators are maintained within the target range), and fatigue state (indicators have reached or exceeded the upper limit).

[0061] Then, based on different state determinations, the system invokes preset adjustment strategies. For example, if "insufficient activation" is determined, an instruction is generated centered on gradually increasing the intensity of the stimulation current; if "fatigue" is determined, an instruction is generated centered on reducing the stimulation frequency and adjusting the pulse waveform. These strategies specify which stimulation parameters(s) to adjust, and the basic direction of the adjustment. Finally, the system quantifies the abstract adjustment strategy into specific parameter adjustments. For example, "increasing stimulation intensity" is transformed into "increasing the current amplitude by 0.5 mA". This quantification process may be based on simple linear rules or on more complex mapping relationships that include historical data. The final output adjustment instructions are a set of explicit electrical stimulation parameter modification values ​​that can be executed by the stimulator, thus completing the closed-loop control from physiological feedback to physical intervention.

[0062] S106: In the second cycle, the parameters of the electrical stimulation waveform of the subsequent output are adjusted based on the adjustment command, wherein the second cycle is a continuous cycle following the first cycle.

[0063] Specifically, a three-dimensional parameter lookup table containing stimulation intensity, frequency, and pulse width can be established. The three-dimensional parameter lookup table stores optimized parameter combinations corresponding to different muscle state index intervals. Based on the interval in which the current comprehensive muscle state index is located, the corresponding parameter combination is retrieved from the lookup table as the adjustment target; and the current electrical stimulation parameters are adjusted to the adjustment target.

[0064] Understandably, in the next consecutive electrical stimulation output cycle, i.e. the second cycle, the system immediately applies the adjustment command calculated at the end of the first cycle to update the electrical stimulation parameters output by the waveform generator.

[0065] One specific implementation method employs an intelligent mapping mechanism based on a three-dimensional parameter lookup table. The system pre-constructs a parameter space lookup table with three dimensions: stimulation intensity, stimulation frequency, and pulse width. This table stores a large number of pre-optimized or learned parameter combinations, each corresponding to a specific range of comprehensive muscle state indicators. For example, an indicator value in the 60-70 range corresponds to one set of parameters, while a value in the 71-80 range corresponds to another set.

[0066] When an adjustment command needs to be generated, the system determines the index range to which the calculated comprehensive muscle state index belongs based on the current value. Subsequently, the system directly retrieves the set of optimized parameters pre-bound to that range from the lookup table and sets this set as the target parameter for this adjustment.

[0067] Finally, the system controls the electrical stimulation output module to smoothly transition the current stimulation parameters or directly set them to the target parameters. In this way, the system achieves a rapid and stable mapping from continuous physiological state assessment to a discrete, optimized set of stimulation parameters, thus completing a closed loop and ensuring the precision and consistency of therapeutic intervention.

[0068] This application proposes an artificial intelligence-based electrical stimulation method. First, during the first cycle, electromyographic (EMG) and kinematic (K) signals induced by electrical stimulation applied to the target muscle are simultaneously acquired. Then, based on EMG signal analysis, a first type of feature parameter characterizing the dynamics of nerve recruitment is obtained, and based on Kinematic signal analysis, a second type of feature parameter characterizing the muscle contraction mechanical state is obtained. Next, the first and second type feature parameters are dynamically weighted and fused to obtain a comprehensive muscle state index. Then, based on the deviation of this comprehensive index from a preset target range, adjustment instructions for adjusting the electrical stimulation parameters are generated. Finally, during the subsequent second cycle, the output electrical stimulation waveform parameters are adjusted in real time according to the adjustment instructions. This invention achieves adaptive and precise control of electrical stimulation parameters through multimodal signal fusion and closed-loop feedback.

[0069] Based on the same inventive concept, this application also proposes an artificial intelligence-based electrical stimulation system, which is configured as follows: During the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle were collected. First-class feature parameters for characterizing neural recruitment dynamics are obtained based on electromyographic signals; The second type of feature parameters for characterizing the mechanical state of muscle contraction are obtained based on kinematic signal analysis; Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index; Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions are obtained for adjusting the electrical stimulation parameters; In the second cycle, the parameters of the electrical stimulation waveform output are adjusted based on the adjustment command. The second cycle is a continuous cycle following the first cycle.

[0070] Optionally, in some implementations, a first type of feature parameter for characterizing neural recruitment dynamics is obtained based on electromyographic signals, including: The electromyographic signal is denoised and decomposed, and the power spectral density ratio of the target frequency band is obtained from the decomposed signal components. Extract the firing sequence of the motor unit of the target muscle and calculate the coefficient of variation corresponding to the firing interval of the motor unit; The power spectral density ratio and the coefficient of variation are combined to form the first type of characteristic parameter.

[0071] Optionally, in some implementations, kinematic signal analysis obtains a second type of feature parameters characterizing the mechanical state of muscle contraction, including: Velocity information is obtained based on kinematic signals, and the peak rate of target muscle contraction and the time required to reach the peak rate are obtained based on the velocity information. The peak rate and the time required are used as parameters to characterize the explosive force of contraction. Displacement information is obtained based on kinematic signals, and the effective mechanical work completed by the target muscle contraction is obtained based on a preset load model. The effective mechanical work is used as a parameter to characterize the contraction efficiency. The peak rate, the time required to reach the peak rate, and the effective mechanical work together constitute the second type of characteristic parameters.

[0072] Optionally, in some implementations, dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index, including: Obtain the signal-to-noise ratio of the first type of feature parameters and the second type of feature parameters respectively; Initial fusion weights are assigned to the two types of feature parameters based on the data signal-to-noise ratio; Based on the consistency of the changing trends of the first type of feature parameters and the second type of feature parameters within adjacent time windows, the initial fusion weights are dynamically corrected. The modified weights are used to perform a weighted summation and normalization of the two types of feature parameters, outputting a comprehensive muscle state index.

[0073] Optionally, in some embodiments, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of the comprehensive muscle state index from a preset target range, including: The comprehensive muscle status index is compared with the preset upper and lower thresholds. Based on the comparison results, it is determined whether the target muscle is in an under-activated state, an ideal state, or a fatigued state. If the target muscle is underactivated, the first adjustment strategy is determined, which is to primarily increase the stimulation intensity linearly and secondarily adjust the stimulation frequency. If the target muscle is in a state of fatigue, then the second adjustment strategy is determined. The second adjustment strategy is to reduce the stimulation frequency and adjust the ratio of pulse width to stimulation intensity. The first or second adjustment strategy is quantified into specific parameter adjustment amounts.

[0074] Optionally, in some embodiments, adjustment instructions for adjusting electrical stimulation parameters are obtained based on the deviation of the comprehensive muscle state index from a preset target range, including: Establish a three-dimensional parameter lookup table containing stimulus intensity, frequency, and pulse width. The three-dimensional parameter lookup table stores the optimized parameter combinations corresponding to different muscle state index ranges. Based on the current range of the comprehensive muscle status index, the corresponding parameter combination is retrieved from the lookup table as the adjustment target; Adjust the current electrical stimulation parameters to the target level.

[0075] Optionally, in some embodiments, prior to acquiring electromyographic and kinematic signals induced by electrical stimulation of the target muscle in the first cycle, the following steps are included: Apply a set of low-intensity standard test electrical stimulation sequences; Obtain the basic physiological response signals corresponding to the test electrical stimulation sequence; Based on basic physiological response signals, individualized baseline values ​​of the first type of feature parameters and the second type of feature parameters are determined; Based on individualized baseline values, an initial target range for comprehensive muscle status indicators is set.

[0076] This application proposes an artificial intelligence-based electrical stimulation system. First, in the first cycle, electromyographic (EMG) and kinematic (K) signals induced by electrical stimulation applied to the target muscle are simultaneously acquired. Then, based on EMG signal analysis, a first type of feature parameter characterizing the dynamics of nerve recruitment is obtained, and based on Kinematic signal analysis, a second type of feature parameter characterizing the muscle contraction mechanical state is obtained. Next, the first and second type feature parameters are dynamically weighted and fused to obtain a comprehensive muscle state index. Then, based on the deviation of this comprehensive index from a preset target range, adjustment instructions for adjusting the electrical stimulation parameters are generated. Finally, in the subsequent second cycle, the output electrical stimulation waveform parameters are adjusted in real time according to the adjustment instructions. This invention achieves adaptive and precise control of electrical stimulation parameters through multimodal signal fusion and closed-loop feedback.

[0077] Based on the same inventive concept, embodiments of this application also propose an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the artificial intelligence-based electrostimulation method of the present application embodiments.

[0078] Furthermore, to achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the artificial intelligence-based electrostimulation method of this application.

[0079] The following is a detailed introduction to the various components of the electronic device: In this context, the processor is the control center of the electronic device. It can be a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0080] Alternatively, the processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0081] The memory is used to store the software program that executes the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can be referred to the above method embodiment, which will not be repeated here.

[0082] Optionally, the memory can be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory can be integrated with the processor or exist independently and coupled to the processor through the interface circuit of the electronic device; the embodiments of the present invention do not specifically limit this.

[0083] A transceiver is used to communicate with network devices or with terminal devices.

[0084] Optionally, the transceiver may include a receiver and a transmitter. The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0085] Optionally, the transceiver can be integrated with the processor or exist independently and coupled to the processor through the router's interface circuit. This embodiment of the invention does not specifically limit this.

[0086] Furthermore, the technical effects of the electronic device can be referred to the technical effects of the data transmission method in the above method embodiments, and will not be repeated here.

[0087] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0088] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0089] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0090] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0091] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0092] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0093] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

Claims

1. An artificial intelligence-based electrical stimulation method, characterized by, The method is applicable to an electrical stimulation system, the electrical stimulation system including a controller, and the method is applicable to the controller, including: During the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle were collected. Based on the electromyographic signals, a first type of feature parameter is obtained to characterize the dynamics of neural recruitment; Based on the kinematic signal analysis, a second type of feature parameter is obtained to characterize the mechanical state of muscle contraction. Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index; Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions for adjusting the electrical stimulation parameters are obtained; In the second cycle, the parameters of the output electrical stimulation waveform are adjusted based on the adjustment command, wherein the second cycle is a continuous cycle following the first cycle.

2. The method of claim 1, wherein the method is based on artificial intelligence. Based on the electromyographic signals, a first type of feature parameters for characterizing neural recruitment dynamics are obtained, including: The electromyographic signal is subjected to noise reduction and signal decomposition processing, and the power spectral density ratio of the target frequency band is obtained from the decomposed signal components. Extract the discharge sequence of the motor unit of the target muscle and calculate the coefficient of variation corresponding to the discharge interval of the motor unit; The power spectral density ratio and the coefficient of variation are combined to form the first type of feature parameter.

3. The method of claim 1, wherein the method is based on artificial intelligence. The kinematic signal analysis obtains a second type of feature parameters used to characterize the mechanical state of muscle contraction, including: Velocity information is obtained based on the kinematic signals, and the peak rate of contraction of the target muscle and the time required to reach the peak rate are obtained based on the velocity information, wherein the peak rate and the time required are used as parameters characterizing the explosive force of contraction. Displacement information is obtained based on the kinematic signals, and the effective mechanical work completed by the target muscle contraction is obtained based on a preset load model. The effective mechanical work is used as a parameter characterizing the contraction efficiency. The peak rate, the time required to reach the peak rate, and the effective mechanical work together constitute the second type of characteristic parameters.

4. The method of claim 1, wherein the method is based on artificial intelligence. Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index, including: Obtain the signal-to-noise ratio of the first type of feature parameters and the second type of feature parameters respectively; Initial fusion weights are assigned to the two types of feature parameters based on the data signal-to-noise ratio; Based on the consistency of the changing trends of the first type of feature parameters and the second type of feature parameters within adjacent time windows, the initial fusion weights are dynamically corrected. The modified weights are used to perform a weighted summation and normalization of the two types of feature parameters, and the comprehensive muscle state index is output.

5. The artificial intelligence-based electrical stimulation method according to claim 1, characterized in that, Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions for adjusting electrical stimulation parameters are obtained, including: The comprehensive muscle state index is compared with preset upper and lower thresholds. Based on the comparison results, it is determined whether the target muscle is in an under-activated state, an ideal state, or a fatigued state. If the target muscle is in the state of under-activation, then the first adjustment strategy is determined, wherein the first adjustment strategy is mainly to linearly increase the stimulation intensity and secondarily to fine-tune the stimulation frequency; If the target muscle is in a state of fatigue, then the second adjustment strategy is determined, wherein the second adjustment strategy is supplemented by reducing the stimulation frequency and adjusting the ratio of pulse width to stimulation intensity. The first adjustment strategy or the second adjustment strategy is quantified into specific parameter adjustment amounts.

6. The method of claim 5, wherein the artificial intelligence is based on a neural network. Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions for adjusting electrical stimulation parameters are obtained, including: A three-dimensional parameter lookup table containing stimulation intensity, frequency, and pulse width is established, wherein the three-dimensional parameter lookup table stores optimized parameter combinations corresponding to different muscle state index ranges; Based on the current range of the comprehensive muscle status index, the corresponding parameter combination is retrieved from the lookup table as the adjustment target; Adjust the current electrical stimulation parameters to the adjustment target.

7. The method of claim 1, wherein the method is based on artificial intelligence. Prior to the acquisition of electromyographic and kinematic signals induced by electrical stimulation of the target muscle in the first cycle, the following is included: Apply a set of low-intensity standard test electrical stimulation sequences; Obtain the basic physiological response signals corresponding to the test electrical stimulation sequence; Based on the aforementioned basic physiological response signals, individualized baseline values ​​for the first type of feature parameters and the second type of feature parameters are determined; Based on the individualized baseline value, the initial target range of the comprehensive muscle status index is set.

8. An artificial intelligence-based electrical stimulation system, characterized by, For performing an artificial intelligence-based electrical stimulation method as described in any one of claims 1-7, the system is configured to: During the first cycle, electromyographic and kinematic signals induced by electrical stimulation of the target muscle were collected. Based on the electromyographic signals, a first type of feature parameter is obtained to characterize the dynamics of neural recruitment; Based on the kinematic signal analysis, a second type of feature parameter is obtained to characterize the mechanical state of muscle contraction. Dynamic weight fusion is performed on the first type of feature parameters and the second type of feature parameters to obtain a comprehensive muscle state index; Based on the deviation between the comprehensive muscle state index and the preset target range, adjustment instructions for adjusting the electrical stimulation parameters are obtained; In the second cycle, the parameters of the output electrical stimulation waveform are adjusted based on the adjustment command, wherein the second cycle is a continuous cycle following the first cycle.

9. An electronic device, comprising: include: At least one processor; And, a memory communicatively connected to at least one of the processors; The memory stores instructions executable by at least one of the processors, which are executed to enable at least one of the processors to perform an artificial intelligence-based electrostimulation method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements an artificial intelligence-based electrostimulation method as described in any one of claims 1-7.

Images