Dynamic control optimization method based on digital traditional Chinese medicine acupuncture

Abstract

The invention relates to the technical field of digital traditional Chinese medicine acupuncture and moxibustion, and discloses a dynamic control optimization method based on digital traditional Chinese medicine acupuncture and moxibustion. The method comprises the following steps: training to obtain an acupuncture response prediction model according to acupuncture point characteristics and meridian response parameters of traditional Chinese medicine acupuncture; predicting responses under different acupuncture stimulation parameters by using the model, and screening an optimal stimulation parameter; simulating an acupuncture stimulation process based on the optimal parameters, analyzing propagation and response behaviors of stimulation in channels and collaterals, and obtaining simulation data; and finally, according to treatment target optimization analog data, obtaining a digital traditional Chinese medicine acupuncture dynamic control scheme. The method gets rid of dependence of traditional acupuncture on experience of doctors, stimulation parameter selection is more scientific, the dynamic process of stimulation in meridians and collaterals can be visually presented, dynamic adjustment of acupuncture stimulation is achieved, the method adapts to individual differences of patients, the stability and pertinence of the treatment effect are improved, and acupuncture standardized popularization and subject modernization development are assisted.

Description

Technical Field

[0001] This invention relates to the field of digital TCM acupuncture technology, specifically to a dynamic control optimization method based on digital TCM acupuncture. Background Technology

[0002] Acupuncture, as an important treatment method in traditional Chinese medicine, has been passed down for thousands of years and widely applied in clinical practice due to its unique role in regulating bodily functions and alleviating various symptoms. In traditional acupuncture treatment, physicians mainly rely on their clinical experience and tactile sense to determine acupoint selection, needle depth, needle retention time, and stimulation parameters such as twisting and lifting. The entire procedure is highly subjective and experience-dependent. Because different physicians have different experience levels and operating habits, even for patients with the same symptoms and constitution, the acupuncture stimulation plans used may be completely different. This makes it difficult to maintain consistent acupuncture treatment effects and poses a significant obstacle to the standardization and promotion of acupuncture techniques.

[0003] With the development of modern medical technology, although some studies have attempted to combine modern technology with acupuncture, using some simple instruments and equipment to assist in monitoring physiological indicators during the acupuncture process, such as skin resistance and electromyography signals, most of these technologies remain at the data acquisition level. They cannot conduct in-depth analysis and effective utilization of the acquired data, let alone achieve dynamic control and optimization of the acupuncture stimulation process.

[0004] In clinical treatment, due to significant individual differences in patients' physical condition, disease severity, and sensitivity to acupuncture, fixed acupuncture stimulation parameters often fail to achieve ideal therapeutic effects. For example, for patients with weak constitutions, excessively strong acupuncture stimulation may cause discomfort such as dizziness and fatigue; while for patients with severe conditions or low sensitivity to acupuncture, insufficient stimulation may not achieve the desired therapeutic effect. Furthermore, the transmission and response of acupuncture stimulation in the meridians is a complex dynamic process that traditional methods cannot directly observe and analyze. Physicians can only make judgments and adjustments based on patient feedback and their own experience, which often involves lag and uncertainty, affecting treatment efficiency and potentially delaying treatment due to inappropriate parameters.

[0005] In the fields of acupuncture teaching and research, the lack of effective dynamic control and simulation methods makes it difficult for students to intuitively understand the relationship between acupuncture stimulation and meridian response, and researchers also struggle to accurately conduct research on the mechanisms of acupuncture action, thus hindering the modernization of traditional Chinese medicine acupuncture. Therefore, how to overcome the limitations of traditional acupuncture and leverage digital technology to achieve accurate prediction, dynamic control, and optimization of the acupuncture stimulation process has become a crucial issue urgently needing to be addressed in the field of traditional Chinese medicine acupuncture. Summary of the Invention

[0006] The purpose of this invention is to provide a dynamic control optimization method for digital TCM acupuncture to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a dynamic control optimization method for digital TCM acupuncture, the method comprising: Based on the characteristics of acupoints and meridian response parameters in traditional Chinese medicine acupuncture, an acupuncture response prediction model was trained and obtained. The trained acupuncture response prediction model is used to predict acupuncture responses under different acupuncture stimulation parameters, and the optimal stimulation parameters are selected based on the prediction results. Based on the optimal stimulation parameters, the acupuncture stimulation process is simulated to analyze the propagation and response behavior of acupuncture stimulation in the meridians and obtain simulation data. The simulation data is optimized according to the treatment goals to obtain a dynamic control scheme for digital TCM acupuncture.

[0008] Preferably, the step of training an acupuncture response prediction model based on the acupoint characteristics and meridian response parameters of traditional Chinese medicine acupuncture includes: Collect acupoint feature data and meridian response parameter data. The acupoint feature data includes acupoint type, location, depth, and acupoint conductivity. The meridian response parameter data includes response intensity and response time. Based on the acupoint feature data and meridian response parameter data, a response feature correlation analysis is performed to construct a training dataset; The acupuncture response prediction model is obtained by training the model to converge using the training dataset.

[0009] Preferably, a training dataset is constructed by performing response feature correlation analysis based on the acupoint feature data and meridian response parameter data, including: Based on the acupoint feature data, the response characteristics of acupuncture stimulation and acupoint configuration are analyzed to determine the stimulation-acupoint configuration response energy. Based on the fusion analysis of the meridian response parameter data and the stimulation-acupoint configuration response energy, the synergistic response coefficient of the meridian response parameters to the stimulation-acupoint configuration response energy is determined. A training dataset is constructed based on the synergistic response coefficients of the stimulation-acupoint configuration response energy and the meridian response parameters, wherein the data label in the training dataset is the response energy.

[0010] Preferably, the response characteristics of acupuncture stimulation and acupoint configuration are analyzed based on the acupoint feature data to determine the stimulation-acupoint configuration response energy, including: Based on the acupoint feature data, the response probability of different acupuncture stimuli between different acupoint configurations was analyzed. The measurement response energy of different acupuncture stimuli under different acupoint configurations was determined experimentally. The response probability and response energy are aligned according to acupuncture stimulation and acupoint configuration to determine the stimulation-acupoint configuration response energy.

[0011] Preferably, the analysis is performed by fusing the meridian response parameter data with the stimulation-acupoint configuration response energy, including: Based on the meridian response parameter data, comparative acupoint feature data is constructed, which includes acupoint record data under different meridian response parameters for the same acupoint configuration under the same acupuncture stimulation. Based on the aforementioned acupoint feature data, a response energy experiment was conducted to determine the response energy difference between the acupoint feature data. The synergistic response coefficient is calculated based on the difference in response energy and the corresponding stimulation-acupoint configuration response energy without meridian response parameters.

[0012] Preferably, the acupuncture response prediction model is obtained by training and converging the model using the training dataset, including: The training dataset is divided according to the segmentation ratio to construct the training set and the test set; A machine learning algorithm is selected to train the acupuncture response prediction model. The acupuncture response prediction model is trained and its hyperparameters are adjusted using a training set. The trained model is validated and optimized using a test set until the convergence target is reached, and the acupuncture response prediction model is determined.

[0013] Preferably, the optimal stimulus parameters are selected based on the prediction results, including: The response of acupuncture stimulation to various acupoint configurations can be used to predict results; With the goal of maximizing the response of the acupuncture stimulation, the optimal stimulation parameters are obtained by searching for stimulation parameters based on the response energy prediction results.

[0014] Preferably, the simulation data is optimized according to the treatment goal, including: Obtain simulated data of acupuncture stimulation, including response location, response energy, and propagation migration path; Using the maximization of treatment goals as the evaluation value, parameter schemes for acupuncture stimulation are searched based on the simulation data to optimize the propagation response pathway.

[0015] Preferably, the optimized propagation response pathway includes: Based on the simulated data of the acupuncture stimulation, a multi-stimulus characteristic quantitative analysis was performed to obtain a multi-stimulus response quantitative list. Based on the multi-stimulus response quantification list, an external field compensation gradient analysis is performed. According to the compensation gradient, multiple levels are divided to construct a multi-level acupoint compensation network. The multi-level acupoint compensation network includes multiple levels of captured acupoint parameters. Obtain dynamic external field coordinated control parameters; Based on the dynamic external field collaborative control parameters, the parameters of the multi-level acupoint compensation network are adjusted layer by layer to obtain the external field collaborative control strategy. The evaluation value is adjusted according to the aforementioned external field coordinated control strategy; Based on the adjustment objectives, parameter schemes for acupuncture stimulation are searched to optimize the propagation response pathway.

[0016] Preferably, the treatment goals include health status parameter goals and treatment threshold goals.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This method trains an acupuncture response prediction model based on the characteristics of acupoints and meridian response parameters in Traditional Chinese Medicine (TCM) acupuncture, freeing the selection of acupuncture stimulation parameters from the subjective experience of physicians. In traditional acupuncture, physicians determine parameters based on experience, which is easily influenced by individual differences in experience. However, with a well-trained prediction model, the acupuncture response under different stimulation parameters can be accurately predicted, thus selecting the optimal stimulation parameters. This makes the determination of acupuncture stimulation parameters more scientific and objective. Regardless of the physician's experience level, treatment can be conducted based on the optimal parameters selected by the model, reducing fluctuations in treatment effects caused by insufficient physician experience or differences in operating habits. This leads to more stable acupuncture treatment effects and creates favorable conditions for the standardized promotion of acupuncture techniques.

[0018] Simulating the acupuncture stimulation process based on optimal stimulation parameters allows for a direct analysis of the propagation and response behavior of acupuncture stimulation in the meridians, yielding simulated data. Traditional acupuncture cannot directly observe the dynamic propagation process of stimulation in the meridians; physicians can only make indirect judgments based on patient feedback, which has significant limitations. However, through simulation, details such as the conduction path, conduction speed, and response intensity at different points in different meridians after the stimulation enters the acupoint can be clearly presented. This allows physicians to comprehensively and deeply understand the intrinsic connection between acupuncture stimulation and meridian response. Based on this simulation data, physicians can make clear predictions about the treatment process before actual treatment, avoiding the situation in traditional treatment where parameters are blindly adjusted due to a lack of understanding of the stimulation propagation patterns, thus improving the targeted nature of treatment.

[0019] Optimizing simulation data based on treatment goals yields a dynamic control scheme, enabling dynamic adjustments to the acupuncture stimulation process. In clinical practice, a patient's condition and physical state change as treatment progresses, and traditional fixed acupuncture protocols struggle to adapt to these dynamic changes. This method's dynamic control scheme can flexibly optimize acupuncture stimulation parameters based on the patient's real-time treatment response and adjustments to treatment goals. For example, it can appropriately reduce the stimulation intensity as the patient's condition improves, or adjust the needle depth and retention time according to changes in the condition. This dynamic adjustment ensures that acupuncture treatment always aligns with the patient's actual condition, better meeting the individual needs of different patients. Whether the patient is weak and requires gentle stimulation, or has a more severe condition requiring stronger stimulation, a suitable treatment plan can be obtained, further enhancing the effectiveness of treatment.

[0020] In the field of acupuncture teaching, this method's simulation process can visualize the abstract meridian response and stimulus propagation process. Students can intuitively understand the relationship between acupuncture stimulation and meridian response by observing simulation data and dynamic control processes, reducing learning difficulty and improving teaching effectiveness. In the field of scientific research, the accurate prediction model, detailed simulation data, and dynamic control scheme provided by this method offer strong technical support for researchers to study the mechanism of acupuncture action and explore optimal treatment plans for different diseases. This helps promote the modernization of traditional Chinese medicine acupuncture and facilitates the application of acupuncture techniques in a wider range of fields. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the working principle of the dynamic control optimization method for digital TCM acupuncture described in this invention. Figure 2 A sub-flowchart for training an acupuncture response prediction model; Figure 3 A sub-flowchart constructed for the training dataset. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1This invention provides a dynamic control optimization method for digital TCM acupuncture. The method involves transforming TCM acupuncture experience into a quantifiable and predictable digital model, and achieving precise treatment through dynamic control. The system collects acupoint feature data and meridian response parameters through a high-precision sensor network and a digital human body model. This multi-source data is cleaned, standardized, and stored in a central database. The core component is constructing an acupuncture response prediction model. A hybrid architecture of machine learning and deep learning is used to train the model based on historical and experimental data. The training process uses feature association analysis to uncover the nonlinear relationship between stimulus parameters and responses, with response energy as the core output indicator. The model can accurately predict the biological response of specific stimulus parameters under individualized conditions. In the optimization stage, the prediction model is used to perform parallel simulations of stimulus parameter combinations. Guided by the treatment goal, intelligent algorithms are used to search for the optimal parameters. These parameters drive a biophysical model-based digital twin system to simulate the propagation process of stimuli in the meridians, outputting dynamic data such as energy distribution and propagation paths. Closed-loop optimization of the simulation data is performed based on real-time treatment goals. Parameters are dynamically adjusted through reinforcement learning to optimize the response pathway, outputting personalized control schemes to drive intelligent devices, achieving a leap from traditional experience to digital precision treatment. The solution forms a closed-loop technology that combines data-driven and model-driven approaches and maps the physical and digital worlds, fully supporting the dynamic control and optimization of digital TCM acupuncture.

[0024] Example 1: See Figure 2Based on the characteristics of acupoints and meridian response parameters in Traditional Chinese Medicine (TCM), an acupuncture response prediction model was trained and integrated with the definition of treatment goals. The entire process began with the data collection phase. Acupoint characteristic data included acupoint category, location, depth, and electrical conductivity. This information was acquired using clinical measurement tools such as high-precision acupoint detectors and conductivity sensors. Standardized protocols were followed during the measurement process to reduce human error. Supplementary data was extracted from historical TCM databases to ensure coverage of diverse acupoint configurations. Meridian response parameter data included response intensity and response time. Response intensity was recorded using a bioelectric signal acquisition system to capture voltage changes after acupuncture stimulation. Response time was precisely measured using a timer module from the application of stimulation to the initial response. Data collection was repeated in multiple experimental environments to enhance reliability. All data was stored in a structured database for subsequent processing. The response feature correlation analysis is based on collected acupoint feature data and meridian response parameter data. Statistical methods such as Pearson correlation analysis are used to assess the linear relationship between acupoint conductivity and response intensity, or machine learning techniques such as decision tree algorithms are used to identify the influence pattern of acupoint depth on response time. During the analysis, a feature correlation matrix is ​​generated, where rows represent acupoint features and columns represent response parameters. Highly correlated feature pairs are selected by calculating covariance and significance level. When constructing the training dataset, these correlation results are integrated into a feature vector set. Each vector contains acupoint category code, location coordinates, depth value, conductivity reading, and corresponding response intensity and response time. The dataset undergoes missing value processing and normalization scaling to ensure input consistency. The label field is set to the derived response energy index to support model learning.

[0025] Model training is performed using a constructed training dataset. First, the data is randomly split into training and test sets at a ratio of 70%-30%. The training set is used to fit the model parameters, and the test set is used to evaluate the generalization ability. A machine learning algorithm, such as a multilayer perceptron neural network, is selected for training. The network structure includes an input layer that processes acupoint features and response parameters, a hidden layer that uses the ReLU activation function to extract nonlinear features, and an output layer that generates the predicted response energy. During the training process, the loss function, such as mean squared error, is optimized using the gradient descent algorithm. The weight parameters are iteratively updated until the training error stabilizes. Hyperparameter tuning involves learning rate scheduling and batch size optimization, which is finely tuned using a grid search method. In the test set validation phase, the prediction accuracy and error index are calculated to confirm model convergence. Finally, the model can accept new acupuncture stimulation parameters and acupoint inputs and outputs predicted response values.

[0026] The treatment objectives encompass health status parameter targets and treatment threshold targets. Health status parameter targets are set based on individual patient physiological indicators such as heart rate and blood pressure. These target values ​​are extracted from electronic health record systems or real-time monitoring devices and dynamically adjusted according to clinical assessments to reflect treatment progress. Treatment threshold targets are defined as the minimum response intensity or maximum response time boundaries that acupuncture responses must meet. These thresholds are set with reference to TCM treatment guidelines and safety protocols. For example, minimum response intensity ensures sufficient stimulation to trigger meridian activity, and maximum response time prevents excessively delayed responses. These objectives are integrated into the model optimization framework as constraints to guide parameter search, ensuring that the treatment plan is both effective and tailored to the patient's specific needs. The entire implementation process emphasizes data quality and model robustness. Iterative improvements and validation ensure that the acupuncture response prediction model provides reliable output in practical applications. Furthermore, the clear definition of treatment objectives enhances the personalization and safety dimensions of the method.

[0027] Example 2: See Figure 3 The process of determining the stimulus-acupoint configuration response energy and synergistic response coefficient in response characteristic correlation analysis begins with the analysis of the response characteristics of acupuncture stimuli and acupoint configurations. Based on collected acupoint feature data, including acupoint category, location, depth, and conductivity, the analysis uses statistical modeling methods to evaluate the interaction between different acupuncture stimulus parameters such as frequency and intensity and acupoint configurations. When calculating the response probability, a conditional probability model is used, where the stimulus parameters are input variables and the acupoint configuration is a conditional variable. The model is trained using historical data to output the probability that a specific stimulus will elicit a response under a specific acupoint configuration. The analysis results generate a probability distribution matrix, where rows represent acupuncture stimulus types, columns represent acupoint configuration combinations, and matrix elements represent the response probability values ​​of the corresponding combinations. In the experimental determination phase, the response energy was obtained through acupuncture experiments under controlled conditions. The experiment was set up using standardized acupuncture equipment to apply preset stimulation parameters, such as electrical pulses or mechanical vibrations of different frequencies, while monitoring changes in acupoint configuration, such as depth adjustment or positional shift. The response energy was measured using a bioenergy harvesting system to record the energy release value after stimulation, in microjoules or equivalent voltage values. The measurement process was repeated multiple times to eliminate random errors. Each experiment recorded stimulation parameters, acupoint configuration data, and the measured response energy value. All data were stored in the experimental database for subsequent processing.

[0028] The alignment operation integrates response probability with measured response energy. A data matching algorithm, based on stimulus parameters and acupoint configuration as key fields, associates entries in the probability matrix with response energy records in the experimental database. During alignment, interpolation methods are used to handle missing data points; for example, when experimental data for a stimulus-acupoint configuration combination is lacking, a weighted average of neighboring data points is used for estimation. This ultimately forms a stimulus-acupoint configuration response energy dataset containing complete fields for stimulus parameters, acupoint configuration, response probability, and measured response energy, serving as the basis for fusion analysis. In the fusion analysis phase, combining meridian response parameter data with stimulus-acupoint configuration response energy, meridian response parameters such as response intensity and response time are first extracted. A comparison dataset corresponding to acupoint feature data is constructed. This comparison dataset focuses on records of the same acupuncture stimulus and the same acupoint configuration under different meridian response parameters, such as varying response intensity values ​​at a fixed stimulus frequency and acupoint depth. These comparison records are obtained by querying historical databases or through real-time measurements.

[0029] The response energy experiments were performed on the comparison dataset. The same stimulus and acupoint configuration conditions were reproduced in a controlled environment, but the meridian response parameters were changed. Energy sensors were used to measure the response energy value for each experiment. When calculating the response energy difference, the response energy with meridian response parameters was compared to the baseline response energy without parameters. The difference was calculated using absolute value or percentage difference methods, and the results were stored as a difference vector associated with the corresponding meridian response parameters. The synergistic response coefficient was calculated based on the response energy difference and the baseline stimulus-acupoint configuration response energy. A ratio formula was used to divide the difference by the baseline response energy to obtain a normalized coefficient. Positive or negative coefficient values ​​indicate enhancement or inhibition effects. The calculation process iteratively covered all comparison data entries, generating a coefficient mapping table linking meridian response parameter types with synergistic effect strengths. This coefficient was ultimately used to adjust the response energy values ​​in the training dataset, enhancing the accuracy of model predictions. The entire implementation emphasized data consistency and computational reliability. Multi-stage analysis ensured that the stimulus-acupoint configuration response energy and synergistic response coefficient reflected the real physical interactions, providing high-quality input for subsequent model training.

[0030] Taking a specific acupuncture treatment scenario as an example, assuming that palpitations are treated by targeting the Neiguan (PC6) acupoint on the Pericardium Meridian of Hand-Jueyin, the response characteristics of acupuncture stimulation and acupoint configuration are analyzed based on the acupoint feature data. The acupoint features of Neiguan include a category of Luo-connecting point, a location 2 cun above the wrist crease, a depth of 0.5-1 cun, and an electrical conductivity of 4.1 μS. During the analysis, different acupuncture stimulation parameters such as electrical pulse frequency (1Hz, 2Hz, 3Hz) and needling techniques (twisting, lifting and thrusting) are selected to evaluate the response probability between these stimuli and acupoint configuration. A conditional probability model is used to calculate the probability of a specific stimulus triggering a response under a specific acupoint configuration. For example, the response probability of a frequency of 2Hz combined with the twisting technique at a depth of 1 cun is 0.78. The measured response energy of different acupuncture stimuli under different acupoint configurations was obtained through experiments. The experiment used an electroacupuncture device with preset parameters, and an energy detector recorded the response energy values. For example, three measurements were performed under the conditions of a frequency of 2Hz, a twisting technique, and a depth of 1 inch, yielding response energy readings of 18.3μJ, 19.1μJ, and 17.9μJ, respectively. The average value of 18.4μJ was taken as the measured response energy. The response probability and measured response energy were aligned according to the acupuncture stimulation and acupoint configuration. The calculated response probability and the experimentally measured response energy value were correlated using a data matching algorithm to form a stimulus-acupoint configuration response energy dataset. For example, the response probability of 0.78 for a frequency of 2Hz, a twisting technique, and a depth of 1 inch was paired and stored with the measured response energy of 18.4μJ.

[0031] A fusion analysis was performed based on meridian response parameter data and stimulus-acupoint configuration response energy. Meridian response parameters included response intensity (1.2V) and response time (0.3s). When constructing comparative acupoint feature data, records under the same stimulus parameters and acupoint configuration were extracted, such as a fixed frequency of 2Hz, a twisting technique, and a depth of 1 inch, collecting acupoint data under different response intensities (0.8V, 1.2V, 1.6V). Response energy experiments were conducted to obtain the response energy difference between the comparative acupoint feature data. The same stimulus and acupoint configuration conditions were reproduced in a controlled environment, but the response intensity was changed, and the response energy value at each response intensity was measured. For example, the response energy measured at a response intensity of 1.2V was 20.5μJ, which was compared with the baseline response energy of 18.4μJ under no meridian response parameter intervention. The difference ΔE = 20.5 - 18.4 = 2.1μJ was calculated. Based on the response energy difference and the corresponding stimulation-acupoint configuration response energy without meridian response parameters, the synergistic response coefficient is calculated using the formula κ=ΔE / E_b=2.1 / 18.4≈0.114. The coefficient value of 0.114 indicates that the response intensity parameter has an enhancement effect of 11.4% on the response energy. This coefficient will be used to adjust the response energy values ​​in the training dataset to improve the accuracy of the prediction model.

[0032] Example 3: Fusion analysis to obtain synergistic response coefficients and model training. This process begins with constructing comparative acupoint feature data. Based on meridian response parameter data, records under the same acupuncture stimulation and acupoint configuration conditions are extracted. For example, when the electrical pulse frequency is fixed at 2Hz and the acupoint depth is 1.5cm, the acupoint conductivity and location information under different response intensity values ​​are collected. Data extraction is performed using a database query tool, and SQL statements are used to filter matching entries to ensure data consistency. The extracted records include stimulation parameters, acupoint configuration fields, and corresponding meridian response parameter values, forming a structured dataset for subsequent analysis. Response energy experiments are performed on each record in the comparative dataset. The exact same stimulation and acupoint configuration conditions are reproduced in the experimental environment, but the meridian response parameters are actively adjusted, such as gradually increasing the response intensity from 0.5V to 3.0V. After each adjustment, a high-precision energy sensor is used to measure the response energy value of the acupoint. The measurement process is repeated three times and the average value is taken to reduce random errors. Experimental records include response energy readings, meridian response parameter values, and environmental variables such as temperature and humidity. Finally, the response energy dataset for each comparative entry is obtained.

[0033] The response energy difference was calculated by comparing the response energy with the baseline response energy determined in a comparative experiment. The baseline response energy refers to the stimulus-acupoint configurational response energy value without intervention from meridian response parameters, which was obtained from the stimulus-acupoint configurational response energy dataset generated in Example 2. The difference was calculated using the following formula:

[0034] in: Indicates the difference in response energy. This indicates the response energy measured in the current experiment. The baseline response energy is represented by a positive value indicating an enhancing effect and a negative value indicating an inhibiting effect. All differential results are stored in a data table and associated with the corresponding meridian response parameters.

[0035] The co-response coefficient is calculated based on the response energy difference and the baseline response energy. The difference is normalized using a ratio method. The calculation formula is as follows:

[0036] in: Represents the collaborative response coefficient. Indicates the difference in response energy. This represents the baseline response energy, with coefficient values ​​ranging from -1 to 1. The positive or negative sign indicates the direction of synergy. The calculation process iterates through all comparison data entries, generating a coefficient mapping table for subsequent training dataset adjustments. Model training is performed using a training dataset containing synergistic response coefficients. First, the data is randomly split according to a 70%-30% ratio. The training set is used for model fitting, and the test set is used for performance evaluation. The splitting process employs stratified sampling to maintain data distribution consistency, ensuring that both the training and test sets cover all acupoint categories and stimulation types.

[0037] The machine learning algorithm uses a support vector machine regression model for training. Input features include acupoint category encoding, location coordinates, depth values, conductivity readings, stimulation parameters, and co-response coefficients. The output target is the predicted response energy. The training process uses a radial basis function kernel to handle nonlinear relationships, and gradient descent optimizes the loss function to minimize prediction error. Hyperparameter tuning includes fine-tuning the kernel parameters C and gamma. A grid search method is used to test parameter combinations within a preset range to select the optimal configuration. In the model validation phase, the trained model is evaluated using a test set. Root mean square error and coefficient of determination are calculated to assess prediction accuracy. The validation process is iterative; if performance does not meet expectations, hyperparameters are readjusted or the amount of training data is increased until the model converges and achieves stable predictive ability. Ultimately, the model can integrate the effects of meridian response parameters and accurately output acupuncture response prediction values. The entire implementation process emphasizes data quality and computational repeatability, ensuring the reliability of the co-response coefficients and the prediction model through systematic experiments and algorithm optimization.

[0038] Taking a specific acupuncture treatment scenario as an example, assuming that the Hegu acupoint (LI4) of the Large Intestine Meridian of Hand Yangming is used to treat headache symptoms, firstly, acupoint feature data is constructed for comparison. Based on the meridian response parameter data, records under the same acupuncture stimulation and the same acupoint configuration are extracted. A fixed electrical pulse frequency of 2Hz and acupuncture depth of 0.8 cun are selected as the baseline stimulation parameters. Acupoint feature data under different response time parameters (0.2s, 0.4s, 0.6s) are collected, including information such as the conductivity of Hegu acupoint of 3.8μS and its location in the depression between the first and second metacarpal bones. Data extraction is completed through database query. Conditional filtering is used to obtain completely matching records to form a comparison dataset. When conducting the response energy experiment, the exact same stimulation parameters and acupoint configuration were reproduced in a controlled environment, but the system response time parameter was adjusted. After each adjustment, the response energy value of Hegu acupoint was recorded using a bioenergy meter. For example, when the response time was 0.4s, three measurements were taken to obtain response energy readings of 21.3μJ, 22.1μJ, and 20.9μJ. The average value of 21.4μJ was taken as the measured value. At the same time, the baseline response energy of 19.6μJ without meridian response parameter intervention was obtained from the stimulation-acupoint configuration response energy dataset of Example 2.

[0039] The difference method is used to calculate the response energy difference. The difference value is obtained by subtracting the baseline response energy from the current measured response energy. For example, the difference under the condition of a response time of 0.4 s. All differential calculation results are stored in association with the corresponding response time parameters. The co-response coefficient is calculated using a ratio method, dividing the response energy difference by the baseline response energy to obtain the normalized coefficient, for example... This coefficient indicates that the response time parameter has a 9.2% enhancement effect on the response energy. The calculation process covers all response time parameter conditions and forms a complete coefficient mapping table.

[0040] The model training phase utilizes a training dataset containing co-response coefficients. First, the data is randomly split according to a 70%-30% ratio, with the training set containing 105 records and the test set containing 45 records. The distribution of acupoint categories and stimulation types is balanced during splitting. Support Vector Machine (SVM) regression is selected for model training. Input features include acupoint category encoding, location coordinates, depth values, conductivity readings, stimulation frequency parameters, and co-response coefficients. The output target is the predicted response energy value. The training process uses a radial basis function (RBF) kernel to handle the nonlinear relationships between features, and iterative optimization minimizes the prediction error. Hyperparameter tuning includes optimizing the penalty parameter C and the kernel coefficient gamma. A grid search method is used to test different combinations within the ranges of C (0.1, 10) and gamma (0.01, 1), selecting the parameter configuration with the smallest root mean square error. Model validation uses a test set to evaluate performance, calculating the correlation coefficient and mean absolute error between the predicted and actual response energies. Once the validation results meet the expected standards, the final model is determined. This model can accurately predict acupuncture response energies under different stimulation parameters and meridian response conditions, providing data support for treatment optimization.

[0041] Example 4: Screening Optimal Stimulation Parameters and Optimizing Simulation Data. This process is illustrated using a treatment case at the Zusanli acupoint (ST36) as a specific example. First, the predicted response energy of acupuncture stimulation under various acupoint configurations is obtained. Different combinations of stimulation parameters are input into the acupuncture response prediction model, such as varying the electrical pulse frequency (1Hz, 2Hz, 3Hz) and intensity (0.5mA, 1.0mA, 1.5mA), while fixing acupoint configuration parameters such as depth 1.5cm and conductivity 3.2μS. The model outputs the predicted response energy value for each parameter combination, and the prediction results form a parameter-response energy mapping table for subsequent analysis. With the goal of maximizing the response of acupuncture stimulation, stimulation parameters are searched based on the predicted response energy results. The search uses a genetic algorithm to traverse the parameter space. The algorithm is set with a population size of 50 and 100 iterations. The fitness function directly uses the predicted response energy value. The search process evaluates the response energy of each parameter combination, and finally selects the parameter combination with the highest response energy as the optimal stimulation parameters. For example, the predicted response energy reaches its peak when the frequency is 2Hz and the intensity is 1.0mA.

[0042] The acupuncture stimulation process is simulated based on optimal stimulation parameters. A meridian biophysical model is constructed using a computer simulation platform. The model input includes stimulation parameters, acupoint configuration, and individual meridian topology data. The simulation output includes response location, response energy distribution, and propagation migration path. The response location is marked with energy release points using three-dimensional coordinates, the response energy is visualized using color gradients, and the propagation migration path uses vector lines to represent the energy flow direction. The simulation data is stored as a structured dataset for optimization. Maximizing the treatment goal is used as the evaluation value. Based on the simulation data, parameter scheme search is performed for acupuncture stimulation. The treatment goal is set as a health state parameter target, such as reducing the pain index to below 3. The evaluation value function integrates the matching degree between the simulated response energy distribution and the target value. The parameter scheme search uses a particle swarm optimization algorithm to adjust the stimulation parameters. The algorithm is set with 30 particles and 50 iterations. Each iteration updates the parameter combination and re-simulates, evaluating the fit between the new simulation data and the treatment goal.

[0043] When optimizing the propagation response pathway, the energy propagation patterns in the simulation data are analyzed to identify the main response paths and potential blockage points. For example, it was found that the energy attenuates relatively quickly in the segment from Zusanli to Shangjuxu acupoints. By adjusting the stimulation parameters, such as increasing the frequency to 2.5Hz or fine-tuning the intensity to 1.2mA, the path improvement is observed during resimulation. The optimization process is repeated until the propagation path is continuous and the response energy is evenly distributed. The entire implementation process relies on a data-driven approach. See Table 1, which shows the predicted response energy of some stimulation parameters for the Zusanli acupoint.

[0044] Table 1: Predicted response energy of Zusanli acupoint under different stimulation parameters

[0045] During the optimization process, the achievement of treatment goals is continuously monitored. By comparing the simulated data after each iteration with the target health status parameters, the search direction is dynamically adjusted, and the dynamic control scheme is finally output, integrating the optimal stimulation parameters and the optimized propagation pathway configuration.

[0046] Example 5: Taking the case of treating chronic gastritis with the Zusanli acupoint (ST36) as a specific example, the quantitative analysis of multiple stimulation features is based on the response points of multiple acupoints in the simulation data. The simulation data comes from the simulation output of Example 4 and includes the response location, response energy and propagation path information of acupoints such as Zusanli, Shangjuxu, and Xiajuxu. The analysis process extracts the stimulation feature parameters of each acupoint, such as the peak response energy, response duration and energy diffusion radius. The feature parameters are converted into numerical scores using the quantitative index calculation method. For example, the peak response energy is standardized to a score of 0-100, and the response duration is quantified in seconds. Finally, a quantitative list of multiple stimulation responses is formed. The rows of the list represent the acupoint names and the columns represent the values ​​of various quantitative indicators.

[0047] The external field compensation gradient analysis is performed based on a multi-stimulus response quantification list. The analysis focuses on the degree of difference in the quantification indicators of each acupoint, and calculates the response energy gradient value between adjacent acupoints. For example, the response energy difference between Zusanli and Shangjuxu is 5.2 μJ. The gradient calculation adopts the difference method, and acupoints are divided into three levels according to the gradient magnitude: high gradient level (difference > 5 μJ), medium gradient level (difference 2-5 μJ), and low gradient level (difference < 2 μJ). When constructing a multi-level acupoint compensation network, capture acupoint parameters are defined for each level, including priority stimulation order, compensation intensity coefficient, and response threshold. For example, high gradient level acupoints are stimulated first, and the compensation intensity coefficient is set to 1.5. Dynamic external field collaborative regulation parameters are obtained from real-time monitoring data, including the patient's current physiological state such as gastric peristalsis frequency and skin conductance changes. These parameters are continuously collected by sensors and updated once per second. The parameter processing uses moving average filtering to eliminate noise, forming a time series dataset for regulation decision-making.

[0048] Based on dynamic external field collaborative regulation parameters, the parameters of a multi-level acupoint compensation network are adjusted layer by layer. The regulation process starts from the high gradient level, adjusting the compensation intensity coefficient according to the real-time gastric peristalsis frequency. For example, the coefficient is increased by 0.1 when the frequency increases. Subsequently, the coefficients of the medium and low levels are adjusted sequentially. The regulation strategy generates a set of control rules, which adopt if-then logic, such as "if the gastric peristalsis frequency > 3 times / minute, then increase the stimulation intensity of the high-level acupoints". The evaluation value is adjusted according to the external field collaborative regulation strategy. The initial evaluation value is based on the treatment target, such as the pain index. The adjustment incorporates the real-time regulation effect, such as including the predicted value of the compensated response energy in the evaluation value calculation, and updating the objective function using a weighted summation method. Based on the adjustment target, a parameter scheme search is performed for acupuncture stimulation. The search adopts an adaptive particle swarm optimization algorithm, and the parameter range expands around the initial optimal stimulation parameters. The algorithm evaluates the performance of each parameter combination under the updated evaluation value. The search outputs optimized stimulation parameters, such as the frequency fine-tuned to 2.2Hz and the intensity adjusted to 1.1mA. Finally, the re-simulation of the propagation response pathway shows that the energy distribution is more uniform and the path continuity is improved. The entire implementation process embodies dynamic adaptability, optimizing treatment precision through real-time data feedback and hierarchical control.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic control optimization method for digital TCM acupuncture, characterized in that, include: Based on the characteristics of acupoints and meridian response parameters in traditional Chinese medicine acupuncture, an acupuncture response prediction model was trained and obtained. The trained acupuncture response prediction model is used to predict acupuncture responses under different acupuncture stimulation parameters, and the optimal stimulation parameters are selected based on the prediction results. Based on the optimal stimulation parameters, the acupuncture stimulation process is simulated to analyze the propagation and response behavior of acupuncture stimulation in the meridians and obtain simulation data. The simulation data is optimized according to the treatment goals to obtain a dynamic control scheme for digital TCM acupuncture.

2. The method for dynamic control optimization based on digital TCM acupuncture according to claim 1, characterized in that, The acupuncture response prediction model, trained based on the acupoint characteristics and meridian response parameters of traditional Chinese medicine acupuncture, includes: Collect acupoint feature data and meridian response parameter data. The acupoint feature data includes acupoint type, location, depth, and acupoint conductivity. The meridian response parameter data includes response intensity and response time. Based on the acupoint feature data and meridian response parameter data, a response feature correlation analysis is performed to construct a training dataset; The acupuncture response prediction model is obtained by training the model to converge using the training dataset.

3. The method for dynamic control optimization based on digital TCM acupuncture according to claim 2, characterized in that, Based on the acupoint feature data and meridian response parameter data, a response feature correlation analysis is performed to construct a training dataset, including: Based on the acupoint feature data, the response characteristics of acupuncture stimulation and acupoint configuration are analyzed to determine the stimulation-acupoint configuration response energy. Based on the fusion analysis of the meridian response parameter data and the stimulation-acupoint configuration response energy, the synergistic response coefficient of the meridian response parameters to the stimulation-acupoint configuration response energy is determined. A training dataset is constructed based on the synergistic response coefficients of the stimulation-acupoint configuration response energy and the meridian response parameters, wherein the data label in the training dataset is the response energy.

4. The method for dynamic control optimization based on digital TCM acupuncture according to claim 3, characterized in that, Based on the acupoint feature data, the response characteristics of acupuncture stimulation and acupoint configuration are analyzed to determine the stimulation-acupoint configuration response energy, including: Based on the acupoint feature data, the response probability of different acupuncture stimuli between different acupoint configurations was analyzed. The measurement response energy of different acupuncture stimuli under different acupoint configurations was determined experimentally. The response probability and response energy are aligned according to acupuncture stimulation and acupoint configuration to determine the stimulation-acupoint configuration response energy.

5. The method for dynamic control optimization based on digital TCM acupuncture according to claim 4, characterized in that, The analysis is performed by fusing the meridian response parameter data with the stimulation-acupoint configuration response energy, including: Based on the meridian response parameter data, comparative acupoint feature data is constructed, which includes acupoint record data under different meridian response parameters for the same acupoint configuration under the same acupuncture stimulation. Based on the aforementioned acupoint feature data, a response energy experiment was conducted to determine the response energy difference between the acupoint feature data. The synergistic response coefficient is calculated based on the difference in response energy and the corresponding stimulation-acupoint configuration response energy without meridian response parameters.

6. The method for dynamic control optimization based on digital TCM acupuncture according to claim 2, characterized in that, The acupuncture response prediction model is obtained by training and converging the model using the training dataset, including: The training dataset is divided according to the segmentation ratio to construct the training set and the test set; A machine learning algorithm is selected to train the acupuncture response prediction model. The acupuncture response prediction model is trained and its hyperparameters are adjusted using a training set. The trained model is validated and optimized using a test set until the convergence target is reached, and the acupuncture response prediction model is determined.

7. The method for dynamic control optimization based on digital TCM acupuncture according to claim 1, characterized in that, The optimal stimulus parameters are selected based on the prediction results, including: The response of acupuncture stimulation to various acupoint configurations can be used to predict results; With the goal of maximizing the response of the acupuncture stimulation, the optimal stimulation parameters are obtained by searching for stimulation parameters based on the response energy prediction results.

8. The method for dynamic control optimization based on digital TCM acupuncture according to claim 1, characterized in that, The simulation data is optimized according to the treatment goals, including: Obtain simulated data of acupuncture stimulation, including response location, response energy, and propagation migration path; Using the maximization of treatment goals as the evaluation value, parameter schemes for acupuncture stimulation are searched based on the simulation data to optimize the propagation response pathway.

9. The method for dynamic control optimization based on digital TCM acupuncture according to claim 8, characterized in that, Optimize the propagation response pathway, including: Based on the simulated data of the acupuncture stimulation, a multi-stimulus characteristic quantitative analysis was performed to obtain a multi-stimulus response quantitative list. Based on the multi-stimulus response quantification list, an external field compensation gradient analysis is performed. According to the compensation gradient, multiple levels are divided to construct a multi-level acupoint compensation network. The multi-level acupoint compensation network includes multiple levels of captured acupoint parameters. Obtain dynamic external field coordinated control parameters; Based on the dynamic external field collaborative control parameters, the parameters of the multi-level acupoint compensation network are adjusted layer by layer to obtain the external field collaborative control strategy. The evaluation value is adjusted according to the aforementioned external field coordinated control strategy; Based on the adjustment objectives, parameter schemes for acupuncture stimulation are searched to optimize the propagation response pathway.

10. The method for dynamic control optimization based on digital TCM acupuncture according to claim 1, characterized in that, The treatment goals include health status parameter goals and treatment threshold goals.

Images