Traditional Chinese medicine lung disease symptom classification system based on Bayesian model

Abstract

The invention relates to the technical field of disease classification, in particular to a traditional Chinese medicine lung disease symptom classification system based on a Bayesian model, and the system comprises a symptom data collection module which collects the lung disease symptom information of a current patient through the symptom record and inquiry data in the traditional Chinese medicine diagnosis process; the model building and training module is used for building a lung disease symptom classification model by adopting a Bayesian classification algorithm based on historical lung disease symptom information and corresponding clinical manifestation data, and the Bayesian algorithm is used for maximizing a posterior probability and combining a conditional dependency relationship among lung disease symptoms; the symptom inference module is used for inferring by utilizing the trained lung disease symptom classification model to obtain a classification result of lung disease symptoms; and a result output module. According to the method, the classification model can be automatically adjusted according to the mutual influence among the symptoms, so that more personalized and accurate diagnosis support is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of symptom classification, and in particular to a TCM lung disease symptom classification system based on a Bayesian model. Background Art

[0002] As the core of traditional Chinese medicine, TCM has a diagnostic system that deeply integrates the comprehensive understanding of human physiology, pathology, symptoms and signs. In the TCM diagnosis process, doctors conduct a comprehensive analysis based on the patient's main symptoms, signs and medical information, combined with the TCM theory of "differentiation and treatment" to determine the cause, location and nature of the disease. The diagnosis of lung disease in TCM usually relies on the manifestation of symptoms, such as cough, sputum, shortness of breath, chest tightness, etc. These symptoms may have different manifestation characteristics in different types of lung disease, and the symptoms are interdependent and intertwined, resulting in traditional diagnostic methods often having certain limitations when facing complex symptom combinations.

[0003] In existing technologies, many diagnostic methods for lung diseases rely on the experience and intuition of clinicians, combined with auxiliary means such as imaging and laboratory tests. However, this method is often affected by the subjective judgment and experience of doctors, and in a large amount of case data, the relationship between symptoms and diseases is difficult to quantitatively analyze in a standardized and systematic way. With the development of big data and artificial intelligence technology, data-driven automatic diagnosis methods have begun to be widely used.

[0004] However, one of the challenges faced by traditional artificial intelligence technology in medical diagnosis is how to deal with the dependencies between complex symptoms. In the classification process of lung disease symptoms, a single model such as the Bayesian model is difficult to effectively consider the complex correlations and potential multi-level structures between symptoms. Summary of the invention

[0005] The present invention provides a TCM lung disease symptom classification system based on a Bayesian model.

[0006] A classification system for TCM lung disease symptoms based on Bayesian model, including:

[0007] The symptom data collection module collects the current patient's lung disease symptom information through symptom records and consultation data during the TCM diagnosis process;

[0008] The model building and training module uses the Bayesian classification algorithm to establish a lung disease symptom classification model based on historical lung disease symptom information and its corresponding clinical manifestation data. The Bayesian algorithm maximizes the posterior probability and combines the conditional dependency relationship between lung disease symptoms to classify TCM lung disease symptoms at multiple levels, and optimizes the accuracy of symptom classification based on the prior probability and posterior probability obtained from model training.

[0009] The symptom inference module uses the trained lung disease symptom classification model to infer the current patient's lung disease symptom information and obtain the classification results of lung disease symptoms;

[0010] The result output module outputs the classification results based on the inference results, providing reference for clinicians and assisting in diagnosis.

[0011] Optionally, the collecting of the current patient's lung disease symptom information specifically includes:

[0012] Through the electronic consultation system, doctors conduct consultations based on TCM diagnostic standards (such as TCM Internal Medicine), record the patient's main symptoms, including cough, sputum, shortness of breath, chest tightness, and fatigue, and record the duration of symptoms, aggravating or relieving factors, and environmental factors that induce or aggravate symptoms based on TCM's syndrome differentiation and treatment method;

[0013] Auxiliary symptoms and signs data related to lung disease are collected through sensors, including body temperature, pulse, and respiratory rate data.

[0014] Optionally, the model building and training module specifically includes:

[0015] Data preprocessing unit: standardize the historical lung disease symptom information and the corresponding clinical manifestation data, and convert the symptom and clinical manifestation data into a format suitable for Bayesian classification algorithm;

[0016] Prior probability estimation unit: calculates the prior probability of various lung disease symptoms according to historical data, wherein the prior probability is estimated based on the symptom frequency distribution in the historical patient data, and is used to reflect the probability of occurrence of each symptom in different types of lung diseases;

[0017] Conditional probability modeling unit: By analyzing the relationship between lung disease symptoms and clinical manifestations in historical data, the conditional probability between symptoms is calculated. Combined with the multi-level interrelationships of TCM lung disease symptoms, the dependency of symptoms is learned through dynamic adaptive learning, and an improved graph model algorithm (multi-level graph neural network) is used to adjust the dependency between symptoms in real time.

[0018] Posterior probability optimization unit: Through the Bayesian classification algorithm, combined with the conditional dependency between prior probability and symptoms, the posterior probability is maximized to generate a trained lung disease symptom classification model, which can accurately classify lung disease symptoms at multiple levels, thereby providing accurate probability distribution for subsequent symptom inference and diagnosis.

[0019] Optionally, the prior probability estimation unit includes statistically analyzing the frequency of each symptom and lung disease type in the historical data to estimate the prior probability P(C i ), where C iFor the i-th type of lung disease (such as "tuberculosis" or "chronic bronchitis"), the calculation is expressed as:

[0020] The prior probability reflects the frequency of occurrence of a certain type of lung disease in the overall cases.

[0021] Optionally, the graph model algorithm in the conditional probability modeling unit includes using a graph neural network (GNN) to deeply learn the multi-level dependencies of lung disease symptoms, construct a graph model including nodes and edges, and each symptom S j In a given lung disease category C i The conditional probability P(S j |C i ) are represented as nodes in the graph, the relationship between symptoms is represented by edges, and the weight of the edge represents the strength of the conditional dependency between symptoms.

[0022] Optionally, the graph model is represented as a weighted undirected graph G=(S,E), where S is a set of nodes, i.e., a set of symptoms, and E is an edge between nodes, i.e., an edge between symptoms, and each edge e kl Connection Symptoms k and symptoms l , the edge weight w kl Representative symptoms k With S l The strength of conditional dependence between them reflects the degree of mutual influence of symptoms. The edge weights are estimated through training data and learned using graph neural networks:

[0023] Where f(S k ,S l ,C n ) is a symptom k and S l In the nth category of lung disease C n The common occurrence frequency in len(C n ) is lung disease category C n The standardized value of the number of symptoms in , Softmax is a mathematical function used in classification problems to convert a set of numerical values ​​into a probability distribution, N means there are a total of N lung disease categories, and n means the nth lung disease category currently being processed.

[0024] Optionally, the method of using the graph neural network to deeply learn the multi-level dependency relationship of lung disease symptoms specifically includes using the graph neural network to update the node S j The characteristic representation of:

[0025] in, Symptom S j The feature representation at the tth layer is Is related to symptoms jThe adjacent symptom set, w jk is the edge weight, b j is the bias term, σ(·) is the activation function,

[0026] Combined with the dependencies learned by the graph neural network, the given symptom S is calculated j The conditional probability of: Among them, P(C n |S j ) is the conditional probability calculated according to the Bayesian algorithm, P(S j ) is a symptom j The prior probability, P(C n ) is lung disease category C n The prior probability, C n Represents the nth lung disease category.

[0027] Optionally, the posterior probability optimization unit includes: using a Bayesian algorithm to combine the prior probability and the conditional probability to maximize the posterior probability P(C i |S1,S2,...,S u ).

[0028] Optionally, the posterior probability P(C i |S1,S2,...,S u ) is calculated as:

[0029] in, For a given lung disease category C n The product of the conditional probabilities of all symptoms under i |S1,S2,...,S u ) is the joint probability of all symptoms, obtained by normalization, and the most likely lung disease category C is selected by maximizing the posterior probability. n As a result of the classification:

[0030] Optionally, the symptom inference module specifically includes:

[0031] Symptom input: The pulmonary disease symptom information S1, S2, ..., S u Input into the symptom inference module, the symptom information includes the patient's main symptoms, auxiliary symptoms and physical signs data;

[0032] Classification model inference: Using the trained lung disease symptom classification model, the Bayesian algorithm is used to calculate the C value of each type of lung disease under given symptom information. n The posterior probability P(C n |S1,S2,...,Su ), select the lung disease category C with the highest probability by maximizing the posterior probability pred .

[0033] Beneficial effects of the present invention:

[0034] The present invention, by combining the Bayesian classification algorithm and the multi-dimensional data of TCM lung disease symptoms, can accurately classify lung disease symptoms. In particular, it introduces a multi-level conditional dependency relationship between symptoms into the Bayesian model, so that the system can further improve the classification accuracy and reliability on the basis of traditional symptom classification, and solves the problem of lack of modeling of complex relationships between symptoms in the prior art. It can automatically adjust the classification model according to the mutual influence between symptoms, thereby providing more personalized and accurate diagnostic support.

[0035] The present invention analyzes the complex relationship between lung disease symptoms and clinical manifestations in the training data and introduces dynamic adaptive adjustment. It can dynamically adjust the conditional probability in the Bayesian model according to the symptom information of different patients, so that the classification model can be optimized as new data is continuously added. It not only improves the classification accuracy of the system, but also can adjust the diagnosis strategy in real time according to the specific symptom changes of the patient, providing doctors with the most accurate auxiliary diagnosis basis.

[0036] The present invention realizes symptom inference and classification through a trained lung disease symptom classification model, which can not only accurately infer the current patient's lung disease type, but also dynamically update the model in combination with the patient's real-time symptoms and medical history information, making the system highly flexible and scalable. In addition, the inference process based on Bayesian theory enables the system to process multi-dimensional and multi-level symptom information, and has high fault tolerance. It can also make reasonable inferences for missing or incomplete information of symptom data, ensuring the stability and availability of classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 A schematic diagram of functional modules of a classification system according to an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of the model building and training module of an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0041] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0042] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0043] like Figure 1-Figure 2 As shown, a TCM lung disease symptom classification system based on a Bayesian model includes:

[0044] The symptom data collection module collects the current patient's lung disease symptom information through symptom records and consultation data during the TCM diagnosis process;

[0045] The model building and training module uses the Bayesian classification algorithm to establish a lung disease symptom classification model based on historical lung disease symptom information and its corresponding clinical manifestation data. The Bayesian algorithm maximizes the posterior probability and combines the conditional dependency relationship between lung disease symptoms to classify TCM lung disease symptoms at multiple levels. It also optimizes the accuracy of symptom classification based on the prior probability and posterior probability obtained from model training.

[0046] The symptom inference module uses the trained lung disease symptom classification model to infer the current patient's lung disease symptom information and obtain the classification results of lung disease symptoms;

[0047] The result output module outputs the classification results based on the inference results, providing reference for clinicians and assisting in diagnosis.

[0048] The information collected about the current patient's lung disease symptoms specifically includes:

[0049] Through the electronic consultation system, doctors conduct consultations based on TCM diagnostic standards (such as TCM Internal Medicine), record the patient's main symptoms, including cough, sputum, shortness of breath, chest tightness, and fatigue, and record the duration of symptoms, aggravating or relieving factors, and environmental factors that induce or aggravate symptoms based on TCM's syndrome differentiation and treatment method;

[0050] Auxiliary symptoms and signs data related to lung disease are collected through sensors, including body temperature, pulse, and respiratory rate data.

[0051] The model building and training modules specifically include:

[0052] Data preprocessing unit: standardize the historical lung disease symptom information and the corresponding clinical manifestation data, and convert the symptom and clinical manifestation data into a format suitable for Bayesian classification algorithm, for example, converting symptom information into discrete feature values ​​to ensure data consistency and processability;

[0053] Prior probability estimation unit: Calculate the prior probability of various lung disease symptoms based on historical data. The prior probability is estimated based on the symptom frequency distribution in historical patient data to reflect the probability of each symptom occurring in different types of lung diseases.

[0054] Conditional probability modeling unit: By analyzing the relationship between lung disease symptoms and clinical manifestations in historical data, calculating the conditional probability between symptoms, combining the multi-level interrelationships of TCM lung disease symptoms, learning through dynamic adaptive symptom dependencies, and using an improved graph model algorithm (multi-level graph neural network), the dependencies between symptoms are adjusted in real time, overcoming the limitation of fixed symptom dependencies in traditional models, thereby more accurately reflecting the changing patterns and mutual influences of symptoms in different patients and clinical scenarios;

[0055] Posterior probability optimization unit: Through the Bayesian classification algorithm, combined with the conditional dependency between prior probability and symptoms, the posterior probability is maximized to generate a trained lung disease symptom classification model, which can accurately classify lung disease symptoms at multiple levels, thereby providing accurate probability distribution for subsequent symptom inference and diagnosis.

[0056] The prior probability estimation unit includes statistical analysis of the frequency of each symptom and lung disease type in the historical data, and estimates the prior probability P(C i ), where C i For the i-th type of lung disease (such as "tuberculosis" or "chronic bronchitis"), the calculation is expressed as:

[0057] The prior probability reflects the frequency of occurrence of a certain type of lung disease in the overall cases.

[0058] The graph model algorithm in the conditional probability modeling unit includes using the graph neural network (GNN) to deeply learn the multi-level dependencies of lung disease symptoms and construct a graph model, including nodes and edges. j In a given lung disease category C i The conditional probability P(S j |C i ) are represented as nodes in the graph, the relationship between symptoms is represented by edges, and the weight of the edge represents the strength of the conditional dependency between symptoms.

[0059] The graph model is represented as a weighted undirected graph G = (S, E), where S is the set of nodes, i.e., the set of symptoms, and E is the edge between nodes, i.e., the edge between symptoms. Each edge e kl Connection Symptoms k and symptoms l , the edge weight w kl Representative symptoms k With S l The strength of conditional dependence between them reflects the degree of mutual influence of symptoms. The edge weights are estimated through training data and learned using graph neural networks:

[0060] Where f(S k ,S l ,C n ) is a symptom k and S l In the nth category of lung disease C n The common occurrence frequency in len(C n ) is lung disease category C n The standardized value of the number of symptoms in , Softmax is a mathematical function used in classification problems to convert a set of numerical values ​​into a probability distribution, N means there are a total of N lung disease categories, and n means the nth lung disease category currently being processed.

[0061] Using graph neural network to deeply learn the multi-level dependency relationship of lung disease symptoms specifically includes using graph neural network to update node S j The characteristic representation of:

[0062] in, Symptom S j The feature representation at the tth layer is Is related to symptoms j The adjacent symptom set, w jk is the edge weight, b j is the bias term, σ(·) is the activation function,

[0063] Bayesian conditional probability calculation: Combine the dependencies learned by the graph neural network to calculate the probability of a given symptom S jThe conditional probability of: Among them, P(C n |S j ) is the conditional probability calculated according to the Bayesian algorithm, P(S j ) is a symptom j The prior probability, P(C n ) is lung disease category C n The prior probability, C n Represents the nth lung disease category, optimizes the conditional dependency between symptoms, and further improves the classification accuracy.

[0064] The posterior probability optimization unit includes: using the Bayesian algorithm to combine the prior probability and conditional probability to maximize the posterior probability P(C i |S1,S2,...,S u ).

[0065] The posterior probability P(C i |S1,S2,...,S u ) is calculated as:

[0066] in, For a given lung disease category C n The product of the conditional probabilities of all symptoms under i |S1,S2,...,S u ) is the joint probability of all symptoms, obtained by normalization, and the most likely lung disease category C is selected by maximizing the posterior probability. n As a result of the classification:

[0067] The symptom inference module specifically includes:

[0068] Symptom input: The pulmonary disease symptom information S1, S2, ..., S u Input into the symptom inference module, the symptom information includes the patient's main symptoms, auxiliary symptoms and physical signs data;

[0069] Classification model inference: Using the trained lung disease symptom classification model, the Bayesian algorithm is used to calculate the C value of each type of lung disease under given symptom information. n The posterior probability P(C n |S1,S2,...,S u ), select the lung disease category C with the highest probability by maximizing the posterior probability pred :

[0070] Based on the inference results, output the most likely lung disease category C pred And provide diagnostic advice to clinicians.

[0071] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0072] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A TCM lung disease symptom classification system based on a Bayesian model, characterized in that: include: The symptom data collection module collects the current patient's lung disease symptom information through symptom records and consultation data during the TCM diagnosis process; The model building and training module uses the Bayesian classification algorithm to establish a lung disease symptom classification model based on historical lung disease symptom information and its corresponding clinical manifestation data. The Bayesian algorithm maximizes the posterior probability and combines the conditional dependency relationship between lung disease symptoms to classify TCM lung disease symptoms at multiple levels, and optimizes the accuracy of symptom classification based on the prior probability and posterior probability obtained from model training. The symptom inference module uses the trained lung disease symptom classification model to infer the current patient's lung disease symptom information and obtain the classification results of lung disease symptoms; The result output module outputs the classification results based on the inference results, providing reference for clinicians and assisting in diagnosis.

2. A TCM lung disease symptom classification system based on a Bayesian model according to claim 1, characterized in that: The collecting of the current patient's lung disease symptom information specifically includes: Through the electronic consultation system, doctors record the patient's main symptoms, including cough, sputum, shortness of breath, chest tightness, fatigue, and the duration of symptoms, aggravating or relieving factors, and environmental factors that induce or aggravate symptoms according to TCM diagnostic standards; Auxiliary symptoms and signs data related to lung disease are collected through sensors, including body temperature, pulse, and respiratory rate data.

3. A TCM lung disease symptom classification system based on a Bayesian model according to claim 1, characterized in that: The model building and training module specifically includes: Data preprocessing unit: standardize the historical lung disease symptom information and the corresponding clinical manifestation data, and convert the symptom and clinical manifestation data into a format suitable for Bayesian classification algorithm; Prior probability estimation unit: calculates the prior probability of various lung disease symptoms according to historical data, wherein the prior probability is estimated based on the symptom frequency distribution in the historical patient data, and is used to reflect the probability of occurrence of each symptom in different types of lung diseases; Conditional probability modeling unit: By analyzing the relationship between lung disease symptoms and clinical manifestations in historical data, the conditional probability between symptoms is calculated. Combined with the multi-level interrelationships of TCM lung disease symptoms, the dependency relationship between symptoms is learned through dynamic adaptive symptom dependency, and the dependency relationship between symptoms is adjusted in real time using a graph model algorithm. Posterior probability optimization unit: Through the Bayesian classification algorithm, combined with the conditional dependency between the prior probability and the symptoms, the posterior probability is maximized to generate a trained lung disease symptom classification model, which can accurately classify lung disease symptoms at multiple levels.

4. A TCM lung disease symptom classification system based on a Bayesian model according to claim 3, characterized in that: The prior probability estimation unit includes statistically analyzing the frequency of each symptom and lung disease type in the historical data, estimating the prior probability P(C i ), where C i For the i-th type of lung disease, the calculation is expressed as: The prior probability reflects the frequency of occurrence of a certain type of lung disease in the overall cases.

5. A TCM lung disease symptom classification system based on a Bayesian model according to claim 4, characterized in that: The graph model algorithm in the conditional probability modeling unit includes using a graph neural network to deeply learn the multi-level dependencies of lung disease symptoms, constructing a graph model including nodes and edges, and each symptom S j In a given lung disease category C i The conditional probability P(S j |C i ) are represented as nodes in the graph, the relationship between symptoms is represented by edges, and the weight of the edge represents the strength of the conditional dependency between symptoms.

6. A TCM lung disease symptom classification system based on a Bayesian model according to claim 5, characterized in that: The graph model is represented as a weighted undirected graph G = (S, E), where S is a node set, i.e., a symptom set, and E is an edge between nodes. kl Connection Symptoms k and symptoms l , the edge weight w kl Representative symptoms k With S l The conditional dependency strength between them, the edge weights are estimated through training data, and graph neural network learning is adopted: Where f(S k ,S l ,C n ) is a symptom k and S l In the nth category of lung disease C n The common occurrence frequency in len(C n ) is lung disease category C n The standardized value of the number of symptoms in , Softmax is a mathematical function used in classification problems to convert a set of numerical values ​​into a probability distribution, N means there are a total of N lung disease categories, and n means the nth lung disease category currently being processed.

7. A TCM lung disease symptom classification system based on a Bayesian model according to claim 6, characterized in that: The method of using the graph neural network to deeply learn the multi-level dependency relationship of lung disease symptoms specifically includes using the graph neural network to update the node S j The characteristic representation of: in, Symptom S j The feature representation at the tth layer is Is related to symptoms j The adjacent symptom set, w jk is the edge weight, b j is the bias term, σ(·) is the activation function, Combined with the dependencies learned by the graph neural network, the given symptom S is calculated j The conditional probability of: Among them, P(C n |S j ) is the conditional probability calculated according to the Bayesian algorithm, P(S j ) is a symptom j The prior probability, P(C n ) is lung disease category C n The prior probability, C n Represents the nth lung disease category.

8. A TCM lung disease symptom classification system based on a Bayesian model according to claim 7, characterized in that: The posterior probability optimization unit comprises: using the Bayesian algorithm to combine the prior probability and the conditional probability to maximize the posterior probability P(C i |S1,S2,...,S u ).

9. A TCM lung disease symptom classification system based on a Bayesian model according to claim 8, characterized in that: The posterior probability P(C i |S1,S2,...,S u ) is calculated as: in, For a given lung disease category C n The product of the conditional probabilities of all symptoms under i |S1,S2,...,S u ) is the joint probability of all symptoms, obtained by normalization, and the most likely lung disease category C is selected by maximizing the posterior probability. n As a result of classification: C pred = arg maxC n P(C n |S1,S2,...,S u ).

10. A TCM lung disease symptom classification system based on a Bayesian model according to claim 9, characterized in that: The symptom inference module specifically includes: Symptom input: The pulmonary disease symptom information S1, S2, ..., S u Input into the symptom inference module, the symptom information includes the patient's main symptoms, auxiliary symptoms and physical signs data; Classification model inference: Using the trained lung disease symptom classification model, the Bayesian algorithm is used to calculate the C value of each type of lung disease under given symptom information. n The posterior probability P(C n |S1,S2,...,S u ), select the lung disease category C with the highest probability by maximizing the posterior probability pred .

Images