An intelligent diagnosis auxiliary system based on a medical knowledge graph

Abstract

The application discloses an intelligent diagnosis auxiliary system based on a medical knowledge graph and belongs to the medical diagnosis field. In order to solve the problem that a traditional medical diagnosis system is difficult to integrate information, a large amount of medical information is wasted, effective auxiliary diagnosis cannot be performed, the diagnosis efficiency is low, effective intelligent triage prediction cannot be performed, the patients are easily caused to have a burden of diagnosis, the hospital is easily caused to have a burden of operation, the work efficiency is low, corresponding diagnosis results cannot be given in time and quickly, and the diagnosis accuracy is influenced. The existing medical data island problem is overcome, the intelligent auxiliary diagnosis of common diseases is realized, the diagnosis efficiency is improved, the doctors are provided with high-credibility auxiliary diagnosis, the intelligent triage prediction is realized, the probability of wrong triage is reduced, the patients are provided with efficient and convenient pre-diagnosis consultation, the burden of the patients and the operation load of the hospital are reduced, the diagnosis accuracy of the diagnosis auxiliary system for comprehensive data analysis of patient information is gradually improved through human error correction.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostics, and in particular to an intelligent diagnostic assistance system based on medical knowledge graphs. Background Technology

[0002] The medical field is characterized by its data-intensive, knowledge-intensive, and intellectually intensive nature. The massive amounts of data generated by the medical industry are mostly unstructured. While my country's medical system is continuously improving, and medical resources, including medical equipment and healthcare personnel, are gradually expanding, shortages of medical resources and low hospital operational efficiency still exist. The uneven distribution of domestic medical resources, coupled with the fact that traditional disease diagnosis methods are often constrained by the decision-makers' personal experience and easily influenced by their subjective opinions and external circumstances, can lead to diagnostic biases. Traditional hospital diagnostic systems also suffer from the following shortcomings:

[0003] 1. The existence of medical data silos within hospitals makes it difficult to integrate information, resulting in a significant waste of medical information, inability to provide effective auxiliary diagnosis, and low efficiency in patient care.

[0004] 2. When patients undergo initial diagnosis, the lack of effective intelligent triage prediction can easily burden patients and hospitals, leading to low work efficiency.

[0005] 3. When diagnosing patients, the inability to provide timely and rapid results necessitates numerous examinations, which not only incurs financial losses for patients but also wastes medical resources and affects the accuracy of diagnosis. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent diagnostic assistance system based on a medical knowledge graph. By establishing electronic medical records for patients, the system reduces the burden of data entry for doctors. Through the establishment of a medical knowledge graph, it integrates medical knowledge, overcoming the problem of isolated medical data and improving the utilization rate of medical data. This enables intelligent assisted diagnosis of common diseases, facilitating the timely detection of characteristic diseases during patient visits, improving the efficiency of medical care, and bringing convenience to patients. The automatic diagnosis module performs corresponding diagnoses, analyzing and obtaining weighted suspected diseases to provide doctors with highly reliable assisted diagnoses, helping to improve doctors' diagnostic efficiency and accuracy. This system improves accuracy, enables intelligent triage prediction, reduces the probability of misdiagnosis, provides patients with efficient and convenient pre-diagnosis consultation, alleviates the burden on patients and the operational load on hospitals, and brings patients an efficient and accurate medical experience. It ensures that after considering more influencing factors, it can provide more objective and reliable diagnostic opinions compared to existing artificial intelligence diagnostic systems. It avoids the complexity of user input information and the problem of insufficient equipment in the user's environment. Users can receive intelligent diagnosis and treatment anytime, anywhere as long as there is a network connection. Through human error correction, it gradually improves the diagnostic accuracy of the comprehensive data analysis and diagnostic assistance system for patient information, thereby solving the problems mentioned in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An intelligent diagnostic assistance system based on medical knowledge graph includes an intelligent diagnostic system. The intelligent diagnostic system has a patient data layer, an entity extraction layer, an auxiliary diagnostic layer, and a recommended treatment layer. The output of the patient data layer is connected to the input of the entity extraction layer, the output of the entity extraction layer is connected to the input of the auxiliary diagnostic layer, and the output of the auxiliary diagnostic layer is connected to the input of the recommended treatment layer.

[0009] The patient data layer includes an information acquisition module, an information storage module, and a medical record generation module.

[0010] The entity extraction layer includes an information extraction module, an information recognition module, an information output module, and a medical knowledge graph.

[0011] The auxiliary diagnostic layer includes an information processing module, an automatic diagnostic module, and a recommended examination module.

[0012] The recommended treatment layer includes a treatment plan generation module, a personalized recommendation module, and an information integration module.

[0013] Furthermore, the output of the information acquisition module is connected to the input of the information storage module, the output of the information storage module is connected to the output of the medical record generation module, and the output of the medical record generation module is connected to the input of the entity extraction layer. The information acquisition module includes a data input module, which generates text by allowing manual or voice input of the patient's complaints. The output of the entity extraction layer consists of the disease type, disease symptoms, and triggering factors.

[0014] Furthermore, the information collection module collects patient ID, gender, age, disease type, medical history, contraindications, and key living environment. The key living environment is used as an evaluation criterion to quantify the probability of a patient having the target disease into three levels: extremely likely, average, and unlikely. Patient information is the "threshold" for the mapping relationship between patient information and the target disease in the medical record learning system, as well as the training set for initially constructing the mapping relationship.

[0015] Furthermore, the output end of the information extraction module is connected to the input end of the information recognition module, the output end of the information recognition module is connected to the input end of the information output module, the output end of the information output module is connected to the input end of the medical knowledge graph, and the output end of the medical knowledge graph is connected to the input end of the auxiliary diagnosis layer.

[0016] Furthermore, the information extraction module is used to extract the patient's medical record information from the patient data layer, and the information recognition module is used to perform diagnostic analysis on the information extracted from the patient data layer. The analysis results are output to the medical knowledge graph through the information output module. The medical knowledge graph includes a disease knowledge base, an examination and testing knowledge base, a symptom knowledge base, a drug knowledge base, a body part knowledge base, and a surgical knowledge base.

[0017] Furthermore, the medical knowledge graph is based on a triple representation, including basic entity information such as symptoms, diseases, locations, drugs, departments, and populations, and is also classified by gender. It also includes relationships such as location-symptom, location-disease, symptom-disease, disease-department, drug-disease, drug-symptom, and drug-population. The extraction of entities and relationships adopts a method based on entity dictionaries, entity rules, and pattern matching, and realizes the quantification of symptom-disease relationships. The medical knowledge graph must be clear and easy to understand, so it is necessary to filter out some secondary information, extract the main information, and randomly sort the results.

[0018] Furthermore, the output of the information processing module is connected to the input of the automatic diagnosis module, the output of the automatic diagnosis module is connected to the input of the recommended examination module, and the output of the recommended examination module is connected to the input of the recommended treatment layer.

[0019] Furthermore, the information processing module includes an information comparison module and an information analysis module. The output end of the information comparison module is connected to the input end of the information analysis module. The information comparison module compares the patient information extracted by the entity extraction layer and analyzes the patient information through the information analysis module. The automatic diagnosis module diagnoses the patient's suspected diseases and provides a corresponding list of recommended examinations through the recommended examination module.

[0020] Furthermore, the output of the treatment plan generation module is connected to the input of the personalized recommendation module, and the output of the personalized recommendation module is connected to the input of the information integration module.

[0021] Furthermore, the treatment plan generation module is used to generate a preliminary treatment plan based on the results of the examination report and the doctor's diagnosis. The personalized recommendation module is used to find treatment patterns of similar patient groups based on the patient's basic information and the doctor's diagnosis to generate personalized recommendations. The information integration module is used to integrate the treatment plan and personalized recommendations using linear model fusion technology to obtain the final treatment plan.

[0022] Furthermore, the treatment plan generation module generates treatment plans through the following steps:

[0023] Obtain the patient's examination results and final treatment plan;

[0024] The examination results are examined and analyzed to determine the number of items in the examination and the disease characteristics of each item.

[0025] The treatment plan is divided into sub-plans based on the number of items, and the disease characteristics and treatment sub-plans are matched to determine the matching value corresponding to the disease characteristics of each test result.

[0026] The fitting value is used as the first treatment standard parameter;

[0027] Based on the patient's examination results, recommended treatment plans are obtained through big data analysis; among them,

[0028] The recommended treatment options shall be no less than two;

[0029] The recommended treatment plan is divided into sub-plans according to the items, and the disease characteristics and recommended sub-plans are matched to determine the secondary radiotherapy plan matching value corresponding to the disease characteristics of each examination result; wherein,

[0030] Each disease feature corresponds to no fewer than two secondary scheme adaptation values;

[0031] By comparing the secondary treatment adaptation value for each disease feature with the first treatment standard parameter, a target treatment sub-plan exceeding the first treatment standard parameter is obtained; wherein,

[0032] When there are multiple target treatment sub-routines for each disease feature, the same target treatment sub-routines are fused together.

[0033] Based on the target treatment sub-plan, the plans are fused to determine the final treatment plan; wherein, the plan fusion adopts multi-dimensional fusion, and after fusion, each disease feature has only one unique treatment sub-plan.

[0034] In current technologies, most companies and systems recommend treatment plans based on deep learning networks using large datasets to match the most suitable option. However, big data also has certain limitations. Specifically, the models built on large datasets are susceptible to errors in training. If the original data samples are incorrect, or if the training iterations are insufficient or the training loss is too high, the recommended treatment plan will be flawed. Furthermore, in hospitals, we use pre-trained models, making it impossible to determine the accuracy of the training process and the original data.

[0035] Furthermore, after the treatment plan generation module performs plan fusion, it also includes calculating the trust level of the fused plans. The specific steps are as follows:

[0036] Step 1: Based on the target treatment sub-plan, calculate the grey relational degree between different target treatment sub-plans using the following formula:

[0037]

[0038] Where X(K) represents the treatment feature of the Kth treatment sub-scheme; X(m) represents the treatment feature of the mth treatment sub-scheme; K≠m, K, m∈positive integers; ρ∈[0~1], ρ represents the normalized value of the treatment feature;

[0039] Step 2: Based on the correlation degree, calculate the fusion expectation, fusion entropy, and fusion hyperentropy among any target treatment sub-schemes:

[0040]

[0041]

[0042]

[0043] Where n represents the total number of target treatment sub-treatments; X represents the characteristic average of the target treatment sub-regimen; i Let represent the treatment characteristics of the i-th target treatment sub-scheme; Q represents the expected fusion among the target treatment sub-schemes; and S represents the fusion entropy among the target treatment sub-schemes. This represents the fusion hyperentropy between target treatment sub-plans;

[0044] Step 3: Determine the trust value based on the fusion expectation, fusion entropy, and fusion hyperentropy.

[0045]

[0046] Where Z represents the level of trust in the scheme integration.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] 1. This invention discloses an intelligent diagnostic assistance system based on a medical knowledge graph. The system collects patient information through an information acquisition module, establishing electronic medical records to reduce the burden of data entry for doctors. It also facilitates the unified and standardized management of medical data. By establishing a medical knowledge graph, it integrates medical knowledge, overcoming the problem of isolated medical data and improving data utilization. This enables intelligent auxiliary diagnosis of common diseases. A feature disease database is established to store and analyze the system-extracted patient disease characteristics and the range of possible diseases mapped by these characteristics. This allows for timely detection of characteristic diseases during patient visits, improving efficiency and convenience. The system analyzes potential patient diseases through information comparison and analysis modules, and performs corresponding diagnoses through an automatic diagnosis module. This analysis yields weighted suspected diseases, providing doctors with highly reliable auxiliary diagnoses and improving diagnostic efficiency and accuracy.

[0049] 2. The present invention provides an intelligent diagnostic assistance system based on medical knowledge graphs. Through a recommended examination module, it recommends departments and examination items, enabling intelligent triage prediction, reducing the probability of misdiagnosis, providing patients with efficient and convenient pre-diagnosis consultation, reducing the burden on patients and the operational load on hospitals, and bringing patients an efficient and accurate medical experience. It ensures that, after considering more influencing factors, it can provide more objective and reliable diagnostic opinions compared to existing artificial intelligence diagnostic systems. It avoids the complexity of user input information and the problem of insufficient equipment in the user's environment, and allows users to conduct intelligent diagnosis and treatment of diseases anytime, anywhere with internet access.

[0050] 3. The present invention provides an intelligent diagnostic assistance system based on a medical knowledge graph. The expert database is used to store the final output results of the comprehensive data analysis system for artificial intelligence disease diagnosis, as well as the establishment of supplementary retrieval tools. The establishment of retrieval tools facilitates doctors to review and judge the accuracy of the patient's disease process and results through the system, and gradually improves the diagnostic accuracy of the comprehensive data analysis system for patient information through human error correction. The supplementary retrieval tools facilitate outpatient doctors to investigate and correct doubtful diagnostic results, thereby improving the accuracy of diagnosis. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the overall system of an intelligent diagnostic assistance system based on medical knowledge graph according to the present invention;

[0052] Figure 2 This is a diagram showing the main modules of an intelligent diagnostic assistance system based on a medical knowledge graph according to the present invention.

[0053] Figure 3 This is a schematic diagram of the module connection of an intelligent diagnostic assistance system based on medical knowledge graph according to the present invention;

[0054] Figure 4 This is a schematic diagram showing the connection of some modules of an intelligent diagnostic assistance system based on medical knowledge graph according to the present invention;

[0055] Figure 5 This is a schematic diagram of some modules of an intelligent diagnostic assistance system based on medical knowledge graph according to the present invention;

[0056] Figure 6 This is a schematic diagram of the workflow of an intelligent diagnostic assistance system based on medical knowledge graph according to the present invention.

[0057] In the diagram: 1. Patient Data Layer; 11. Information Acquisition Module; 12. Information Storage Module; 13. Medical Record Generation Module; 2. Entity Extraction Layer; 21. Information Extraction Module; 22. Information Recognition Module; 23. Information Output Module; 24. Medical Knowledge Graph; 3. Auxiliary Diagnosis Layer; 31. Information Processing Module; 311. Information Comparison Module; 312. Information Analysis Module; 32. Automatic Diagnosis Module; 33. Recommended Examination Module; 4. Recommended Treatment Layer; 41. Treatment Plan Generation Module; 42. Personalized Recommendation Module; 43. Information Integration Module; 5. Intelligent Diagnosis System. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1

[0060] See Figures 1-4An intelligent diagnostic assistance system based on medical knowledge graph is disclosed, comprising an intelligent diagnostic system 5. The intelligent diagnostic system 5 includes a patient data layer 1, an entity extraction layer 2, an auxiliary diagnostic layer 3, and a recommended treatment layer 4. The output of the patient data layer 1 is connected to the input of the entity extraction layer 2, the output of the entity extraction layer 2 is connected to the input of the auxiliary diagnostic layer 3, and the output of the auxiliary diagnostic layer 3 is connected to the input of the recommended treatment layer 4. The patient data layer 1 includes an information acquisition module 11, an information storage module 12, and a medical record generation module 13. The output of the information acquisition module 11 is connected to the input of the information storage module 12, the output of the information storage module 12 is connected to the output of the medical record generation module 13, and the output of the medical record generation module 13 is connected to the input of the entity extraction layer 2. The information acquisition module 11 includes a data input module. This module uses manual or voice input of the patient's complaints to generate text. The output of the entity extraction layer 2 consists of disease type, symptoms, and triggering factors. The information acquisition module 11 collects patient ID, gender, age, disease type, medical history, contraindications, and key living environment information. The key living environment information is used as an evaluation criterion to quantify the probability of the patient having the target disease into three levels: extremely likely, moderate, and unlikely. Patient information serves as the "threshold" for the mapping relationship between patient information and the target disease in the medical record learning system, as well as the initial training set for constructing the mapping relationship. The entity extraction layer 2 includes an information extraction module 21, an information recognition module 22, an information output module 23, and a medical knowledge graph 24. The output of the information extraction module 21 is connected to the input of the information recognition module 22, the output of the information recognition module 22 is connected to the input of the information output module 23, the output of the information output module 23 is connected to the input of the medical knowledge graph 24, and the output of the medical knowledge graph 24 is connected to the input of the auxiliary diagnosis layer 3. 1 is used to extract patient medical record information from patient data layer 1. Information recognition module 22 is used to perform diagnostic analysis on the information extracted from patient data layer 1. The analysis results are output to the medical knowledge graph 24 through information output module 23. Medical knowledge graph 24 includes disease knowledge base, examination and test knowledge base, symptom knowledge base, drug knowledge base, body part knowledge base, and surgical knowledge base. Medical knowledge graph 24 is based on a triplet representation, including basic entity information of symptoms, diseases, parts, drugs, departments, and populations, and is also classified by gender. It also includes relationships between parts and symptoms, parts and diseases, symptoms and diseases, diseases and departments, drugs and diseases, drugs and symptoms, and drugs and populations. The extraction of entities and relationships adopts a method based on entity dictionary, entity rules, and pattern matching, and realizes the quantification of the relationship between symptoms and diseases. Medical knowledge graph 24 must be clear and easy to understand, so it is necessary to filter some secondary information, extract the main information, and randomly sort the results. The auxiliary diagnosis layer 3 includes information processing module 31, automatic diagnosis module 32, and recommended examination module 33.The output of the information processing module 31 is connected to the input of the automatic diagnosis module 32. The output of the automatic diagnosis module 32 is connected to the input of the recommended examination module 33. The output of the recommended examination module 33 is connected to the input of the recommended treatment layer 4. The information processing module 31 includes an information comparison module 311 and an information analysis module 312. The output of the information comparison module 311 is connected to the input of the information analysis module 312. The information comparison module 311 compares the patient information extracted by the entity extraction layer 2 and analyzes the patient information through the information analysis module 312. The automatic diagnosis module 32 diagnoses the patient's suspected diseases and provides corresponding recommendations through the recommended examination module 33. The examination list and recommended treatment layer 4 include a treatment plan generation module 41, a personalized recommendation module 42, and an information integration module 43. The output of the treatment plan generation module 41 is connected to the input of the personalized recommendation module 42, and the output of the personalized recommendation module 42 is connected to the input of the information integration module 43. The treatment plan generation module 41 generates a preliminary treatment plan based on the examination report results and the doctor's diagnosis. The personalized recommendation module 42 generates personalized recommendations by finding treatment patterns for similar patient groups based on the patient's basic information and the doctor's diagnosis. The information integration module 43 integrates the treatment plan and personalized recommendations using linear model fusion technology to obtain the final treatment plan.

[0061] Please see Figures 5-6 A medical knowledge graph-based intelligent diagnostic assistance system includes the following operational steps:

[0062] S1: Establish patient information and create corresponding electronic medical records based on the patient information.

[0063] S101: The information collection module 11 collects information including patient ID, gender, age, disease type, medical history, contraindications, and key living environment. The key living environment is used as the evaluation basis to quantify the probability of the patient having the target disease into three levels: extremely easy, average, and difficult. The collected data is stored through the information storage module 12.

[0064] S102: Generate corresponding electronic medical records from the patient's medical information through the medical record generation module 13.

[0065] S2: Build a medical knowledge graph framework and analyze medical information.

[0066] S201: The construction process of the medical knowledge graph 24 involves three steps: establishing a knowledge base schema diagram, defining the schema diagram, knowledge extraction, and knowledge fusion. The schema diagram definition includes the concepts in the knowledge base, the attributes of the concepts, and the hierarchical relationships between the concepts. The knowledge base schema diagram commonly used in traditional Chinese medicine knowledge bases mainly includes upper-level concepts such as Chinese medicinal materials, Chinese medicine syndromes, and Chinese medicine diseases, as well as the attributes of these concepts. The schema diagram is constructed. Knowledge extraction mainly includes entities related to traditional Chinese medicine in the network, entity types, synonym relationships, and attribute value relationships. Knowledge fusion is the fusion of the extracted knowledge content. The medical knowledge base used to construct the medical knowledge graph 24 needs to identify drug-related entities from professional traditional Chinese medicine texts, learn a labeling model using large-scale corpora, and then label the sentences.

[0067] S202: Establish a characteristic disease database. The characteristic disease database is used to store and analyze the characteristic information of patients' diseases extracted by the system and the possible range of diseases mapped by the characteristic information.

[0068] S203: The information extraction module 21 extracts the patient medical record information collected from the information acquisition module 11, stores the possible disease range mapped by the extracted patient information and feature information in the feature disease database, and uses the information identification module 22 to perform diagnostic analysis on the information extracted from the information acquisition module 11. The analysis results are output to the medical knowledge graph 24 through the information output module 23.

[0069] S3: Use the extracted patient information to assist in diagnosis and provide a list of recommended examinations.

[0070] S301: Establish an expert database. The expert database is used to store the final output results of the comprehensive data analysis system for artificial intelligence disease diagnosis, as well as to establish supplementary search tools. The supplementary search tools are used to facilitate outpatient doctors to investigate and correct doubtful diagnostic results. The expert database is for doctors to operate and refer to, and is the core of the system's reference value.

[0071] S302: The patient information extracted by the entity extraction layer 2 is compared with the information in the medical knowledge graph 24 by the information comparison module 311, and the information analysis module 312 performs data analysis on the compared information.

[0072] S303: The inference engine is used to apply the rules to the extracted patient information, infer the information provided by the patient, and diagnose the suspected diseases of the patient through the automatic diagnosis module 32.

[0073] S303: Call the final output of the comprehensive data analysis system of the characteristic disease database in the expert database, and establish a supplementary search tool. The establishment of the search tool makes it convenient for doctors to review and judge whether the patient's disease process and results are accurate through the system, and gradually improve the diagnostic accuracy of the comprehensive data analysis diagnostic auxiliary system through human error correction. The recommended examination module 33 provides a corresponding recommended examination list, which includes suspected disease types, recommended departments, and recommended examination items.

[0074] S4: Recommend treatment options and generate personalized recommendations.

[0075] S401: Treatment plan generation module 41 generates a preliminary treatment plan based on association rule analysis of clinical guidelines, medical knowledge images and clinical data, and based on the results of examination reports and doctors' diagnoses.

[0076] S402: Based on the patient's basic information and the doctor's diagnosis, find clinically similar patient groups, and generate personalized recommendations by combining the treatment patterns of similar patient groups and common treatment patterns.

[0077] S403: The treatment plan and personalized recommendations are integrated through the information integration module 43 using linear model fusion technology to obtain the final treatment plan.

[0078] In summary, the intelligent diagnostic assistance system based on a medical knowledge graph of the present invention collects patient information through the information collection module 11, establishes electronic medical records for patients, reduces the burden of medical record entry for doctors, and facilitates the unified and standardized management of medical data. By establishing a medical knowledge graph 24, it integrates medical knowledge, overcomes the problem of existing medical data silos, improves the utilization rate of medical data, and realizes intelligent auxiliary diagnosis of common diseases. The established characteristic disease database is used to store and analyze the characteristic information of patients' diseases extracted by the system and the range of possible diseases mapped by the characteristic information, making it convenient for patients to promptly identify characteristic diseases during medical visits, improving the efficiency of medical treatment and bringing convenience to patients. The information comparison module 311 and the information analysis module 312 analyze the possible diseases of patients, and the automatic diagnosis module 32 performs corresponding diagnoses, thereby analyzing and obtaining weighted suspected diseases, providing doctors with highly reliable auxiliary diagnoses, helping to improve the efficiency and accuracy of doctors' diagnoses. The system also includes a recommended examination module. The system includes a list of 33 recommended departments and examination items, enabling intelligent triage prediction, reducing the probability of misdiagnosis, providing patients with efficient and convenient pre-diagnosis consultation, reducing the burden on patients and the operational load on hospitals, and bringing patients an efficient and accurate medical experience. The expert database stores the final output results of the comprehensive data analysis system for AI-powered disease diagnosis, as well as supplementary search tools. These search tools allow doctors to review and judge the accuracy of patients' disease processes and outcomes through the system, and gradually improve the diagnostic accuracy of the comprehensive data analysis system for patient information through human error correction. The supplementary search tools facilitate outpatient doctors in investigating and correcting questionable diagnostic results. The expert database, for doctors' operation and reference, is the core of the system's reference value. After considering more influencing factors, it can provide more objective and reliable diagnostic opinions compared to existing AI diagnostic systems, avoiding the complexity of user input information and the problem of insufficient equipment in the user's environment. Users can receive intelligent diagnosis and treatment anytime, anywhere with internet access.

[0079] Furthermore, the treatment plan generation module generates treatment plans through the following steps:

[0080] Obtain the patient's test results and the treatment plan issued by the doctor;

[0081] The examination results are examined and analyzed to determine the number of items in the examination and the disease characteristics of each item.

[0082] The treatment plan is divided into sub-plans based on the number of items, and the disease characteristics and treatment sub-plans are matched to determine the matching value corresponding to the disease characteristics of each test result.

[0083] The fitting value is used as the first treatment standard parameter;

[0084] Based on the patient's examination results, recommended treatment plans are obtained through big data analysis; among them,

[0085] The recommended treatment options shall be no less than two;

[0086] The recommended treatment plan is divided into sub-plans according to the items, and the disease characteristics and recommended sub-plans are matched to determine the secondary radiotherapy plan matching value corresponding to the disease characteristics of each examination result; wherein,

[0087] Each disease feature corresponds to no fewer than two secondary scheme adaptation values;

[0088] By comparing the secondary treatment adaptation value for each disease feature with the first treatment standard parameter, a target treatment sub-plan exceeding the first treatment standard parameter is obtained; wherein,

[0089] When there are multiple target treatment sub-routines for each disease feature, the same target treatment sub-routines are fused together.

[0090] Based on the target treatment sub-plan, the plans are fused to determine the final treatment plan; wherein, the plan fusion adopts multi-dimensional fusion, and after fusion, each disease feature has only one unique treatment sub-plan.

[0091] In existing technologies, most enterprises and systems recommend treatment plans based on deep learning networks using big data to match the most suitable solution. However, big data also has certain limitations. Specifically, if the original data samples are incorrect, or if the training iterations are insufficient or the training loss is too high, the recommended treatment plan will be flawed. In hospitals, we use pre-trained models, making it impossible to determine the accuracy of their training and the original data. Therefore, this invention compares and integrates treatment plans provided by doctors and those generated by big data to generate a unique treatment plan. In this fusion process, the present invention includes the following steps: First, based on the technical plan provided by the doctor, each treatment objective for the disease is evaluated. For example, when treating a cold, some medications are for reducing fever, which is one sub-plan, while others are for treating cough, which is another sub-plan. Each treatment objective corresponds to a specific disease characteristic. In this way, we can determine the fit value of each treatment objective in the treatment plan for each disease characteristic. We use the doctor's plan as a standard. Then, we determine the treatment plan through big data. If the treatment effect of a plan exceeds this fit value, it means it is better than the doctor's plan, and we can replace the treatment plan. However, if multiple treatment plans are better than the doctor's plan, we can use multi-dimensional fusion, that is, to take into account the advantages of each treatment sub-plan to generate a treatment sub-plan for each disease characteristic, which is to fuse the treatment plans. Finally, after fusing all the sub-plans, we determine the final treatment plan. Because this plan is based on the doctor's plan, there will not be too much deviation, solving the problem of huge bias in big data. Moreover, by fusing only the better plans, we can also solve the problem of insufficient training times.

[0092] Furthermore, after the treatment plan generation module 41 performs plan fusion, it also includes calculating the trust level of the fused plans. The specific steps are as follows:

[0093] Step 1: Based on the target treatment sub-plan, calculate the grey relational degree between different target treatment sub-plans using the following formula:

[0094]

[0095] Where X(K) represents the treatment feature of the Kth treatment sub-scheme; X(m) represents the treatment feature of the mth treatment sub-scheme; K≠m, K, m∈positive integers; ρ∈[0~1], ρ represents the normalized value of the treatment feature;

[0096] The present invention calculates the grey relational degree by means of the correlation between different target treatment sub-schemes. The grey relational degree is the correlation between different target treatment sub-schemes. [X(K) and X(m)] represent different treatment sub-schemes. Our calculation based on normalized values ​​is mainly to ensure that the correlation between the schemes can be in (0~1), so as to ensure that the correlation is not 100%.

[0097] Step 2: Based on the correlation degree, calculate the fusion expectation, fusion entropy, and fusion hyperentropy among any target treatment sub-schemes:

[0098]

[0099]

[0100]

[0101] Where n represents the total number of target treatment sub-treatments; X represents the characteristic average of the target treatment sub-regimen; i Let represent the treatment characteristics of the i-th target treatment sub-scheme; Q represents the expected fusion among the target treatment sub-schemes; and S represents the fusion entropy among the target treatment sub-schemes. This represents the fusion hyperentropy between target treatment sub-plans;

[0102] The main purpose of the above three values ​​is to determine the effectiveness of each treatment sub-treatment:

[0103] First, we integrate the expected values ​​to determine the final result after integration.

[0104] Fusion entropy reflects the degree of chaos after fusion, which in turn determines the effectiveness of treatment.

[0105] Finally, hyperentropy is used to determine the degree of deviation from the normal treatment plan after the fusion of multiple plans. In other words, after fusion, the result of not being able to fuse is an incorrect treatment plan, and it will not deviate from the normal treatment plan.

[0106] Step 3: Determine the trust value based on the fusion expectation, fusion entropy, and fusion hyperentropy.

[0107]

[0108] Where Z represents the level of trust in the scheme integration.

[0109] Using the three values ​​mentioned above, we determine the final trust value based on an exponential function. The trust value is determined by the likelihood that each treatment plan will produce a beneficial therapeutic effect.

[0110] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent diagnostic assistance system based on medical knowledge graph, comprising an intelligent diagnostic system (5), characterized in that... The intelligent diagnostic system (5) includes a patient data layer (1), an entity extraction layer (2), an auxiliary diagnostic layer (3), and a recommended treatment layer (4). The output of the patient data layer (1) is connected to the input of the entity extraction layer (2), the output of the entity extraction layer (2) is connected to the input of the auxiliary diagnostic layer (3), and the output of the auxiliary diagnostic layer (3) is connected to the input of the recommended treatment layer (4). The patient data layer (1) includes an information acquisition module (11), an information storage module (12), and a medical record generation module (13). The entity extraction layer (2) includes an information extraction module (21), an information recognition module (22), an information output module (23), and a medical knowledge graph (24). The auxiliary diagnostic layer (3) includes an information processing module (31), an automatic diagnostic module (32), and a recommended examination module (33). The recommended treatment layer (4) includes a treatment plan generation module (41), a personalized recommendation module (42), and an information integration module (43). The treatment plan generation module (41) generates a treatment plan by including the following steps: Obtain the patient's test results and the doctor's treatment plan; The examination results were examined and analyzed to determine the number of items in the examination and the disease characteristics of each item. The treatment plan is divided into sub-plans based on the number of items, and the disease characteristics and treatment sub-plans are matched to determine the matching value corresponding to the disease characteristics of each examination result; The fitting value is used as the first treatment standard parameter; Based on the patient's examination results, recommended treatment plans are obtained through big data analysis; wherein, there are no fewer than two recommended treatment plans. The recommended treatment plan is divided into sub-plans based on the number of items in the examination results, and the disease characteristics and recommended treatment sub-plans are matched to determine the secondary plan matching value corresponding to the disease characteristics of each examination result; wherein, the secondary plan matching value corresponding to each disease characteristic is not less than two. By comparing the secondary treatment adaptation value of each disease feature with the first treatment standard parameter, a target treatment sub-plan that exceeds the first treatment standard parameter is obtained; wherein, when there are multiple target treatment sub-plans for each disease feature, the same target treatment sub-plans are fused. Based on the target treatment sub-plans, the plans are fused to determine the final treatment plan; wherein, the plan fusion adopts multi-dimensional fusion, and after fusion, each disease feature has only one uniquely determined treatment sub-plan; After the treatment plan generation module (41) performs plan fusion, it also includes calculating the trust level of the plan fusion. The specific steps are as follows: Step 1: Based on the target treatment sub-plan, calculate the grey relational degree between different target treatment sub-plans using the following formula: ； Where X(K) represents the treatment feature of the Kth treatment sub-scheme; X(m) represents the treatment feature of the mth treatment sub-scheme; K≠m, K, m∈positive integers; ρ∈[0~1], ρ represents the normalized value of the treatment feature; Step 2: Based on the correlation degree, calculate the fusion expectation, fusion entropy, and fusion hyperentropy among any target treatment sub-schemes: ； ； ； Where n represents the total number of target treatment sub-treatments; Let X represent the average feature of the target treatment sub-scheme; X represent the treatment feature of the i-th target treatment sub-scheme; Q represent the expected fusion among the target treatment sub-schemes; and S represent the fusion entropy among the target treatment sub-schemes. This represents the fusion hyperentropy between target treatment sub-plans; Step 3: Determine the trust value based on the fusion expectation, fusion entropy, and fusion hyperentropy. ； Where Z represents the level of trust in the scheme integration.

2. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 1, characterized in that... The output of the information acquisition module (11) is connected to the input of the information storage module (12), the output of the information storage module (12) is connected to the output of the medical record generation module (13), the output of the medical record generation module (13) is connected to the input of the entity extraction layer (2), the information acquisition module (11) is equipped with a data input module, the data input module generates text by manually inputting or voice inputting the patient's complaints, and the output of the entity extraction layer (2) consists of disease type, disease symptoms and causes.

3. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 1, characterized in that... The information collection module (11) collects patient ID, gender, age, disease type, medical history, contraindications, and key living environment. The key living environment is used as the evaluation basis to quantify the probability that the patient has the target disease. The patient information is the "threshold" of the mapping relationship between patient information and target disease in the medical record learning system and the training set for the initial construction of the mapping relationship.

4. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 1, characterized in that... The output of the information extraction module (21) is connected to the input of the information recognition module (22), the output of the information recognition module (22) is connected to the input of the information output module (23), the output of the information output module (23) is connected to the input of the medical knowledge graph (24), and the output of the medical knowledge graph (24) is connected to the input of the auxiliary diagnosis layer (3).

5. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 4, characterized in that... The information extraction module (21) is used to extract the patient's medical record information from the patient data layer (1). The information recognition module (22) is used to perform diagnostic analysis on the information extracted from the patient data layer (1). The analysis results are output to the medical knowledge graph (24) through the information output module (23). The medical knowledge graph (24) includes a disease knowledge base, an examination and testing knowledge base, a symptom knowledge base, a drug knowledge base, a body part knowledge base, and a surgical knowledge base. The medical knowledge graph (24) is based on a triplet representation, including basic entity information such as symptoms, diseases, parts, drugs, departments, and populations. It is also classified according to gender and includes relationships such as part-symptom relationships, part-disease relationships, symptom-disease relationships, disease-department relationships, drug-disease relationships, drug-symptom relationships, and drug-population relationships. The extraction of entities and relationships adopts a method based on entity dictionary, entity rules, and pattern matching, and realizes the quantification of the relationship between symptoms and diseases. The medical knowledge graph (24) must be clear and easy to understand. Therefore, it is necessary to filter some secondary information, extract the main information, and randomly sort the results.

6. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 1, characterized in that... The output of the information processing module (31) is connected to the input of the automatic diagnosis module (32), the output of the automatic diagnosis module (32) is connected to the input of the recommended examination module (33), and the output of the recommended examination module (33) is connected to the input of the recommended treatment layer (4). The information processing module (31) includes an information comparison module (311) and an information analysis module (312). The output of the information comparison module (311) is connected to the input of the information analysis module (312). The information comparison module (311) compares the patient information extracted by the entity extraction layer (2) and analyzes the patient information through the information analysis module (312). The automatic diagnosis module (32) diagnoses the suspected diseases of the patient and provides a corresponding list of recommended examinations through the recommended examination module (33).

7. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 1, characterized in that... The output of the treatment plan generation module (41) is connected to the input of the personalized recommendation module (42), and the output of the personalized recommendation module (42) is connected to the input of the information integration module (43).

8. The intelligent diagnostic assistance system based on medical knowledge graph as described in claim 7, characterized in that... The treatment plan generation module (41) is used to generate a preliminary treatment plan based on the results of the examination report and the doctor's diagnosis. The personalized recommendation module (42) is used to find treatment patterns of similar patient groups based on the patient's basic information and the doctor's diagnosis to generate personalized recommendations. The information integration module (43) is used to integrate the treatment plan and personalized recommendations using linear model fusion technology to obtain the final treatment plan.

Images