Intelligent disease diagnosis and differential diagnosis system based on knowledge graph

Abstract

The invention relates to the technical field of medical diagnosis processing, and discloses an intelligent disease diagnosis and differential diagnosis system based on a knowledge graph, and the system comprises a data input module which is used for receiving and standardizing clinical symptoms, signs, laboratory examination data and historical medical record data of a patient; the knowledge graph construction and updating module is used for constructing and updating a knowledge graph of diseases and symptoms according to the medical literature and the clinical data, and the knowledge graph automatically extracts an incidence relation between the symptoms and the diseases from the medical literature through a natural language processing technology; and the reasoning and diagnosis module is used for performing intelligent disease diagnosis and differential diagnosis. According to the method, by optimizing reasoning path selection and information gain calculation, the path with the most information content is selected from multiple reasoning paths for diagnosis reasoning, and the most representative path is selected by calculating the correlation degree between each symptom and the disease.

Description

Technical Field

[0001] The present invention relates to the field of medical diagnosis, and in particular to a disease intelligent diagnosis and differential diagnosis system based on knowledge graphs. Background Art

[0002] In the medical field, the development of intelligent diagnostic systems has long been a core research direction in medical artificial intelligence. With the growth of data volumes and advancements in computing power, many traditional medical diagnostic methods are gradually being replaced by intelligent diagnostic systems. Traditional medical diagnostic methods typically rely on physician experience and simple symptom analysis. This approach has numerous limitations, including human interference, low efficiency, and insufficient capacity to handle complex cases. This is particularly true for complex diseases, rare conditions, or the coexistence of multiple illnesses, where traditional methods often struggle to provide accurate and timely diagnoses.

[0003] With advances in computer technology and artificial intelligence, intelligent diagnostic systems have gradually been introduced into clinical medicine and have achieved considerable success. These systems are typically trained on large amounts of historical medical records and utilize machine learning, rule-based reasoning, or other AI techniques to analyze the correlation between symptoms and diseases, thereby assisting doctors in making diagnoses. However, existing intelligent diagnostic systems still face numerous challenges, particularly in effectively handling complex symptom-disease relationships, coping with data incompleteness and inconsistency, and providing sufficiently explainable reasoning processes. Summary of the Invention

[0004] In order to make up for the above shortcomings, the present invention provides a disease intelligent diagnosis and differential diagnosis system based on knowledge graph, which aims to improve the problem that the existing technology cannot correctly identify the relationship between complex symptoms and diseases.

[0005] In a first aspect, the present invention provides the following technical solution: a disease intelligent diagnosis and differential diagnosis system based on a knowledge graph, the system comprising:

[0006] A data input module is used to receive and standardize the patient's clinical symptoms, physical signs, laboratory test data and historical medical record data;

[0007] A knowledge graph construction and update module is used to construct and update a knowledge graph of diseases and symptoms based on medical literature and clinical data. The knowledge graph automatically extracts the association between symptoms and diseases from medical literature using natural language processing technology;

[0008] A reasoning and diagnosis module, configured to perform intelligent diagnosis and differential diagnosis of diseases based on the patient data received by the data input module and the knowledge graph provided by the knowledge graph construction and update module;

[0009] The inference result display and explainability module is used to visualize the inference results and show each step of the inference path, each decision basis, and the information flow transmission process;

[0010] The feedback and optimization module is used to provide feedback based on the doctor's diagnosis results of the system to further optimize the knowledge graph and reasoning process.

[0011] Preferably, the reasoning and diagnosis module includes:

[0012] Information gain and optimal path selection module, used to calculate the information gain between symptoms and diseases and select the reasoning path with the largest information gain;

[0013] A graph neural network reasoning optimization module is used to optimize the reasoning path based on the knowledge graph through multiple rounds of information transmission and update the representation of each node;

[0014] The game theory decision optimization module is used to optimize the reasoning path and select the optimal diagnostic result based on the Nash equilibrium principle in game theory.

[0015] In a second aspect, the present invention provides the following technical solution: a method for intelligent disease diagnosis and differential diagnosis based on a knowledge graph, the method comprising:

[0016] S1. Receive and standardize the patient's clinical data;

[0017] S2. Build and update the knowledge graph of diseases and symptoms based on natural language processing technology;

[0018] S3. Calculate the information gain between symptoms and diseases and select the reasoning path with the largest information gain;

[0019] S4, reasoning optimization of symptom and disease nodes based on graph neural network;

[0020] S5. Optimize the reasoning path based on the Nash equilibrium principle in game theory and select the optimal diagnostic result;

[0021] S6. Demonstrate the reasoning process and provide each step of the reasoning process.

[0022] Preferably, the clinical data include symptoms, signs, laboratory test data and historical medical records.

[0023] Preferably, the S2 knowledge graph construction process includes:

[0024] S21. Extract the association between diseases and symptoms from medical literature;

[0025] S22. Use natural language processing technology to process sentences and words of text;

[0026] S23, extracting disease and symptom entities from the text and establishing preliminary disease-symptom associations;

[0027] S24, constructing a disease symptom map and assigning weights to the association relationships;

[0028] S25. Dynamically update the atlas content based on new data and medical research results.

[0029] Preferably, the S3 information gain calculation process includes:

[0030] S31. Calculate the mutual information between each symptom and the potential disease to determine the strength of the association between the symptom and the disease;

[0031] S32. Based on the calculated mutual information, select the symptom-disease association path with the largest information gain for inference;

[0032] S33, assigning information gain weights to the relationships between each symptom and disease to optimize the reasoning path;

[0033] S34. Based on the selected reasoning path, analyze the sequence or combination of symptoms and select the path with the most representativeness and maximum information gain for subsequent reasoning.

[0034] Preferably, the S4 graph neural network reasoning optimization process includes:

[0035] S41. Represent symptom and disease nodes as nodes in a graph structure, define the relationships between nodes as edges of the graph, and transfer node information based on a graph neural network model.

[0036] S42. Update the representation of each node through a multi-round iterative information transmission mechanism;

[0037] S43. Calculate the relationship strength between each symptom and the disease based on the updated node representation, and select the reasoning path that best reflects the disease diagnosis information;

[0038] S44. Based on the output results of the graph neural network, select the optimal path for subsequent reasoning.

[0039] Preferably, the S5 game theory optimization process includes:

[0040] S51. Establish a game model between symptom nodes and disease nodes, consider each symptom node and disease node as a participant in the game, and define their strategy set;

[0041] S52. Based on the Nash equilibrium principle in game theory, calculate the optimal strategy for each symptom and disease node, that is, choose the strategy that can maximize its own benefits;

[0042] S53. By calculating the Nash equilibrium of each node in the game process, the optimal reasoning path is selected, in which the strategy equilibrium of the symptom node and the disease node maximizes the accuracy of diagnosis;

[0043] S54. Based on the result of Nash equilibrium, the final diagnostic path is selected, and the optimal diagnostic result is finally obtained.

[0044] In the third aspect, the invention provides the following technical solution: a computer device comprising a memory, a processor, and a computer program stored on the memory and runnable on the processor; when the processor executes the computer program, it implements the above-mentioned method of intelligent disease diagnosis and differential diagnosis based on knowledge graph.

[0045] In a fourth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for intelligent disease diagnosis and differential diagnosis based on a knowledge graph.

[0046] The present invention has the following beneficial effects:

[0047] 1. In the present invention, by optimizing the reasoning path selection and information gain calculation, the most informative path is selected from multiple reasoning paths for diagnostic reasoning. By calculating the correlation between each symptom and the disease and selecting the most representative path, the accuracy of the diagnostic results is maximized and the possibility of misdiagnosis and missed diagnosis is reduced.

[0048] 2. In the present invention, the problem of untimely updates of traditional knowledge graphs is solved by automatically extracting the latest correlations between symptoms and diseases from medical literature and dynamically updating the knowledge graph using natural language processing technology. In this way, the system can obtain the latest medical research results in a timely manner, thereby ensuring that the information in the disease diagnosis process always remains accurate and up-to-date.

[0049] 3. In the present invention, by combining graph neural networks with optimized calculations of information gain, the optimal diagnostic path can be efficiently selected among a large number of symptom and disease nodes. By continuously optimizing the information flow transmission and node representation in the reasoning process, unnecessary calculations and redundant reasoning steps are reduced, the efficiency of the diagnostic process is improved, and the system can provide high-precision diagnostic results in a short time. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a method for intelligent disease diagnosis and differential diagnosis based on knowledge graph proposed by the present invention;

[0051] Figure 2 This is a system architecture diagram for intelligent disease diagnosis and differential diagnosis based on knowledge graph proposed in the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0053] Example 1

[0054] Reference Figure 1 In a first embodiment of the present invention, the present invention provides a disease intelligent diagnosis and differential diagnosis system based on a knowledge graph, the system comprising:

[0055] A data input module is used to receive and standardize the patient's clinical symptoms, physical signs, laboratory test data and historical medical record data;

[0056] The knowledge graph construction and update module is used to build and update the knowledge graph of diseases and symptoms based on medical literature and clinical data. The knowledge graph automatically extracts the association between symptoms and diseases from medical literature through natural language processing technology;

[0057] The reasoning and diagnosis module is used to perform intelligent disease diagnosis and differential diagnosis based on the patient data received by the data input module and the knowledge graph provided by the knowledge graph construction and update module; it includes:

[0058] Information gain and optimal path selection module, used to calculate the information gain between symptoms and diseases and select the reasoning path with the largest information gain;

[0059] The graph neural network reasoning optimization module is used to optimize the reasoning path through multiple rounds of information transmission based on the knowledge graph and update the representation of each node;

[0060] The game theory decision optimization module is used to optimize the reasoning path and select the optimal diagnostic result based on the Nash equilibrium principle in game theory.

[0061] The inference result display and explainability module is used to visualize the inference results and show each step of the inference path, each decision basis, and the information flow transmission process;

[0062] The feedback and optimization module is used to provide feedback based on the doctor's diagnosis results of the system to further optimize the knowledge graph and reasoning process.

[0063] Specifically, at the initial stage of the computational process, the system first receives the patient's clinical symptoms, physical signs, laboratory test data, and historical medical records. Using natural language processing techniques, it performs word segmentation and entity recognition on the raw text data. It then converts data from various sources into structured information using standardized coding (such as ICD-10) to ensure data consistency and standardization. Subsequently, based on medical literature and clinical data, the system automatically extracts the correlations between symptoms and diseases through natural language processing, constructing and dynamically updating a knowledge graph of diseases and symptoms.

[0064] Whenever new medical literature or clinical data provides new information, the atlas is incrementally updated to ensure the real-time and accuracy of the knowledge base. Next, the system enters the inference and diagnosis module. It first calculates information gain to assess the strength of the association between each symptom and the underlying disease, selecting the path with the highest information gain for inference. During this process, using the formula for calculating mutual information, the system models the joint probability between each symptom and the underlying disease, selecting the symptom-disease path with the highest information gain, ensuring the inference process is as concise as possible and highly diagnostically relevant.

[0065] The system then optimizes the symptom and disease nodes using a graph neural network. Based on the graph neural network model, the node representations are updated through multiple rounds of information transfer, iteratively optimizing the relationship between symptoms and diseases. In each round, the graph neural network updates the node state based on the current node representation and the associated information of adjacent nodes, thereby more comprehensively capturing the complex relationship between diseases and symptoms.

[0066] Furthermore, during the graph neural network optimization process, the representation of each node is continuously refined to ensure that the final reasoning path encompasses the most representative symptom-disease associations. Next, the game theory decision optimization module further optimizes the reasoning path based on the Nash equilibrium principle. Through the game model, the system sets each symptom and disease as a participant in the game, defines a strategy set and a payoff function, and uses an iterative method to calculate the optimal strategy for each node. This allows each symptom and disease node to select the strategy that maximizes its payoff in the reasoning path, ultimately achieving the globally optimal diagnostic path.

[0067] During this process, the system comprehensively considers the interplay between symptoms and diseases based on the payoff function of each node to ensure the optimal diagnostic path. After reasoning is complete, the system displays the results through a visualization module, providing detailed visualization of each step of the reasoning path, the decision basis, and the information flow. This allows doctors to clearly understand the system's decision-making logic and enhance their confidence in the diagnosis. Finally, the feedback and optimization module receives diagnostic feedback from doctors and dynamically optimizes the system's reasoning process and knowledge graph.

[0068] Example 2:

[0069] Reference Figure 2 In a second embodiment of the present invention, the present invention provides a disease intelligent diagnosis and differential diagnosis method based on a knowledge graph, the method comprising:

[0070] S1. Receive and standardize the patient's clinical data, including symptoms, signs, laboratory test data, and medical history;

[0071] Specifically, in the implementation process of the present invention, the first step is S1, which is to receive and standardize the patient's clinical data. This clinical data includes symptoms, signs, laboratory test results, and historical medical history. The key task here is to ensure that all input data can be provided to subsequent processing modules in a unified and standardized format. This process is very important because the accuracy and standardization of the input data directly affect the reliability of the final diagnosis.

[0072] First, the patient's symptoms, signs, and laboratory data are usually presented in text, tables, or free formats. In order to be able to perform computational processing, these raw data need to be standardized. In this step, natural language processing technology is used to extract specific information about symptoms and signs, and each symptom and sign is coded according to internationally standardized medical coding systems such as ICD-10 (International Classification of Diseases) and SNOMED-CT (Clinical Terminology System). For example, if a doctor records the symptom of "chest pain", the system will automatically recognize it and convert it into the corresponding standardized code. For example, the code for "chest pain" in ICD-10 is R07.9. This ensures the consistency of symptom data and facilitates subsequent reasoning and diagnosis.

[0073] Next, laboratory test data is often stored in a table format, containing values such as blood sugar, blood pressure, and cholesterol. These numerical data need to be converted into specific numerical ranges during standardization and normalized through calculations or standardization methods to make the test results of different laboratories comparable. For example, blood sugar values may be affected by different laboratory equipment and methods. The standardization process ensures that these results can be processed and used uniformly, eliminating the influence of different data sources.

[0074] Furthermore, the patient's historical medical history also needs to be standardized. This information often includes a patient's past diagnosis history, treatment plans, and known disease risks. This data must not only be extracted to identify relevant medical history, but also converted into structured data so that the system can efficiently extract historical medical history relevant to current symptoms. For example, the system can automatically identify and extract whether a patient has previously had hypertension, diabetes, or other conditions, so that it can be correlated with current symptoms and examination results.

[0075] The standardized data will enter the knowledge graph construction and update module, as well as the reasoning and diagnosis module.

[0076] S2. Build and update a knowledge graph of diseases and symptoms based on natural language processing technology; including:

[0077] S21. Extract the association between diseases and symptoms from medical literature;

[0078] S22. Use natural language processing technology to process sentences and words of text;

[0079] S23, extracting disease and symptom entities from the text and establishing preliminary disease-symptom associations;

[0080] S24, constructing a disease symptom map and assigning weights to the association relationships;

[0081] S25. Dynamically update the atlas content based on new data and medical research results.

[0082] Specifically, first, step S21 extracts the association between diseases and symptoms from medical literature. Medical literature contains a large amount of research data on the relationship between diseases and symptoms. Manually extracting this information is very time-consuming and prone to errors. Therefore, this solution uses natural language processing (NLP) technology to process medical literature. By identifying sentences related to diseases and symptoms in the literature, syntactic analysis and semantic analysis techniques are used to extract clear disease-symptom relationships. For example, if a study mentions that "high blood pressure is often accompanied by headaches", the system will automatically identify the association between "high blood pressure" and "headaches" and convert this association into an edge in the knowledge graph, indicating that the two are related.

[0083] In step S22, this solution uses natural language processing technology to perform sentence and word segmentation on the text. This step is a basic operation in NLP, which helps this solution break down long texts in medical literature into separate sentences and words that are easier to handle. Through sentence segmentation, this solution divides a large continuous medical paper into smaller units, which can be further analyzed separately. The word segmentation operation splits each sentence into words, which facilitates the identification of medical terms, symptoms, disease names, etc. For terms specific to the medical field, the word segmentation tool will be specially trained so that it can correctly identify proper nouns such as "coronary heart disease" and "angina pectoris" instead of breaking them down into unrelated words.

[0084] Next, in step S23, the system extracts the disease and symptom entities in the text and establishes preliminary disease-symptom associations. This stage is very critical, as the system needs to accurately identify the entities of diseases and symptoms from medical literature and then understand the relationship between them. For example, for the text "Diabetic patients often experience thirst symptoms", the system will identify "diabetes" and "thirst" as two entities and automatically associate them. Through entity recognition and relationship extraction, the system can construct this information into a preliminary disease-symptom graph. This graph can be viewed as a directed graph, with diseases and symptoms as nodes and the associations between diseases and symptoms as edges, initially forming a network that reflects the relationship between diseases and symptoms.

[0085] In step S24, the system further constructs a disease-symptom map and assigns a weight to each association. The weight reflects the strength of the relationship between the disease and the symptom. For example, the association between some diseases and symptoms may be more direct (such as "cold" and "cough"), while others may be indirect or weaker (such as "diabetes" and "weight change").

[0086] In the graph, these relationships are weighted based on their co-occurrence frequency in medical literature or the strength of expert evaluations. For example, if "cough" frequently appears in literature related to "cold," this relationship will be weighted higher. This step, by assigning weights to related relationships, makes the graph more refined and accurate, facilitating subsequent reasoning and diagnosis.

[0087] Step S25 is to dynamically update the graph content based on new data and medical research results. Knowledge in the medical field is constantly evolving, and new discoveries and new research results will continue to emerge. In order to ensure the timeliness and accuracy of the knowledge graph, the system will regularly extract new information from the latest medical literature through incremental learning or automated crawlers to update the content in the graph. For example, when new research shows that "depression" and "sleep disorders" have a closer relationship, the system will automatically extract and update this new related information and add relevant nodes and edges to the graph. This update process is automated, which can ensure that the graph can continue to keep pace with the medical frontier during long-term use.

[0088] S3. Calculate the information gain between symptoms and diseases and select the inference path with the largest information gain; including:

[0089] S31. Calculate the mutual information between each symptom and the potential disease to determine the strength of the association between the symptom and the disease;

[0090] S32. Based on the calculated mutual information, select the symptom-disease association path with the largest information gain for inference;

[0091] S33, assigning information gain weights to the relationships between each symptom and disease to optimize the reasoning path;

[0092] S34. Based on the selected reasoning path, analyze the sequence or combination of symptoms and select the path with the most representativeness and maximum information gain for subsequent reasoning.

[0093] Specifically, first, in step S31, the system calculates the mutual information between each symptom and the underlying disease. This step is based on the core concept of information theory, namely mutual information. Mutual information is used to measure the dependency between two variables (in this case, symptoms and diseases). The formula is as follows: ;

[0094] here, represents the mutual information between symptoms and diseases, is the joint probability distribution of symptoms and diseases, and are the marginal probability distributions of symptoms and diseases, respectively. By calculating the mutual information, this approach can assess the strength of the association between symptoms and diseases. A larger mutual information indicates a closer relationship between symptoms and diseases, and vice versa. For example, if a patient exhibits high fever and cough, and these two symptoms are highly correlated with pneumonia, then their mutual information will also be large.

[0095] Next, in step S32, the system selects the symptom-disease association path with the highest information gain based on the calculated mutual information. In this step, the system no longer analyzes all possible symptom-disease pairings one by one, but instead prioritizes those symptom-disease relationship paths with high mutual information. This approach improves diagnostic efficiency by reducing the number of symptom-disease pairs that need to be analyzed. By selecting the path with the highest information gain, the system maximizes the information gained from each symptom, helping doctors quickly identify possible diseases.

[0096] In step S33, the system assigns information gain weights to the relationships between each symptom and disease. The information gain weights directly determine the priority of the reasoning path. If the relationship weights between certain symptoms and certain diseases are higher, these paths will be given priority. This weight assignment process is based on the following calculation formula: Information gain weight ;

[0097] This formula shows that the information gain weight is obtained by comparing the mutual information of each pair of symptoms and diseases with the sum of the total mutual information of all possible relationships. The purpose of this is to ensure that the most important symptom-disease paths are prioritized during the inference process by quantifying the importance of each relationship.

[0098] This allocation method helps the system avoid invalid reasoning on less relevant paths, further improving the accuracy and efficiency of reasoning.

[0099] Finally, in step S34, based on the selected inference path, the system analyzes the order or combination of symptoms and selects the most representative path with the greatest information gain for subsequent inference. At this stage, the system does not simply select a single path linking symptoms to the disease; instead, it comprehensively considers the order and combination of symptoms.

[0100] Different symptom combinations can lead to different reasoning paths and results. Choosing a path with the right order and maximum information gain will improve diagnostic accuracy. For example, cough and high fever are usually typical symptoms of pneumonia, but if shortness of breath appears first, the reasoning path may need to be adjusted to prioritize respiratory-related diseases.

[0101] This stage of reasoning is not based solely on the intensity of individual symptoms, but also considers the relationships and combined effects of symptoms. In this way, the system can gradually screen out the most likely diseases, ensuring that the reasoning path is more consistent with actual clinical diagnostic rules.

[0102] S4. Optimize reasoning of symptom and disease nodes based on graph neural networks; including:

[0103] S41. Represent symptom and disease nodes as nodes in a graph structure, define the relationships between nodes as edges of the graph, and transfer node information based on a graph neural network model.

[0104] S42. Update the representation of each node through a multi-round iterative information transmission mechanism;

[0105] S43. Calculate the relationship strength between each symptom and the disease based on the updated node representation, and select the reasoning path that best reflects the disease diagnosis information;

[0106] S44. Based on the output results of the graph neural network, select the optimal path for subsequent reasoning.

[0107] Specifically, step S41 first represents symptom and disease nodes as nodes in a graph structure and defines the relationships between nodes as edges. A core concept of graph neural networks is to transform complex relationships into a graph structure, in which nodes represent symptoms or diseases, and edges represent the relationships between them. For example, the symptom "headache" may be associated with the disease "migraine," forming an edge in the graph.

[0108] In this way, the relationship between symptoms and diseases is clearly represented and can be further reasoned and optimized through graph neural networks. In order to effectively transmit this information, it is first necessary to initialize the representation of each node in the graph, which is usually achieved by mapping symptom and disease codes into vectors.

[0109] Next, in step S42, the system updates the representation of each node through a multi-round iterative information transmission mechanism. In a graph neural network, the representation of a node is not only determined by the characteristics of the node itself, but also influenced by its neighboring nodes. In each iteration, the node transmits information to its neighboring nodes, and the neighboring nodes also update their own states after receiving the information. This process is implemented using the following update formula:

[0110] ;

[0111] In the formula, represents the representation of node v in the kth iteration, represents the neighbor nodes of node v, represents the relationship feature of edge vu, is a fusion function, is the learning weight of the kth round, is the activation function.

[0112] Through this iterative process, the representation of each node is continuously optimized, making it increasingly accurate in reflecting the complex relationships between symptoms and diseases. Multiple rounds of iteration help capture the deep connections between nodes and gradually improve the reliability of the inference results.

[0113] In step S43, the system calculates the strength of the relationship between each symptom and disease based on the updated node representations. This calculation is based on the similarity of the node representations and the strength of the connections. In graph neural networks, a node's representation reflects its relative relationship to surrounding nodes, and the strength of this relationship can be calculated using the inner product or similarity metric of the node representations.

[0114] For example, when calculating the strength of the relationship between the symptom "headache" and the disease "migraine," the system evaluates the similarity of the node representations of the two. In this way, the system can quantify the strength of the association between each symptom and the disease and select the inference path that best reflects the disease diagnosis information.

[0115] Finally, in step S44, the system selects the optimal path for subsequent reasoning based on the output of the graph neural network. After multiple rounds of iteration and node representation updates, the system can accurately identify the optimal reasoning path between symptoms and diseases. This path is typically selected based on the similarity and relationship strength of the node representations, ensuring that the final diagnostic path best reflects the patient's actual condition. During the reasoning process, the system calculates the shortest distance or minimum loss for each path and selects an optimal path that meets the diagnostic requirements, further improving diagnostic accuracy.

[0116] S5. Optimize the reasoning path based on the Nash equilibrium principle in game theory and select the optimal diagnostic result; including:

[0117] S51. Establish a game model between symptom nodes and disease nodes, consider each symptom node and disease node as a participant in the game, and define their strategy set;

[0118] S52. Based on the Nash equilibrium principle in game theory, calculate the optimal strategy for each symptom and disease node, that is, choose the strategy that can maximize its own benefits;

[0119] S53. By calculating the Nash equilibrium of each node in the game process, the optimal reasoning path is selected, in which the strategy equilibrium of the symptom node and the disease node maximizes the accuracy of diagnosis;

[0120] S54. Based on the result of Nash equilibrium, the final diagnostic path is selected, and the optimal diagnostic result is finally obtained.

[0121] Specifically, first, in step S51, this solution establishes a game model between symptom nodes and disease nodes. In this model, symptom nodes and disease nodes are considered two participants in the game, each with its own strategy set. The symptom node's strategy set includes different symptom choices, while the disease node's strategy set includes different disease diagnosis results.

[0122] Each node optimizes its own payoff based on its own strategy. Symptom nodes select a manifested symptom as their strategy, while disease nodes choose the most appropriate diagnosis based on the combination and intensity of symptoms. In this way, the system abstracts the relationship between symptoms and diseases into a game theory, laying the foundation for subsequent optimal strategy selection.

[0123] Next, in step S52, the system calculates the optimal strategy for each symptom and disease node based on the Nash equilibrium principle in game theory. Nash equilibrium is a core concept in game theory, which refers to the situation in which all participants in a game have chosen an optimal strategy, leaving no participant with any incentive to unilaterally change their strategy.

[0124] In the application of this invention, the system calculates the strategy of each symptom node and disease node to find the strategy combination that maximizes the benefits of all participants. To calculate this optimal strategy, the following formula is used to express the benefit function of each node: ;

[0125] in, represents the revenue of the i-th node, is the strategy chosen by the node, is the strategy set of other nodes, is the payoff function of each node. In the payoff function of each node, the system considers the interaction between the strategy selected by the node and the strategies of other nodes, and calculates the optimal strategy for each node.

[0126] Through this calculation, the solution determines the optimal behavior for each symptom and disease node, thereby helping select the optimal reasoning path. In step S53, the system selects the optimal reasoning path by calculating the Nash equilibrium for each node in the game. In this step, the goal of the game model is to find an equilibrium point where the strategies of the symptom and disease nodes reach equilibrium.

[0127] In this equilibrium state, the choice of symptom nodes and disease nodes remains unchanged, as changing any single node would result in a decrease in benefits. By solving the Nash equilibrium, the system can select the reasoning path that best reflects diagnostic accuracy. This Nash equilibrium is typically calculated using iterative optimization algorithms such as backpropagation or gradient descent. This process enables the system to find an ideal balance between symptoms and diseases, maximizing diagnostic accuracy.

[0128] Finally, in step S54, based on the Nash equilibrium results, the system selects the final diagnostic path and ultimately obtains the optimal diagnostic result. Through the optimization of the game model, the final diagnostic path can effectively integrate the optimal strategy between symptoms and diseases, thereby achieving accurate diagnosis.

[0129] The beneficial effect of this approach is that, through the framework of game theory, the relationship between symptoms and diseases is not simply determined through simple reasoning, but rather through an optimized game decision-making process. Ultimately, the system can provide a diagnosis that is most representative and reflects the patient's true condition.

[0130] S6. Demonstrate the reasoning process and provide each step of the reasoning process.

[0131] Specifically, step S6 demonstrates the reasoning process and provides detailed information for each step. The core of this step is to ensure transparency of the reasoning process, allowing doctors to clearly understand the reasoning path and basis between each symptom and disease.

[0132] By displaying the reasoning process clearly and easily understood, doctors can better understand the system's decision-making logic, thereby increasing their trust in the diagnosis results. This not only improves the system's interpretability but also provides doctors with more references and feedback, helping them make more accurate clinical judgments.

[0133] Example 3

[0134] The third embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of a disease intelligent diagnosis and differential diagnosis method based on a knowledge graph in the above embodiment.

[0135] Example 4

[0136] The fourth embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer device, which includes: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory, and execute a disease intelligent diagnosis and differential diagnosis method based on a knowledge graph in the above embodiment.

[0137] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0138] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A disease intelligent diagnosis and differential diagnosis system based on knowledge graph, characterized by: The system comprises: A data input module is used to receive and standardize the patient's clinical symptoms, physical signs, laboratory test data and historical medical record data; A knowledge graph construction and update module is used to construct and update a knowledge graph of diseases and symptoms based on medical literature and clinical data. The knowledge graph automatically extracts the association between symptoms and diseases from medical literature using natural language processing technology; A reasoning and diagnosis module, configured to perform intelligent diagnosis and differential diagnosis of diseases based on the patient data received by the data input module and the knowledge graph provided by the knowledge graph construction and update module; The inference result display and explainability module is used to visualize the inference results and show each step of the inference path, each decision basis, and the information flow transmission process; The feedback and optimization module is used to provide feedback based on the doctor's diagnosis results of the system to further optimize the knowledge graph and reasoning process.

2. The disease intelligent diagnosis and differential diagnosis system based on knowledge graph according to claim 1 is characterized in that: The reasoning and diagnosis module includes: Information gain and optimal path selection module, used to calculate the information gain between symptoms and diseases and select the reasoning path with the largest information gain; A graph neural network reasoning optimization module is used to optimize the reasoning path based on the knowledge graph through multiple rounds of information transmission and update the representation of each node; The game theory decision optimization module is used to optimize the reasoning path and select the optimal diagnostic result based on the Nash equilibrium principle in game theory.

3. A disease intelligent diagnosis and differential diagnosis method based on knowledge graph, according to the disease intelligent diagnosis and differential diagnosis system based on knowledge graph according to any one of claims 1-2, characterized in that: The method comprises: S1. Receive and standardize the patient's clinical data; S2. Build and update the knowledge graph of diseases and symptoms based on natural language processing technology; S3. Calculate the information gain between symptoms and diseases and select the reasoning path with the largest information gain; S4, reasoning optimization of symptom and disease nodes based on graph neural network; S5. Optimize the reasoning path based on the Nash equilibrium principle in game theory and select the optimal diagnostic result; S6. Demonstrate the reasoning process and provide each step of the reasoning process.

4. The method for intelligent disease diagnosis and differential diagnosis based on knowledge graph according to claim 3, characterized in that: The clinical data include symptoms, signs, laboratory test data and historical medical records.

5. The method for intelligent disease diagnosis and differential diagnosis based on knowledge graph according to claim 3, characterized in that: The S2 knowledge graph construction process includes: S21. Extract the association between diseases and symptoms from medical literature; S22. Use natural language processing technology to process sentences and words of text; S23, extracting disease and symptom entities from the text and establishing preliminary disease-symptom associations; S24, constructing a disease symptom map and assigning weights to the association relationships; S25. Dynamically update the atlas content based on new data and medical research results.

6. The method for intelligent disease diagnosis and differential diagnosis based on knowledge graph according to claim 3, characterized in that: The S3 information gain calculation process includes: S31. Calculate the mutual information between each symptom and the potential disease to determine the strength of the association between the symptom and the disease; S32. Based on the calculated mutual information, select the symptom-disease association path with the largest information gain for inference; S33, assigning information gain weights to the relationships between each symptom and disease to optimize the reasoning path; S34. Based on the selected reasoning path, analyze the sequence or combination of symptoms and select the path with the most representativeness and maximum information gain for subsequent reasoning.

7. The method for intelligent disease diagnosis and differential diagnosis based on knowledge graph according to claim 3, characterized in that: The S4 graph neural network inference optimization process includes: S41. Represent symptom and disease nodes as nodes in a graph structure, define the relationships between nodes as edges of the graph, and transfer node information based on a graph neural network model. S42. Update the representation of each node through a multi-round iterative information transmission mechanism; S43. Calculate the relationship strength between each symptom and the disease based on the updated node representation, and select the reasoning path that best reflects the disease diagnosis information; S44. Based on the output results of the graph neural network, select the optimal path for subsequent reasoning.

8. The method for intelligent disease diagnosis and differential diagnosis based on knowledge graph according to claim 3, characterized in that: The S5 game theory optimization process includes: S51. Establish a game model between symptom nodes and disease nodes, consider each symptom node and disease node as a participant in the game, and define their strategy set; S52. Based on the Nash equilibrium principle in game theory, calculate the optimal strategy for each symptom and disease node, that is, choose the strategy that can maximize its own benefits; S53. By calculating the Nash equilibrium of each node in the game process, the optimal reasoning path is selected, in which the strategy equilibrium of the symptom node and the disease node maximizes the accuracy of diagnosis; S54. Based on the result of Nash equilibrium, the final diagnostic path is selected, and the optimal diagnostic result is finally obtained.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, it implements a disease intelligent diagnosis and differential diagnosis method based on knowledge graph as described in any one of claims 3 to 8.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which, when executed by a processor, implements a disease intelligent diagnosis and differential diagnosis method based on a knowledge graph as described in any one of claims 3 to 8.

Images