Traditional Chinese medicine intelligent syndrome differentiation system based on yin-yang qi-blood theory

Abstract

The invention discloses a traditional Chinese medicine intelligent dialectical system based on a yin-yang qi-blood theory, which realizes accurate matching of symptoms and pathological directions through a machine learning algorithm, eliminates subjectivity in a traditional method, improves objectivity and accuracy of diagnosis, quantifies and standardizes traditional Chinese medicine symptoms, eliminates subjective differences, and improves diagnosis accuracy. And the consistency and reliability of diagnosis are improved. Secondly, a traditional dialectical system has extensive dimensions and static limitation and cannot comprehensively describe the pathological state and dynamically track the change of an illness state, and the dialectical granularity is improved and the comprehensiveness and accuracy of pathological state description are ensured through refined eight-dimensional pathological dimensions and twelve pathological directions. Meanwhile, in combination with a machine learning algorithm, the diagnosis model can be dynamically optimized on the basis of continuous accumulation of clinical data, it is ensured that the accuracy of the model is continuously improved along with data increase, the dynamic adaptability of the syndrome differentiation method is achieved, and the limitation of traditional static classification is broken through.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent syndrome differentiation in traditional Chinese medicine based on the Yin-Yang, Qi and Blood theory, and in particular to an intelligent syndrome differentiation system in traditional Chinese medicine based on the Yin-Yang, Qi and Blood theory. Background Art

[0002] With the continuous development of modern medicine, especially the increasing demand for Traditional Chinese Medicine (TCM) diagnosis and treatment, the integration of traditional TCM theories with modern technologies to achieve more accurate and efficient diagnosis and treatment has become a key topic in academia and clinical practice. As one of the essences of traditional Chinese culture, TCM emphasizes the regulation and balance of the human body's overall state, offering unique advantages in the prevention and treatment of disease. However, traditional TCM syndrome differentiation methods face many challenges in modern applications. With the rapid development of artificial intelligence and machine learning technologies in recent years, leveraging these technologies to quantify, refine, and dynamically analyze TCM syndrome differentiation has become a key research direction for improving TCM diagnosis and treatment.

[0003] Currently, traditional Chinese medicine (TCM) syndrome differentiation methods primarily include Eight-Principle Syndrome Differentiation, Six-Jing Syndrome Differentiation, Wei-Qi-Ying-Xue Syndrome Differentiation, Triple-Energizer Syndrome Differentiation, and Zang-Fu Syndrome Differentiation. These methods, centered on yin and yang, qi, and blood, possess a profound theoretical foundation. However, existing syndrome differentiation systems generally suffer from certain limitations. First, existing syndrome differentiation systems are relatively independent and fragmented, lacking a unified framework and connections, making effective integration and dynamic analysis difficult. Second, traditional TCM syndrome differentiation methods lack refinement. While each system maintains a certain degree of self-consistency overall, the lack of detail makes it difficult to precisely grasp the syndrome differentiation process, impacting the accuracy and effectiveness of diagnosis. Furthermore, traditional syndrome differentiation methods primarily rely on the experience and subjective judgment of TCM practitioners, lacking objective quantitative standards and difficulty translating symptoms into calculable parameters. This often leads to problems such as poor reproducibility and unstable results in clinical application.

[0004] Therefore, in traditional Chinese medicine syndrome differentiation methods, how to improve the accuracy of syndrome differentiation, achieve intelligent matching between symptoms and pathology, and introduce modern data-driven dynamic analysis has become an urgent problem to be solved. Summary of the Invention

[0005] The present application provides an intelligent TCM syndrome differentiation system based on the Yin-Yang Qi and Blood theory, aiming to solve the problems of poor repeatability and unstable effects in clinical applications of traditional syndrome differentiation methods, which mainly rely on the experience and subjective judgment of TCM practitioners, lack objective quantitative standards, and have difficulty converting symptoms into calculable parameters.

[0006] In a first aspect, a TCM intelligent syndrome differentiation system based on the Yin-Yang Qi and Blood theory is provided, characterized in that the system comprises:

[0007] Symptom data collection module, which is used to extract patient symptom descriptions through natural language processing technology and quantify the frequency, intensity and duration of symptoms into standardized numerical parameters;

[0008] The symptom-syndrome association matrix construction module is used to establish a mapping relationship between symptoms and syndromes based on the classical Chinese medicine literature and expert experience knowledge base, and calculate the initial weight value of each symptom in the target syndrome through a feature selection algorithm; wherein, the mapping relationship between symptoms and syndromes based on the classical Chinese medicine literature and expert experience knowledge base includes constructing an eight-dimensional twelve-item syndrome differentiation system, wherein the eight-dimensional pathological dimensions are divided into Yin deficiency, Yin excess, Yang deficiency, Yang excess, Qi deficiency, Qi gathering, blood stasis, and blood dispersion, a total of eight dimensions, covering the pathological states of the human body The twelve pathological directions are divided into: basic pathological directions, including yang deficiency, yin deficiency, water dampness, qi stagnation, qi deficiency, wind, and sweating; complex pathological directions, including excess heat, damp heat, blood heat, phlegm and blood stasis, and food stagnation and intestinal obstruction, and the pathological directions include visible tangible pathological products such as phlegm and blood stasis or food stagnation and intestinal obstruction, and invisible imbalances in the body state, among which tangible pathological products include phlegm and blood stasis, food stagnation and intestinal obstruction, water dampness, and sweating, and invisible imbalances in the body state include qi deficiency, yang deficiency, qi accumulation, yin deficiency, blood stasis, etc.;

[0009] A dynamic weight optimization module, which is used to input the numerical parameters into a trained machine learning model. The model dynamically adjusts the weights of each symptom using a weighted sum algorithm, wherein the initial weights are set based on Traditional Chinese Medicine theory, and subsequent weights are iteratively updated using a backpropagation algorithm or reinforcement learning combined with real-time symptom data;

[0010] The syndrome diagnosis output module is used to output the patient's syndrome diagnosis results through a deep neural network, support vector machine or integrated learning model based on the dynamically adjusted weight values ​​and symptom numerical parameters, and generate an explanatory report including the symptom weight distribution.

[0011] In the above solution, optionally, the symptom quantification method includes:

[0012] Symptom frequency coding: symptom frequency is divided into "occasional", "intermittent" and "persistent" and assigned values ​​of 1-3 respectively;

[0013] Symptom intensity is quantified by using fuzzy set value statistical methods or visual analog scale method to map the symptom intensity into a numerical interval [0,1];

[0014] Tongue and pulse parameter extraction: The texture, color, and morphological characteristics of the tongue, as well as the frequency, rhythm, and intensity characteristics of the pulse are extracted through image recognition algorithms, and are normalized with the symptom numerical parameters to form a unified multimodal input vector.

[0015] In the above solution, optionally, the dynamic weight optimization module is specifically used to:

[0016] Initialization of TCM expert experience weights, based on the TCM Syndrome Efficacy Evaluation Scale or the Delphi method to determine the initial weights of symptoms;

[0017] Machine learning weights are iteratively updated, and by comparing the model output with the diagnosis results of traditional Chinese medicine experts, the weights are dynamically adjusted using gradient descent or genetic algorithms;

[0018] The two-layer optimization of weight distribution assigns weights to the contribution of syndrome factors and syndrome types in a hierarchical manner according to the principle of high frequency and low weight of symptoms and low frequency and high weight of symptoms.

[0019] In the above solution, optionally, the training process of the machine learning model includes:

[0020] Data preprocessing: training is performed using a labeled TCM medical record dataset containing symptom-syndrome annotated pairs, and rare syndrome samples are expanded using a generative adversarial network.

[0021] Model training and validation: cross-validation was used to optimize model hyperparameters. Model output results must meet the following conditions: the accuracy of syndrome differentiation for the main symptom is ≥90%, the accuracy of syndrome differentiation for the secondary symptom is ≥78%, and the F1 value of syndrome type diagnosis is ≥0.6;

[0022] The model's interpretability is enhanced, and the symptom weights are visualized and analyzed through the attention mechanism, outputting the contribution ranking of each symptom to the target syndrome and the trajectory of weight changes.

[0023] In the above solution, optionally, the dynamic adjustment of the symptom weight includes:

[0024] Real-time symptom feedback mechanism: when a patient's symptoms change, the model recalculates the weights based on the latest intensity and frequency of the symptoms;

[0025] Multimodal feature fusion combines non-symptomatic features such as tongue and pulse conditions, assigns independent weight modules through the attention mechanism, and performs weighted fusion with the symptom weight module to ultimately output a comprehensive syndrome diagnosis result.

[0026] In the above solution, optionally, the system further includes:

[0027] The symptom data collection interface is used to connect to the patient's terminal device or smart wearable device to collect the patient's physiological data, motion data and other related symptom information in real time, and transmit it to the symptom data collection module for analysis.

[0028] In the above solution, optionally, the algorithms used by the dynamic weight optimization module include random forest, neural network and support vector machine machine learning methods.

[0029] In the above scheme, optionally, the deep neural network is a multilayer perceptron network, which is used to perform nonlinear mapping based on the input symptom data.

[0030] In the above solution, optionally, the syndrome type diagnosis output module further includes:

[0031] The comprehensive syndrome evaluation module is used to output the optimal diagnostic result based on the matching degree of multiple syndromes and the patient's comprehensive symptom data.

[0032] Compared with the prior art, this application has at least the following beneficial effects:

[0033] This application, based on further analysis and research of existing technical issues, recognizes that traditional syndrome differentiation methods rely primarily on the experience and subjective judgment of traditional Chinese medicine practitioners, lack objective quantitative standards, and have difficulty converting symptoms into calculable parameters. This often leads to poor repeatability and unstable results in clinical applications. This system converts patient symptoms into standardized numerical parameters and uses machine learning algorithms to accurately match symptoms with pathological directions, eliminating the subjectivity in traditional methods and improving the objectivity and accuracy of diagnosis. Secondly, existing syndrome differentiation methods lack effective tracking of dynamic changes in the disease state and are generally based only on static classification. The dynamic weight optimization module of the present invention, by updating symptom weights in real time and tracking pathological changes, enables the diagnostic system to adjust according to the patient's symptom changes, enhancing the system's dynamic adaptability. Finally, traditional Chinese medicine syndrome differentiation systems lack a unified framework and detailed processing of pathological relationships, making it difficult for the syndrome differentiation process to comprehensively and accurately cover various pathological conditions. By constructing an eight-dimensional, twelve-pathological direction framework, the present invention provides a multi-dimensional, unified syndrome differentiation system that not only covers a variety of pathological characteristics but also effectively associates the relationship between symptoms and pathological directions, improving the comprehensiveness and accuracy of syndrome differentiation. In summary, the technical solution of the present invention effectively solves the deficiencies in the background technology and improves the scientificity, flexibility and accuracy of TCM syndrome differentiation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 An eight-dimensional twelve-item system structure diagram for intelligent TCM syndrome differentiation based on the Yin-Yang Qi and Blood theory provided for one embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0036] In one embodiment, a TCM intelligent syndrome differentiation system based on the Yin-Yang Qi-Blood theory is provided, the system comprising:

[0037] Symptom data collection module, which is used to extract patient symptom descriptions through natural language processing technology and quantify the frequency, intensity and duration of symptoms into standardized numerical parameters;

[0038] The symptom-syndrome association matrix construction module is used to establish a mapping relationship between symptoms and syndromes based on the classical Chinese medicine literature and expert experience knowledge base, and calculate the initial weight value of each symptom in the target syndrome through a feature selection algorithm; wherein, the mapping relationship between symptoms and syndromes based on the classical Chinese medicine literature and expert experience knowledge base includes constructing an eight-dimensional twelve-item syndrome differentiation system, wherein the eight-dimensional pathological dimensions are divided into Yin deficiency, Yin excess, Yang deficiency, Yang excess, Qi deficiency, Qi gathering, blood stasis, and blood dispersion, a total of eight dimensions, covering the pathological states of the human body The twelve pathological directions are divided into: basic pathological directions, including yang deficiency, yin deficiency, water dampness, qi stagnation, qi deficiency, wind, and sweating; complex pathological directions, including excess heat, damp heat, blood heat, phlegm and blood stasis, and food stagnation and intestinal obstruction, and the pathological directions include visible tangible pathological products such as phlegm and blood stasis or food stagnation and intestinal obstruction, and invisible imbalances in the body state, among which tangible pathological products include phlegm and blood stasis, food stagnation and intestinal obstruction, water dampness, and sweating, and invisible imbalances in the body state include qi deficiency, yang deficiency, qi accumulation, yin deficiency, blood stasis, etc.;

[0039] A dynamic weight optimization module, which is used to input the numerical parameters into a trained machine learning model. The model dynamically adjusts the weights of each symptom using a weighted sum algorithm, wherein the initial weights are set based on Traditional Chinese Medicine theory, and subsequent weights are iteratively updated using a backpropagation algorithm or reinforcement learning combined with real-time symptom data;

[0040] The syndrome diagnosis output module is used to output the patient's syndrome diagnosis results through a deep neural network, support vector machine or integrated learning model based on the dynamically adjusted weight values ​​and symptom numerical parameters, and generate an explanatory report including the symptom weight distribution.

[0041] In this embodiment, the symptom quantification method includes:

[0042] Symptom frequency coding: symptom frequency is divided into "occasional", "intermittent" and "persistent" and assigned values ​​of 1-3 respectively;

[0043] Symptom intensity is quantified by using fuzzy set value statistical methods or visual analog scale method to map the symptom intensity into a numerical interval [0,1];

[0044] Tongue and pulse parameter extraction: The texture, color, and morphological characteristics of the tongue, as well as the frequency, rhythm, and intensity characteristics of the pulse are extracted through image recognition algorithms, and are normalized with the symptom numerical parameters to form a unified multimodal input vector.

[0045] In this embodiment, the dynamic weight optimization module is specifically used to:

[0046] Initialization of TCM expert experience weights, based on the TCM Syndrome Efficacy Evaluation Scale or the Delphi method to determine the initial weights of symptoms;

[0047] Machine learning weights are iteratively updated, and by comparing the model output with the diagnosis results of traditional Chinese medicine experts, the weights are dynamically adjusted using gradient descent or genetic algorithms;

[0048] The two-layer optimization of weight distribution assigns weights to the contribution of syndrome factors and syndrome types in a hierarchical manner according to the principle of high frequency and low weight of symptoms and low frequency and high weight of symptoms.

[0049] In this embodiment, the training process of the machine learning model includes:

[0050] Data preprocessing: training is performed using a labeled TCM medical record dataset containing symptom-syndrome annotated pairs, and rare syndrome samples are expanded using a generative adversarial network.

[0051] Model training and validation: cross-validation was used to optimize model hyperparameters. Model output results must meet the following conditions: the accuracy of syndrome differentiation for the main symptom is ≥90%, the accuracy of syndrome differentiation for the secondary symptom is ≥78%, and the F1 value of syndrome type diagnosis is ≥0.6;

[0052] The model's interpretability is enhanced, and the symptom weights are visualized and analyzed through the attention mechanism, outputting the contribution ranking of each symptom to the target syndrome and the trajectory of weight changes.

[0053] In this embodiment, the dynamic adjustment of the symptom weight includes:

[0054] Real-time symptom feedback mechanism: when a patient's symptoms change, the model recalculates the weights based on the latest intensity and frequency of the symptoms;

[0055] Multimodal feature fusion combines non-symptomatic features such as tongue and pulse conditions, assigns independent weight modules through the attention mechanism, and performs weighted fusion with the symptom weight module to ultimately output a comprehensive syndrome diagnosis result.

[0056] In this embodiment, the system further comprises:

[0057] The symptom data collection interface is used to connect to the patient's terminal device or smart wearable device to collect the patient's physiological data, motion data and other related symptom information in real time, and transmit it to the symptom data collection module for analysis.

[0058] In this embodiment, the algorithms used by the dynamic weight optimization module include random forest, neural network and support vector machine machine learning methods.

[0059] In this embodiment, the deep neural network is a multilayer perceptron network, which is used to perform nonlinear mapping based on the input symptom data.

[0060] In this embodiment, the syndrome diagnosis output module further includes:

[0061] The comprehensive syndrome evaluation module is used to output the optimal diagnostic result based on the matching degree of multiple syndromes and the patient's comprehensive symptom data.

[0062] In one embodiment, although the traditional Chinese medicine syndrome differentiation system (such as Eight Principles Syndrome Differentiation, Six Channel Syndrome Differentiation, Wei Qi Ying Xue Syndrome Differentiation, Qi, Blood and Body Fluid Syndrome Differentiation, etc.) is centered on Yin and Yang, Qi and blood, it has the following limitations:

[0063] The dialectical systems are independent of each other and are difficult to relate to each other and cannot be unified;

[0064] Coarse dimensionality: Each dialectical system is self-consistent as a whole, but the degree of refinement is insufficient, which makes it difficult to grasp the dialectical system and thus affects the accuracy of the dialectical system.

[0065] Lack of quantitative standards: It relies more on the subjective experience of physicians and is difficult to convert symptoms into calculable parameters.

[0066] Static limitations: The existing systems, especially the Six-Channel Syndrome Differentiation and Wei-Qi-Ying-Xue Syndrome Differentiation systems, have clear explanations of the dynamic changes of diseases. However, in actual application, they are mostly based on static classifications, making it difficult to dynamically track pathological evolution.

[0067] Insufficient data-driven: Failure to fully integrate modern clinical data and machine learning technology.

[0068] This example builds a refined, quantifiable, eight-dimensional, twelve-item syndrome differentiation framework based on the theory of yin and yang, qi, and blood. Through machine learning, a dynamic mathematical model is established to intelligently match symptom combinations with pathological characteristics. This provides an iteratively optimized syndrome differentiation tool, supporting personalized diagnosis and treatment, as well as disease early warning.

[0069] This embodiment takes yin and yang, qi and blood as the theoretical basis. Figure 1 As shown in the figure, an eight-dimensional twelve-item dialectical system is constructed, and a dynamic mathematical model is established by combining machine learning algorithms, specifically including:

[0070] There are eight dimensions in total, including Yin deficiency (dryness), Yin excess (dampness), Yang deficiency (cold), Yang excess (heat), Qi deficiency (stagnation), Qi accumulation (stagnation), blood stasis (wind), and blood dispersion (sweat), covering the core characteristics of the human body's pathological state.

[0071] Twelve pathological categories are detailed: Basic pathology: Yang deficiency, Yin deficiency, dampness, Qi stagnation, Qi deficiency, wind, and sweating. Complex pathology: Yang excess-related: excess heat, damp heat, and blood heat; Yin excess-related: phlegm and blood stasis, food stagnation, and intestinal obstruction.

[0072] Specifically, yang deficiency manifests as coldness in the human body, which is a basic pathological direction; yang excess manifests as heat in the human body, which includes three pathological directions: blood heat, damp heat and real heat, which is a complex pathological direction; yin deficiency manifests as dryness in the human body, which is a basic pathological direction; yin excess manifests as dampness in the human body, and blood stasis manifests as stasis in the human body, each forming a basic pathological direction, but dampness gathers into phlegm, and forms a pathological direction of phlegm and stasis with blood stasis, which is a complex pathological direction; qi gathering in the human body manifests as stagnation, which is a basic pathological direction; qi deficiency manifests as sinking in the human body, which is a basic pathological direction; blood dispersion in the human body manifests as sweating, which is a basic pathological direction; qi gathering, yang deficiency and yin excess will produce a complex pathological direction of food accumulation and intestinal obstruction; thus, twelve pathological directions are formed; among them, phlegm and stasis, food accumulation and intestinal obstruction, as well as dampness and sweating are tangible pathological products, while the remaining eight are intangible states. It is the imbalance of the body's intangible state that produces tangible pathological products.

[0073] Mathematical model construction method:

[0074] Symptom quantification: converting TCM symptoms into numerical parameters;

[0075] Matching degree calculation: The cosine similarity algorithm is used to calculate the matching degree between the patient's symptom vector and the standard vector of each pathological direction;

[0076] Dynamic optimization: Based on clinical feedback data, model weights are iteratively updated through random forest or neural network.

[0077] This embodiment expands a single dimension (such as Yang Sheng) into a multi-level complex pathology (excess heat, damp heat, blood heat), improving the granularity of syndrome differentiation. It quantifies the synergistic effect of symptom combinations (such as the contribution of "dry mouth + bad breath + constipation" to food stagnation and intestinal stagnation is higher than that of a single symptom). When the matching degree of a certain pathological direction continues to exceed the threshold, a disease progression warning is triggered (such as the transformation of Yin deficiency to blood heat).

[0078] The eight-dimensional twelve-item framework of this embodiment reduces the pathology classification error rate by 35% (clinical trial data).

[0079] Dynamic adaptability: The model can be automatically optimized based on new case data, and the accuracy of syndrome differentiation increases as the amount of data grows.

[0080] Compatibility of Chinese and Western medicine: supports the fusion analysis of traditional four diagnostic data and laboratory indicators (such as cyst and nodule examination).

[0081] Strong clinical practicality: The output results are actionable pathological directions (such as "damp-heat syndrome"), which directly guide the selection of prescriptions and medications.

[0082] Based on the traditional Chinese medicine theory of yin and yang, qi and blood, this example constructs eight core pathological dimensions: yin deficiency, yin excess, yang deficiency, yang excess, qi deficiency, qi accumulation, blood stasis, and blood dispersion. These dimensions cover the main pathological characteristics of the human body and can comprehensively describe the patient's pathological state.

[0083] The construction of each pathological dimension is based on the classical theories of traditional Chinese medicine to ensure that it can accurately reflect the symptoms of patients under different pathological conditions.

[0084] Based on the eight-dimensional pathology framework, the present invention further refines the pathological direction of each dimension, expanding a single dimension (such as Yang Sheng) into multiple composite pathological directions. For example, the Yang Sheng dimension can be further subdivided into composite pathological directions such as real heat, damp heat, and blood heat, ensuring that the granularity of syndrome differentiation is more refined and closer to actual clinical needs. This refinement of pathological direction makes the syndrome differentiation process more layered, improves the accuracy of syndrome differentiation, and avoids the problem that a single dimension cannot cover all pathological conditions.

[0085] In order to overcome the problem that traditional TCM syndrome differentiation methods rely on experience and subjective judgment, the present invention converts TCM symptoms into numerical parameters and adopts coding and standardization methods to make the symptoms quantitative and operational.

[0086] By setting standardized coding for symptoms, different TCM doctors are ensured to use the same quantitative standards during the diagnosis process, thereby eliminating human differences and improving the consistency and repeatability of the diagnosis. By using the cosine similarity algorithm, the present invention compares the patient's symptom vector with the standard vector of each pathological direction and calculates the matching degree between the symptom vector and the standard vector of the pathological direction. This algorithm can quantify the similarity between the patient's symptoms and different pathological directions, thereby more accurately judging the patient's pathological state. Using cosine similarity helps to avoid the ambiguity and uncertainty of pathological judgment in traditional methods, making the diagnosis more accurate.

[0087] The diagnostic method of this embodiment combines machine learning technology and performs dynamic optimization based on clinical feedback data. In particular, machine learning algorithms such as random forests or neural networks are used to iteratively optimize the model, continuously adjusting the model weights to adapt to new case data. As the amount of data increases, the model can be continuously updated, thereby improving the accuracy and flexibility of syndrome differentiation. This dynamic optimization method breaks through the static limitations of traditional Chinese medicine syndrome differentiation methods, allowing the diagnostic system to be continuously optimized and improved as actual data accumulates.

[0088] This embodiment, by constructing an eight-dimensional twelve-item syndrome differentiation framework and its intelligent diagnosis method, can effectively solve several problems of the conventional traditional Chinese medicine syndrome differentiation method in the prior art and bring about significant technical effects:

[0089] By constructing an eight-dimensional pathology dimension and refining it into twelve pathology directions, the present invention's syndrome differentiation framework offers a refined improvement over the traditional syndrome differentiation system. The combination of the eight-dimensional dimension and twelve pathology directions increases the granularity of syndrome differentiation, enabling a more comprehensive and detailed description of pathological conditions. This refined pathology direction reduces errors in syndrome differentiation, effectively avoiding the misdiagnosis and missed diagnosis problems inherent in traditional methods, and further enhancing the accuracy of syndrome differentiation.

[0090] By quantifying and standardizing TCM symptoms, the present invention transforms TCM syndrome differentiation from reliance on subjective experience into an objective quantitative operation. Symptoms are converted into numerical parameters, which can achieve consistent diagnosis among different doctors, eliminating the influence of subjective differences and personal experience in traditional TCM diagnosis. This quantification method makes the syndrome differentiation process more scientific and systematic, and provides reliable data support for subsequent data analysis and model optimization. Through the introduction of machine learning technology, the TCM syndrome differentiation method has dynamic adaptability. With the continuous accumulation of clinical data, the diagnostic model can be updated and optimized in real time to ensure that the model can always adapt to new case data and continuously improve the accuracy of syndrome differentiation. This dynamic optimization function makes the syndrome differentiation results have a high degree of personalization, which can be adjusted according to the specific situation of the patient, thereby providing patients with personalized treatment plans.

[0091] The intelligent diagnostic method of this embodiment also includes a real-time early warning mechanism. When the matching degree of a specific pathological direction continuously exceeds a preset threshold, the system automatically triggers a warning of disease progression, helping physicians to promptly identify potential pathological changes and implement early intervention. This early warning mechanism allows patients to receive effective monitoring and intervention at the early stages of disease, improving disease prevention effectiveness.

[0092] This embodiment supports the integration and analysis of data from the traditional four diagnostic methods with modern laboratory test data, allowing TCM syndrome differentiation methods to be combined with modern medical methods, further improving the comprehensiveness and accuracy of diagnosis. This integrated diagnostic approach of Chinese and Western medicine helps to leverage the strengths of traditional Chinese medicine while addressing certain shortcomings of modern medicine, forming a more comprehensive diagnosis and treatment system.

[0093] The intelligent diagnostic method of this embodiment can directly provide clinically actionable pathological indicators (such as "damp-heat syndrome") and, combined with the support of modern technology, guide prescription and medication selection, greatly enhancing the practical application value of Traditional Chinese Medicine diagnosis and treatment. In clinical practice, physicians can formulate more accurate treatment plans based on the pathological indicators provided by this method, thereby improving treatment efficacy and patient satisfaction.

[0094] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A TCM intelligent syndrome differentiation system based on the Yin-Yang Qi-Blood theory, characterized by: The system comprises: Symptom data collection module, which is used to extract patient symptom descriptions through natural language processing technology and quantify the frequency, intensity and duration of symptoms into standardized numerical parameters; A symptom-syndrome association matrix construction module is used to establish a mapping relationship between symptoms and syndrome types based on classical Chinese medicine literature and expert experience knowledge bases, and calculate the initial weight value of each symptom in the target syndrome type through a feature selection algorithm; wherein, the establishment of the mapping relationship between symptoms and syndrome types is specifically to construct an eight-dimensional twelve-item syndrome differentiation system, wherein the eight-dimensional pathological dimensions are divided into yin deficiency, yin excess, yang deficiency, yang excess, qi deficiency, qi gathering, blood stasis, and blood dispersion; the twelve pathological directions are divided into: basic pathological directions, including yang deficiency, yin deficiency, water dampness, qi stagnation, qi deficiency, wind, and sweating; compound pathological directions, including excess heat, damp heat, blood heat, phlegm and blood stasis, and food stagnation and intestinal stagnation, and the pathological directions include visible tangible pathological products such as phlegm and blood stasis or food stagnation and intestinal stagnation and invisible body state imbalance, wherein the tangible pathological products include phlegm and blood stasis, food stagnation and intestinal stagnation, as well as water dampness and sweating, and the invisible body state imbalance includes qi deficiency, yang deficiency, qi gathering, yin deficiency, and blood stasis; A dynamic weight optimization module, which is used to input the numerical parameters into a trained machine learning model. The model dynamically adjusts the weights of each symptom using a weighted sum algorithm, wherein the initial weights are set based on Traditional Chinese Medicine theory, and subsequent weights are iteratively updated using a backpropagation algorithm or reinforcement learning combined with real-time symptom data; The syndrome diagnosis output module is used to output the patient's syndrome diagnosis results through a deep neural network, support vector machine or integrated learning model based on the dynamically adjusted weight values ​​and symptom numerical parameters, and generate an explanatory report including the symptom weight distribution.

2. The system according to claim 1, wherein: Methods for quantifying symptoms include: Symptom frequency coding: symptom frequency is divided into "occasional", "intermittent" and "persistent" and assigned a value of 1-3 respectively; Symptom intensity is quantified by using fuzzy set value statistical methods or visual analog scale method to map the symptom intensity into a numerical interval [0,1]; Tongue and pulse parameter extraction: The texture, color, and morphological characteristics of the tongue, as well as the frequency, rhythm, and intensity characteristics of the pulse are extracted through image recognition algorithms, and are normalized with the symptom numerical parameters to form a unified multimodal input vector.

3. The system according to claim 1, wherein: The dynamic weight optimization module is specifically used for: Initialization of TCM expert experience weights, based on the TCM Syndrome Efficacy Evaluation Scale or the Delphi method to determine the initial weights of symptoms; Machine learning weights are iteratively updated, and by comparing the model output with the diagnosis results of traditional Chinese medicine experts, the weights are dynamically adjusted using gradient descent or genetic algorithms; The two-layer optimization of weight distribution assigns weights to the contribution of syndrome factors and syndrome types in a hierarchical manner according to the principle of high frequency and low weight of symptoms and low frequency and high weight of symptoms.

4. The system according to claim 1, wherein: The training process of the machine learning model includes: Data preprocessing: training is performed using a labeled TCM medical record dataset containing symptom-syndrome annotated pairs, and rare syndrome samples are expanded using a generative adversarial network. Model training and validation: cross-validation was used to optimize model hyperparameters. Model output results must meet the following conditions: the accuracy of syndrome differentiation for the main symptom is ≥90%, the accuracy of syndrome differentiation for the secondary symptom is ≥78%, and the F1 value of syndrome type diagnosis is ≥0.6; The model's interpretability is enhanced, and the symptom weights are visualized and analyzed through the attention mechanism, outputting the contribution ranking of each symptom to the target syndrome and the trajectory of weight changes.

5. The system according to claim 1, wherein: The dynamic adjustment of the symptom weight includes: Real-time symptom feedback mechanism: when a patient's symptoms change, the model recalculates the weights based on the latest intensity and frequency of the symptoms; Multimodal feature fusion combines non-symptomatic features such as tongue and pulse conditions, assigns independent weight modules through the attention mechanism, and performs weighted fusion with the symptom weight module to ultimately output a comprehensive syndrome diagnosis result.

6. The system according to claim 1, wherein: The system further comprises: The symptom data collection interface is used to connect to the patient's terminal device or smart wearable device to collect the patient's physiological data, motion data and other related symptom information in real time, and transmit it to the symptom data collection module for analysis.

7. The system according to claim 1, wherein: The algorithms used in the dynamic weight optimization module include random forest, neural network and support vector machine machine learning methods.

8. The system according to claim 1, wherein: The deep neural network is a multi-layer perceptron network, which is used to perform nonlinear mapping based on input symptom data.

9. The system according to claim 1, wherein: The syndrome diagnosis output module also includes: The comprehensive syndrome evaluation module is used to output the optimal diagnostic result based on the matching degree of multiple syndromes and the patient's comprehensive symptom data.

Images