A type 2 prediabetes diagnosis and treatment method based on traditional Chinese medicine state science, electronic equipment and computer readable storage medium

Abstract

This invention relates to the intersection of traditional Chinese medicine modernization and intelligent healthcare, specifically falling under the category of pre-chronic disease screening and health management technology. More particularly, it relates to a method for diagnosing and treating pre-type 2 diabetes based on traditional Chinese medicine state theory, along with electronic devices and computer-readable storage media, constructing a complete, human-machine collaborative closed-loop diagnostic and treatment process. This method achieves objective data collection through portable devices using the four diagnostic methods of traditional Chinese medicine, utilizes a multi-algorithm fusion model to accurately identify the "potential pathological state" of pre-T2DM, and intelligently generates personalized intervention prescriptions based on this. Finally, it quantifies the effects and dynamically optimizes the treatment plan through data re-collection and model re-evaluation. This solves the problems of strong subjectivity in traditional Chinese medicine diagnosis and treatment, disconnection from modern technology, and lack of quantitative feedback and continuous adjustment after intervention. It achieves full-cycle, intelligent, and personalized health management for pre-T2DM, from identification and intervention to follow-up management.

Description

Technical Field

[0001] This invention relates to the intersection of the modernization of traditional Chinese medicine and intelligent medical care, specifically to the field of early screening and health management technology for chronic diseases, and particularly to a method for the diagnosis and treatment of prediabetes type 2 based on traditional Chinese medicine state theory, an electronic device, and a computer-readable storage medium. Background Technology

[0002] Prediabetes of type 2 is a critical window for the development of diabetes and is reversible. Western medicine relies on biochemical indicators such as blood glucose for diagnosis, but lacks an effective early intervention system. The traditional Chinese medicine theory of "treating disease before it manifests" provides an important approach, but traditional diagnosis relies on physician experience, which suffers from strong subjectivity, inconsistent standards, and difficulty in large-scale implementation.

[0003] In existing technologies, such as patent document CN201710955624.2, a TCM preventive medicine auxiliary diagnosis and treatment system is disclosed. This system collects information through a TCM four diagnostic instruments and combines it with a constitution identification scale for health assessment. However, this system does not focus on specific pre-disease states (such as the "pre-disease state" of T2DM). Its identification is based on a general constitution classification, and the intervention plan requires remote participation from doctors to generate. It fails to achieve a fully automated and intelligent closed loop from state identification to personalized intervention.

[0004] Patent document CN202210200206.3 discloses a TCM classification system for diabetes based on Western medicine examination indicators, which uses a logistic regression model to correlate Western medicine clinical indicators with TCM syndrome types. This technology mainly targets TCM classification for diagnosed diabetes, rather than the specific "early stage," and does not involve the objective collection and integration of TCM four diagnostic methods, nor does it integrate portable hardware devices for home management.

[0005] Patent document CN202311258179.6 discloses a Traditional Chinese Medicine (TCM) health status identification system, which analyzes state elements through preprocessing steps such as data cleaning, labeling and encoding, and feature scaling, as well as a complex multi-label learning algorithm. This system focuses on general health status identification algorithms and does not construct a dedicated model for the pathophysiological characteristics of the early stages of T2DM or the TCM theory of "potential pathological state," nor does it have deep integration with physical data collection equipment and closed-loop intervention management.

[0006] In summary, the existing technologies have the following shortcomings: 1) There is a lack of a dedicated and quantitative TCM state identification system for the early "pathological state" of T2DM; 2) The diagnosis (or identification) is disconnected from the generation of personalized intervention plans, and no automated closed loop is formed; 3) Professional TCM equipment is not portable, while portable equipment has limited functions and cannot meet the intelligent diagnosis and treatment needs of the four diagnostic methods in the home setting; 4) The system has poor compatibility and serious data silos. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a method for the diagnosis and treatment of prediabetes type 2 based on traditional Chinese medicine state theory, an electronic device, and a computer-readable storage medium to solve the technical problems involved in the above background art.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A diagnostic and treatment method for prediabetes type 2 based on traditional Chinese medicine state theory, executed collaboratively by a portable intelligent diagnostic and treatment device and a data processing platform, includes the following steps: S1. The portable intelligent diagnostic and treatment device collects the user's objective data of the four diagnostic methods of traditional Chinese medicine and basic health data; the objective data of the four diagnostic methods of traditional Chinese medicine includes at least tongue and facial images collected by a high-definition camera, pulse signals collected by a high-precision pulse sensor, and auscultation voice information collected by an audio unit. S2. The data collected in step S1 is transmitted to the data processing platform. After preprocessing, it is input into a multi-algorithm fusion TCM state identification model. The TCM state identification model is based on a predefined set of T2DM pre-TCM "pathogenic state" and integrates at least a multi-state Markov model, a Logistic regression model and a Cox proportional hazards regression model for analysis, and outputs the user's current TCM state classification and corresponding risk score. S3. Based on the TCM status classification and corresponding risk score output in step S2, call the pre-built TCM intervention knowledge base for T2DM and generate a personalized health intervention prescription through an intelligent matching algorithm. S4. After the user executes the health intervention prescription generated in step S3, repeat steps S1 to S2 to obtain the user's follow-up data and updated TCM status; compare and analyze the status before and after the update with the risk score to evaluate the intervention effect, and dynamically adjust or regenerate the intervention prescription based on the intervention results to complete the diagnosis and treatment operation.

[0009] Furthermore, the predefined set of T2DM pre-TCM "pathological state" states in step S2 includes: spleen deficiency with dampness, qi and yin deficiency, liver stagnation and spleen deficiency, phlegm and blood stasis, and balance with imbalance.

[0010] Furthermore, the multi-state Markov model in step S2 is used to quantify the dynamic transition risk between different TCM states, and its state transition strength matrix... Defined as: in This indicates that at time t, the state changes from... Transition to state The instantaneous intensity, K is the total number of states, and satisfies .

[0011] Furthermore, the fusion strategy of the multi-algorithm fusion model in step S2 is as follows: Let the output state probability of the Markov model be... The output probability of the Logistic regression model is The Cox model outputs a risk score of The fusion weights are dynamically determined through intelligent optimization algorithms. Final status score Calculated using the following formula: The intelligent optimization algorithm is either particle swarm optimization or ant colony optimization.

[0012] Furthermore, the intelligent matching algorithm in step S3 is a hybrid algorithm that integrates a retrieval-enhanced sequence-to-sequence learning model with a hypergraph neural network.

[0013] Furthermore, the T2DM early intervention knowledge base called in step S3 integrates the content from national standards, clinical guidelines and experience data of famous veteran TCM doctors, and is structured to include at least two types of intervention measures from TCM conditioning, dietary therapy, acupoint massage and lifestyle guidance.

[0014] Furthermore, the basic health data collected in step S1 is entered through a mobile terminal application or obtained from a connected third-party health monitoring device, including blood glucose levels, BMI, lifestyle information, and symptom self-report.

[0015] Furthermore, in step S4, the effect is evaluated by calculating the probability changes of the user's TCM state before and after the intervention, and the state transition intensity matrix. Changes in relevant elements or risk scores This can be achieved through at least one of the methods used to reduce the magnitude of the reduction.

[0016] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps performed by the data processing platform as described in the "A Method for Diagnosing and Treating Prediabetes Type 2 Based on Traditional Chinese Medicine State Theory".

[0017] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps performed by a data processing platform as described in a method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory.

[0018] The beneficial effects of this invention are: This invention provides a method, electronic device, and computer-readable storage medium for the diagnosis and treatment of prediabetes type 2 (T2DM) based on Traditional Chinese Medicine (TCM) state theory, constructing a complete, human-machine collaborative closed-loop diagnosis and treatment process. The method achieves objective data collection through portable devices using the four diagnostic methods of TCM, accurately identifies the "pathological state" of pre-T2DM using a multi-algorithm fusion model, intelligently generates personalized intervention prescriptions based on this, and finally quantifies the effects and dynamically optimizes the treatment plan through data re-collection and model re-evaluation. This solves the problems of strong subjectivity in traditional TCM diagnosis and treatment, disconnection from modern technology, and lack of quantitative feedback and continuous adjustment after intervention, realizing full-cycle, intelligent, and personalized health management for pre-T2DM, from identification and intervention to follow-up management. Attached Figure Description

[0019] Figure 1 The diagram shows a flowchart of the steps in a method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to the present invention. Figure 2 The diagram shown is a block diagram of an electronic device according to the present invention; Explanation of icon numbers: 1-Memory; 2-Processor. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments: like Figure 1 As shown, the present invention provides a method for the diagnosis and treatment of prediabetes type 2 based on traditional Chinese medicine state theory, which is executed collaboratively by a portable intelligent diagnostic and treatment device and a data processing platform, and includes the following steps: S1. The portable intelligent diagnostic and treatment device collects the user's objective data of the four diagnostic methods of traditional Chinese medicine and basic health data; the objective data of the four diagnostic methods of traditional Chinese medicine includes at least tongue and facial images collected by a high-definition camera, pulse signals collected by a high-precision pulse sensor, and auscultation voice information collected by an audio unit. S2. The data collected in step S1 is transmitted to the data processing platform. After preprocessing, it is input into a multi-algorithm fusion TCM state identification model. The TCM state identification model is based on a predefined set of T2DM pre-TCM "pathogenic state" and integrates at least a multi-state Markov model, a Logistic regression model and a Cox proportional hazards regression model for analysis, and outputs the user's current TCM state classification and corresponding risk score. S3. Based on the TCM status classification and corresponding risk score output in step S2, call the pre-built TCM intervention knowledge base for T2DM and generate a personalized health intervention prescription through an intelligent matching algorithm. S4. After the user executes the health intervention prescription generated in step S3, repeat steps S1 to S2 to obtain the user's follow-up data and updated TCM status; compare and analyze the status before and after the update with the risk score to evaluate the intervention effect, and dynamically adjust or regenerate the intervention prescription based on the intervention results to complete the diagnosis and treatment operation.

[0021] As can be seen from the above description, the present invention has the following beneficial effects: This invention provides a method for diagnosing and treating prediabetes type 2 diabetes based on Traditional Chinese Medicine (TCM) state theory, constructing a complete, human-machine collaborative closed-loop treatment process. This method uses a portable device to objectively collect data from the four diagnostic methods of TCM, employs a multi-algorithm fusion model to accurately identify the "pathological state" of pre-T2DM, and intelligently generates personalized intervention prescriptions based on this. Finally, through data re-collection and model re-evaluation, the method quantifies the effects and dynamically optimizes the treatment plan. This solves the problems of strong subjectivity in traditional TCM diagnosis and treatment, disconnection from modern technology, and lack of quantitative feedback and continuous adjustment after intervention. It achieves full-cycle, intelligent, and personalized health management for pre-T2DM, from identification and intervention to follow-up management.

[0022] Furthermore, the predefined set of T2DM pre-TCM "pathological state" states in step S2 includes: spleen deficiency with dampness, qi and yin deficiency, liver stagnation and spleen deficiency, phlegm and blood stasis, and balance with imbalance.

[0023] As described above, the core TCM "pathological state" classification for the early stages of T2DM is defined. This specific set (such as spleen deficiency with dampness, qi and yin deficiency, etc.) gives the state identification model clear identification targets that are highly relevant to the disease stage, rather than broad constitution classifications. This significantly improves the clinical guidance significance of the identification results and the targeting of intervention plans, providing a key theoretical and classification foundation for achieving "precision intervention".

[0024] Furthermore, the multi-state Markov model in step S2 is used to quantify the dynamic transition risk between different TCM states, and its state transition strength matrix... Defined as: in This indicates that at time t, the state changes from... Transition to state The instantaneous intensity, K is the total number of states, and satisfies .

[0025] As described above, by introducing a multi-state Markov model and its state transition strength matrix Q(t), the system can not only identify the current state but also quantify the probability and risk of the dynamic evolution of different TCM states over time and with intervention. This provides a forward-looking predictive capability for disease progression, upgrading risk assessment from static scoring to dynamic process analysis, and providing a more scientific basis for early warning and intervention timing selection.

[0026] Furthermore, the fusion strategy of the multi-algorithm fusion model in step S2 is as follows: Let the output state probability of the Markov model be... The output probability of the Logistic regression model is The Cox model outputs a risk score of The fusion weights are dynamically determined through intelligent optimization algorithms. Final status score Calculated using the following formula: The intelligent optimization algorithm is either particle swarm optimization or ant colony optimization.

[0027] As described above, a specific multi-model fusion and weight optimization strategy was proposed. By weighting and fusing the dynamic evolution information of the Markov model, the classification probability of Logistic regression, and the temporal risk information of Cox regression, and dynamically adjusting the weights using intelligent optimization algorithms (such as PSO), the advantages of different models are effectively integrated, overcoming the limitations of a single algorithm, thereby comprehensively improving the overall accuracy, robustness, and predictive ability of state identification.

[0028] Furthermore, the intelligent matching algorithm in step S3 is a hybrid algorithm that integrates a retrieval-enhanced sequence-to-sequence learning model with a hypergraph neural network.

[0029] As described above, the core algorithm for generating personalized prescriptions is defined as a hybrid algorithm combining retrieval-enhanced Seq2Seq and hypergraph neural networks. This technical solution can simultaneously utilize the precise retrieval of the knowledge base (ensuring the standardization of the prescription) and the generation and reasoning capabilities of the neural network (handling complex and personalized matching relationships), thereby efficiently and accurately matching the most suitable and combinable personalized health prescriptions for the user's current multidimensional state from massive, structured TCM intervention knowledge.

[0030] Furthermore, the T2DM early intervention knowledge base called in step S3 integrates the content from national standards, clinical guidelines and experience data of famous veteran TCM doctors, and is structured to include at least two types of intervention measures from TCM conditioning, dietary therapy, acupoint massage and lifestyle guidance.

[0031] As described above, the authoritative sources (national standards, guidelines, and renowned physicians' experience) and structured content (covering drug therapy, dietary therapy, and non-drug therapies) of the TCM intervention knowledge base are limited. This ensures that the generated intervention prescriptions are scientific, standardized, and clinically feasible, integrating standardized knowledge with individualized experience, making the suggestions output by the intelligent system both reliable and comprehensive, covering multimodal intervention methods.

[0032] Furthermore, the basic health data collected in step S1 is entered through a mobile terminal application or obtained from a connected third-party health monitoring device, including blood glucose levels, BMI, lifestyle information, and symptom self-report.

[0033] As described above, this expands the dimensions and methods of health data collection. By integrating data entered via mobile devices and from third-party devices (such as blood glucose meters), it combines information from the four diagnostic methods of traditional Chinese medicine with modern medical indicators and lifestyle data, constructing a more complete personal health profile. This multi-source data fusion provides a richer and more comprehensive data foundation for subsequent precise identification and personalized intervention, achieving complementarity and integration of health information from both traditional Chinese and Western medicine.

[0034] Furthermore, in step S4, the effect is evaluated by calculating the probability changes of the user's TCM state before and after the intervention, and the state transition intensity matrix. Changes in relevant elements or risk scores This can be achieved through at least one of the methods used to reduce the magnitude of the reduction.

[0035] As described above, various specific technical means are provided for quantitatively assessing the effectiveness of interventions. By calculating changes in state probability, analyzing alterations in the intensity of state transitions, or observing the decline in risk scores, the abstract concept of "effect" is transformed into measurable and comparable mathematical model indicators. This makes efficacy assessment more objective and refined, providing clear, data-driven decision-making basis for the closed-loop optimization step of "dynamically adjusting or regenerating prescriptions," rather than subjective feelings.

[0036] like Figure 2 As shown, the present invention provides an electronic device including a memory 1, a processor 2, and a computer program stored in the memory 1 and executable on the processor 2. When the processor 2 executes the program, it implements the steps performed by the data processing platform as described in the "A Method for Diagnosing and Treating Prediabetes Type 2 Based on Traditional Chinese Medicine State Theory".

[0037] As described above, the aforementioned diagnostic and treatment methods are protected as "electronic devices." This expands the scope of patent protection, ensuring that any hardware device implementing the data processing platform logic described in the method, or any medium storing the program software, falls within the protection of the claims, thus enhancing the practical value and commercial protection of the patent.

[0038] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps performed by the data processing platform in the described method for the diagnosis and treatment of prediabetes type 2 based on traditional Chinese medicine state theory.

[0039] As described above, the aforementioned diagnostic and treatment methods are protected in the form of a "computer-readable storage medium." This expands the scope of patent protection, ensuring that any hardware device implementing the data processing platform logic described in the method, or any medium storing the program software, falls within the protection of the claims, thus enhancing the practical value and commercial protection of the patent.

[0040] The following are several preferred embodiments or application embodiments to help those skilled in the art better understand the technical content of the present invention and the technical contributions made by the present invention compared with the prior art: Preferred embodiment 1: like Figure 1 As shown, the "Portable Intelligent Diagnosis and Treatment Method for Prediabetes Type 2 Based on Traditional Chinese Medicine State Theory" provided by this invention is based on the construction of a complete system of "device data acquisition - cloud analysis - intelligent matching - closed-loop optimization". The following provides a detailed and practical explanation of each step, along with relevant experimental verification data.

[0041] S1: Specific Implementation of Data Acquisition Steps This step aims to collect users' health information in a standardized and objective manner through a combination of hardware and software, laying a data foundation for subsequent analysis.

[0042] Hardware implementation: Portable intelligent diagnostic and treatment device: It adopts an integrated design, with dimensions of approximately 180mm*80mm*30mm and a weight of approximately 450g. The outer shell is made of medical-grade antibacterial material.

[0043] Traditional Chinese Medicine Four Diagnostic Methods Data Collection Unit: Visual inspection unit: Integrates a 5-megapixel fixed-focus high-definition camera with a ring light. Following on-screen guidance, the user extends their tongue to the designated area, and the device automatically captures high-definition images (2592x1944 resolution) of the front of the tongue and the veins under the tongue. Simultaneously, it captures a frontal image of the user's face under standard white light.

[0044] Auscultation Unit: Equipped with a built-in high-sensitivity microphone. Guide the user to read a standard text (such as "The weather is very nice today") and take several calm breaths, collecting a total of 10 seconds of speech and breathing sound signals.

[0045] Palpation Unit: Integrates a high-precision pulse wave sensor (sampling rate 1kHz) based on a piezoelectric thin-film sensor. The user places their index finger in the sensor groove, and the device acquires a 90-second radial artery pulse wave signal.

[0046] Data processing and communication unit: The device has a built-in low-power ARM Cortex-A53 processor, 4GB RAM, and is equipped with Wi-Fi (802.11ac) and Bluetooth 5.0 BLE modules for initial data processing and uploading.

[0047] Software and data flow implementation: Users initiate the test via the device's touchscreen or a paired mobile app. The app guides users to fill out a structured electronic questionnaire (basic health data), which includes: dietary preferences over the past week (such as frequency of oily and sweet foods), sleep quality (PSQI simplified scale), emotional state (such as whether they are irritable or depressed), fatigue, and 30 other symptoms; and obtains current BMI, fasting blood glucose / 2-hour postprandial blood glucose values ​​by manually inputting or authorizing the reading of data from a smart scale and blood glucose meter (such as a Bluetooth blood glucose meter).

[0048] After the device completes the four diagnostic methods, it packages and uploads the tongue image (JPEG format), raw pulse wave signal (WAV format), audio file (WAV format), and encrypted questionnaire data to the data receiving server of the cloud data processing platform via HTTPS protocol.

[0049] Data preprocessing implementation (conducted on a cloud platform): Image processing: The U-Net network was used to segment the tongue body and background of the tongue image, and 12 features were extracted, including tongue body color (mean in RGB and HSV spaces), tongue coating color, tongue coating coverage, and crack index. After face alignment of the face image, features such as facial gloss and lip color were extracted.

[0050] Signal processing: Wavelet denoising and period segmentation were performed on the pulse wave signal to extract eight time-domain and frequency-domain features, including pulse rate, pulse rate standard deviation, main wave height, and diphtheria depth. MFCC (Mel-frequency cepstral coefficients) feature extraction was performed on the audio signal to analyze the frequency-domain energy distribution of speech and the fundamental frequency of breath sounds.

[0051] Text processing: The questionnaire text was semantically embedded using a word vector model fine-tuned based on the "BERT-wwm-ext" pre-trained model. Combined with the BiGRU-CRF model, key symptom entities such as "excessive appetite", "dry mouth", and "fatigue" and their degree adverbs were identified and converted into structured feature vectors.

[0052] After integration, all features are imputed by the median of missing values ​​and Z-score normalized, ultimately forming a structured feature vector X containing approximately 60 dimensions.

[0053] Experimental data supports this: In a preliminary experiment involving 300 participants with pre-T2DM, this data collection protocol was compared with the independent diagnostic results of two associate chief physicians of traditional Chinese medicine. In judging tongue coating color, the consistency rate between the device and the physicians was 89.3% (Kappa=0.81); in distinguishing between "wiry" and "slippery" pulse characteristics, the accuracy rate of the feature-based support vector machine classifier and the physicians' consensus reached 82.5%. This demonstrates the high reliability of this objective data collection protocol.

[0054] S2: Specific implementation of the state identification step This step is the core algorithm module, responsible for converting feature vectors into TCM state diagnoses.

[0055] Model building and training: State Definition: Define the state set S = {S1 (Spleen deficiency with dampness), S2 (Qi and Yin deficiency), S3 (Liver stagnation and Spleen deficiency), S4 (Phlegm and blood stasis), S5 (Balanced but unbalanced)}. The annotation work was conducted by an expert group of three senior TCM physicians. Based on the theory of "TCM State Theory" and clinical guidelines, they performed double-blind annotation on the complete four diagnostic methods and follow-up data (including whether diabetes developed) of 5000 historical patients in the early stages of T2DM, forming the gold standard training set.

[0056] Multi-state Markov model: The frequency of state transitions between patients in the training set was statistically analyzed at six-month intervals. A non-homogeneous Markov model was fitted using the msm package in R language to estimate the state transition intensity matrix Q. For example, the study found that the intensity q of the transition from S1 (spleen deficiency and dampness) to S4 (phlegm and blood stasis) was... 14 The significantly higher percentage of metastases suggests a possible pathway for disease progression. The model can output the probability of being in each state at a given time. .

[0057] Logistic regression model: Using the TCM state as the dependent variable (e.g., whether it is S2) and the feature vector X as the independent variable, L2 regularized logistic regression is used for training, resulting in 5 binary classification models, which output the probability of belonging to each state. .

[0058] Cox proportional hazards model: A Cox model is constructed using the time from progression to diagnosis of type 2 diabetes as the endpoint event and the aforementioned feature X as a covariate. The model outputs a risk score. This represents the long-term risk of disease transformation based on current characteristics.

[0059] Fusion and Optimization: Building a Fusion Model .

[0060] Initialize the weights to equal values. Optimize using Particle Swarm Optimization (PSO): set the particle swarm size to 30 and iterate 100 times. The fitness function is the macro-average F1 score of the model on the independent validation set (1000 examples). After PSO search, a set of optimal weights (e.g., ...) are obtained. =0.35, =0.45, =0.20), this weight configuration optimizes the overall performance of the model.

[0061] Online identification process: When a new user's feature vector X is input, the system calls three sub-models in parallel to calculate... , , Then, substituting the optimized fusion formula, the five fusion scores for S1 to S5 are calculated. Take the state with the highest rating as the primary state category and output that rating and its corresponding value. As a comprehensive risk score, the results are presented in the form of a visual report, such as: "Primary condition: Liver stagnation and spleen deficiency (probability 72%); Comprehensive risk score: 65 (percentile, indicating a risk higher than 65% of the peers)".

[0062] Experimental data support the validation in a prospective cohort (followed for 2 years) of 1200 pre-T2DM participants. The single logistic regression model achieved an accuracy of 83.5% in identifying the primary state. The multi-algorithm fusion model of this invention improves the accuracy to 91.2%. More importantly, the high-risk group output by the fusion model ( The proportion of the top 30% of the risk score who developed diabetes within 2 years was 28.7%, which was significantly higher than the conversion rate of 5.1% in the low-risk group (bottom 30%) (HR=5.8, 95%CI: 3.2-10.5, P<0.001), demonstrating its excellent risk warning capability.

[0063] S3: Specific Implementation of Intervention Matching Steps This step translates the status diagnosis into a personalized action plan.

[0064] Knowledge base construction: Using natural language processing and knowledge graph technology, a knowledge base for T2DM pre-treatment TCM intervention containing approximately 100,000 relationships was constructed.

[0065] Data source: Authoritative documents such as the "Guidelines for the Prevention and Treatment of Diabetes with Traditional Chinese Medicine" and the "Guiding Principles for Clinical Research of New Traditional Chinese Medicine Drugs" are entered in a structured manner; using information extraction technology, the "syndrome-treatment-prescription-lifestyle" quadruple is automatically extracted from 133 digital works by famous doctors, such as "Shi Jinmo's Medical Cases" and "Li Zhenhua's Clinical Experience in Spleen and Stomach Diseases", and then entered into the database after being reviewed by experts.

[0066] Knowledge representation: Hypergraphs are used for storage. For example, a "yam and barley porridge" recipe is a hyperedge, connecting multiple nodes such as "spleen deficiency and dampness", "qi deficiency", "medicinal diet", "ingredients: yam, barley", and "effects: strengthening the spleen and removing dampness". A "Zusanli" acupoint node is connected to multiple hyperedges such as "spleen deficiency", "qi deficiency", "massage", and "moxibustion".

[0067] The intelligent matching algorithm operates as a two-stage pipeline that integrates the two technologies described in this solution.

[0068] The first stage (retrieval-enhanced Seq2Seq): User status (e.g., "liver stagnation and spleen deficiency"), risk score (65), and symptom keywords (e.g., "irritability" and "hypochondriac distension") are used as input sequences. A Transformer-based Seq2Seq model (PreGenerator) is trained to generate a prescription text draft, such as "soothe the liver and regulate qi, while also strengthening the spleen." Simultaneously, the system rapidly retrieves the Top-K (K=10) historical medical records or standard solutions most relevant to the input conditions from the knowledge base in parallel.

[0069] The second stage (Hypergraph Neural Network Inference): The generated draft and search results are jointly transformed into a subgraph in the knowledge base hypergraph. A Hypergraph Neural Network (HGNN) is used for information propagation and inference on this subgraph. HGNN can capture higher-order associations between "liver stagnation" and "Taichong acupoint" and "Bupleurum," as well as the relationships between "spleen deficiency" and "yam" and "avoiding raw and cold foods." Through neural network calculations, the intervention measures in the draft are weighted, compatibility-checked, and supplemented. For example, it might strengthen "it is recommended to massage Taichong acupoint for 3 minutes daily" and supplement "a small amount of Buddha's Hand flower can be added to the diet as a tea substitute," while weakening a search result for a medicinal herb that slightly conflicts with the user's constitution.

[0070] Output: The final output is a structured, personalized health prescription that clearly lists: Traditional Chinese medicine treatment suggestions (e.g., modifications to the classic formula "Xiaoyao San" are recommended; it is advised to consult a doctor online for a prescription).

[0071] Dietary therapy options (e.g., this week's recommended "Rose and Tangerine Peel Drink", recipe and preparation method).

[0072] Acupressure (e.g., press Taichong and Zusanli acupoints for 3 minutes each day, see attached acupoint diagram).

[0073] Lifestyle guidance (e.g., it is recommended to take a 30-minute relaxing walk every afternoon and practice the "Shh" sound to soothe the liver before bed).

[0074] Experimental data support: Fifteen TCM experts were invited to conduct a double-blind evaluation of the prescriptions generated by the system for 100 virtual cases (using a 5-point Likert scale: 1 - completely irrelevant, 5 - very accurate). Evaluation criteria included: 1. Relevance to syndrome differentiation; 2. Feasibility of the measures; 3. Comprehensiveness of the solutions. The system's prescriptions achieved average scores of 4.42, 4.65, and 4.28 on the three criteria, significantly higher than the baseline system based solely on rule matching (average scores of 3.70, 3.80, and 3.50). In a pilot program with 50 real users, 90% of the users found the prescriptions "easy to understand and suitable for their own situations."

[0075] S4: Specific Implementation of Effect Evaluation and Closed-Loop Optimization Steps This step ensures that health management is a dynamic, self-optimizing process.

[0076] Effectiveness evaluation implementation: Triggering mechanism: Users can set reminders in the app (once every 4 weeks by default), or actively trigger a retest when they notice changes in their symptoms. The system will push a message guiding the user to complete a full S1 data collection using a portable device.

[0077] Quantitative assessment: The new data undergoes the S2 process to obtain a new state probability distribution P_new and a risk score S_f_new. The system automatically calculates the following indicators: State improvement: Calculate the probability increment of transition to S5 (neutral bias) or clinical target state in the new distribution.

[0078] Risk reduction rate: (S_f_old-S_f_new) / S_f_old*100%.

[0079] Markov trajectory analysis: By comparing the old and new features, we estimate the changes in the state transition strength matrix and observe whether the intensity of the transition from the current state to the unfavorable state has weakened.

[0080] Visualized report: Generates a comparative radar chart or line graph to clearly show the changing trends of key indicators such as "TCM status score" and "comprehensive risk index".

[0081] Closed-loop optimization implementation: Decision-making rules: The system has built-in optimization thresholds. For example, if the risk reduction rate is less than 10% and the primary state remains unchanged, it is judged as "insignificant effect" and the solution optimization is automatically triggered.

[0082] The optimized process involves taking the user's initial state, historical intervention prescriptions, current state, and effect evaluation data as new inputs and sending them back to the S3 intelligent matching module. At this point, the knowledge base and matching algorithm will learn from "tried but limitedly effective solutions" as negative feedback, re-searching and reasoning to generate a revised prescription. Adjustments may include: changing dietary therapy, increasing the frequency of acupoint stimulation, or strengthening guidance on a specific lifestyle.

[0083] Physician Intervention Interface: If the system still shows "insignificant effect" after two consecutive optimization assessments, or if the risk score rises sharply, the system will generate a "high-risk user warning" in the physician management backend, along with a complete data chain, and recommend that physicians conduct remote or offline proactive intervention.

[0084] Experimental data supports this finding: In a 6-month randomized controlled trial, 120 participants with pre-T2DM were divided into two groups: the experimental group was managed using the closed-loop system of this invention; the control group received only a one-time health assessment and standardized advice manual. Results showed that after 6 months, the average decrease in the comprehensive risk score in the experimental group was 34.2%, significantly higher than the 8.7% in the control group (P<0.01). In the experimental group, the proportion of participants whose TCM status improved to "balanced but slightly imbalanced" or showed significant improvement reached 68.3%, compared to only 26.7% in the control group. More importantly, the improvement in fasting blood glucose and 2-hour postprandial blood glucose in the experimental group was also significantly better than that in the control group. This fully demonstrates the effectiveness and superiority of closed-loop optimized management.

[0085] Preferred Embodiment Two like Figure 2 As shown, the present invention provides an electronic device including a memory 1, a processor 2, and a computer program stored in the memory 1 and executable on the processor 2. When the processor 2 executes the program, it implements the steps performed by the data processing platform as described in the "A Method for Diagnosing and Treating Prediabetes Type 2 Based on Traditional Chinese Medicine State Theory".

[0086] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps performed by the data processing platform in the described method for the diagnosis and treatment of prediabetes type 2 based on traditional Chinese medicine state theory.

[0087] As described above, the aforementioned diagnostic and treatment methods are protected in the form of "electronic devices" and "computer-readable storage media." This expands the scope of patent protection, ensuring that any hardware device implementing the data processing platform logic described in the method, or any medium storing the program software, falls within the protection of the claims, thus enhancing the practical value and commercial protection of the patent.

[0088] The present invention has been described with reference to the foregoing embodiments and accompanying drawings; however, the foregoing embodiments are merely examples for implementing the present invention. It must be noted that the disclosed embodiments do not limit the scope of the present invention. Conversely, modifications and equivalents included within the spirit and scope of the claims are all included within the scope of the present invention.

Claims

1. A method for diagnosing and treating prediabetes type 2 based on Traditional Chinese Medicine (TCM) state theory, characterized in that, The process, executed collaboratively by portable intelligent diagnostic and treatment devices and a data processing platform, includes the following steps: S1. The portable intelligent diagnostic and treatment device collects the user's objective data of the four diagnostic methods of traditional Chinese medicine and basic health data; the objective data of the four diagnostic methods of traditional Chinese medicine includes at least tongue and facial images collected by a high-definition camera, pulse signals collected by a high-precision pulse sensor, and auscultation voice information collected by an audio unit. S2. The data collected in step S1 is transmitted to the data processing platform. After preprocessing, it is input into a TCM state identification model that integrates multiple algorithms. The TCM state identification model is based on a predefined set of TCM "pathogenic state" in the early stage of T2DM. It integrates at least a multi-state Markov model, a Logistic regression model, and a Cox proportional hazards regression model for analysis, and outputs the user's current TCM state classification and corresponding risk score. S3. Based on the TCM status classification and corresponding risk score output in step S2, call the pre-built TCM intervention knowledge base for T2DM and generate a personalized health intervention prescription through an intelligent matching algorithm. S4. After the user executes the health intervention prescription generated in step S3, repeat steps S1 to S2 to obtain the user's follow-up data and updated TCM status; compare and analyze the status before and after the update with the risk score to evaluate the intervention effect, and dynamically adjust or regenerate the intervention prescription based on the intervention results to complete the diagnosis and treatment operation.

2. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The predefined set of T2DM pre-TCM "pathological state" in step S2 includes: spleen deficiency with dampness, qi and yin deficiency, liver stagnation and spleen deficiency, phlegm and blood stasis, and balance with imbalance.

3. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The multi-state Markov model in step S2 is used to quantify the dynamic transition risk between different TCM states, and its state transition strength matrix... Defined as: in Indicates the state at time t. Transition to state The instantaneous intensity, K is the total number of states, and satisfies .

4. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The fusion strategy of the multi-algorithm fusion model in step S2 is as follows: Let the output state probability of the Markov model be... The output probability of the Logistic regression model is The Cox model outputs a risk score of The fusion weights are dynamically determined through intelligent optimization algorithms. Final status score Calculated using the following formula: The intelligent optimization algorithm is either particle swarm optimization or ant colony optimization.

5. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The intelligent matching algorithm in step S3 is a hybrid algorithm that combines a sequence-to-sequence learning model with enhanced retrieval and a hypergraph neural network.

6. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The T2DM early intervention knowledge base called in step S3 integrates the content from national standards, clinical guidelines and experience data of famous veteran TCM doctors. It is structured to include at least two types of intervention measures from TCM conditioning, dietary therapy, acupoint massage and lifestyle guidance.

7. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, The basic health data collected in step S1 is entered through a mobile terminal application or obtained from a connected third-party health monitoring device, including blood glucose levels, BMI, lifestyle information, and symptom self-report.

8. The method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory according to claim 1, characterized in that, In step S4, the effect evaluation is carried out by calculating the probability changes of the user's TCM state before and after the intervention, and the state transition intensity matrix. Changes in relevant elements or risk scores This can be achieved through at least one of the methods used to reduce the magnitude of the reduction.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps performed by the data processing platform in the method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps performed by the data processing platform in the method for diagnosing and treating prediabetes type 2 based on traditional Chinese medicine state theory as described in any one of claims 1 to 8.

Images