A multi-physical stimulation regulation method combined with biofeedback

Abstract

The application discloses a kind of multi-physical stimulation regulation methods combined with biofeedback, it is related to medical instrument technical field, including, based on multimodal physiological signal set, muscle contraction state is identified by convolutional neural network classifier, and user intention is judged in combination with historical pelvic floor muscle state label, and target stimulation mode instruction is generated;According to target stimulation mode instruction, call parameter mapping table to obtain basic stimulation parameter set, and weight distribution is carried out, and stimulation parameter set is generated in combination with current value;Actual output electric stimulation signal is generated according to stimulation parameter set control matrix electrode array discharge, and the paste of solid lubrication matrix is driven piezoelectric vibrator by using Joule heat and is gelatinized, and comprehensive physical stimulation signal is generated;Based on physical stimulation signal, real-time impedance monitoring value is calculated, and safety state flag bit is judged, and work cycle report is generated.The application realizes intention-driven stimulation strategy dynamic matching, and improves the intelligent level of regulation.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a multi-physical stimulation regulation method that combines biofeedback. Background Technology

[0002] In the field of rehabilitation treatment for pelvic floor dysfunction, biofeedback combined with electrical stimulation therapy has become a routine clinical approach. Current technologies typically use surface electrodes to collect electromyographic signals, assess muscle contraction status, and apply electrical pulses of a fixed frequency according to a pre-programmed sequence. This method aims to strengthen pelvic floor muscles and improve neuromuscular control. With the development of sensor technology, some solutions have incorporated temperature monitoring modules to assist in assessing tissue contact status. These technologies are widely used in postpartum rehabilitation and the treatment of pelvic floor muscle relaxation in middle-aged and elderly individuals, forming a relatively mature diagnostic and treatment process. They provide patients with a basic non-invasive intervention approach, comply with medical device safety regulations, and constitute an important technological foundation for current rehabilitation engineering.

[0003] However, conventional technical solutions have limitations in terms of stimulation mode diversity and intention recognition accuracy. On the one hand, traditional devices mostly use a single electrical stimulation mode, lacking the synergistic regulation of thermal effects and mechanical vibration, making it difficult to dynamically allocate energy proportions according to different physiological states, thus limiting rehabilitation effects. On the other hand, existing control strategies are mostly based on fixed-sequence output, failing to combine historical muscle state tags for intention judgment, and unable to distinguish between voluntary defecation intention and muscle fatigue state, resulting in insufficient matching between stimulation parameters and user needs. This affects energy utilization efficiency and neuromuscular remodeling effects during treatment. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-physical stimulus regulation method that combines biofeedback to solve the problems of coordinated regulation of multiple physical stimuli and accurate recognition of user intent.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a multi-physical stimulation regulation method combining biofeedback, comprising: acquiring electromyographic signals of the perianal sphincter and temperature sensor data, and packaging them into a multimodal physiological signal set; based on the multimodal physiological signal set, identifying muscle contraction states through a convolutional neural network classifier, and determining user intent by combining historical pelvic floor muscle state labels, generating a target stimulation mode instruction; according to the target stimulation mode instruction, calling a parameter mapping table to obtain a basic stimulation parameter set, performing weight allocation, and generating a stimulation parameter set by combining current values; controlling the discharge of a matrix electrode array according to the stimulation parameter set to generate an actual output electrical stimulation signal, using Joule heating to accelerate the gelatinization of a solid lubricating matrix and drive a piezoelectric oscillator to generate a comprehensive physical stimulation signal; calculating real-time impedance monitoring values ​​based on the physical stimulation signal, determining safety status flags, and generating a work cycle report.

[0007] As a preferred embodiment of the multi-physical stimulation regulation method combining biofeedback described in this invention, the steps of collecting electromyographic signals of the perianal sphincter and temperature sensor data, and packaging them into a multimodal physiological signal set, are as follows: Raw electromyographic signals of the perianal sphincter were acquired using a matrix electrode array; The original electromyography (EMG) signal is filtered out by a bandpass filter to remove power frequency interference and baseline drift, and then analog-to-digital conversion is performed to generate a preprocessed EMG signal. The system synchronously collects internal temperature sensor data and packages the pre-processed electromyographic signals with the temperature sensor data to generate a multimodal physiological signal set.

[0008] As a preferred embodiment of the multi-physical stimulation regulation method combining biofeedback described in this invention, the step of identifying muscle contraction state based on a multimodal physiological signal set using a convolutional neural network classifier includes the following steps: Based on a set of multimodal physiological signals, the signals are converted into a time-frequency distribution matrix through short-time Fourier transform and arranged into electromyographic feature vectors. Based on an electromyographic feature vector dataset labeled with muscle contraction states, a convolutional neural network classifier is trained by iteratively optimizing the weight parameters through backpropagation algorithm to obtain a pre-trained convolutional neural network classifier. The electromyographic feature vector is input into a pre-trained convolutional neural network classifier to identify the current muscle contraction state and output a pelvic floor muscle state label.

[0009] As a preferred embodiment of the multi-physical stimulation modulation method combining biofeedback described in this invention, the steps of determining the user's intention and generating the target stimulation pattern instruction are as follows: The pelvic floor muscle status labels are compared with historical pelvic floor muscle status labels to obtain the status change trend value. Based on the state change trend value, the intention recognition algorithm is used to determine whether the user has an intention to defecate actively and whether the user is in a state of muscle fatigue, and an intention recognition sequence is generated. Based on the intent recognition sequence, a specific electrode group in the matrix electrode array is selected, and the combination of electrical stimulation, thermal stimulation, and vibration stimulation that needs to be performed is determined to generate a target stimulation pattern instruction.

[0010] As a preferred embodiment of the multi-physical stimulation modulation method combining biofeedback described in this invention, the steps of obtaining the basic stimulus parameter set by calling the parameter mapping table according to the target stimulus mode instruction and performing weight allocation are as follows: Based on the target stimulus pattern instruction, the corresponding basic frequency, pulse width and intensity values ​​are obtained by combining the parameter mapping table, and a basic stimulus parameter set is generated. Based on the set of basic stimulation parameters, the proportion of electrical stimulation, thermal stimulation and mechanical vibration in the total energy is calculated through a weighting strategy, generating voltage, current, temperature and vibration frequency parameters, which are then combined to form a set of multiple physical stimulation sub-parameters. The total energy is calculated from the stimulation duration, output current, electrode impedance, and pulse duty cycle. The set of multiple physical stimuli parameters also includes electrode impedance parameters and energizing time.

[0011] As a preferred embodiment of the multi-physical stimulus regulation method combining biofeedback described in this invention, the steps for generating the stimulus parameter set are as follows: Based on the electrode impedance and energizing time in the multi-physical stimuli parameter set, and according to the conductivity characteristics of the solid lubricating matrix, the current value required to activate the surface state transition of the solid lubricating matrix is ​​calculated, and the current value is incorporated into the multi-physical stimuli parameter set. Data integrity is verified on the merged multi-physical stimulus sub-parameter set to generate a stimulus parameter set.

[0012] As a preferred embodiment of the multi-physical stimulation modulation method combining biofeedback described in this invention, the step of controlling the discharge of the matrix electrode array according to the set of stimulation parameters to generate an actual output electrical stimulation signal, and using Joule heating to accelerate the gelatinization of the solid lubricating matrix, includes the following steps: Based on the set of stimulation parameters, the discharge of a specific group of electrodes in the matrix electrode array is controlled to generate the actual output electrical stimulation signal. Joule heating is generated by the actual output of electrical stimulation signals, which drives the solid lubricating matrix to change from solid to semi-solid, thereby generating the lubricating medium release rate.

[0013] As a preferred embodiment of the multi-physical stimulation modulation method combining biofeedback described in this invention, the steps for generating the comprehensive physical stimulation signal are as follows: By controlling the current throughput of the internal solid lubricating matrix layer, the local temperature is adjusted to the target value, thereby generating an actual output thermal stimulation signal. The micro piezoelectric vibrator is driven by the vibration frequency parameters to generate mechanical waves of a specific frequency. Combined with the release rate of the lubricating medium and the actual output thermal stimulation signal, these waves act on the perianal tissue to generate a comprehensive physical stimulation signal.

[0014] As a preferred embodiment of the multi-physical stimulation modulation method combining biofeedback described in this invention, the steps of calculating the real-time impedance monitoring value based on the physical stimulation signal and determining the safety status flag are as follows: Based on the comprehensive physical stimulus signal, the voltage-to-current ratio of the sampling loop is measured in real time, and the real-time impedance monitoring value is calculated. The real-time impedance monitoring value is compared with the preset safety threshold to identify contact abnormalities and safety abnormalities, and a safety status flag is generated.

[0015] As a preferred embodiment of the multi-physical stimulation regulation method combining biofeedback described in this invention, the steps for generating the work cycle report are as follows: When the safety status flag indicates an abnormality, all stimulus source drive circuits are immediately cut off. When the safety status flag indicates a normality, the duration of the current stimulation process is accumulated to obtain the cumulative stimulation duration value. The system identifies the cumulative stimulation duration that reaches the preset single work cycle threshold, stops all outputs, records the data, and generates a work cycle report.

[0016] The beneficial effects of this invention are as follows: by identifying muscle contraction states through a convolutional neural network classifier and combining historical pelvic floor muscle state labels to determine user intent, dynamic matching of intent-driven stimulation strategies is achieved, improving the level of intelligent regulation; by controlling the discharge of the matrix electrode array and using Joule heating to accelerate the gelatinization of the solid lubricating matrix to drive the piezoelectric oscillator, the linkage regulation of electro-thermal-vibration multi-field synergy and material phase change is achieved, thereby enhancing the therapeutic synergy effect and improving patient comfort. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a multi-physical stimulus regulation method that incorporates biofeedback.

[0019] Figure 2 This is a flowchart of signal acquisition and preprocessing.

[0020] Figure 3 A flowchart for intent recognition and parameter generation.

[0021] Figure 4 A flowchart for stimulus execution and safety monitoring. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a multi-physical stimulus modulation method combining biofeedback, comprising the following steps: S1. Collect electromyographic signals and temperature sensor data of the perianal sphincter, and package them to generate a multimodal physiological signal set.

[0026] Raw electromyographic signals of the perianal sphincter were acquired using a matrix electrode array.

[0027] Furthermore, the solid lubricating matrix of the matrix electrode array is inserted into the perianal region to establish physical contact between the electrodes and the mucosal tissue, and output a contact impedance signal. When the contact impedance signal is lower than the impedance threshold, the matrix electrode array is activated to capture the potential changes generated by the contraction of the perianal sphincter and generate the original electromyographic signal. For example, the matrix electrode array has a spacing of 2 mm, covering the annular area of ​​the perianal sphincter, to achieve multi-point synchronous acquisition.

[0028] It should be noted that the matrix electrode array is a multi-point electrode group integrated on the surface of a solid lubricating matrix and arranged in a ring (e.g., with a 2 mm spacing). It is used to simultaneously collect electromyographic signals of the perianal sphincter and output multi-physical field stimulation such as electrical stimulation, Joule heating, and mechanical vibration as needed. The solid lubricating matrix refers to a biodegradable carrier made of soap raw materials, pressed or 3D printed into a spherical or ellipsoidal shape, with a matrix electrode array integrated on its surface. The impedance threshold is set based on the physiological impedance characteristics of human perianal mucosa, the conductivity of the electrode-tissue interface, and the safety stimulation current limit, and the value range is usually from 1 kΩ to 10 kΩ.

[0029] The original electromyography (EMG) signal is filtered out by a bandpass filter to remove power frequency interference and baseline drift, and then analog-to-digital conversion is performed to generate a preprocessed EMG signal.

[0030] Furthermore, the raw electromyography (EMG) signal is input into a bandpass filter for bandpass filtering (e.g., a cutoff frequency of 10 Hz to 500 Hz) to filter out power frequency interference and baseline drift (baseline drift refers to the slow, non-periodic shift of the DC component or low-frequency component of a bioelectrical signal (e.g., EMG signal) over time), and a filtered EMG signal is output. The filtered EMG signal is then sampled discretely by a sample-and-hold circuit at a sampling frequency of 1000 Hz to obtain the sampling amplitude. The sampling amplitude is quantized into digital levels by an analog-to-digital converter (e.g., 12-bit resolution) and encoded into binary code to output a digital EMG signal. The digital EMG signal is then combined with the contact impedance value of the solid lubricating matrix (e.g., 5 kΩ) for gain compensation to eliminate impedance change noise caused by fluctuations in the contact interface state, and a preprocessed EMG signal is output.

[0031] The system synchronously collects internal temperature sensor data and packages the pre-processed electromyographic signals with the temperature sensor data to generate a multimodal physiological signal set.

[0032] Furthermore, based on the preprocessed electromyographic signal, a temperature sensor (e.g., an NTC thermistor) is triggered to acquire the tissue interface temperature (e.g., at a frequency of 10 Hz), generating a temperature digital sequence. The temperature digital sequence and the preprocessed electromyographic signal are synchronized with each other by a hardware clock and marked with the same timestamp. The electromyographic values ​​and temperature quantization values ​​are alternately arranged according to a preset frame structure (e.g., a 32-bit timestamp, a 12-bit electromyographic value, and a 10-bit temperature quantization value) to form a composite data frame. The composite data frame is encapsulated into a multimodal physiological signal set after being appended with a CRC checksum.

[0033] S2. Based on a set of multimodal physiological signals, a convolutional neural network classifier is used to identify muscle contraction states and combine historical pelvic floor muscle state labels to determine user intent and generate target stimulation pattern instructions.

[0034] Based on a set of multimodal physiological signals, the signals are converted into a time-frequency distribution matrix through short-time Fourier transform and arranged into electromyographic feature vectors.

[0035] Furthermore, electromyographic value sequences and temperature quantization values ​​are extracted from the multimodal physiological signal set. First, a short-time Fourier transform is performed on the electromyographic value sequence to generate a time-frequency distribution matrix. Then, temperature compensation is performed on the time-frequency distribution matrix using the temperature quantization values. A segment is obtained by sliding a Hamming window (e.g., the window length is set to 256 points and the window overlap rate is set to 50%). A fast Fourier transform is performed on the segment to obtain the spectral energy density. The spectral energy density of each sliding window is arranged in chronological order to generate a time-frequency distribution matrix. Each energy value in the time-frequency distribution matrix is ​​normalized to eliminate conductivity fluctuation errors caused by tissue temperature changes, resulting in a temperature-compensated time-frequency map. The temperature-compensated time-frequency map is expanded into a single-column array in row-major order from low to high frequency points. The array values ​​are truncated to unsigned integers (e.g., 8 bits) and arranged to form an electromyographic feature vector.

[0036] Based on an electromyographic feature vector dataset labeled with muscle contraction states, a pre-trained convolutional neural network classifier is trained by iteratively optimizing the weight parameters through backpropagation algorithm.

[0037] Furthermore, the electromyographic feature vector dataset labeled with muscle contraction state is divided into a training subset and a validation subset. The training subset is input into the input layer of the convolutional neural network classifier. Local features are extracted through the convolutional layer and dimensionality is reduced through the pooling layer. The fully connected layer outputs the predicted probability and calculates the cross-entropy loss value with the labeled muscle contraction state. The cross-entropy loss value is added to the L2 regularization term to form the total loss. The backpropagation algorithm updates the weight parameters of the convolutional layer and the fully connected layer according to the gradient of the total loss. The algorithm iterates (e.g., 1000 times) until the classification accuracy of the validation subset reaches the convergence threshold. The weight parameters are saved to generate a pre-trained convolutional neural network classifier.

[0038] It should be noted that the backpropagation algorithm is a supervised learning strategy used to train multi-layer neural networks. Its core idea is to propagate the prediction error from the output back to each layer, thereby calculating the gradient and updating the network parameters to minimize the loss function. The convergence threshold is determined based on the stable value of the classification accuracy of the validation subset, set by monitoring the accuracy fluctuations in consecutive iterations; an exemplary value range is 95% to 99%. The backpropagation algorithm is a neural network training method. It obtains the output value through forward propagation and compares it with the target value to obtain the loss value. Then, it uses the chain rule to obtain the gradient of the loss value with respect to the weight parameters and propagates the error signal from the output layer to the input layer layer by layer to iteratively update the weight parameters. The chain rule is a rule for calculating the derivative of a composite function in calculus. In neural networks, it is the mathematical basis for decomposing and propagating the error gradient from the output layer to the input layer layer by layer, ensuring that each weight is accurately corrected. The convolutional neural network classifier consists of an input layer, convolutional layers, pooling layers, and fully connected layers, and is a multi-layer neural network structure used to output the predicted probability of muscle contraction state. The L2 regularization term is a penalty term to prevent overfitting.

[0039] The expression for the objective function of neural network regularization optimization is: ; in, Total loss; It is a set of neural network weight parameters, including iterative optimization weight parameters and convolution kernel weight parameters; The total number of samples in the training subset; Index for the training subset; For the first Each sample was labeled with a muscle contraction status tag; For the first The predicted probability of each sample; is the regularization coefficient. Based on cross-validation, it is selected as the minimum effective value under the premise that the classification accuracy of the validation set reaches the convergence threshold. It is used to control the proportion of the penalty term in the total loss and prevent the convolutional neural network classifier from overfitting. For the first A specific weight value; This is the index of the network weight parameters.

[0040] The electromyographic feature vector is input into a pre-trained convolutional neural network classifier to identify the current muscle contraction state and output a pelvic floor muscle state label.

[0041] Furthermore, the electromyography (EMG) feature vector is input into the input layer of a pre-trained convolutional neural network classifier. The convolutional layer performs convolution operations on the EMG feature vector to generate a feature map (e.g., a convolution kernel size of 3×3 and a stride of 1). The feature map is then passed to a pooling layer for max pooling downsampling to generate a dimensionality-reduced feature vector (e.g., a pooling window of 2×2). The dimensionality-reduced feature vector is then passed to a fully connected layer to map to a state category probability distribution. The probability value of each state category is obtained through a Softmax layer, and the category number corresponding to the maximum probability is selected as the pelvic floor muscle state label.

[0042] The pelvic floor muscle status labels are compared with historical pelvic floor muscle status labels to obtain the status change trend value.

[0043] Furthermore, the pelvic floor muscle status labels are stored in a circular buffer as the current labels. The most recent (e.g., 10) historical pelvic floor muscle status labels are read from the historical label memory to form a label sequence. The number and direction of state transitions between the current label and adjacent labels in the label sequence are identified. The frequency change rate of contraction status labels (the category number of the muscle in a contraction state) is statistically analyzed (e.g., within a 5-minute time window). The frequency change rate is combined with the time decay coefficient to obtain a weighted change value. The weighted change value is then summed with the standard deviation of the label sequence to generate a state change trend value.

[0044] Based on the state change trend value, the intent recognition algorithm is used to determine whether the user has an intention to defecate and whether the user is in a state of muscle fatigue, and an intent recognition sequence is generated.

[0045] Furthermore, the state change trend value is input into the intention recognition algorithm. The algorithm compares the state change trend value with the defecation intention threshold. When the state change trend value is greater than the defecation intention threshold, the defecation intention flag is set to 1. When the defecation intention flag is 1 for multiple consecutive time windows (e.g., 3 consecutive time windows), the active defecation intention is confirmed. The intention recognition algorithm also compares the state change trend value with the fatigue threshold. When the state change trend value is greater than the fatigue threshold, the fatigue state flag is set to 1. When the fatigue state flag is 1 for multiple consecutive time windows (e.g., 2 consecutive time windows), the fatigue state is confirmed. The confirmed active defecation intention state and muscle fatigue state are arranged in chronological order to generate an intention recognition sequence composed of binary code. If the state change trend value exceeds two thresholds at the same time, it is preferentially determined to be a muscle fatigue state, and the active defecation intention is excluded.

[0046] It should be noted that the defecation intention threshold is statistically determined based on a dataset of electromyographic feature vectors labeled with active defecation intentions. It is set by calculating the fixed quantile (e.g., 95%) of the corresponding state change trend value and fine-tuning it through cross-validation. An exemplary value range is 0.7 to 0.9. The fatigue threshold is determined based on the mean and standard deviation of the state change trend value sequence under muscle fatigue state. It is set to increase the standard deviation by the mean (e.g., by 1.5 times the standard deviation). An exemplary value range is 0.4 to 0.6. The intention recognition algorithm includes temporal logic judgment and multi-condition verification process. It monitors the status of the flag bit through a continuous time window to distinguish between active contraction behavior and abnormal fluctuations caused by muscle fatigue. The intention recognition algorithm integrates hysteresis comparison and state machine logic to ensure the continuous confirmation of defecation intention and avoid triggering erroneous stimulation by a single signal peak.

[0047] Based on the intent recognition sequence, a specific electrode group in the matrix electrode array is selected, and the combination of electrical stimulation, thermal stimulation, and vibration stimulation that needs to be performed is determined to generate a target stimulation pattern instruction.

[0048] Furthermore, based on the intent recognition sequence composed of binary code, when the high bit of the binary code is 1, the electrode group in the central region of the matrix electrode array (e.g., indexes 1 to 4) is selected, and when the low bit of the binary code is 1, the electrode group in the peripheral region (e.g., indexes 5 to 8) is selected, generating a selected electrode index list. Based on the selected electrode index list, the state bits of the intent recognition sequence are identified. When the active defecation intent state bit is 1, the electrical stimulation and thermal stimulation activation flags are set, and when the muscle fatigue state bit is 1, the vibration stimulation activation flag is set, generating a set of stimulation type flag bits. The set of stimulation type flag bits is merged and encoded with the selected electrode index list, and start and stop bits are added to generate a target stimulation mode instruction, which is used to guide the matrix electrode array to execute a specific multi-physics field stimulation combination.

[0049] S3. Based on the target stimulus mode instruction, call the parameter mapping table to obtain the basic stimulus parameter set, perform weight allocation, and generate the stimulus parameter set by combining the current value.

[0050] Based on the target stimulus pattern instruction, the corresponding baseline frequency, pulse width, and intensity values ​​are obtained by combining the parameter mapping table, and a set of baseline stimulus parameters is generated.

[0051] Furthermore, based on the target stimulation mode instruction, the parameter mapping table is queried using the stimulation type flag set as the address offset. The corresponding basic frequency value (e.g., 50 Hz), pulse width value (e.g., 200 microseconds), and intensity value (e.g., 5 mA) are read. The read values ​​are bound to the selected electrode index list item by item to form electrode-parameter pairs. The pairs are then arranged in timestamp order and a check bit is added to generate the basic stimulation parameter set.

[0052] It should be noted that the parameter mapping table is determined based on the electromyographic feature vector dataset labeled with muscle contraction state and the historical statistics of stimulus response effects. It is set through cross-validation and includes the key-value correspondence between stimulus type flag bits and baseline frequency, pulse width and intensity.

[0053] Based on the set of basic stimulation parameters, the proportion of electrical stimulation, thermal stimulation and mechanical vibration in the total energy is calculated through a weighting strategy, generating voltage, current, temperature and vibration frequency parameters, which are then combined to form a set of multiple physical stimulation sub-parameters.

[0054] Furthermore, intensity values ​​are extracted from the basic stimulus parameter set and converted into total energy values ​​(total energy values ​​are calculated jointly by stimulus duration, output current, electrode impedance, and pulse duty cycle). The state bits of the intent recognition sequence are read using a weighted allocation strategy. When the active defecation intent state bit is 1, an electrical stimulation weight (e.g., 0.6), a thermal stimulation weight (e.g., 0.3), and a mechanical vibration weight (e.g., 0.1) are set. When the muscle fatigue state bit is 1, the mechanical vibration weight is adjusted (e.g., 0.5). The total energy value is combined with the electrical stimulation weight, thermal stimulation weight, and mechanical vibration weight respectively to obtain the corresponding allocated energy values. Based on the electrical stimulation allocated energy value, voltage and current values ​​are obtained according to Joule's law, electrode impedance, and energizing time. Temperature values ​​are obtained based on the thermal stimulation allocated energy value and the specific heat capacity, latent heat of phase change, mass, and gelatinization temperature threshold of the solid lubricating matrix. Vibration frequency values ​​are obtained based on the mechanical vibration allocated energy value and the resonant characteristics of the piezoelectric oscillator. The voltage, current, temperature, and vibration frequency values ​​are sequentially encoded and combined to form a multi-physical stimulus sub-parameter set.

[0055] It should be noted that the weighting strategy is a rule-based algorithm that dynamically allocates the total energy value to three physical fields: electrical stimulation, thermal stimulation, and mechanical vibration. It dynamically determines the energy proportion coefficients of electrical stimulation, thermal stimulation, and mechanical vibration based on the state position of the intent recognition sequence through a lookup table. The allocated energy value is then converted into voltage / current values, temperature values, and vibration drive values, respectively, by combining electrode impedance, energizing time, thermal parameters of the solid lubricating matrix, and the resonant characteristics of the piezoelectric oscillator. Electrode impedance is the electrical impedance characteristic of the interface between the electrode and the tissue, reflecting the degree of resistance to current flow through the interface. Solid lubricating matrix... The heat capacity of a lubricating matrix is ​​a physical quantity that a solid lubricating matrix absorbs or releases heat under a unit temperature change (e.g., 1 degree Celsius), including sensible heat capacity and latent heat capacity. The resonant characteristics of a piezoelectric oscillator are the inherent properties of a piezoelectric oscillator that produces the maximum amplitude response at a specific frequency (e.g., 40 kHz), which are determined by the material's elastic modulus, geometric dimensions, and boundary conditions. The latent heat of phase transition refers to the critical heat energy absorbed per unit mass during the transition of a solid lubricating matrix from a solid to a semi-solid gelatinized state. It is obtained by measuring the endothermic peak area of ​​the lubricating matrix during the phase transition process using differential scanning calorimetry and combining it with the sample mass.

[0056] Based on the electrode impedance and energizing time in the multi-physical stimuli parameter set, and according to the conductivity characteristics of the solid lubricating matrix, the current value required to activate the surface state transition of the solid lubricating matrix is ​​calculated, and the current value is incorporated into the multi-physical stimuli parameter set.

[0057] Furthermore, the electrode impedance and energizing time are read from the multi-physical stimuli parameter set. Based on the conductivity characteristics of the solid lubricating matrix, the electrode impedance is corrected to the equivalent resistance value. The phase change activation heat value is calculated by combining the gelatinization temperature threshold and the heat capacity of the solid lubricating matrix. The activation current value is obtained by combining Joule's law and the equivalent resistance value. The activation current value is truncated to (e.g., 12 bits) an unsigned integer format and appended to the end of the multi-physical stimuli parameter set to generate a complete parameter set containing the activation current.

[0058] It should be noted that the electrical conductivity of solid lubricating matrix refers to the resistivity variation of solid lubricating matrix at different temperatures and phases. This includes the transition characteristics between the solid high-resistivity state (the high resistivity state exhibited by the solid lubricating matrix before reaching the gelatinization temperature threshold, where limited ion migration channels result in low conductivity) and the gelatinized semi-solid low-resistivity state (the phenomenon of conductivity jump where the solid lubricating matrix undergoes a phase transition when the tissue interface temperature rises to the gelatinization temperature threshold, rapidly opening ion migration channels and causing a sharp decrease in resistivity within a narrow temperature range). This is used to provide accurate resistivity in Joule heating calculations. Parameters; The gelatinization temperature threshold of the solid lubricating matrix is ​​set based on the phase transition characteristics of the lubricating matrix (e.g., soap). It is determined by differential scanning calorimetry (DSC) at the critical temperature (e.g., 42 degrees Celsius) at which the lubricating matrix transitions from solid to semi-solid state. The value is then determined after reserving a safety margin by combining the upper limit of tissue safety tolerance (e.g., 48 degrees Celsius). An exemplary value range is typically 40 to 50 degrees Celsius. Differential scanning calorimetry (DSC) precisely quantifies the energy absorbed or released when the material undergoes a phase transition by measuring the heat difference between the sample and the reference material during the temperature control process, thereby determining the gelatinization temperature threshold.

[0059] Data integrity is verified on the merged multi-physical stimulus sub-parameter set to generate a stimulus parameter set.

[0060] Furthermore, the CRC checksum of the multi-physical stimuli parameter set header containing the activation current is read and the CRC-16 value of the parameter body is recalculated and compared. If the comparison is consistent, the voltage value (e.g., 0 to 50 volts), current value (e.g., 0 to 20 mA), temperature value (e.g., 35 to 50 degrees Celsius), and vibration frequency value (e.g., 20 to 60 kHz) are checked one by one to see if they are within the safe boundaries. The Joule heating relationship between the activation current value and the electrode impedance value is verified to be matched. The data is then re-encoded into a fixed-length binary frame according to the preset frame structure to generate the stimulation parameter set.

[0061] S4. Based on the set of stimulation parameters, control the discharge of the matrix electrode array to generate the actual output electrical stimulation signal. Use Joule heating to accelerate the gelatinization of the solid lubricating matrix and drive the piezoelectric oscillator to generate a comprehensive physical stimulation signal.

[0062] Based on the set of stimulation parameters, the discharge of a specific group of electrodes in the matrix electrode array is controlled to generate the actual output electrical stimulation signal.

[0063] Furthermore, based on the set of stimulation parameters, the corresponding electrode group in the matrix electrode array is located according to the selected electrode index list. The voltage and current values ​​are converted into driving signal amplitudes. The period duration is obtained based on the fundamental frequency value. The pulse width value is used as the high-level duration within the period duration to generate a periodic square wave triggering sequence. According to the periodic triggering sequence, the driving signal amplitude is controlled to output pulse current to the corresponding electrode group at a set time interval (e.g., 20 milliseconds) to generate the actual output electrical stimulation signal.

[0064] Joule heating is generated by the actual output of electrical stimulation signals, which drives the solid lubricating matrix to change from solid to semi-solid, thereby generating the lubricating medium release rate.

[0065] Furthermore, the actual output electrical stimulation signal is applied to the matrix electrode array. The current flows through the solid lubricating matrix and generates Joule heat through the contact impedance. The tissue interface temperature value (referring to the actual temperature value of the interface between the perianal mucosa and the electrode) is collected in real time. When the tissue interface temperature value rises to the gelatinization temperature threshold of the solid lubricating matrix, the surface of the solid lubricating matrix undergoes a phase change from solid to semi-solid paste. The heat value per unit time is obtained according to Joule's law, and the release rate of the lubricating medium is obtained according to the latent heat of phase change, thereby realizing interface lubrication, current homogenization, thermal buffering, and adaptive protection.

[0066] By controlling the current throughput of the internal solid lubricating matrix layer, the local temperature is adjusted to the target value, thereby generating the actual output thermal stimulation signal.

[0067] Furthermore, temperature values ​​are read from the multi-physical stimuli parameter set as the target temperature value. The interface temperature value of the solid lubricating matrix layer (referring to the temperature of the interface between the solid lubricating matrix layer and the electrode inside the solid lubricating matrix) is collected in real time. The difference between the target temperature value and the interface temperature value of the solid lubricating matrix layer is calculated to obtain the temperature deviation value. According to the PID control logic, the temperature deviation value is converted into the current throughput adjustment quantity, and combined with the current value to obtain the actual output current value. The actual output current value is applied to the electrode of the solid lubricating matrix layer to generate Joule heating. Joule heating drives the temperature of the solid lubricating matrix layer to approach the target temperature value (e.g., 42 degrees Celsius). The temperature-stable solid lubricating matrix layer transfers a constant heat flow to the tissue interface, generating the actual output thermal stimulation signal.

[0068] The micro piezoelectric vibrator is driven by the vibration frequency parameters to generate mechanical waves of a specific frequency. Combined with the release rate of the lubricating medium and the actual output thermal stimulation signal, these waves act on the perianal tissue to generate a comprehensive physical stimulation signal.

[0069] Furthermore, the vibration frequency value (e.g., 40 kHz) is read from the set of stimulation parameters. The upper limit of the period counter of the digital waveform generator is set according to the vibration frequency value. When the counter accumulates to the upper limit of the period counter, the corresponding frequency AC drive signal is flipped and applied to the excitation end of the micro piezoelectric vibrator. The micro piezoelectric vibrator generates a mechanical vibration wave of a specific frequency. The mechanical vibration wave penetrates the semi-solid paste layer of the solid lubricating matrix and is synchronously superimposed with the constant heat flow transmitted by the actual output thermal stimulation signal and the liquid film layer formed by the release rate of the lubricating medium. The mechanical vibration wave, constant heat flow and liquid film layer contact the perianal tissue surface together. The tissue surface simultaneously receives the actual output electrical stimulation signal, constant heat flow and mechanical vibration energy to form a comprehensive physical stimulation signal.

[0070] S5. Calculate the real-time impedance monitoring value based on the physical stimulus signal, determine the safety status flag, and generate a work cycle report.

[0071] Based on the comprehensive physical stimulus signal, the voltage-to-current ratio of the sampling circuit is sampled in real time, and the real-time impedance monitoring value is calculated.

[0072] Furthermore, based on the comprehensive physical stimulus signal, the instantaneous voltage and current values ​​at both ends of the circuit are synchronously acquired through a real-time sampling circuit, and digital voltage and current sequences are obtained through analog-to-digital conversion. The digital voltage and current sequences are then subjected to sliding window averaging filtering at the sampling frequency (e.g., 1000 Hz). Based on the filtered voltage and current values, the real-time impedance monitoring value is obtained.

[0073] The real-time impedance monitoring value is compared with the preset safety threshold to identify contact abnormalities and safety abnormalities, and a safety status flag is generated.

[0074] Furthermore, the real-time impedance monitoring value and the reference impedance value (e.g., 5 kΩ) are subjected to a sliding difference operation to generate an impedance offset sequence, and the variance of the impedance offset sequence is calculated according to a fixed window (e.g., 100 milliseconds). When the variance of the impedance offset sequence exceeds the preset safety threshold, a poor contact flag is output (0 represents normal, 1 represents abnormal), and a safety status flag bit is generated in the order of timestamps.

[0075] It should be noted that the reference impedance value is set based on the conductivity of the electrode-tissue contact interface and the safety stimulation current limit. It is determined by statistically analyzing the central tendency of impedance measurements under normal contact conditions. An exemplary value range is typically 1 kΩ to 10 kΩ. The preset safety threshold is set based on the critical impedance values ​​of abnormal contact and safety abnormality and the boundary for poor contact judgment. It is determined by statistically analyzing the extreme value range of impedance measurement data under abnormal conditions and combining it with a safety factor for fine-tuning. An exemplary value range is typically 1 kΩ to 10 kΩ.

[0076] When the safety status flag indicates an abnormality, all stimulus source drive circuits are immediately cut off. When the safety status flag indicates a normal status, the duration of the current stimulus process is accumulated to obtain the cumulative stimulus duration value.

[0077] Furthermore, the binary code is read from the safety status flag. When the code indicates an abnormality, the drive signal output of electrical stimulation, thermal stimulation, and vibration stimulation is cut off. When the code indicates a normal status, the difference between the current timestamp and the stimulation start timestamp is obtained as the duration of the current stimulation. The duration of the current stimulation is added to the cumulative value of the historical stimulation duration in the memory to obtain the updated cumulative stimulation duration value.

[0078] The system identifies the cumulative stimulation duration that reaches the preset single work cycle threshold, stops all outputs, records the data, and generates a work cycle report.

[0079] Furthermore, the cumulative stimulation duration value is compared with the preset single work cycle threshold. When the cumulative stimulation duration value reaches the preset single work cycle threshold, the drive signal output of electrical stimulation, thermal stimulation and vibration stimulation is cut off. The intention recognition sequence, stimulation parameter set, real-time impedance monitoring value and safety status flag bit in the stimulation process are read and encoded into binary data frames, with timestamps and CRC-16 check codes added. The data is then output to the external interface to generate a work cycle report.

[0080] It should be noted that the threshold for a single work cycle is set based on historical work cycle reports and is determined through hierarchical preset and dynamic correction methods. The value range is usually from 10 minutes to 30 minutes.

[0081] In summary, this invention achieves dynamic matching of intent-driven stimulation strategies and improves the intelligence level of regulation by: recognizing muscle contraction states using a convolutional neural network classifier and judging user intent by combining historical pelvic floor muscle state labels; and by controlling the discharge of the matrix electrode array and using Joule heating to accelerate the gelatinization of the solid lubricating matrix to drive the piezoelectric oscillator, it achieves the linkage regulation of electro-thermal-vibration multi-field synergy and material phase change, thereby enhancing the therapeutic synergy and improving patient comfort.

[0082] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-physical stimulus regulation method combining biofeedback, characterized in that, include: Collect electromyographic signals and temperature sensor data from the perianal sphincter, and package them to generate a multimodal physiological signal set; Based on a multimodal physiological signal set, a convolutional neural network classifier is used to identify muscle contraction states, and historical pelvic floor muscle state labels are combined to determine user intent and generate target stimulation pattern instructions. According to the target stimulus mode instruction, the parameter mapping table is called to obtain the basic stimulus parameter set, and weights are assigned. The stimulus parameter set is then generated by combining the current value. The actual output electrical stimulation signal is generated by controlling the discharge of the matrix electrode array according to the set of stimulation parameters. Joule heating is used to accelerate the gelatinization of the solid lubricating matrix and drive the piezoelectric oscillator to generate a comprehensive physical stimulation signal. The system calculates real-time impedance monitoring values ​​based on physical stimulus signals, determines safety status flags, and generates a work cycle report.

2. The multi-physical stimulus regulation method combining biofeedback as described in claim 1, characterized in that, The steps for collecting electromyographic signals of the perianal sphincter and temperature sensor data, and packaging them into a multimodal physiological signal set, are as follows: Raw electromyographic signals of the perianal sphincter were acquired using a matrix electrode array; The original electromyography (EMG) signal is filtered out by a bandpass filter to remove power frequency interference and baseline drift, and then analog-to-digital conversion is performed to generate a preprocessed EMG signal. The system synchronously collects internal temperature sensor data and packages the pre-processed electromyographic signals with the temperature sensor data to generate a multimodal physiological signal set.

3. The multi-physical stimulus regulation method combining biofeedback as described in claim 1, characterized in that, The steps for identifying muscle contraction states based on a multimodal physiological signal set using a convolutional neural network classifier are as follows: Based on a set of multimodal physiological signals, the signals are converted into a time-frequency distribution matrix through short-time Fourier transform and arranged into electromyographic feature vectors. Based on an electromyographic feature vector dataset labeled with muscle contraction states, a convolutional neural network classifier is trained by iteratively optimizing the weight parameters through backpropagation algorithm to obtain a pre-trained convolutional neural network classifier. The electromyographic feature vector is input into a pre-trained convolutional neural network classifier to identify the current muscle contraction state and output a pelvic floor muscle state label.

4. The multi-physical stimulus regulation method combining biofeedback as described in claim 1, characterized in that, The steps for determining the user's intent and generating the target stimulus pattern instruction are as follows: The pelvic floor muscle status labels are compared with historical pelvic floor muscle status labels to obtain the status change trend value. Based on the state change trend value, the intention recognition algorithm is used to determine whether the user has an intention to defecate actively and whether the user is in a state of muscle fatigue, and an intention recognition sequence is generated. Based on the intent recognition sequence, a specific electrode group in the matrix electrode array is selected, and the combination of electrical stimulation, thermal stimulation, and vibration stimulation that needs to be performed is determined to generate a target stimulation pattern instruction.

5. The multi-physical stimulus regulation method combining biofeedback as described in claim 4, characterized in that, The steps for obtaining the basic stimulus parameter set by calling the parameter mapping table according to the target stimulus pattern instruction and assigning weights are as follows: Based on the target stimulus pattern instruction, the corresponding basic frequency, pulse width and intensity values ​​are obtained by combining the parameter mapping table, and a basic stimulus parameter set is generated. Based on the set of basic stimulation parameters, the proportion of electrical stimulation, thermal stimulation and mechanical vibration in the total energy is calculated through a weighting strategy, generating voltage, current, temperature and vibration frequency parameters, which are then combined to form a set of multiple physical stimulation sub-parameters. The total energy is calculated from the stimulation duration, output current, electrode impedance, and pulse duty cycle. The set of multiple physical stimuli parameters also includes electrode impedance parameters and energizing time.

6. The multi-physical stimulus regulation method combining biofeedback as described in claim 5, characterized in that, The steps for generating the stimulus parameter set are as follows: Based on the electrode impedance and energizing time in the multi-physical stimuli parameter set, and according to the conductivity characteristics of the solid lubricating matrix, the current value required to activate the surface state transition of the solid lubricating matrix is ​​calculated, and the current value is incorporated into the multi-physical stimuli parameter set. Data integrity is verified on the merged multi-physical stimulus sub-parameter set to generate a stimulus parameter set.

7. The multi-physical stimulus regulation method combining biofeedback as described in claim 1, characterized in that, The process of controlling the discharge of the matrix electrode array based on the set of stimulation parameters to generate the actual output electrical stimulation signal, and using Joule heating to accelerate the gelatinization of the solid lubricating matrix, is as follows: Based on the set of stimulation parameters, the discharge of a specific group of electrodes in the matrix electrode array is controlled to generate the actual output electrical stimulation signal. Joule heating is generated by the actual output of electrical stimulation signals, which drives the solid lubricating matrix to change from solid to semi-solid, thereby generating the lubricating medium release rate.

8. The multi-physical stimulus regulation method combining biofeedback as described in claim 7, characterized in that, The steps for generating the comprehensive physical stimulation signal are as follows: By controlling the current throughput of the internal solid lubricating matrix layer, the local temperature is adjusted to the target value, thereby generating an actual output thermal stimulation signal. The micro piezoelectric vibrator is driven by the vibration frequency parameters to generate mechanical waves of a specific frequency. Combined with the release rate of the lubricating medium and the actual output thermal stimulation signal, these waves act on the perianal tissue to generate a comprehensive physical stimulation signal.

9. The multi-physical stimulus regulation method combining biofeedback as described in claim 1, characterized in that, The steps for calculating the real-time impedance monitoring value based on the physical stimulus signal and determining the safety status flag are as follows: Based on the comprehensive physical stimulus signal, the voltage-to-current ratio of the sampling loop is measured in real time, and the real-time impedance monitoring value is calculated. The real-time impedance monitoring value is compared with the preset safety threshold to identify contact abnormalities and safety abnormalities, and a safety status flag is generated.

10. The multi-physical stimulus regulation method combining biofeedback as described in claim 9, characterized in that, The steps for generating the work cycle report are as follows: When the safety status flag indicates an abnormality, all stimulus source drive circuits are immediately cut off. When the safety status flag indicates a normality, the duration of the current stimulation process is accumulated to obtain the cumulative stimulation duration value. The system identifies the cumulative stimulation duration that reaches the preset single work cycle threshold, stops all outputs, records the data, and generates a work cycle report.

Images