Pulse signal adaptive control method and system for electrotherapy equipment

Abstract

The invention relates to the technical field of pulse electrotherapy, in particular to a pulse signal self-adaptive control method and system for electrotherapy equipment. A pulse signal adaptive control system for electrotherapy equipment comprises a data acquisition module, a pulse signal control strategy output model updating module and a pulse signal control strategy output module. The pulse signals of the electrotherapy equipment are adjusted through the pulse signal control strategy output model, the physiological parameters fed back by the patient in real time are obtained at the same time interval, and the pulse signal control strategy output model is updated in real time based on the physiological parameters fed back in real time. The pulse signal control strategy output by the pulse signal control strategy output model can better meet the treatment requirement and the physiological bearing requirement of the patient, and the use effect of the electrotherapy equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pulse electrotherapy, and in particular to a pulse signal adaptive control method and system for electrotherapy equipment. Background Art

[0002] Electrotherapy equipment has been widely used in clinical rehabilitation, pain management and physical therapy. Its core working principle is to stimulate nerves or muscles by applying pulse current to specific parts of the patient to achieve the purpose of treatment. At present, the pulse signal control of most electrotherapy equipment relies on pre-set fixed parameters, such as pulse frequency, pulse width, amplitude, etc. Although this fixed parameter setting can meet the needs of some patients, it has many inconveniences in practical applications.

[0003] First, fixed parameter settings lack flexibility and cannot be dynamically adjusted according to individual differences in patients (such as tolerance, changes in condition, and sensitivity of the treatment site). Each patient has different physiological conditions and responses to electrical stimulation. Fixed parameter settings may result in some patients not being able to achieve the best treatment effect, and may even cause discomfort or side effects due to excessive or weak stimulation. Secondly, since the patient's condition may change over time during treatment, fixed parameters cannot adapt to these changes in real time, thus affecting the treatment effect.

[0004] In addition, the complexity of manual control further increases the inconvenience of use. In traditional electrotherapy equipment, medical staff or patients need to manually adjust the parameters of the equipment, which not only increases the complexity of operation, but also may lead to inaccurate parameter settings due to human operation errors, affecting the treatment effect and safety. Especially during the treatment process, if the patient feels uncomfortable, the treatment must be suspended and the equipment parameters must be manually adjusted, which not only affects the continuity of the treatment, but also increases the pain of the patient and the workload of the medical staff. Summary of the invention

[0005] The present invention adjusts the pulse signal of the electrotherapy equipment through a pulse signal control strategy output model, obtains the patient's real-time feedback of physiological parameters at the same time interval, and then updates the pulse signal control strategy output model in real time based on the real-time feedback of physiological parameters, so that the pulse signal control strategy output by the pulse signal control strategy output model can better meet the patient's treatment needs and physiological tolerance needs, thereby improving the use effect of the electrotherapy equipment.

[0006] A pulse signal adaptive control method for an electrotherapy device, comprising: Acquire a pulse signal data set and a patient's physiological parameter set at a current monitoring time point, wherein the pulse signal data set includes pulse signal data, and the patient's physiological parameter set includes physiological parameters; The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point; The pulse signal data set and physiological parameter set obtained at the current monitoring time point are combined into a state data set, and the state data set is sent to the updated pulse signal control strategy output model for processing, and the pulse signal control strategy is output, and then the pulse signal output by the electrotherapy equipment is controlled by the pulse signal control strategy.

[0007] As a preferred aspect of the present invention, it comprises: The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point, specifically including the following steps: The pulse signal data set and the physiological parameter set obtained at the current monitoring time point constitute the latter real-time analysis part, the pulse signal data set and the physiological parameter set obtained at the previous monitoring time point and the pulse signal control strategy output at the previous monitoring time point constitute the former real-time analysis part, and the former real-time analysis part and the latter real-time analysis part constitute the real-time training sample; Get the first N real-time training samples; select the real-time training samples in chronological order, and for each selected nth real-time training sample, n=1, 2, 3, ..., N, where n is the sequence number of the real-time training sample, send the previous real-time analysis part of the selected nth real-time training sample to the strategy evaluation network model for processing, and output the real-time strategy value QS n ; Send the latter real-time analysis part of the selected n-th real-time training sample to the target pulse signal control strategy output model for processing, output the target simulation pulse signal control strategy, then combine the latter real-time analysis part of the selected n-th real-time training sample with the target simulation pulse signal control strategy to form a target simulation analysis part, and send the target simulation analysis part to the target strategy evaluation network model for processing, and output the long-term target strategy value T n , based on the real-time training samples, calculate the reward value R corresponding to the real-time training samples n , calculate the comprehensive target strategy value QT n =R n +γT n , where γ is the discount factor; until all real-time training samples are selected, the strategy evaluation loss value of the strategy evaluation network model is calculated , by minimizing the policy evaluation loss value U of the policy evaluation network model, the parameters of the policy evaluation network model are updated in real time using the gradient descent method; The real-time training samples are selected in chronological order. For each selected n-th real-time training sample, the subsequent real-time analysis part of the selected n-th real-time training sample is sent to the pulse signal control strategy output model for processing, and the simulated pulse signal control strategy is output. Then, the pulse signal data set and the physiological parameter set inside the subsequent real-time analysis part of the selected n-th real-time training sample are combined with the simulated pulse signal control strategy to form a simulation analysis part, and the simulation analysis part is sent to the updated strategy evaluation network model for processing, and the simulated strategy value QE is output. n ; Until all real-time training samples are selected, calculate the pulse signal control strategy loss value of the pulse signal control strategy output model , by maximizing the pulse signal control strategy loss value Y of the pulse signal control strategy output model, the parameters of the pulse signal control strategy output model are updated using the gradient ascent method; It also includes updates to the target pulse signal control strategy output model and the target strategy evaluation network model. The initial target pulse signal control strategy output model is consistent with the pulse signal control strategy output model, and the target strategy evaluation network model is consistent with the strategy evaluation network model.

[0008] As a preferred aspect of the present invention, the updating of the target pulse signal control strategy output model and the target strategy evaluation network model specifically includes the following steps: The parameter in the target pulse signal control strategy output model is denoted as θ i , i=1, 2, 3, …, I, I is the total number of parameters in the target pulse signal control strategy output model. Whenever the parameters of the pulse signal control strategy output model are updated, the parameters in the target pulse signal control strategy output model are updated by the following formula, θ i =μη i +(1-μ)θ i , where η i is the parameter θ in the updated pulse signal control strategy output model and the target pulse signal control strategy output model i Corresponding parameters; The parameter in the target strategy evaluation network model is denoted as δ i , i=1, 2, 3, …, I, I is the total number of parameters in the target policy evaluation network model. Whenever the parameters of the policy evaluation network model are updated, the parameters in the target policy evaluation network model are updated by the following formula, δ i =μζ i +(1-μ)δ i , where ζ i The updated policy evaluation network model and the target policy evaluation network model parameter δ i The corresponding parameters.

[0009] As a preferred aspect of the present invention, calculating the reward value corresponding to the real-time training sample based on the real-time training sample specifically includes the following steps: For the real-time training samples, the two physiological parameter sets in the real-time training samples are subjected to a difference operation to construct a physiological parameter change set, and the physiological parameter change value in the physiological parameter change set is recorded as D m , m=1, 2, 3, ..., M, M is the total number of physiological parameter change values ​​in the physiological parameter change set, calculate the reward value , where α m is the reward weight corresponding to the mth physiological parameter, L m is the safe range of change corresponding to the mth physiological parameter, C is the penalty constant, and all reward weights α m The sum of is 1, which is obtained by processing the physiological parameter change set by the reward weight output model.

[0010] As a preferred aspect of the present invention, training is performed on the pulse signal control strategy output model and the strategy evaluation network model, specifically including the following steps: Step S1: Initialize the pulse signal control strategy output model and the strategy evaluation network model, and set the parameters of the target pulse signal control strategy output model and the target strategy evaluation network model to be consistent with the parameters of the corresponding pulse signal control strategy output model and the strategy evaluation network model; Step S2: construct N training samples, which are stored in the same format as the real-time training samples, and also include the pulse signal data set and physiological parameter set obtained at two adjacent monitoring time points, as well as the pulse signal control strategy output at the former monitoring time point; Step S3: train the pulse signal control strategy output model, the strategy evaluation network model, the target pulse signal control strategy output model and the target strategy evaluation network model through N training samples, the training method is consistent with the method of updating the pulse signal control strategy output model, and output the strategy evaluation loss value; Step S4: Determine whether the strategy evaluation loss value is less than the preset threshold. If the strategy evaluation loss value is less than the preset threshold, output the trained pulse signal control strategy output model and strategy evaluation network model, and use the trained pulse signal control strategy output model and strategy evaluation network model as the initialized target pulse signal control strategy output model and target strategy evaluation network model; otherwise, repeat steps S2-S3 to continue training the pulse signal control strategy output model and strategy evaluation network model.

[0011] As a preferred aspect of the present invention, constructing N training samples specifically includes the following steps: Step T1: randomly setting the pulse signal data set and physiological parameter set at the initial monitoring time point; Step T2: at the current monitoring time point, the set pulse signal data set and physiological parameter set are sent to the pulse signal control strategy output model, the training pulse signal control strategy is output, and random noise is added to the training pulse signal control strategy. The random noise can be Gaussian noise. Adding random noise can expand the search range of the pulse signal control strategy; Step T3: Execute the training pulse signal control strategy with added noise through simulation software, obtain the pulse signal data set and physiological parameter set at the current monitoring time point, and obtain the reward value of the training pulse signal control strategy; the pulse signal data set and physiological parameter set at the current monitoring time point, the pulse signal data set and physiological parameter set obtained at the previous monitoring time point, and the pulse signal control strategy output at the previous monitoring time point form a training sample; Step T4: Repeat steps T2 to T3 until N training samples are constructed.

[0012] As a preferred aspect of the present invention, training the reward weight output model includes the following steps: In the process of training the pulse signal control strategy output model and the strategy evaluation network model through N training samples each time, all reward weights α are optimized through the swarm optimization algorithm. m Perform simulation update, the corresponding fitness is the pulse signal control strategy loss value of the pulse signal control strategy output model and the strategy evaluation network model trained by N training samples, the output of the difference operation of the two physiological parameter sets in the training samples is recorded as the training physiological parameter change set, and the training physiological parameter change set is marked by the reward weight combination corresponding to the maximum pulse signal control strategy loss value; All labeled training physiological parameter change sets are combined into a reward weight training set, and the reward weight training set is sent to the reward weight output model with parameter initialization for training, the reward weight loss value is calculated, and it is determined whether the reward weight loss value is within the preset range. If the reward weight loss value is within the preset range, the trained reward weight output model is output; otherwise, the reward weight output model is continued to be trained through the reward weight training set.

[0013] A pulse signal adaptive control system for electrotherapy equipment, comprising: A data acquisition module, used for acquiring a pulse signal data set and a patient's physiological parameter set at a current monitoring time point, wherein the pulse signal data set includes pulse signal data, and the patient's physiological parameter set includes physiological parameters; A pulse signal control strategy output model updating module is used to update the pulse signal control strategy output model based on the pulse signal data set and physiological parameter set acquired at the current monitoring time point; The pulse signal control strategy output module is used to form a state data set from the pulse signal data set and the physiological parameter set obtained at the current monitoring time point, and send the state data set to the updated pulse signal control strategy output model for processing, output the pulse signal control strategy, and then control the pulse signal output by the electrotherapy equipment through the pulse signal control strategy.

[0014] The present invention has the following advantages: 1. The present invention adjusts the pulse signal of the electrotherapy equipment through the pulse signal control strategy output model, and obtains the physiological parameters of the patient's real-time feedback at the same interval, and then updates the pulse signal control strategy output model in real time based on the real-time feedback physiological parameters, so that the pulse signal control strategy output by the pulse signal control strategy output model can better meet the patient's treatment needs and physiological tolerance needs, thereby improving the use effect of the electrotherapy equipment.

[0015] 2. The present invention updates the target pulse signal control strategy output model and the target strategy evaluation network model by means of soft update, which can make the target pulse signal control strategy output model and the target strategy evaluation network model approach the parameters of the corresponding pulse signal control strategy output model and the strategy evaluation network model without drastic changes, thereby ensuring that when the pulse signal control strategy output model is updated, the comprehensive target strategy value will not fluctuate drastically, thereby improving the stability of the pulse signal control strategy output model when updating.

[0016] 3. The present invention determines the reward value of the pulse signal control strategy based on the changes in the patient's physiological parameters, and sets a safety mechanism to avoid excessive changes in the patient's physiological parameters caused by the output pulse signal control strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the structure of a pulse signal adaptive control system for electrotherapy equipment adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0019] Embodiment 1, a pulse signal adaptive control method for an electrotherapy device, comprising: A pulse signal data set and a patient's physiological parameter set are obtained at the current monitoring time point. The pulse signal data set includes pulse signal data such as the frequency, pulse width, intensity and waveform type of the current pulse signal. Specifically, the frequency of the pulse signal refers to the number of pulses generated per second. Generally, low frequency is used for deep muscle stimulation, and high frequency is used for pain management. The pulse width indicates the duration of each pulse. The longer the pulse width, the stronger the stimulation to the tissue. The intensity refers to the amplitude of the pulse current. The waveform type refers to the shape of the pulse current, such as square wave and triangle wave. The patient's physiological parameter set includes physiological parameters such as skin resistance value, electromyographic signal amplitude and heart rate. The skin resistance value can reflect the patient's pain sensation, the electromyographic signal amplitude can reflect the muscle activity, such as muscle recovery or muscle reaction, and the heart rate can reflect the patient's physiological load. Too high a heart rate may mean that the patient feels uncomfortable. The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point. During the updating process, the parameters in the pulse signal control strategy output model can be adjusted according to the patient's real-time feedback, so that the pulse signal control strategy output by the pulse signal control strategy output model is more in line with the patient's actual situation. For example, when pain is relieved through transcutaneous electrical nerve stimulation, if the skin resistance value decreases (indicating that the patient feels discomfort or pain), the device will reduce the intensity or frequency of the pulse current; The pulse signal data set and the physiological parameter set obtained at the current monitoring time point are combined into a state data set, and the state data set is sent to the updated pulse signal control strategy output model for processing, and the pulse signal control strategy is output, and then the pulse signal output by the electrotherapy device is controlled by the pulse signal control strategy; The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point, specifically including the following steps: The pulse signal data set and physiological parameter set obtained at the current monitoring time point are composed of the latter real-time analysis part, the pulse signal data set and physiological parameter set obtained at the previous monitoring time point and the pulse signal control strategy output at the previous monitoring time point are composed of the former real-time analysis part, and the former real-time analysis part and the latter real-time analysis part are composed of real-time training samples. In the real-time training samples, the pulse signal data set and physiological parameter set obtained at the current monitoring time point can reflect the changes in the pulse signal data set and physiological parameter set obtained at the previous monitoring time point after the pulse signal control strategy output at the previous monitoring time point is executed, and the benefits obtained by executing the pulse signal control strategy output at the previous monitoring time point can be further analyzed, based on which the advantages and disadvantages of the pulse signal control strategy can be judged, and then the pulse signal control strategy output model can be adjusted; Get the first N real-time training samples. It should be noted that each time the pulse signal control strategy is executed, a real-time training sample will be constructed. Each time the pulse signal control strategy output model is updated, the real-time training samples corresponding to the current monitoring time point and the previous monitoring time point will be selected; the real-time training samples are selected in chronological order. For each selected nth real-time training sample, n=1, 2, 3, ..., N, where n is the sequence number of the real-time training sample, the previous real-time analysis part of the selected nth real-time training sample is sent to the strategy evaluation network model for processing, and the real-time strategy value QS is output. n , the real-time strategy value can reflect the pros and cons of the selected pulse signal control strategy under the state corresponding to the previous real-time analysis part, that is, whether it is suitable for the patient's treatment needs and physiological tolerance needs; the latter real-time analysis part in the selected n-th real-time training sample is sent to the target pulse signal control strategy output model for processing, and the target simulated pulse signal control strategy is output, and then the latter real-time analysis part in the selected n-th real-time training sample and the target simulated pulse signal control strategy are combined to form a target simulation analysis part, and the target simulation analysis part is sent to the target strategy evaluation network model for processing, and the long-term target strategy value T is output. n , the long-term target strategy value reflects the cumulative reward of the pulse signal control strategy output model in the long-term selection of the pulse signal control strategy. Since the pulse current itself is a gradual process, the patient's treatment needs and physiological tolerance needs also change over time. Therefore, considering the reward of the long-term selection of the pulse signal control strategy, the adaptability of the pulse signal control strategy output can be improved. The reward value R corresponding to the real-time training sample is calculated based on the real-time training sample. n , calculate the comprehensive target strategy value QT n =R n +γT n , where γ is a discount factor, which is used to characterize the emphasis of the reward value and the target policy value, and is determined by the developer based on experience; until all real-time training samples are selected, the policy evaluation loss value of the policy evaluation network model is calculated The policy evaluation loss value of the policy evaluation network model reflects the consistency of the pulse signal control strategy output by the pulse signal control strategy output model under different conditions. The lower the policy evaluation loss value, the better the consistency. The pulse signal control strategy output by the pulse signal control strategy output model is more stable and accurate. By minimizing the policy evaluation loss value U of the policy evaluation network model, the parameters of the policy evaluation network model are updated in real time using the gradient descent method. The gradient descent method can use the Adam optimizer. The real-time training samples are selected in chronological order. For each selected n-th real-time training sample, the subsequent real-time analysis part of the selected n-th real-time training sample is sent to the pulse signal control strategy output model for processing, and the simulated pulse signal control strategy is output. Then, the pulse signal data set and the physiological parameter set inside the subsequent real-time analysis part of the selected n-th real-time training sample are combined with the simulated pulse signal control strategy to form a simulation analysis part, and the simulation analysis part is sent to the updated strategy evaluation network model for processing, and the simulated strategy value QE is output. n ; Until all real-time training samples are selected, calculate the pulse signal control strategy loss value of the pulse signal control strategy output model , by maximizing the pulse signal control strategy loss value Y of the pulse signal control strategy output model, the parameters of the pulse signal control strategy output model are updated using the gradient ascent method; It also includes the update of the target pulse signal control strategy output model and the target strategy evaluation network model. The initial target pulse signal control strategy output model is consistent with the pulse signal control strategy output model, and the target strategy evaluation network model is consistent with the strategy evaluation network model, that is, when it is first used, the trained pulse signal control strategy output model is directly used as the target pulse signal control strategy output model, and the trained strategy evaluation network model is used as the target strategy evaluation network model; and the pulse signal control strategy output model, strategy evaluation network model, target pulse signal control strategy output model and target strategy evaluation network model used in this embodiment are all established based on the BP neural network; The present application adjusts the pulse signal of the electrotherapy equipment through a pulse signal control strategy output model, and obtains the patient's real-time feedback of physiological parameters at the same interval, and then updates the pulse signal control strategy output model in real time based on the real-time feedback of physiological parameters, so that the pulse signal control strategy output by the pulse signal control strategy output model can be more in line with the patient's treatment needs and physiological tolerance needs, thereby improving the use effect of the electrotherapy equipment.

[0020] The updating of the target pulse signal control strategy output model and the target strategy evaluation network model specifically includes the following steps: The parameter in the target pulse signal control strategy output model is denoted as θ i , i=1, 2, 3, …, I, I is the total number of parameters in the target pulse signal control strategy output model. Whenever the parameters of the pulse signal control strategy output model are updated, the parameters in the target pulse signal control strategy output model are updated by the following formula, θ i =μη i +(1-μ)θ i , where η iis the parameter θ in the updated pulse signal control strategy output model and the target pulse signal control strategy output model i The corresponding parameters, μ is the update constant, which is generally 0.001; The parameter in the target strategy evaluation network model is denoted as δ i , i=1, 2, 3, …, I, I is the total number of parameters in the target policy evaluation network model. Whenever the parameters of the policy evaluation network model are updated, the parameters in the target policy evaluation network model are updated by the following formula, δ i =μζ i +(1-μ)δ i , where ζ i The updated policy evaluation network model and the target policy evaluation network model parameter δ i Corresponding parameters; The present application updates the target pulse signal control strategy output model and the target strategy evaluation network model by means of soft update, which can make the target pulse signal control strategy output model and the target strategy evaluation network model approach the parameters of the corresponding pulse signal control strategy output model and the strategy evaluation network model without drastic changes, thereby ensuring that when the pulse signal control strategy output model is updated, the comprehensive target strategy value will not fluctuate drastically, thereby improving the stability of the pulse signal control strategy output model when updating.

[0021] Calculating the reward value corresponding to the real-time training sample based on the real-time training sample specifically includes the following steps: For the real-time training samples, the two physiological parameter sets in the real-time training samples are subjected to a difference operation to construct a physiological parameter change set, and the physiological parameter change value in the physiological parameter change set is recorded as D m , m=1, 2, 3, ..., M, M is the total number of physiological parameter change values ​​in the physiological parameter change set, which is consistent with the total number of physiological parameters in the physiological parameter set, and the reward value is calculated , where α m is the reward weight corresponding to the mth physiological parameter, L m is the safe variation range corresponding to the mth physiological parameter. During electrotherapy, excessive changes in the patient's physiological parameters will cause serious effects. Therefore, in the process of outputting the pulse signal control strategy, it is not expected that the output will cause the pulse signal control strategy to exceed the range of physiological parameter changes. Therefore, it is necessary to punish the pulse signal control strategy that will cause the pulse signal control strategy to exceed the range of physiological parameter changes, so that the updated pulse signal control strategy model can avoid such behavior that will cause the physiological parameter to exceed the range of change. C is the penalty constant, which is set by the developer according to the corresponding physiological parameters, and all reward weights α m The sum of is 1, which is obtained by the reward weight output model processing the physiological parameter change set.

[0022] The present application determines the reward value of the pulse signal control strategy based on the changes in the patient's physiological parameters, and sets a safety mechanism to prevent the output of the pulse signal control strategy from causing excessive changes in the patient's physiological parameters.

[0023] Training the pulse signal control strategy output model and strategy evaluation network model includes the following steps: Step S1: Initialize the pulse signal control strategy output model and the strategy evaluation network model, and set the parameters of the target pulse signal control strategy output model and the target strategy evaluation network model to be consistent with the parameters of the corresponding pulse signal control strategy output model and the strategy evaluation network model; Step S2: construct N training samples, which are stored in the same format as the real-time training samples, and also include the pulse signal data set and physiological parameter set obtained at two adjacent monitoring time points, as well as the pulse signal control strategy output at the former monitoring time point; Step S3: train the pulse signal control strategy output model, the strategy evaluation network model, the target pulse signal control strategy output model and the target strategy evaluation network model through N training samples, the training method is consistent with the method of updating the pulse signal control strategy output model, and output the strategy evaluation loss value; Step S4: Determine whether the strategy evaluation loss value is less than a preset threshold. The preset threshold is determined by the developer and is used to determine whether the pulse signal control strategy output by the pulse signal control strategy output model is stable, that is, the pulse signal control strategy output model will select the same or similar actions under the same state. If the strategy evaluation loss value is less than the preset threshold, output the trained pulse signal control strategy output model and strategy evaluation network model, and use the trained pulse signal control strategy output model and strategy evaluation network model as the initialized target pulse signal control strategy output model and target strategy evaluation network model; otherwise, repeat steps S2-S3 to continue training the pulse signal control strategy output model and strategy evaluation network model.

[0024] Constructing N training samples includes the following steps: Step T1: randomly setting the pulse signal data set and physiological parameter set at the initial monitoring time point; Step T2: at the current monitoring time point, the set pulse signal data set and physiological parameter set are sent to the pulse signal control strategy output model, the training pulse signal control strategy is output, and random noise is added to the training pulse signal control strategy. The random noise can be Gaussian noise. Adding random noise can expand the search range of the pulse signal control strategy; Step T3: Execute the training pulse signal control strategy with added noise through simulation software, obtain the pulse signal data set and physiological parameter set at the current monitoring time point, and obtain the reward value of the training pulse signal control strategy. The simulation software can use the HumMod model; the pulse signal data set and physiological parameter set at the current monitoring time point, the pulse signal data set and physiological parameter set obtained at the previous monitoring time point, and the pulse signal control strategy output at the previous monitoring time point form a training sample; Step T4: Repeat steps T2 to T3 until N training samples are constructed.

[0025] Training the reward weight output model includes the following steps: In the process of training the pulse signal control strategy output model and the strategy evaluation network model through N training samples each time, all reward weights α are optimized through the swarm optimization algorithm. m For simulation update, the swarm optimization algorithm can adopt the sparrow search algorithm, and the corresponding fitness is the pulse signal control strategy loss value of the pulse signal control strategy output model and the strategy evaluation network model trained by N training samples. The output of the difference operation of the two physiological parameter sets in the training samples is recorded as the training physiological parameter change set, and the training physiological parameter change set is marked by the reward weight combination corresponding to the maximum pulse signal control strategy loss value; All the labeled training physiological parameter change sets are combined into a reward weight training set, and the reward weight training set is sent to the reward weight output model with parameter initialization for training. The reward weight output model is established based on the BP neural network, calculates the reward weight loss value, and determines whether the reward weight loss value is within the preset range. The preset range is determined by the developer. If the reward weight loss value is within the preset range, the trained reward weight output model is output; otherwise, the reward weight output model is continued to be trained through the reward weight training set.

[0026] Embodiment 2, a pulse signal adaptive control system for electrotherapy equipment, such as Figure 1 As shown, including: The data acquisition module is used to obtain the pulse signal data set and the patient's physiological parameter set at the current monitoring time point. The pulse signal data set includes pulse signal data such as the frequency, pulse width, intensity and waveform type of the current pulse signal. Specifically, the frequency of the pulse signal refers to the number of pulses generated per second. Generally, low frequency is used for deep muscle stimulation, and high frequency is used for pain management. The pulse width indicates the duration of each pulse. The longer the pulse width, the stronger the stimulation to the tissue. The intensity refers to the amplitude of the pulse current. The waveform type refers to the shape of the pulse current, such as square wave and triangle wave. The patient's physiological parameter set includes physiological parameters such as skin resistance value, electromyographic signal amplitude and heart rate. The skin resistance value can reflect the patient's pain, the electromyographic signal amplitude can reflect the muscle activity, such as muscle recovery or muscle reaction, and the heart rate can reflect the patient's physiological load. Too high a heart rate may mean that the patient feels uncomfortable. The pulse signal control strategy output model update module is used to update the pulse signal control strategy output model based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point. During the update process, the parameters in the pulse signal control strategy output model can be adjusted according to the patient's real-time feedback, so that the pulse signal control strategy output by the pulse signal control strategy output model is more in line with the patient's actual situation. For example, if the skin resistance value decreases (indicating that the patient feels discomfort or pain) to relieve pain through transcutaneous electrical nerve stimulation, the device will reduce the intensity or frequency of the pulse current; The pulse signal control strategy output module is used to form a state data set from the pulse signal data set and the physiological parameter set obtained at the current monitoring time point, and send the state data set to the updated pulse signal control strategy output model for processing, output the pulse signal control strategy, and then control the pulse signal output by the electrotherapy equipment through the pulse signal control strategy.

[0027] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A pulse signal adaptive control method for electrotherapy equipment, characterized in that: include: Acquire a pulse signal data set and a patient's physiological parameter set at a current monitoring time point, wherein the pulse signal data set includes pulse signal data, and the patient's physiological parameter set includes physiological parameters; The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point; The pulse signal data set and physiological parameter set obtained at the current monitoring time point are combined into a state data set, and the state data set is sent to the updated pulse signal control strategy output model for processing, and the pulse signal control strategy is output, and then the pulse signal output by the electrotherapy equipment is controlled by the pulse signal control strategy.

2. A pulse signal adaptive control method for electrotherapy equipment according to claim 1, characterized in that: include: The pulse signal control strategy output model is updated based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point, specifically including the following steps: The pulse signal data set and the physiological parameter set obtained at the current monitoring time point constitute the latter real-time analysis part, the pulse signal data set and the physiological parameter set obtained at the previous monitoring time point and the pulse signal control strategy output at the previous monitoring time point constitute the former real-time analysis part, and the former real-time analysis part and the latter real-time analysis part constitute the real-time training sample; Get the first N real-time training samples; Select real-time training samples in chronological order. For each selected nth real-time training sample, n=1, 2, 3, ..., N, where n is the sequence number of the real-time training sample, send the previous real-time analysis part of the selected nth real-time training sample to the strategy evaluation network model for processing, and output the real-time strategy value QS n ; Send the latter real-time analysis part of the selected n-th real-time training sample to the target pulse signal control strategy output model for processing, output the target simulation pulse signal control strategy, then combine the latter real-time analysis part of the selected n-th real-time training sample with the target simulation pulse signal control strategy to form a target simulation analysis part, and send the target simulation analysis part to the target strategy evaluation network model for processing, and output the long-term target strategy value T n , based on the real-time training samples, calculate the reward value R corresponding to the real-time training samples n , calculate the comprehensive target strategy value QT n =R n +γT n , where γ is the discount factor; until all real-time training samples are selected, the strategy evaluation loss value of the strategy evaluation network model is calculated , by minimizing the policy evaluation loss value U of the policy evaluation network model, the parameters of the policy evaluation network model are updated in real time using the gradient descent method; The real-time training samples are selected in chronological order. For each selected n-th real-time training sample, the subsequent real-time analysis part of the selected n-th real-time training sample is sent to the pulse signal control strategy output model for processing, and the simulated pulse signal control strategy is output. Then, the pulse signal data set and the physiological parameter set inside the subsequent real-time analysis part of the selected n-th real-time training sample are combined with the simulated pulse signal control strategy to form a simulation analysis part, and the simulation analysis part is sent to the updated strategy evaluation network model for processing, and the simulated strategy value QE is output. n ; Until all real-time training samples are selected, calculate the pulse signal control strategy loss value of the pulse signal control strategy output model , by maximizing the pulse signal control strategy loss value Y of the pulse signal control strategy output model, the parameters of the pulse signal control strategy output model are updated using the gradient ascent method; It also includes updates to the target pulse signal control strategy output model and the target strategy evaluation network model. The initial target pulse signal control strategy output model is consistent with the pulse signal control strategy output model, and the target strategy evaluation network model is consistent with the strategy evaluation network model.

3. A pulse signal adaptive control method for electrotherapy equipment according to claim 2, characterized in that: The updating of the target pulse signal control strategy output model and the target strategy evaluation network model specifically includes the following steps: The parameter in the target pulse signal control strategy output model is denoted as θ i , i=1, 2, 3, …, I, I is the total number of parameters in the target pulse signal control strategy output model. Whenever the parameters of the pulse signal control strategy output model are updated, the parameters in the target pulse signal control strategy output model are updated by the following formula, θ i =μη i +(1-μ)θ i , where η i is the parameter θ in the updated pulse signal control strategy output model and the target pulse signal control strategy output model i Corresponding parameters; The parameter in the target strategy evaluation network model is denoted as δ i , i=1, 2, 3, …, I, I is the total number of parameters in the target policy evaluation network model. Whenever the parameters of the policy evaluation network model are updated, the parameters in the target policy evaluation network model are updated by the following formula, δ i =μζ i +(1-μ)δ i , where ζ i The updated policy evaluation network model and the target policy evaluation network model parameter δ i The corresponding parameters.

4. A pulse signal adaptive control method for electrotherapy equipment according to claim 3, characterized in that: Calculating the reward value corresponding to the real-time training sample based on the real-time training sample specifically includes the following steps: For the real-time training samples, the two physiological parameter sets in the real-time training samples are subjected to a difference operation to construct a physiological parameter change set, and the physiological parameter change value in the physiological parameter change set is recorded as D m , m=1, 2, 3, ..., M, M is the total number of physiological parameter change values ​​in the physiological parameter change set, calculate the reward value , where α m is the reward weight corresponding to the mth physiological parameter, L m is the safe range of change corresponding to the mth physiological parameter, C is the penalty constant, and all reward weights α m The sum of is 1, which is obtained by processing the physiological parameter change set by the reward weight output model.

5. A pulse signal adaptive control method for electrotherapy equipment according to claim 4, characterized in that: Training the pulse signal control strategy output model and strategy evaluation network model includes the following steps: Step S1: Initialize the pulse signal control strategy output model and the strategy evaluation network model, and set the parameters of the target pulse signal control strategy output model and the target strategy evaluation network model to be consistent with the parameters of the corresponding pulse signal control strategy output model and the strategy evaluation network model; Step S2: construct N training samples, which are stored in the same format as the real-time training samples, and also include the pulse signal data set and physiological parameter set obtained at two adjacent monitoring time points, as well as the pulse signal control strategy output at the former monitoring time point; Step S3: train the pulse signal control strategy output model, the strategy evaluation network model, the target pulse signal control strategy output model and the target strategy evaluation network model through N training samples, the training method is consistent with the method of updating the pulse signal control strategy output model, and output the strategy evaluation loss value; Step S4: Determine whether the strategy evaluation loss value is less than the preset threshold. If the strategy evaluation loss value is less than the preset threshold, output the trained pulse signal control strategy output model and strategy evaluation network model, and use the trained pulse signal control strategy output model and strategy evaluation network model as the initialized target pulse signal control strategy output model and target strategy evaluation network model; otherwise, repeat steps S2-S3 to continue training the pulse signal control strategy output model and strategy evaluation network model.

6. A pulse signal adaptive control method for electrotherapy equipment according to claim 5, characterized in that: Constructing N training samples includes the following steps: Step T1: randomly setting the pulse signal data set and physiological parameter set at the initial monitoring time point; Step T2: at the current monitoring time point, the set pulse signal data set and physiological parameter set are sent to the pulse signal control strategy output model, the training pulse signal control strategy is output, and random noise is added to the training pulse signal control strategy. The random noise can be Gaussian noise. Adding random noise can expand the search range of the pulse signal control strategy; Step T3: Execute the training pulse signal control strategy with added noise through simulation software, obtain the pulse signal data set and physiological parameter set at the current monitoring time point, and obtain the reward value of the training pulse signal control strategy; The pulse signal data set and physiological parameter set at the current monitoring time point, the pulse signal data set and physiological parameter set obtained at the previous monitoring time point, and the pulse signal control strategy output at the previous monitoring time point form a training sample; Step T4: Repeat steps T2 to T3 until N training samples are constructed.

7. A pulse signal adaptive control method for electrotherapy equipment according to claim 6, characterized in that: Training the reward weight output model includes the following steps: In the process of training the pulse signal control strategy output model and the strategy evaluation network model through N training samples each time, all reward weights α are optimized through the swarm optimization algorithm. m Perform simulation update, the corresponding fitness is the pulse signal control strategy loss value of the pulse signal control strategy output model and the strategy evaluation network model trained by N training samples, the output of the difference operation of the two physiological parameter sets in the training samples is recorded as the training physiological parameter change set, and the training physiological parameter change set is marked by the reward weight combination corresponding to the maximum pulse signal control strategy loss value; All labeled training physiological parameter change sets are combined into a reward weight training set, and the reward weight training set is sent to the reward weight output model with parameter initialization for training, the reward weight loss value is calculated, and it is determined whether the reward weight loss value is within the preset range. If the reward weight loss value is within the preset range, the trained reward weight output model is output; otherwise, the reward weight output model is continued to be trained through the reward weight training set.

8. A pulse signal adaptive control system for electrotherapy equipment, characterized in that: The system applies a pulse signal adaptive control method for electrotherapy equipment according to any one of claims 1 to 7, comprising: A data acquisition module, used for acquiring a pulse signal data set and a patient's physiological parameter set at a current monitoring time point, wherein the pulse signal data set includes pulse signal data, and the patient's physiological parameter set includes physiological parameters; A pulse signal control strategy output model updating module is used to update the pulse signal control strategy output model based on the pulse signal data set and physiological parameter set obtained at the current monitoring time point; The pulse signal control strategy output module is used to form a state data set from the pulse signal data set and the physiological parameter set obtained at the current monitoring time point, and send the state data set to the updated pulse signal control strategy output model for processing, output the pulse signal control strategy, and then control the pulse signal output by the electrotherapy equipment through the pulse signal control strategy.

Images