A method and system for diabetes intervention based on meridian and acupoint therapy using complex network resilience

Abstract

This invention relates to the field of integrated traditional Chinese and Western medicine and intelligent computing, and particularly to a method and system for intervening in diabetes treatment based on the resilience of complex networks and meridian acupoints. The method includes: constructing a meridian network topology model with acupoints as nodes and meridian connections as edges, and extracting topological features; establishing a node activity dynamics model, and dynamically simulating the functional state and interactions of acupoints through refined dynamic equations; collecting multi-time-point and multi-acupoint observation data from patients, fusing it with the topological features, and using a resilience assessment model to determine whether the meridian system has the ability to stabilize and recover blood sugar; for non-resilient states, calling an intervention library and generating personalized acupuncture intervention plans through intervention simulation and optimization algorithms. This invention achieves quantitative modeling of traditional Chinese medicine meridian theory and integrates traditional Chinese and Western medicine, solving the problems of difficulty in quantifying traditional Chinese medicine theory and lack of personalization and standardization in intervention plans in traditional diagnosis and treatment, thereby improving the efficiency and effectiveness of diabetes diagnosis and treatment.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of integrated traditional Chinese and Western medicine and intelligent computing, and in particular to a method and system for intervening in diabetes based on meridian acupoints and complex network resilience. Background Technology

[0002] Currently, there are two main technological systems in the research and treatment of diabetes:

[0003] 1. Modern medical system: Focuses on molecular biology and endocrinology, studying linear causal relationships such as insulin secretion and glucose metabolism. Its interventions (such as drugs and insulin) have clear targets, but often only address the symptoms, lacking a systemic view of regulating the body's overall condition, and may be accompanied by side effects.

[0004] 2. Traditional Chinese Medicine System: Based on the holistic concept and syndrome differentiation of "spleen deficiency and stomach heat" and "liver stagnation and kidney deficiency," it systematically regulates the body through acupuncture, herbal medicine, and other methods. However, the theories of this system (such as meridians and qi and blood) are highly abstract, and its mechanisms of action are difficult to quantify and explain using modern scientific language. The efficacy of this system heavily relies on the personal experience of physicians, making it difficult to standardize and promote on a large scale.

[0005] Currently, there is a lack of a bridging technology that can deeply integrate the holistic systemic view of Traditional Chinese Medicine (TCM) with modern computational science. It is impossible to transform the abstract system of "meridians and acupoints" into a computable and analyzable digital model, thus hindering the scientific revelation of the systemic pathological mechanisms of diabetes and the provision of precise, quantitative predictions for interventions such as acupuncture. The existing technology mainly suffers from the following shortcomings:

[0006] 1. Inability to quantify system dynamics: Existing methods cannot describe the dynamic interactions between acupoints and cannot transform "meridian conduction" into a computable and predictable model.

[0007] 2. Unclear Mechanism: There is a lack of quantitative explanations based on network dynamics and information transmission for key questions such as "why acupuncture at distal acupoints can treat visceral diseases".

[0008] 3. Low degree of personalization: Acupuncture treatment plans are mostly based on fixed prescriptions or physicians' subjective experience, lacking data-driven personalized recommendations based on the patient's individual network status.

[0009] 4. Efficiency and standardization issues: The process of developing TCM diagnosis and treatment plans is complex, time-consuming, and difficult to replicate and standardize, which limits its promotion and application. Summary of the Invention

[0010] This invention addresses the technical problems in existing diabetes diagnosis and treatment, namely the difficulty in quantifying traditional Chinese medicine meridian theory, the lack of a systematic regulatory perspective in modern medicine, and the inadequacy of personalized and standardized intervention programs. It proposes a meridian and acupoint intervention method and system based on the resilience of complex networks, which realizes the quantitative modeling of traditional Chinese medicine meridian theory and the integration of traditional Chinese and Western medicine, providing precise, personalized, and standardized acupuncture intervention programs for diabetes, thereby improving the efficiency and effectiveness of diagnosis and treatment.

[0011] To achieve the above objectives, the technical solution adopted is:

[0012] This invention provides a meridian acupoint diabetes intervention method based on complex network resilience, comprising the following steps:

[0013] Step 1: Construct a meridian network topology model: Abstract the TCM meridian and acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and extract network topology features through network coding technology;

[0014] Step 2: Establish a node activity dynamic model: Define quantitative indicators of acupoint functional state, construct refined dynamic equations to describe the evolution law of acupoint function and the dynamic interaction between acupoints, and realize the dynamic simulation of the functional state of the meridian system.

[0015] Step 3: Integrate multi-source data for system resilience analysis: Collect observation data of multiple acupoints at T time points over a period of time from the patient, integrate the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar.

[0016] Step 4: Generate personalized acupuncture intervention plan: Based on the system resilience analysis results, for system states that do not have the ability to stabilize blood sugar, the intervention plans in the preset intervention library are called. Through intervention simulation and optimization algorithms, the effects of different intervention plans on the meridian system are simulated, the optimal intervention plan is selected, and a personalized acupuncture intervention plan is generated.

[0017] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the specific process of constructing the meridian network topology model in step 1 is as follows: Based on classical Chinese medicine literature and expert knowledge, a weighted directed graph G(V,E,W) is constructed, where V is the set of nodes containing core acupoints and auxiliary acupoints, E is the set of edges representing meridian connections, and W is the set of weights representing connection strength; the meridian association relationships include exterior-interior meridian relationships, same meridian relationships, different meridian relationships, and adjacent relationships.

[0018] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the network encoding technology in step 1 is implemented by a graph neural network. The graph neural network is used to encode the constructed weighted directed graph G, learn and output the topological feature vector of each acupoint node; the graph neural network adopts a graph convolutional network that introduces an attention mechanism.

[0019] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the refined kinetic equation in step 2 is further defined as follows:

[0020]

[0021] in, This refers to the self-kinetic function that describes the physiological processes of acupoints. This represents the state variable of each acupoint node i. To describe the coupling dynamics function of the mutual influence between adjacent acupoints, To simulate the noise term of physiological fluctuations and environmental disturbances, It is a random input source for the noise term. For interventions that describe the effects of external treatment interventions, This represents the external intervention signal applied to acupoint node i.

[0022] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, further, the state variable of each acupoint node i... Defined as the normalized functional activity of acupoints, its value ranges from [-1, 1], and satisfies: This indicates that the acupoint is in a state of hyperactivity, corresponding to the excess syndrome and heat syndrome in Traditional Chinese Medicine. This indicates that the acupoint is in a state of functional inhibition, corresponding to deficiency syndrome and cold syndrome in traditional Chinese medicine; This indicates that the acupoint is in a state of functional balance.

[0023] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the data fusion processing in step 3 is further described as follows: multi-acupoint observation data and topological feature vectors obtained by graph neural network encoding are concatenated to form enhanced node features, which are then processed by an attention encoder to obtain the system state representation vector Z; the attention encoder calculates the attention score for each time step t and each node i. Implement weighted aggregation.

[0024] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the resilience assessment model in step 3 is a fully connected neural network classifier. The system state representation vector Z is fed into the classifier, and the output is a binary judgment of Resilient or Non-Resilient. Resilient means that the system can recover to a healthy state with stable blood sugar after being subjected to common disturbances, which corresponds to the "yin-yang balance" in traditional Chinese medicine. Non-Resilient means that the system has lost its self-recovery ability and is in a pathological stable state of hyperglycemia, which corresponds to the pathological state of diabetes.

[0025] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, further, the intervention simulation and optimization algorithm in step 4 specifically includes:

[0026] Intervention simulation: The predefined intervention library contains a variety of acupuncture programs with single or combined acupoints. For the system state of patients in the Non-Resilient state, perturbations are applied to the corresponding acupoint nodes in the refined dynamic equation to simulate the acupuncture sensation. The system evolution trajectory over a period of time is simulated by numerical integration of the system dynamic equation.

[0027] Effect evaluation and optimization: Calculate the degree of similarity between the system state and the resilience state after the simulation evolution is completed, and use it as the effect evaluation value of the intervention scheme; based on the effect evaluation value, select one or more optimal schemes from the intervention library as recommended schemes.

[0028] According to the meridian acupoint diabetes intervention method based on complex network resilience of the present invention, the acupuncture program further includes a list of recommended acupoints, a suggested acupuncture sequence, and an estimated clinical effectiveness score.

[0029] Furthermore, the present invention also provides a meridian acupoint diabetes intervention system based on complex network resilience, comprising:

[0030] The network topology construction module is used to abstract the TCM meridian and acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and to extract network topology features through network coding technology.

[0031] The dynamic simulation module is used to define quantitative indicators of acupoint functional status, construct refined dynamic equations to describe the evolution of acupoint functions and the dynamic interactions between acupoints, and realize the dynamic simulation of the functional status of the meridian system.

[0032] The resilience analysis module is used to collect multi-acupoint observation data of patients at T time points over a period of time, fuse the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar.

[0033] The results output module is used to, based on the system resilience analysis results, call the intervention plans in the preset intervention library for system states that do not have the ability to stabilize blood sugar, and simulate the effects of different intervention plans on the meridian system through intervention simulation and optimization algorithms, select the optimal intervention plan and generate a personalized acupuncture intervention plan.

[0034] The beneficial effects achieved by adopting the above technical solution are:

[0035] 1. Deep integration and upgrading of traditional Chinese and Western medicine: Deeply couple the abstract theories of meridians, qi and blood, deficiency and excess in traditional Chinese medicine with modern complex networks, artificial intelligence, and dynamic modeling technology to construct a calculable and analyzable digital twin model. This will build a core technology bridge for the scientific and quantitative expression of traditional Chinese medicine theories, and achieve the organic integration of traditional Chinese and Western medicine theories and methods rather than a simple superposition.

[0036] 2. More systematic analysis of pathological mechanisms: Breaking through the limitations of single biomarkers, from the perspective of network dynamics and system resilience, it quantitatively reveals the imbalance mechanism of the human meridian system in the state of diabetes, providing a brand-new system-level explanation for the pathological understanding of diabetes that transcends the traditional perspectives of Chinese and Western medicine.

[0037] 3. Precision and standardization of intervention plans: Through digital twin simulation and optimization algorithms, personalized acupuncture prescriptions based on the individual patient's meridian network status are generated, quantifying and clarifying the selection of acupoints, stimulation sequence and expected efficacy, greatly reducing reliance on the physician's personal experience, and achieving precision, standardization and replicability of acupuncture intervention.

[0038] 4. Improved diagnostic and treatment efficiency and practicality: The system automates the fusion of multi-source data, system resilience assessment, and intervention plan generation, significantly shortening the diagnostic and treatment decision-making cycle. It provides an efficient and easy-to-use intelligent tool for large-scale diabetes health management, clinical diagnosis and treatment, and the modernization of traditional Chinese medicine research, taking into account both clinical practicality and scientific research value. Attached Figure Description

[0039] 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. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.

[0040] Figure 1 This is a flowchart illustrating the meridian acupoint diabetes intervention method based on complex network resilience, according to an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the process for establishing a node activity dynamics model according to an embodiment of the present invention. Detailed Implementation

[0042] The exemplary solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art.

[0043] This invention discloses a meridian acupoint diabetes intervention method based on complex network resilience, such as... Figure 1 As shown, it includes the following steps:

[0044] Step S1: Construct a meridian network topology model: Abstract the TCM meridian acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and extract network topology features through network coding technology.

[0045] Based on classical Chinese medicine texts (such as the *Huangdi Neijing* and *Zhenjiu Jiayi Jing*) and expert knowledge, a weighted directed graph G(V,E,W) is constructed, with acupoints as nodes and meridian pathways and specific relationships (such as exterior-interior meridian relationship, same meridian relationship, separate meridian relationship, and adjacent relationship) as edges. Here, V is the set of nodes containing core acupoints (such as Zusanli ST36, Yishu EX-B5, and Sanyinjiao SP6) and auxiliary acupoints, E is the set of edges representing meridian connections, and W is the set of weights representing connection strength, which can be initialized with the association strength in classical texts.

[0046] Network coding techniques employ Graph Neural Networks (GNNs). These GNNs encode the constructed weighted directed graph G into a low-dimensional feature tensor, which encapsulates the network's topological properties (such as clustering coefficients and feature path lengths). Graph Neural Networks transform the complex topological relationships of a network into machine-understandable feature representations, specifically including:

[0047] (I) Multi-relationship graph neural network architecture

[0048] (1) A relational graph convolutional network (R-GCN) is used to specifically handle various types of connections in the meridian network. The core objective of this model is to update the feature representation of each acupoint in the meridian network. Through a mechanism called "message passing," each acupoint not only retains its own information but also intelligently and discriminatively aggregates information from neighboring acupoints with different types of connections. Ultimately, each acupoint obtains a completely new and richer feature vector, which deeply encodes its structural role and function in the entire meridian network.

[0049] (2) The meaning of the model input and input elements

[0050] First, we need to clarify a few basic concepts: ① Acupoint feature vector: using... It can be imagined as the "digital ID card" of acupoint i in the l-th layer of the neural network. This vector contains all the feature information of the acupoint in the current layer, such as initial attributes (meridian to which it belongs, specific effects, etc.) and abstract patterns learned from previous network layers.

[0051] ② Set of relation types: Represented by R. This is a list of relation types predefined according to Traditional Chinese Medicine theory, for example:

[0052] Same meridian relationship: Connecting two acupoints belonging to the same meridian (such as "Shousanli" and "Quchi" on the Large Intestine Meridian of Hand Yangming);

[0053] Exterior-interior relationship: Connecting acupoints on two meridians that have an exterior-interior relationship (such as acupoints on the Foot Taiyin Spleen Meridian and acupoints on the Foot Yangming Stomach Meridian).

[0054] Differential connection: Connecting acupoints on meridians that have a "differential connection" (such as "the liver and large intestine are connected").

[0055] Adjacent relationship: Connecting acupoints that are anatomically adjacent to each other.

[0056] ③ Neighbor Set: For a specific acupoint i and a relation r (e.g., "same meridian"), all acupoints j directly connected to acupoint i through relation r form a neighbor set, denoted as [i, j]. .

[0057] (3) Input element processing

[0058] The processing of input elements can be broken down into the following three key steps:

[0059] ① Aggregate information from various relational neighbors. For each defined relation type r (e.g., first process "same meridian", then process "internal and external", and so on), perform the following sub-steps: 1) Identify neighbors: Find the set of all direct neighboring acupoints j of acupoint i under the relation r. 2) Transform neighbor information: For the feature vector of each neighbor acupoint j A trainable weight matrix specifically prepared for relation r is used. Perform a linear transformation; this means the model will learn different information transformation rules for different relationships such as "same meridian" and "internal / external". For example, it may learn that the weights of "same meridian" connections are more important; 3) Summation and normalization: Summate the neighbor vectors of all the transformations obtained in step 2), and then divide by a normalization constant. This constant is usually the number of neighbors under relation r. This is done to prevent "pivotal acupoints" with a large number of neighbors from having excessively high feature values, thus ensuring the stability of the training process. This step allows acupoint i to collect information from its different social circles, such as those related to "same meridian" or "exterior / interior".

[0060] ② Add self-connect information: the original feature vector of acupoint i itself It will also use a dedicated self-connection weight matrix. This transformation is to ensure that acupoints can "remember themselves" during updates, preventing their own characteristics from being completely overwhelmed by information from their neighbors and preserving their individual characteristics.

[0061] ③ Merging and Activation: The sum of all aggregated relational neighbor information obtained in step ① is added to the transformed information obtained in step ②. This merged result is then passed to a non-linear activation function. (Such as the ReLU function), the addition operation combines external influences with its own characteristics. Nonlinear activation functions are key to the ability of neural networks to learn complex patterns in R-GCN. They introduce nonlinear transformations, enabling the model to capture the complex patterns in the interactions between acupoints that are not simple addition, subtraction, multiplication, or division.

[0062] (4) Final output

[0063] Based on the above calculations, the new feature vector of acupoint i in the (l+1)th layer can be obtained. When the model is constructed from multiple layers of such a structure, each acupoint ultimately acquires a high-order feature vector. This final feature vector is a highly refined "acupoint embedding." It not only represents the acupoint itself but also a low-dimensional representation that integrates its own characteristics, its multi-level neighbors (neighbors of neighbors, etc.), and various complex relationships within the meridian network. It reveals the acupoint's functional role and structural importance in the network more profoundly than the original acupoint attributes or simple connectivity relationships, providing a powerful data foundation for subsequent advanced tasks such as acupoint function analysis and treatment recommendations.

[0064] (ii) Attention-enhanced topology learning

[0065] Building upon R-GCN, an attention mechanism is introduced to dynamically weight the importance of each neighbor. The core logic is: after transforming the features of each neighbor node j of the central node i, a weighted sum is calculated based on the importance of each neighbor to i, and finally, an activation function is applied to obtain the new features of node i. This is the "finishing touch" of R-GCN. (R-GCN relation weights) The importance of neighbors is predefined and fixed, but even within the same type of relationship, different neighbors may have different importance. The attention mechanism allows the model to dynamically and adaptively learn the importance of each neighbor. Specifically, it includes the following key steps:

[0066] (1) Transformation: Use a shared weight matrix W to transform the features of the current acupoint i respectively. Characteristics of neighboring acupoints j Perform a transformation; concatenate the two transformed vectors to form a longer vector, which encodes the joint information of acupoint i and acupoint j.

[0067] (2) Calculate the correlation: Transpose a learnable attention vector a, perform a dot product operation with the concatenated vector, and obtain the original attention score after processing with the LeakyReLU activation function; then normalize it with softmax to obtain the normalized attention weight. This makes all attention weights The sum is 1 ( This represents the importance of neighbor j to the current acupoint i after considering all information; it is a value between 0 and 1.

[0068] (3) Determine the contribution ability of a node: use the calculated attention weights. The information of each neighbor is weighted, and important neighbors ( The larger the neighbor, the greater their contribution; the less important the neighbor, the greater their contribution. (Small) Their contribution is small.

[0069] The value of attention mechanisms:

[0070] Dynamic importance: It no longer assumes that all neighbors of a "same meridian" are equally important. For example, for "Zusanli", "Shangjuxu" on the same meridian may have a higher attention weight than "Xiajuxu" on the same meridian but further away.

[0071] Model interpretability: After training, these attention weights can be visualized. This allows us to clearly understand which connections the model prioritizes when assessing the state of a specific acupoint or the disease status of an entire network. This provides quantitative, data-driven evidence for traditional Chinese medicine theories (such as "meridian-based acupoint selection"), greatly enhancing the model's interpretability.

[0072] Step S2: Establish a node activity dynamics model: Define quantitative indicators of acupoint functional state, construct refined dynamic equations to describe the evolution of acupoint functions and the dynamic interactions between acupoints, and realize the dynamic simulation of the functional state of the meridian system. The process of establishing a node activity dynamics model is as follows: Figure 2 As shown.

[0073] This scheme transforms the concepts of "qi and blood" and "deficiency and excess" in traditional Chinese medicine into precise mathematical expressions, and establishes a "digital twin" model of the meridian system.

[0074] (1) Physiological definition of state variables

[0075] The state variable of each acupoint node i Defined as the normalized functional activity of acupoints, its value ranges from [-1, 1] (determined by the properties of the kinetic equation itself; values ​​exceeding the boundary will be automatically detected and corrected to ensure the stability of the function), and satisfies: This indicates that the acupoint is in a state of hyperactivity, corresponding to the excess syndrome and heat syndrome in Traditional Chinese Medicine. This indicates that the acupoint is in a state of functional inhibition, corresponding to deficiency syndrome and cold syndrome in traditional Chinese medicine. This indicates that the acupoint is in a state of functional balance.

[0076] (2) The refined dynamic equations are in the following specific form:

[0077]

[0078] in, This represents the rate of change in the state of acupoint i, and its significance is that it is the core output of the entire equation system; it represents the "functional activity" of acupoint i at a given instant. The rate and direction of change. A positive... A positive value indicates that the acupoint is being "activated" (e.g., changing from deficiency to excess), while a negative value indicates that it is being "inhibited" (e.g., changing from excess to deficiency). By integrating this rate of change, the future state of the acupoint can be predicted. It is the activity level of acupoint function, obtained by solving an equation. It is a function that changes with time t, describing the state of the acupoint at any given moment. This is the main basis for quantifying, visualizing, and analyzing acupoint function.

[0079] A. Self-dynamic function

[0080] Self-dynamic function The output is a scalar value representing the state of acupoint i within a unit of time, solely due to its internal physiological processes. The instantaneous rate of change reflects the acupoint's ability to maintain and regulate itself independently of external stimuli. A positive output value indicates positive self-activation of the acupoint, while a negative output value indicates self-inhibition or attenuation of the acupoint.

[0081] B. Coupled dynamic functions

[0082] The output of the coupled model is a scalar value, representing the state of the target acupoint i per unit time due to the influence of all adjacent acupoints j. The instantaneous rate of change, for example, a positive output value indicates that the acupoint is currently being activated. This will increase; a negative output value indicates that the acupoint is currently being suppressed. It will decrease.

[0083] C. Noise Item

[0084] The noise term output is a scalar representing the state of acupoints caused by physiological fluctuations and environmental disturbances per unit time. Unpredictable, instantaneous changes that occur. This is the random input source for the noise term. The noise term simulates the inherent physiological changes in living systems to prevent the model from becoming too idealized, while also measuring the model's stability and resilience under random disturbances. For example, an output of 0.05 indicates that a small random fluctuation causes acupoints to be briefly activated, while an output of -0.03 indicates that random disturbances cause acupoints to be momentarily inhibited.

[0085] D. Intervention Items

[0086] This represents the external intervention signal applied to acupoint node i; the intervention term output is a scalar value, representing the acupoint state caused by the external treatment intervention per unit time. The instantaneous rate of change (and the efficiency with which external treatment translates into changes in the state of acupoints). For example, the effect of acupuncture on the state of acupoints, enabling automated adjustment of tonifying deficiency and purging excess.

[0087] Step S3: Integrate multi-source data for system resilience analysis: Collect multi-acupoint observation data of patients at T time points over a period of time, integrate the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar.

[0088] (I) Data fusion and trajectory representation

[0089] (1) Input clinical time series data: Collect observation data of multiple acupoints at T time points within a period of time for the patient. (e.g., temperature, conductivity, etc.), forming an observation sequence. .

[0090] (2) Feature enhancement: This involves enhancing the observed data. Topological feature vectors obtained with R-GCN encoding The nodes are spliced ​​together to form enhanced node features. .

[0091] (3) Trajectory aggregation and attention mechanism: The enhanced temporal features are input into an attention encoder (such as a Transformer encoding layer + a fully connected layer. The Transformer encoding layer is used to update the features of each spatiotemporal location to include global information. An attention score is calculated for each updated feature using a trainable fully connected layer). This attention encoder will calculate an attention score for each time step t and each node i. This score represents the contribution of the acupoint at that moment in inferring the current diabetic state of the patient.

[0092] (4) Output: Use attention score As weights, the features of all nodes at all time points are summed in a weighted manner to obtain a fixed-length system state representation vector Z that condenses the spatiotemporal dynamics of the entire network.

[0093] (II) Resilience Inference

[0094] The system state representation vector Z is fed into a fully connected neural network classifier. The classifier outputs a binary judgment: Resilient(1) or Non-Resilient(0). Resilient means that the system (human body) is capable of recovering to a stable blood sugar state on its own after being subjected to common disturbances (such as dietary indiscretion or emotional fluctuations), corresponding to the "yin-yang balance" in traditional Chinese medicine. Non-Resilient means that the system has lost its self-recovery ability and is trapped in a pathological stable state (attractor) with high blood sugar, corresponding to the pathological state of diabetes.

[0095] The model's training data comes from long-term monitoring data of healthy individuals and a cohort of patients diagnosed with diabetes.

[0096] Step S4: Generate personalized acupuncture intervention plan: Based on the system resilience analysis results, for system states that do not have the ability to stabilize blood sugar, the intervention plans in the preset intervention library are called. Through intervention simulation and optimization algorithms, the effects of different intervention plans on the meridian system are simulated, the optimal intervention plan is selected, and a personalized acupuncture intervention plan is generated.

[0097] The system now acts as a "digital twin" testing platform:

[0098] (1) Initialization: The system state representation vector Z of the patient currently diagnosed as Non-Resilient is loaded into the model as the initial state; then a decoder network (such as a fully connected network) is used to map Z back to the state variables of each acupoint. (Functional activity) is used as the initial activity level for each acupoint.

[0099] (2) Intervention Simulation: The system has a predefined intervention library, which contains a variety of acupuncture schemes for single or combined acupoints. For each scheme in the library, the system applies a predefined perturbation to the corresponding acupoint node in the refined dynamic equation. (Simulating acupuncture and obtaining qi). Then, the system evolution trajectory within a future time period Δt is simulated by numerically integrating the system dynamics equations.

[0100] (3) Effect evaluation and optimization

[0101] After the simulation ends, the system state after the simulation is extracted again, and a classifier is used to determine whether it has turned into a resilient state, or to calculate its distance from the resilient state (toughness recovery).

[0102] The system will traverse or use optimization algorithms (such as genetic algorithms) to search the intervention library and find the Top-K optimal acupoint intervention schemes that can most effectively and quickly push the system state back to the Resilient region.

[0103] Output: The final output is a personalized acupuncture prescription to the physician, which includes: a list of recommended acupoints, a suggested acupuncture order (derived by simulating the effects of different orders), and an estimated clinical effectiveness score.

[0104] Corresponding to the above method, embodiments of the present invention also disclose a meridian acupoint diabetes intervention system based on complex network resilience, comprising:

[0105] The network topology construction module is used to abstract the TCM meridian and acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and to extract network topology features through network coding technology.

[0106] The dynamic simulation module is used to define quantitative indicators of acupoint functional status, construct refined dynamic equations to describe the evolution of acupoint functions and the dynamic interactions between acupoints, and realize the dynamic simulation of the functional status of the meridian system.

[0107] The resilience analysis module is used to collect multi-acupoint observation data of patients at T time points over a period of time, fuse the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar.

[0108] The results output module is used to, based on the system resilience analysis results, call the intervention plans in the preset intervention library for system states that do not have the ability to stabilize blood sugar, and simulate the effects of different intervention plans on the meridian system through intervention simulation and optimization algorithms, select the optimal intervention plan and generate a personalized acupuncture intervention plan.

[0109] Alternative technical means in this invention: Graph neural network type: The GNN can be a Graph Convolutional Network (GCN), a Graph Attention Network (GAT), or other variants. Machine learning model: Trajectory aggregator, dimensionality reduction network, and classifier. In addition to Attention, Transformer, and fully connected networks, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), or other suitable temporal models and classification algorithms can also be used. Data source: Observations. In addition to infrared thermography and skin conductance, it can also include multimodal data that can reflect the state of the system, such as fMRI brain functional connectivity and body fluid metabolomics.

[0110] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A meridian-acupoint diabetes intervention method based on complex network resilience, characterized in that, Includes the following steps: Step 1: Construct a meridian network topology model: Abstract the TCM meridian and acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and extract network topology features through network coding technology; Step 2: Establish a node activity dynamic model: Define quantitative indicators of acupoint functional state, construct refined dynamic equations to describe the evolution law of acupoint function and the dynamic interaction between acupoints, and realize the dynamic simulation of the functional state of the meridian system. Step 3: Integrate multi-source data for system resilience analysis: Collect observation data of multiple acupoints at T time points over a period of time from the patient, integrate the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar. Step 4: Generate personalized acupuncture intervention plan: Based on the system resilience analysis results, for system states that do not have the ability to stabilize blood sugar, the intervention plans in the preset intervention library are called. Through intervention simulation and optimization algorithms, the effects of different intervention plans on the meridian system are simulated, the optimal intervention plan is selected, and a personalized acupuncture intervention plan is generated.

2. The meridian acupoint diabetes intervention method based on complex network resilience according to claim 1, characterized in that, The specific process of constructing the meridian network topology model in step 1 is as follows: Based on classical Chinese medicine literature and expert knowledge, a weighted directed graph G(V,E,W) is constructed, where V is the set of nodes containing core acupoints and auxiliary acupoints, E is the set of edges representing meridian connections, and W is the set of weights representing connection strength; the meridian relationships include exterior-interior meridian relationships, same meridian relationships, separate meridian relationships, and adjacent relationships.

3. The meridian acupoint diabetes intervention method based on complex network resilience according to claim 2, characterized in that, The network encoding technique in step 1 is implemented using a graph neural network. The graph neural network is used to encode the constructed weighted directed graph G, learn and output the topological feature vector of each acupoint node; the graph neural network adopts a graph convolutional network that introduces an attention mechanism.

4. The meridian acupoint diabetes intervention method based on complex network resilience according to claim 1, characterized in that, The refined dynamic equations in step 2 are in the following specific form: in, This refers to the self-kinetic function that describes the physiological processes of acupoints. This represents the state variable of each acupoint node i. To describe the coupling dynamics function of the mutual influence between adjacent acupoints, To simulate the noise term of physiological fluctuations and environmental disturbances, It is a random input source for the noise term. For interventions that describe the effects of external treatment interventions, This represents the external intervention signal applied to acupoint node i.

5. The meridian acupoint diabetes intervention method based on complex network resilience according to claim 4, characterized in that, The state variables of each acupoint node i Defined as the normalized functional activity of acupoints, its value ranges from [-1, 1] and satisfies: This indicates that the acupoint is in a state of hyperactivity, corresponding to the excess syndrome and heat syndrome in Traditional Chinese Medicine. This indicates that the acupoint is in a state of functional inhibition, corresponding to deficiency syndrome and cold syndrome in traditional Chinese medicine; This indicates that the acupoint is in a state of functional balance.

6. The method for intervening in diabetes based on meridian acupoints using complex network resilience according to claim 1, characterized in that, The data fusion process in step 3 specifically involves: concatenating the multi-acupoint observation data with the topological feature vector obtained from graph neural network encoding to form enhanced node features, which are then processed by an attention encoder to obtain the system state representation vector Z; the attention encoder calculates the attention score for each node i at each time step t. Implement weighted aggregation.

7. The meridian acupoint diabetes intervention method based on complex network resilience according to claim 6, characterized in that, The resilience assessment model in step 3 is a fully connected neural network classifier. The system state representation vector Z is fed into the classifier, which outputs a binary judgment of Resilient or Non-Resilient. Resilient means that the system can recover to a healthy state with stable blood sugar after being subjected to common disturbances, which corresponds to the "yin-yang balance" in traditional Chinese medicine. Non-Resilient means that the system has lost its self-recovery ability and is in a pathological stable state of hyperglycemia, which corresponds to the pathological state of diabetes.

8. The method for intervening in diabetes based on meridian acupoints according to claim 7, characterized in that, Step 4, the intervention simulation and optimization algorithm specifically includes: Intervention simulation: The predefined intervention library contains a variety of acupuncture programs with single or combined acupoints. For the system state of patients in the Non-Resilient state, perturbations are applied to the corresponding acupoint nodes in the refined dynamic equation to simulate the acupuncture sensation. The system evolution trajectory over a period of time is simulated by numerical integration of the system dynamic equation. Effect evaluation and optimization: Calculate the degree of similarity between the system state and the resilience state after the simulation evolution is completed, and use it as the effect evaluation value of the intervention scheme; based on the effect evaluation value, select one or more optimal schemes from the intervention library as recommended schemes.

9. The method for intervening in diabetes based on meridian acupoints according to claim 8, characterized in that, The acupuncture protocol includes a list of recommended acupoints, a suggested order of needling, and an estimated clinical effectiveness score.

10. A meridian-acupoint diabetes intervention system based on complex network resilience, characterized in that, include: The network topology construction module is used to abstract the TCM meridian and acupoint system into a network structure with acupoints as nodes and meridian relationships as edges, and to extract network topology features through network coding technology. The dynamic simulation module is used to define quantitative indicators of acupoint functional status, construct refined dynamic equations to describe the evolution of acupoint functions and the dynamic interactions between acupoints, and realize the dynamic simulation of the functional status of the meridian system. The resilience analysis module is used to collect multi-acupoint observation data of patients at T time points over a period of time, fuse the data with network topology feature information, and judge the health resilience status of the human meridian system through a preset resilience assessment model to distinguish whether the system has the ability to stabilize and recover blood sugar. The results output module is used to, based on the system resilience analysis results, call the intervention plans in the preset intervention library for system states that do not have the ability to stabilize blood sugar, and simulate the effects of different intervention plans on the meridian system through intervention simulation and optimization algorithms, select the optimal intervention plan and generate a personalized acupuncture intervention plan.

Images