A traditional Chinese medicine introduction intensity control method for a traditional Chinese medicine directional medicine penetration therapeutic instrument

Abstract

This invention discloses a method for controlling the intensity of TCM (Traditional Chinese Medicine) drug delivery in a TCM-targeted transdermal therapy device, relating to the field of medical device control technology. Based on individual patient and treatment information, this method uses a machine learning model to generate personalized initial intensity parameters for the therapy device. It simultaneously collects the patient's objective physiological signals and subjective feelings in real time, constructs a state vector, and inputs it to an adaptive regulator employing a two-layer decision-making architecture. Through an outer layer of deep reinforcement learning intelligence and an inner layer of constrained model predictive controller, the method outputs the final control current value at the current moment. Based on this current value, it drives the treatment electrodes of the therapy device to output corresponding energy to adjust the electric field force and osmotic pressure, achieving precise control of the TCM delivery intensity. This method solves the technical problems of existing methods, such as the difficulty in quantifying the patient's subjective feelings into reliable control signals and the lack of comprehensive benefit optimization in control decisions, achieving dynamic optimal control of the TCM delivery intensity within a safe boundary.

Description

Technical Field

[0001] This application relates to the field of medical device control technology, and in particular to a method for controlling the intensity of Chinese medicine delivery in a Chinese medicine-targeted transdermal drug delivery device. Background Technology

[0002] The TCM-guided transdermal drug delivery device is a physical therapy device that combines TCM meridian theory with modern iontophoresis technology. Its core performance and ultimate efficacy depend highly on the control precision, stability, and adaptability of the "induction intensity" (usually referring to current or voltage intensity) during the treatment process.

[0003] Existing methods employ open-loop programming and simple single-feedback closed-loop control, with the device operating according to a preset, fixed intensity-time curve. This approach ignores individual patient differences and the dynamic changes in skin impedance during treatment. Furthermore, some devices incorporate monitoring of a single electrical variable and use it as feedback for adjustment. However, the dynamic characteristics of the human skin-electrode system change with current magnitude, temperature, and drug composition. Traditional PID controllers have fixed parameters, which are prone to overshoot, oscillation, or slow adjustment when dealing with such nonlinear controlled objects, leading to instability in the treatment process.

[0004] Traditional Chinese medicine (TCM) treatment emphasizes the "feeling of Qi," but the existing control methods of TCM-targeted transdermal drug delivery devices are difficult to quantify the patient's subjective sensations of warmth, acupuncture, etc., into reliable closed-loop control signals. Control decisions are mostly based on the current instantaneous state and lack optimization of the overall patient comfort benefits throughout a single treatment. Safety protection is mostly triggered by post-treatment thresholds and fails to proactively avoid risks in the control decision-making process. Summary of the Invention

[0005] To address the technical problems of the prior art, this application provides a method for controlling the intensity of Chinese medicine delivery in a TCM-targeted transdermal drug delivery device. This method integrates the patient's subjective feelings and objective physiological signals in real time and utilizes intelligent algorithms to achieve dynamic optimal control of the intensity of Chinese medicine delivery within a safe boundary. Under the premise of ensuring patient safety and comfort, it maintains an electric field force sufficient to drive effective drug penetration and can dynamically adapt to the complex changes in skin impedance, drug properties, and patient sensations during the treatment process.

[0006] This application provides a method for controlling the intensity of Chinese medicine delivery in a targeted transdermal drug delivery device, including: Step S10: Based on the received individual patient information and treatment information, apply a pre-trained gradient boosting decision tree model to generate and output initial intensity parameters, setting a safe and personalized initial state for the TCM-targeted transdermal drug delivery device. Step S20: During the operation of the TCM-targeted transdermal drug delivery therapy device, the patient's objective physiological signals and subjective feeling signals are collected simultaneously in real time. The subjective feeling signal is the patient's quantitative feeling score of the treatment intensity obtained in real time through the human-computer interaction interface of the therapy device. Step S30: Based on the real-time collected objective physiological signals and subjective feelings of the patient, a state vector is constructed and input to an adaptive regulator with a two-layer decision architecture. Through the outer layer deep reinforcement learning agent and the inner layer constrained model prediction controller, the final control current value at the current moment is calculated and output. In step S40, based on the final control current value output in step S30, the treatment electrode of the TCM directional transdermal drug delivery device is driven to output corresponding physical energy to adjust the electric field force and osmotic pressure, thereby achieving precise control over the intensity of TCM drug delivery, simultaneously monitoring the hardware safety parameters of the device, and triggering a forced protection action when the parameters exceed the limit.

[0007] Furthermore, the patient's individual information includes at least age, TCM constitution type, treatment site, and skin condition at the treatment site; the treatment information includes at least the conductivity range of the TCM liquid used and the preset treatment duration; based on the above information, a pre-trained gradient boosting decision tree model is used to recommend the initial intensity, target intensity range, and intensity ramp-up rate of TCM targeted transdermal drug delivery for the TCM transdermal drug delivery device.

[0008] Furthermore, after the TCM-targeted transdermal drug delivery device is activated, it simultaneously collects and quantifies multimodal signals: Objective physiological signals: real-time skin impedance of the patient, actual feedback current through the treatment area, and calculation of the rate of change of impedance over time; Subjective perception signal: The patient obtains a quantitative perception score of the treatment intensity through a slider operated in real time. This score is positively correlated with the intensity of the patient's perceived warmth and tingling sensations. The score range is [0,10], and the rate of change of the perception score over time is calculated.

[0009] In addition, the patient's subjective feeling score is not always reliable. When the patient's real-time skin impedance is stable but the subjective feeling score changes by more than a set threshold within a set time, a reliability decay mechanism is activated to reduce the weight of the subjective feeling score in the state vector and give higher weight to the objective physiological signal, so as to avoid control inaccuracy caused by patient misoperation or momentary sensory deviation.

[0010] Furthermore, at each control moment, the collected and calculated multimodal information is combined into a state vector, including real-time feedback current, real-time sensory score, instantaneous skin impedance, impedance change rate, sensory score change rate, treatment time, and target intensity benchmark of Chinese medicine delivery. The outer layer of the adaptive regulator uses a deep reinforcement learning agent to make decisions and provide the adjustment direction for the target current. This agent employs an actor-critic structure, using the actor network as the policy function. It takes the current state vector as input and outputs the suggested control action, i.e., the adjustment amount for the desired target current value. The agent is trained by maximizing a reward function, which is a weighted sum of the immediate perception score, the tracking error between the feedback current and the target current, and the rate of change of the current. The inner layer of the adaptive regulator uses a model predictive controller with safety constraints. It receives control actions suggested by the agent, uses them as reference targets, and solves a constrained optimization problem within a set prediction window. The controller uses a long short-term memory network to predict the feedback current value and subjective feeling score at future moments based on the historical state sequence. The controller solves an optimization problem in each control cycle. The objective function of the optimization problem is to minimize the deviation between the feedback current and the reference target current and the change of control quantity in the prediction time domain. The reference target current is determined by the current target intensity benchmark and the suggested control action. The absolute current safety boundary and the output current change rate limit are used as hard constraints, and the predicted comfort level is not exceeded as a soft constraint. After solving the problem, the optimal control sequence of the therapeutic device output current is obtained.

[0011] This application discloses the following technical effects: This application proposes a method for controlling the intensity of TCM drug delivery in a TCM-targeted transdermal drug delivery device. It employs an adaptive regulator based on a two-layer decision-making architecture. The inner layer, a deep reinforcement learning agent, makes decisions based on the cumulative therapeutic benefit, automatically searching for and determining the optimal intensity curve pattern for a specific patient and specific symptoms, thus improving the penetration efficiency of the TCM drug. The outer layer, a model predictive controller, predicts changes in the patient's skin impedance through a recurrent neural network, adjusting the output in advance to counteract changes in drug penetration resistance caused by skin hydration, pore opening and closing, etc., maintaining a relatively stable electric field force and osmotic pressure acting on the drug. Before the treatment device is run, the method provides an initialization scheme through a decision tree model to ensure that the initial state of the treatment device matches the patient's basic condition. During the operation of the treatment device, the scheme is adjusted online through the strategy of the deep reinforcement learning agent and the prediction model of the model prediction controller, so that the control strategy adapts to the patient's current physical response in real time and achieves deep personalization of the intensity control of Chinese medicine delivery. Furthermore, the constraint optimization mechanism of the model predictive controller actively and smoothly reduces the control current value of the therapeutic device before predicting that the patient's subjective feeling score will exceed the standard or the current will exceed the limit. This fundamentally avoids multiple stimuli and discomfort caused by response lag, ensuring the safety and stability of the treatment process. Taking the patient's subjective feeling as the core control input and optimization target of the method proposed in this application, the treatment process is transformed from machine-led to human-machine collaboration, actively adapting to and seeking the patient's comfort range, and improving the compliance of the intensity control of Chinese medicine delivery. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0013] Figure 1 This is a flowchart illustrating a method for controlling the intensity of Chinese medicine delivery in a targeted transdermal drug delivery device, as provided in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] Example 1: This application provides a method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device, such as... Figure 1 As shown, the method includes: Step S10: Based on the received individual patient information and treatment information, a pre-trained gradient boosting decision tree model is applied to generate and output initial intensity parameters, setting a safe and personalized initial state for the TCM-targeted transdermal drug delivery device.

[0016] In this embodiment, the gradient boosting decision tree model is trained offline to obtain a pre-trained model, including three stages: data preparation and annotation, feature engineering and preprocessing, and model training and validation. Data preparation and annotation: Collect a large amount of historical clinical treatment data as input features for the model. Each data record contains individual patient information and treatment information. Individual patient information includes at least age, TCM constitution type, treatment site and skin condition of the treatment site. Treatment information includes at least the conductivity range of the Chinese herbal medicine solution used and the preset treatment duration. From successful, safe, and comfortable treatment records, ideal steady-state intensity parameters determined by experienced physicians are extracted as label data for model training. These parameters include at least the steady-state target current value during the treatment phase and the maximum safe ramp-up slope allowed from the initial intensity to the target current.

[0017] Feature engineering and preprocessing: Patient body type, treatment site and skin condition are encoded one-hot. Missing features are filled in using median, mode or other simple models. Continuous features (age, weight, conductivity) are standardized to eliminate the influence of dimensions and accelerate model convergence. The preprocessed input features are matched with the label data to form a dataset.

[0018] Model training and validation: The gradient boosting decision tree model is trained using the processed dataset to learn the complex mapping relationship between input features and labeled data; the dataset is divided into training set, validation set and test set, and the hyperparameters of the model (number of trees, maximum depth and learning rate) are optimized by grid search to obtain the best performance on the validation set. The model parameters at this time are retained to form the pre-trained model.

[0019] The pre-trained gradient boosting decision tree model is integrated into the control system of a TCM-based targeted transdermal drug delivery device, and loaded and run using a lightweight inference engine. When initiating treatment for a new patient, the physician inputs the patient's individual information and treatment information through the device's human-computer interaction interface. The treatment device control system encodes and preprocesses the input information according to the same process as the training phase, constructing a feature vector. This feature vector is then input into the pre-trained gradient boosting decision tree model, which outputs the predicted personalized initial intensity parameters, including at least the initial intensity of the Chinese medicine infusion (20% of the predicted target intensity), the target intensity range, and the intensity ramp-up rate.

[0020] Step S20: During the operation of the TCM-oriented transdermal drug delivery therapy device, the patient's objective physiological signals and subjective feeling signals are collected simultaneously in real time. The subjective feeling signal is the patient's quantitative feeling score of the treatment intensity obtained in real time through the human-computer interaction interface of the therapy device.

[0021] In this embodiment, after the TCM-targeted transdermal drug delivery device is activated, multimodal signals are acquired and quantified in parallel: Objective physiological signals include at least the patient's real-time skin impedance and the actual feedback current through the treatment area. The rate of change of impedance over time is calculated, and the measurement process is as follows: The patient's real-time skin impedance was acquired using a four-electrode method. In the treatment electrode pair, one pair of electrodes was used to apply the detection AC excitation signal, and the other pair of electrodes was used to measure the voltage drop generated therefrom. The complex impedance of the skin tissue was obtained by calculating the voltage-to-current ratio and phase difference of the AC excitation signal, and its magnitude was the required skin impedance. A sampling resistor is connected in series in the output circuit of the treatment current. By measuring the voltage drop across the resistor, the actual feedback current flowing through the patient's treatment area is calculated in real time using Ohm's law. The sampling circuit is electrically isolated from the main control system, and the sampling frequency is synchronized with the impedance measurement to capture instantaneous fluctuations in the current.

[0022] The subjective perception signal is a quantitative score of the patient's perception of the treatment intensity. It transforms the patient's subjective and vague sensations of "deqi" (warmth, acupuncture, soreness, numbness, distension, etc.) into quantitative, time-series signals that can be understood by the controller. The signal acquisition process includes: A touch slider is installed on the main unit of the treatment device. Its design conforms to human factors, making it convenient for patients to operate with one hand during treatment. The software interface of the therapeutic device displays a scale bar with visual feedback. The scale bar is a vertical light strip with a length of 10cm. The bottom is 0, indicating no sensation; the top is 10, indicating an unbearable stinging sensation. The scale bar is divided into different color intervals: the green area indicates a rating interval of [0,3], representing a faint sensation; the yellow area indicates a rating interval of [4,7], representing a noticeably comfortable warmth and tingling sensation; and the red area indicates a rating interval of [8,10], representing an excessively strong tingling sensation. The patient slides a touch bar to indicate the intensity of the current sensation by a cursor on the scale bar, reads the real-time sensation score from the scale bar, and calculates the rate of change of the score to determine whether the patient's sensation is comfortable, insufficient, or too strong.

[0023] In addition, the patient's subjective feeling score is not always reliable. When the patient's real-time skin impedance is stable but the subjective feeling score changes by more than a set threshold within a set time, a reliability decay mechanism is activated to reduce the weight of the subjective feeling score in the state vector and give higher weight to the objective physiological signal, so as to avoid control inaccuracy caused by patient misoperation or momentary sensory deviation.

[0024] Step S30: Based on the real-time collected objective physiological signals and subjective feelings of the patient, a state vector is constructed and input to an adaptive regulator with a two-layer decision architecture. Through the outer layer deep reinforcement learning agent and the inner layer constrained model prediction controller, the final control current value at the current moment is calculated and output.

[0025] In this embodiment, at each control moment, the real-time acquired objective physiological signals and subjective feeling signals of the patient are time-synchronized and filtered, and combined into a state vector. The vector consists of real-time feedback current, real-time sensory score, instantaneous skin impedance, impedance change rate, sensory score change rate, treatment time, and target intensity benchmark of Chinese medicine delivery. The outer layer of the adaptive regulator uses a deep reinforcement learning agent to make decisions, providing the adjustment direction for the target current. This agent employs an actor-critic structure, including: Actor network; input is a state vector The output is an action defined in the interval [-1, 1]. The recommended current adjustment amount is obtained through linear mapping. : ,in Indicates the maximum allowable adjustment amount per step; Two critic networks: A dual-network architecture is used to reduce overestimation, with the input being a state vector. and actions Output respectively Value estimation; Target network: This includes a target actor network and a target critic network. Its parameters are slowly tracked by the actor network and the critic network to stabilize the training process.

[0026] The aforementioned network is built upon a deep neural network, consisting of multiple convolutional neural networks and fully connected layers; the parameters of each layer are obtained through training, and a reward function is set. The function that guides the training process of the agent is a multinomial weighted sum, expressed by the formula:

[0027] in, This indicates the comfort reward sub-item. Perform a subjective assessment of the patient's feelings. The median of the preset ideal comfort score, when The reward is maximized when the score is 6 (equivalent to 1), and the reward value decreases according to a Gaussian distribution curve as the score deviates from 6. This indicates a sub-item for rewarding therapeutic effects. Feedback current value measured in real time With the present target current value at any time The smaller the deviation between the two values, the smaller the magnitude of the value and the higher the reward value. This indicates a stability reward sub-item. This represents the change in the feedback current at the current moment relative to the previous moment. This represents the maximum current change in a single step. The smoother the current change, the smaller the negative value of this term, which encourages stable output. This indicates a safety penalty; when the safety red line is crossed, a negative reward is given, and the current training round is terminated. , and This indicates the weighting coefficient of each reward sub-item, used to balance the importance of comfort, efficacy, and stability; During the operation of the TCM-targeted transdermal drug delivery device, the trained actor network freezes its internal parameters and runs in forward inference mode. Based on the current state vector, it outputs suggested control actions, namely the current adjustment amount.

[0028] The inner layer of the adaptive regulator employs a model predictive controller with safety constraints, which receives control actions suggested by the agent and transforms them into smoothly executable control instructions that satisfy all immediate safety and comfort constraints.

[0029] This controller uses a long short-term memory network as a dynamic prediction model to predict based on historical state vector sequences and control sequences. The state vector sequence of the step is used to obtain the predicted feedback current value. and subjective feelings rating The predictive model is pre-trained on historical data and fine-tuned using new data acquired in real time within an online Kalman filter framework to adapt to individual patient differences. The controller solves a rolling optimization problem in each control cycle to obtain the future... The control sequence of steps, which consists of the control variables to be solved. Composition; Objective function of the optimization problem Expressed as a formula:

[0030] in, Indicates a time index. Indicates the future The predicted feedback current value at time [time]. Indicates control quantity exist The change in time compared to the previous time. Represents the linear rectified function. Indicates the future Subjective feeling rating based on the prediction of time. This indicates the upper limit of comfort, corresponding to a feeling score of 7.5; , and These represent the tracking error weight, the control change weight, and the comfort penalty weight, respectively. A quadratic programming solver is used to solve the above optimization problem with the objective function as the goal. During the solution process, the absolute current safety boundary and the output current change rate limit are used as hard constraints, and the predicted comfort level is not exceeded as a soft constraint. After the solution is obtained, the first element of the optimal control sequence is taken as the output of each control cycle.

[0031] In step S40, based on the final control current value output in step S30, the treatment electrode of the TCM directional transdermal drug delivery device is driven to output corresponding physical energy to adjust the electric field force and osmotic pressure, thereby achieving precise control over the intensity of TCM drug delivery, simultaneously monitoring the hardware safety parameters of the device, and triggering a forced protection action when the parameters exceed the limit.

[0032] In this embodiment, the final control current value output in step S30, i.e., the digital control quantity, is converted into actual electrical energy applied to the patient's body surface through the power drive module inside the TCM-oriented transdermal drug delivery device, including: The final control current value output in step S30 It is a digital control signal, which is first converted into an analog voltage reference signal by a digital-to-analog converter within the module. ; Secondly, the voltage-controlled current source circuit within the module is... For input, output and The actual therapeutic current applied to the patient is linearly proportional to the applied current, and its value is equal to... This ensures that the output current remains stable when the skin impedance changes during treatment, thereby maintaining a constant electric field force and achieving precise drug delivery. Finally, the output therapeutic current is passed through a full-bridge power amplifier circuit, modulated into a therapeutic waveform with a fixed frequency and duty cycle, and then output through a filter circuit to ensure that the energy finally applied to the therapeutic electrodes of the TCM directional transdermal drug delivery device is a safe and pure therapeutic waveform.

[0033] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device, characterized in that, The method includes: Step S10: Based on the received individual patient information and treatment information, apply a pre-trained gradient boosting decision tree model to generate and output initial intensity parameters, setting a safe and personalized initial state for the TCM-targeted transdermal drug delivery device. Step S20: During the operation of the TCM-targeted transdermal drug delivery therapy device, the patient's objective physiological signals and subjective feeling signals are collected simultaneously in real time. The subjective feeling signal is the patient's quantitative feeling score of the treatment intensity obtained in real time through the human-computer interaction interface of the therapy device. Step S30: Based on the collected objective physiological signals and subjective feelings of the patient, a state vector is constructed and input to an adaptive regulator using a two-layer decision architecture. Through an outer deep learning agent and an inner constrained model predictive controller, the final control current value at the current moment is calculated and output to adjust the osmotic pressure, thereby achieving closed-loop control of the intensity of traditional Chinese medicine delivery. The outer layer of the adaptive regulator uses a deep reinforcement learning agent to make decisions and provide the adjustment direction of the target current. The agent adopts an actor-critic structure, using the actor network as the policy function, taking the current state vector as input, and outputting the suggested control action, i.e., the adjustment amount of the target current value. The agent is trained by maximizing the reward function, which is a weighted sum function based on the instantaneous perception score, the tracking error between the feedback current and the target current, and the rate of change of the current. When the real-time skin impedance in the patient's objective physiological signal is stable, but the change amplitude of the quantitative perception score exceeds the set threshold within a set time, the reliability decay mechanism is activated to reduce the weight of the subjective perception score in the state vector and give the objective physiological signal a higher weight. The inner layer of the adaptive regulator is a constrained model predictive controller, which receives control actions suggested by the agent, uses them as reference targets, and solves a constrained optimization problem within a set prediction window. The controller uses a long short-term memory network to predict the feedback current value and subjective feeling score of future moments based on the historical state sequence. The controller solves an optimization problem in each control cycle. The objective function of the optimization problem is to minimize the deviation between the feedback current and the reference target current in the predicted time domain and the change of the control quantity. The reference target current is determined by the current target intensity benchmark and the suggested control action. The absolute current safety boundary and the output current change rate limit are used as hard constraints, and the predicted comfort level is not exceeded as a soft constraint. After solving the problem, the optimal control sequence of the output current of the therapeutic device is obtained. The first element of the sequence is taken as the final control current value output in each control cycle. In step S40, based on the final control current value output in step S30, the treatment electrode of the TCM directional transdermal drug delivery device is driven to output corresponding physical energy. The safety monitoring thread runs synchronously to monitor the hardware safety parameters of the device in real time and triggers a forced protection action when the parameters exceed the limit.

2. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 1, characterized in that, In step S10, the patient's individual information includes at least age, TCM constitution type, treatment site, and skin condition of the treatment site; The treatment information includes at least the conductivity range of the Chinese herbal medicine solution used and the preset treatment duration; The initial intensity parameters include at least the initial intensity of the TCM-directed transdermal drug delivery system, the target intensity range, and the intensity ramp-up rate.

3. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 1, characterized in that, In step S20, after the TCM-targeted transdermal drug delivery device is activated, it simultaneously acquires and quantifies multimodal signals, including: Objective physiological signals: real-time skin impedance of the patient, actual feedback current through the treatment area, and calculation of the rate of change of impedance over time; Subjective perception signal: The patient obtains a quantitative perception score of the treatment intensity through a slider operated in real time. This score is positively correlated with the intensity of the patient's perceived warmth and tingling sensations. The score range is [0,10], and the rate of change of the perception score over time is calculated.

4. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 3, characterized in that, The patient's real-time skin impedance was acquired using a four-electrode method. One pair of electrodes on the treatment device was used to apply a detection AC excitation signal, and the other pair of electrodes was used to measure the voltage drop generated therefrom. The complex impedance of the skin tissue was obtained by calculating the voltage-to-current ratio and phase difference of the AC excitation signal, and its magnitude was the skin impedance. A sampling resistor is connected in series in the output circuit of the treatment current. By measuring the voltage drop across the resistor, the actual feedback current flowing through the patient's treatment area can be calculated in real time using Ohm's law.

5. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 1, characterized in that, In step S30, at each control moment, the real-time acquired objective physiological signals and subjective feeling signals of the patient are time-synchronized and filtered, and combined into a state vector. This vector consists of real-time feedback current, real-time feeling score, instantaneous skin impedance, impedance change rate, feeling score change rate, treatment time, and target intensity benchmark of Chinese medicine delivery.

6. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 1, characterized in that, In step S30, the inner constrained model predictive controller solves the rolling optimization problem in each control cycle to obtain the future... The control sequence of steps, which consists of the control variables to be solved. Composition; Objective function of the optimization problem Expressed as a formula: in, Indicates a time index. Indicates the future The feedback current value predicted by the long short-term memory network is applied at all times. Indicates control quantity exist The change in time compared to the previous time. Represents the linear rectified function. Indicates the future Subjective feeling scores predicted by applying long short-term memory networks at all times. Indicates the upper limit of comfort. , and These represent the tracking error weight, control change weight, and comfort penalty weight, respectively.

7. The method for controlling the intensity of traditional Chinese medicine delivery in a targeted transdermal drug delivery device as described in claim 1, characterized in that, In step S40, the final control current value output in step S30 is converted into actual electrical energy applied to the patient's body surface through the power drive module inside the TCM-oriented transdermal drug delivery device, including: The final control current value output in step S30 is a digital control signal, which is first converted into an analog voltage reference signal by the digital-to-analog converter within the module. ; Secondly, the voltage-controlled current source circuit within the module is... For input, output and The actual therapeutic current applied to the patient in a linearly proportional relationship ensures that the output current remains stable when the skin impedance changes during treatment, thereby maintaining a constant electric field force and achieving precise drug delivery. Finally, the output therapeutic current is passed through a full-bridge power amplifier circuit, modulated into a therapeutic waveform with a fixed frequency and duty cycle, and then output through a filter circuit to ensure that the energy finally applied to the therapeutic electrodes of the TCM directional transdermal drug delivery device is a safe and pure therapeutic waveform.

Images