Method, device and equipment for generating auxiliary diagnosis information

Abstract

The invention discloses a method, a device and equipment for generating auxiliary diagnosis information, relates to the technical field of information processing, and can effectively perform data dimension reduction on multi-modal medical data so as to realize correct disease auxiliary judgment and improve the accuracy of the auxiliary diagnosis information. The method comprises the steps of obtaining to-be-diagnosed medical record data and medical resource data, wherein the to-be-diagnosed medical record data comprises subjective medical record data and objective medical record data; performing correlation analysis on the medical resource data and the subjective medical record data by using a pre-trained correlation model so as to convert the subjective medical record data into normalized expression; utilizing a pre-trained diagnosis and treatment model to perform association query on the subjective medical record data subjected to the standardized expression to obtain association data of the subjective medical record; and performing information fusion on the associated data of the subjective medical record, the objective medical record data and the medical resource data, and performing auxiliary diagnosis on the fused multi-modal features by using a pre-trained auxiliary diagnosis model to obtain auxiliary diagnosis information.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a method, device and equipment for generating auxiliary diagnosis information. Background Art

[0002] In clinical medicine, auxiliary diagnosis is auxiliary diagnosis and is an indispensable part of the medical process. It mainly refers to assisting doctors in making accurate judgments on patients' conditions through a series of examinations, tests and evaluation methods. The corresponding auxiliary diagnosis information usually covers multiple fields such as imaging examinations, biochemical tests, and pathological examinations. For example, when a patient has unexplained abdominal pain, the doctor may arrange imaging examinations such as ultrasound or CT scans to observe the morphology and structure of the abdominal organs, so as to assist in determining whether there are lesions such as stones and tumors.

[0003] In related technologies, auxiliary diagnosis information is usually achieved by establishing a series of rule combinations based on an expert system of a knowledge base or knowledge graph. However, this expert system relies heavily on the establishment of rule combinations. When the actual situation of the patient does not fully match the corresponding rule combination, it is difficult for the expert system to clearly describe the relationship between the disease and symptoms, and it is impossible to make a correct auxiliary judgment on the disease, which affects the accuracy of the auxiliary diagnosis information. Summary of the invention

[0004] In view of this, the present application provides a method, device and equipment for generating auxiliary diagnosis information, the main purpose of which is to solve the problem that the existing technology cannot make correct auxiliary judgments on diseases, affecting the accuracy of auxiliary diagnosis information.

[0005] According to a first aspect of the present application, a method for generating auxiliary diagnosis information is provided, comprising:

[0006] Acquiring medical record data to be diagnosed and medical resource data, wherein the medical record data to be diagnosed includes subjective medical record data and objective medical record data;

[0007] Performing correlation analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model to convert the subjective medical record data into a standardized expression;

[0008] Using a pre-trained diagnosis and treatment model to perform an association query on the standardized subjective medical record data to obtain the association data of the subjective medical record, the diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the diagnosis and treatment model is used to describe the association relationship between the disease and the disease diagnosis result in the subjective medical record data;

[0009] The associated data of the subjective medical record, the objective medical record data and the medical resource data are fused, and a pre-trained auxiliary diagnosis model is used to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

[0010] Furthermore, before obtaining the medical record data to be diagnosed and the medical resource data, the method further includes:

[0011] Establishing medical master data of at least one medical service system, wherein the medical data includes a plurality of medical fields describing disease symptoms;

[0012] Based on the medical resource category system, the medical master data is divided into multiple levels to obtain medical master data with a standard level;

[0013] Medical resource data is constructed based on the medical master data with a standard hierarchy.

[0014] Furthermore, after the medical master data is divided into multiple levels based on the medical resource category system to obtain medical master data with a standard level, the method further includes:

[0015] If it is detected that the medical resources corresponding to the medical master data are changed, the fields in the medical master data are updated and maintained so that the fields in the medical master data comply with the latest data specifications and value ranges.

[0016] Furthermore, after establishing the medical master data of at least one medical service system, the method further includes:

[0017] The medical master data of the at least one medical service system are merged according to the consistency of the field descriptions in the medical master data, so that the same medical master data in different medical service systems are combined and stored and different medical master data are stored in parallel.

[0018] Furthermore, after obtaining the medical record data to be diagnosed and the medical resource data, the method further includes:

[0019] Performing standardized cleaning on each medical field in the medical resource data;

[0020] According to the blood relationship recorded in the standard cleaning process of the medical field, the medical resource data is tracked and traced;

[0021] Accordingly, the use of a pre-trained correlation model to perform correlation analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression includes:

[0022] On the basis that the medical resource data has traceability value, a pre-trained correlation model is used to perform correlation analysis on the medical resource data and the subjective medical record data, so as to convert the subjective medical record data into a standardized expression;

[0023] The medical resource data is tracked and traced according to the blood relationship recorded in the standard cleaning process of the medical field, including:

[0024] Determine the data records of the medical resource data on different processing links according to the blood relationship recorded in the standard cleaning process of the medical field;

[0025] If the data records of the medical resource data on different processing links pass the traceability verification, it is determined that the medical resource data has traceability value.

[0026] Furthermore, the use of a pre-trained correlation model to perform correlation analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression includes:

[0027] Extracting structured medical record information for auxiliary diagnosis from the subjective medical record data by using medical record structuring technology;

[0028] The structured medical record information is mapped to a standard vector having a corresponding characteristic dimension with the medical resource data using a pre-trained correlation model, thereby obtaining subjective case data with standardized expression.

[0029] Furthermore, the pre-trained diagnosis and treatment model is used to perform an association query on the standardized subjective medical record data to obtain the association data of the subjective medical record, including:

[0030] Use the pre-trained diagnosis and treatment model to perform association queries on the standardized subjective medical record data to obtain the association relationship between entities in the subjective medical record data;

[0031] Determine, based on the association relationship between entities in the subjective medical record data, an association relationship between a disease in the subjective medical record data and a disease diagnosis result, wherein the disease diagnosis result includes at least one of a disease type, a symptom, a drug, and an examination result;

[0032] According to the association relationship between the disease and the disease diagnosis result in the subjective medical record data, the association data of the subjective medical record is obtained.

[0033] According to a second aspect of the present application, a device for generating auxiliary diagnosis information is provided, comprising:

[0034] An acquisition unit, used to acquire medical record data to be diagnosed and medical resource data, wherein the medical record data to be diagnosed includes subjective medical record data and objective medical record data;

[0035] An analysis unit, configured to perform association analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model, so as to convert the subjective medical record data into a standardized expression;

[0036] A query unit, used to perform an association query on the subjective medical record data in a standardized expression using a pre-trained diagnosis and treatment model to obtain the association data of the subjective medical record, wherein the diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the diagnosis and treatment model is used to describe the association relationship between the disease and the disease diagnosis result in the subjective medical record data;

[0037] The diagnosis unit is used to fuse the associated data of the subjective medical record, the objective medical record data and the medical resource data, and use a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

[0038] Furthermore, the device also includes:

[0039] An establishing unit, used for establishing medical master data of at least one medical service system before acquiring the medical record data to be diagnosed and the medical resource data, wherein the medical data includes a plurality of medical fields describing disease symptoms;

[0040] A division unit, for dividing the medical master data into multiple levels based on the medical resource category system to obtain medical master data with a standard level;

[0041] A construction unit is used to construct medical resource data according to the medical master data with a standard hierarchy.

[0042] Furthermore, the device also includes:

[0043] The update unit is used to divide the medical master data into multiple levels based on the medical resource category system. After obtaining the medical master data with a standard level, if it is detected that the medical resources corresponding to the medical master data have changed, the fields in the medical master data are updated and maintained to make the fields in the medical master data comply with the latest data specifications and value ranges.

[0044] Furthermore, the device also includes:

[0045] A fusion unit is used to fuse the medical master data of at least one medical service system after the medical master data of at least one medical service system is established, based on the consistency of the field descriptions in the medical master data, so that the same medical master data in different medical service systems can be merged and stored and different medical master data can be stored in parallel.

[0046] Furthermore, the device also includes:

[0047] A cleaning unit, used for, after obtaining the medical record data to be diagnosed and the medical resource data, performing standardized cleaning on each medical field in the medical resource data;

[0048] A traceability unit, used to trace the medical resource data according to the blood relationship recorded in the standard cleaning process of the medical field;

[0049] Correspondingly, the analysis unit is further used to perform association analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model on the basis that the medical resource data has traceability value, so as to convert the subjective medical record data into a standardized expression;

[0050] The traceability unit is specifically used for:

[0051] Determine the data records of the medical resource data on different processing links according to the blood relationship recorded in the standard cleaning process of the medical field;

[0052] If the data records of the medical resource data on different processing links pass the traceability verification, it is determined that the medical resource data has traceability value.

[0053] Furthermore, the analysis unit is specifically used for:

[0054] Extracting structured medical record information for auxiliary diagnosis from the subjective medical record data by using medical record structuring technology;

[0055] The structured medical record information is mapped to a standard vector having a corresponding characteristic dimension with the medical resource data using a pre-trained correlation model, thereby obtaining subjective case data with standardized expression.

[0056] Furthermore, the query unit is specifically used to:

[0057] Use the pre-trained diagnosis and treatment model to perform association queries on the standardized subjective medical record data to obtain the association relationship between entities in the subjective medical record data;

[0058] Determine, based on the association relationship between entities in the subjective medical record data, an association relationship between a disease in the subjective medical record data and a disease diagnosis result, wherein the disease diagnosis result includes at least one of a disease type, a symptom, a drug, and an examination result;

[0059] According to the association relationship between the disease and the disease diagnosis result in the subjective medical record data, the association data of the subjective medical record is obtained.

[0060] According to a third aspect of the present application, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described in the first aspect when executing the computer program.

[0061] According to a fourth aspect of the present application, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0062] Through the above-mentioned technical scheme, the present application provides a method, device and equipment for generating auxiliary diagnosis information. Compared with the current prior art method of generating auxiliary diagnosis information by establishing a series of rule combinations based on an expert system of a knowledge base or a knowledge graph, the present application obtains medical record data and medical resource data to be diagnosed, and the medical record data to be diagnosed includes subjective medical record data and objective medical record data; uses a pre-trained correlation model to perform association analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression; uses a pre-trained diagnosis and treatment model to perform association query on the standardized expressed subjective medical record data to obtain the associated data of the subjective medical record; fuses the associated data of the subjective medical record, the objective medical record data and the medical resource data, and uses a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information. The entire process uses medical resource data to express the subjective data of medical records in a structured and standardized manner to obtain the associated data of subjective cases, describe the relationship between disease and symptoms through the associated data of subjective medical records, and further integrate the associated data of subjective medical records, objective medical record data and medical resource data. It can effectively reduce the dimension of multimodal medical data to achieve correct auxiliary judgment of diseases and improve the accuracy of auxiliary diagnosis information.

[0063] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0065] Figure 1 is a flow chart of a method for generating auxiliary diagnostic information in an embodiment of the present application;

[0066] Figure 2 is a flow chart of a method for generating auxiliary diagnostic information in another embodiment of the present application;

[0067] Figure 3 is a flow chart of a method for generating auxiliary diagnostic information in another embodiment of the present application;

[0068] Figure 4 yes Figure 1 A schematic flow chart of a specific implementation of step 102;

[0069] Figure 5 yes Figure 1 A schematic flow chart of a specific implementation of step 103;

[0070] Figure 6 is a schematic diagram of the structure of a device for generating auxiliary diagnostic information in an embodiment of the present application;

[0071] Figure 7 It is a schematic diagram of the device structure of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] Now the content of the present invention will be discussed with reference to several exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and thus implement the content of the present invention, rather than implying any limitation on the scope of the present invention.

[0073] As used herein, the term "including" and variations thereof are to be interpreted as open-ended terms meaning "including but not limited to." The term "based on" is to be interpreted as "based, at least in part, on." The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment." The term "another embodiment" is to be interpreted as "at least one other embodiment."

[0074] In the related art, auxiliary diagnosis information is usually achieved by establishing a series of rule combinations based on an expert system of a knowledge base or knowledge graph. However, such an expert system relies heavily on the establishment of rule combinations. When the actual situation of the patient does not completely match the corresponding rule combination, it is difficult for the expert system to clearly describe the relationship between the disease and the symptoms, and it is impossible to make a correct auxiliary judgment on the disease, which affects the accuracy of the auxiliary diagnosis information. In order to solve this problem, this embodiment provides a method for generating auxiliary diagnosis information, such as Figure 1 As shown, the method can be applied to the server corresponding to the auxiliary diagnosis system, and includes the following steps:

[0075] 101. Obtain the medical record data and medical resource data to be diagnosed.

[0076] Among them, the medical record data to be diagnosed include subjective medical record data and objective medical record data. The subjective medical record data here mainly refers to the subjective analysis and judgment of the patient's condition by medical staff, including the medical history collection content recorded by the doctor during the diagnosis and treatment process. It can be specifically expressed as the patient's description of his or her own symptoms, which can be the nature, location, onset time and other self-perceptions of pain. For example, the patient said, "I feel like my chest is being pressed by a heavy object, and it lasts about 10 minutes each time, and it will get worse after activity." It can also be expressed as the doctor's initial diagnostic ideas, differential diagnosis analysis, etc. For example, the doctor recorded, "It is considered that the patient's chest pain may be caused by myocardial ischemia, but the possibility of pleurisy is not ruled out because the patient has symptoms of coughing." The objective medical record data here mainly refers to the verifiable data obtained by medical staff through observation, examination, etc. during medical activities, including vital signs data, such as body temperature (such as 37.5℃), pulse (such as 70 times per minute), respiratory rate (such as 18 times per minute), blood pressure (such as 120 / 80mmHg), etc., as well as laboratory test data, such as white blood cell count, red blood cell count, hemoglobin content in routine blood tests, etc., and also includes imaging examination results, such as X-rays showing shadows in the lungs, CT scans showing space-occupying lesions in the liver and other specific imaging feature descriptions.

[0077] It is understandable that subjective medical record data can help us understand the doctor's diagnostic ideas and process. However, due to its high subjectivity, it will be questioned when used as evidence. Objective medical record data, as basic data for studying the occurrence and development of the disease, plays a role in assisting in verifying the diagnostic process of subjective medical record data.

[0078] In this embodiment, the medical resource data mainly includes personal health data, medical service data, disease monitoring and public health data, etc. Here, the personal health data can specifically include basic health information, family medical history data, lifestyle data, health examination data, etc. Here, the medical service data can specifically include diagnosis record data, medical expenses data, medical staff data, medical resource utilization data, etc. Here, the disease monitoring and public health data can specifically include infectious disease monitoring data, public health intervention data, environmental health data, etc.

[0079] It can be understood that medical resource data is an information collection of various resources involved in the medical service process, which includes data for quantifying and recording various elements used to provide medical services. Usually, medical resource data comes from different medical service systems, which may include medical institutions, such as hospital information management systems, laboratory information management systems, medical image storage and transmission systems, and medical equipment management systems, such as equipment maintenance and repair records, equipment usage statistics, and medical supplies management systems, such as drug inventory management systems, medical consumables management systems, blood products management systems, etc.

[0080] The executor of this embodiment can be a device or equipment for generating auxiliary diagnostic information, which can be configured on the server side of auxiliary medical diagnosis. The server side can assist the disease diagnosis process by combining medical resource data with the medical record data to be diagnosed. The whole process integrates various information to assist in the diagnosis of the medical record data to be diagnosed, which can quickly plan the diagnosis path and speed up the diagnosis process.

[0081] 102. Use a pre-trained correlation model to perform correlation analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression.

[0082] Considering that subjective medical records contain detailed descriptions of patients' own symptoms and lack corresponding standards and norms for diagnosis and treatment, association analysis can combine patients' subjective feelings with appropriate diagnostic tools and professionals. For example, when a patient describes squeezing pain in the chest that radiates to the left shoulder, association analysis can convert the patient's description into a chest examination using an electrocardiograph and arrange for a cardiovascular physician to make a diagnosis, thereby improving the accuracy of the diagnosis of diseases such as coronary heart disease. Association analysis can also combine patients' subjective feelings with standardized descriptions. For example, when a patient describes "coughing and wheezing 1 day ago", association analysis can convert the patient's description into [coughing, 1 day], [wheezing, 1 day].

[0083] In this embodiment, the correlation model is a mathematical model used to measure the degree of association between two or more variables, and can be used to analyze the relationship between medical resource data and subjective medical record data, for example, the degree of association between patient symptoms and standard symptom descriptions, and for example, the degree of association between patient symptoms and required medical equipment and medical staff.

[0084] The correlation model here can use a deep learning model, such as a transformer architecture model, or a machine learning model, such as a decision tree, support vector machine, etc.

[0085] Specifically, feature extraction can be performed on medical resource data and subjective medical record data respectively, and the features extracted from the medical resource data and subjective medical record data can be combined to construct a correlation feature matrix, which is input into a pre-trained correlation model. The correlation model mainly performs two aspects of correlation analysis on the correlation feature matrix. On the one hand, there is semantic correlation analysis, which is a process of calculating the similarity between the features extracted from the medical resource data in the correlation feature matrix and the features extracted from the subjective medical record data to obtain a semantic similarity score. On the other hand, there is dependency correlation analysis, which is a process of calculating the dependency between the features extracted from the medical resource data in the correlation feature matrix and the features extracted from the subjective medical record data using association rules to obtain a dependency correlation score. Further, according to the semantic correlation analysis process, the subjective medical record data with a semantic similarity score greater than a preset value is converted into a standardized expression of the semantics corresponding to the medical resource data. According to the dependency correlation analysis, the subjective medical record data with a dependency correlation score greater than a preset value is converted into a standardized expression of the dependency corresponding to the medical resource data.

[0086] For example, for the symptom characteristics of patients, semantic association can be used to perform semantic normalization expression on the patient's symptom feature vector according to the feature vector with corresponding symptom expression in the medical resource data to obtain a symptom feature vector with normalized semantic expression. Dependency association expression can be used to perform dependency normalization expression on the patient's symptom feature vector according to the feature vector with corresponding symptom association in the medical resource data to obtain a feature vector combination with normalized dependency expression, which may include a combination of the patient's symptom feature vector and a feature vector of a medical device that may be used to diagnose the symptom, or a combination of the patient's symptom feature vector and a feature vector of a medical staff that is good at treating related diseases.

[0087] 103. Use the pre-trained diagnosis and treatment model to perform associated query on the standardized subjective medical record data to obtain the associated data of the subjective medical record.

[0088] In this embodiment, the pre-trained diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the knowledge graph is a graph structure composed of the relationship between entities. In the process of constructing the medical knowledge graph, an entity system of a certain type of medical concept (such as diagnosis) can be constructed by medical terminology through ontology, and then the relationship between these entities can be further established through medical guidelines, consensus, and textbooks, and improved through actual medical records. The diagnosis and treatment model is used to describe the association between diseases and disease diagnosis results in subjective medical record data, where the disease diagnosis results include at least one of the disease type, symptoms, drugs, and examination results. It can be understood that subjective medical record data with standardized expressions have standardized semantic expressions, and key symptoms and possible disease categories can be identified in subjective medical record data with standardized expressions through the diagnosis and treatment model. For example, for the input "symptoms: chest pain, radiating to the left shoulder; duration: 2 hours; severity: severe", the diagnosis and treatment model can obtain acute myocardial infarction or other cardiovascular emergencies through association queries, and can also obtain resources such as medical equipment for diagnosis and treatment.

[0089] Specifically, in the process of associative query on the standardized subjective medical record data, the diagnosis and treatment model can be used as a feature extractor to extract feature vectors from the subjective case data. For example, information such as symptoms, onset time, severity, etc. is converted into a high-dimensional feature vector, which represents the patient's condition characteristics. The extracted feature vector is then matched with the nodes in the knowledge graph model using the relationship between the medical knowledge stored in the diagnosis and treatment model, and queries are performed along the edges of the knowledge graph model to obtain the associated data of the subjective medical record.

[0090] 104. Fuse the associated data of the subjective medical record, the objective medical record data and the medical resource data, and use a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

[0091] In this embodiment, information fusion can sequentially splice the feature vectors corresponding to subjective medical record-related data, objective medical record data, and medical resource data by vector splicing to form a new multimodal feature vector. For example, the disease possibility probability vector and the examination suggestion coding vector in the subjective medical record-related data are first spliced, and then the vital signs vector and the laboratory test result vector in the objective medical record data are connected, and finally the equipment function vector and the medical staff feature vector in the medical resource data are added. Information fusion can also fuse the feature vectors corresponding to subjective medical record-related data, objective medical record data, and medical resource data by semantic embedding fusion to form a new multimodal feature vector. For example, for subjective medical record data, the text information in the subjective medical record associated data (such as disease name, examination recommendations, etc.) is converted into a semantic embedding vector. For objective medical record data, the non-text information such as examination results in the objective medical record data is converted into a semantic embedding vector by combining numerical values ​​with semantic labels (such as "white blood cell count is 10×10 / L (normal)") using a word embedding model. For medical resource data, the equipment and medical staff information in the medical resource data can also be similarly embedded. The semantic embedding vectors are finally fused by weighted summation, averaging, etc. to obtain the fused multimodal semantic embedding features. Information fusion can also fuse the feature vectors corresponding to the subjective medical record associated data, objective medical record data, and medical resource data based on the established fusion rules to form a new multimodal feature vector. For example, if the subjective medical record associated data suggests that it may be an infectious disease, and the white blood cell count in the objective medical record data is elevated, then the association of "infectious disease-inflammatory index" is highlighted in the fusion feature. For medical resource data, if the hospital has sufficient inventory of special drugs for infectious diseases and medical staff who are good at treating infectious diseases, this information will be integrated into the overall features according to the rules, such as increasing the weight of these resources or marking them as high priority.

[0092] Specifically, in the process of using a pre-trained auxiliary diagnosis model to assist in the diagnosis of the fused multimodal features, the auxiliary diagnosis model can use a deep learning diagnostic model, such as a convolutional neural network, a recurrent neural network, or an integrated model that combines multiple machine learning algorithms. The auxiliary diagnosis model can process the fused multimodal features through its pre-trained weights and network structure. For example, the convolution layer of the model may extract local patterns of image and numerical features in the fused features, the recurrent layer processes time series-related features (such as the development process of symptoms), and the fully connected layer makes a comprehensive judgment. Ultimately, the model outputs auxiliary diagnosis information, which may include disease diagnosis results (such as specific disease names, disease probabilities), treatment recommendations (such as recommended treatment plans, drug use recommendations), and medical resource allocation recommendations (such as the inspection equipment that needs to be used and the arrangement of medical staff).

[0093] Furthermore, in order to ensure the accuracy of auxiliary diagnosis information, the auxiliary diagnosis information can be explained. For example, if the auxiliary diagnosis model outputs a high probability of a certain disease, analyze which fused features contribute more to this diagnosis result, and use the feature importance analysis method of the auxiliary diagnosis model (such as calculating the importance score of the feature, checking the weight of the feature in the decision-making process, etc.) to explain the auxiliary diagnosis information. For treatment suggestions and resource allocation suggestions, combine medical knowledge and actual conditions to explain their rationality.

[0094] The method for generating auxiliary diagnosis information provided in the embodiment of the present application is compared with the current prior art method of generating auxiliary diagnosis information by establishing a series of rule combinations based on an expert system of a knowledge base or a knowledge graph. The present application obtains medical record data to be diagnosed and medical resource data, and the medical record data to be diagnosed includes subjective medical record data and objective medical record data; uses a pre-trained correlation model to perform association analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression; uses a pre-trained diagnosis and treatment model to perform association query on the standardized expressed subjective medical record data to obtain the associated data of the subjective medical record; fuses the associated data of the subjective medical record, the objective medical record data and the medical resource data, and uses a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information. The entire process uses medical resource data to express the subjective data of medical records in a structured and standardized manner to obtain the associated data of subjective cases, describe the relationship between disease and symptoms through the associated data of subjective medical records, and further integrate the associated data of subjective medical records, objective medical record data and medical resource data. It can effectively reduce the dimension of multimodal medical data to achieve correct auxiliary judgment of diseases and improve the accuracy of auxiliary diagnosis information.

[0095] Considering that the medical resource data comes from different medical service systems, it is necessary to perform corresponding standardized processing on the medical resource data during the process of integrating the medical resource data. Furthermore, in this embodiment, Figure 2 As shown, before step 101, the method further includes the following steps:

[0096] 201. Establish medical master data for at least one medical service system.

[0097] 202. Based on the medical resource category system, the medical master data is divided into multiple levels to obtain medical master data with a standard level.

[0098] 203. Construct medical resource data based on the medical master data with standard hierarchy.

[0099] Among them, medical master data is data shared between medical service systems. It is not all business data within the medical service system, but data that needs to be shared between various medical service systems. It is high-value data that can be reused across businesses, mainly including basic patient information, medical service information, medical resource information, clinical data, etc. Basic patient information may include but is not limited to personal identity information, medical insurance information, etc. Medical service information may include but is not limited to medical records, medical expenses information, etc. Medical resource information may include but is not limited to medical staff information, medical equipment and facilities information, etc. Clinical data may include but is not limited to diagnostic data, treatment data, etc.

[0100] Specifically, according to the functional requirements and business processes of the medical service system, the scope and classification of medical master data can be accurately determined, and a data model for storing medical master data can be designed. For medical master data with simple structures, a relational database model can be used to store the medical master data. For medical master data with complex structures, it can be stored in combination with a non-relational database, or saved in a relational database using large text fields. Finally, data migration tools can be used to migrate the corresponding data from different medical service systems to the medical master data for storage.

[0101] In this embodiment, the process of multi-level division of medical master data can be based on the application of medical, management, research and other departments, and the medical master data can be divided into multiple levels with reference to the category data standards of diseases, drugs and environment. For example, the medical master data is divided into subcategories such as diseases and drugs according to the application resource directory system, and each subcategory is divided according to its characteristics. The corresponding disease subcategories may include the following categories: 1. Subtotal of certain infectious diseases and parasitic diseases, 2. Subtotal of tumors, 3. Subtotal of blood and hematopoietic organ diseases and certain diseases involving immune mechanisms, 4. Subtotal of endocrine, nutritional and metabolic diseases, 5. Subtotal of mental and behavioral disorders, 6. Subtotal of nervous system diseases, 7. Subtotal of eye diseases, 8. Subtotal of endocrine diseases, nutritional and metabolic diseases, 9. Subtotal of endocrine diseases, 10. Subtotal of endocrine diseases, 11. Subtotal of endocrine diseases, 12. Subtotal of endocrine diseases, 13. Subtotal of endocrine diseases, 14. Subtotal of endocrine diseases, 15. Subtotal of endocrine diseases, 16. Subtotal of endocrine diseases, 17. Subtotal of endocrine diseases, 18. Subtotal of endocrine diseases, 19. Subtotal of endocrine diseases, 20. Subtotal of endocrine diseases, 21. Subtotal of endocrine diseases, 22. Subtotal of endocrine diseases, 23. Subtotal of endocrine diseases, 24. Subtotal of endocrine diseases, 25. Subtotal of endocrine diseases, 26. Subtotal of endocrine diseases, 27. Subtotal of endocrine diseases, 28. Subtotal of endocrine diseases, 29. Subtotal of endocrine diseases, 30. Subtotal of endocrine diseases, 31. 8. Diseases of the ear and mastoid process; 9. Diseases of the circulatory system; 10. Diseases of the respiratory system; 11. Diseases of the digestive system; 12. Diseases of the skin and subcutaneous tissue; 13. Diseases of the musculoskeletal system and connective tissue; 14. Diseases of the genitourinary system; 15. Pregnancy, childbirth and the puerperium; 16. Certain conditions originating in the perinatal period; 17. Congenital malformations, deformations and chromosomal abnormalities; 18. Symptoms, signs and abnormal clinical and experimental findings that cannot be attributed elsewhere; 19. Injury, poisoning and certain other consequences of external causes; 20. Factors affecting health status and exposure to health care institutions.

[0102] It should be noted that each medical service system has corresponding medical master data, but it is difficult for these medical master data to be provided by not one medical service system. On the one hand, medical master data is often related to the specific design of each database in the medical service system. The corresponding medical master data is maintained by multiple medical service systems, and there is no unified standard for different medical service systems. On the other hand, medical master data is constantly evolving and has a lot to do with the business development of the hospital. For example, after a hospital introduces a new testing instrument, the reference value range of this instrument becomes part of the medical master data. In other words, the medical master data provided by different medical service systems needs to have a field maintenance method. Accordingly, based on the medical resource category system, the medical master data is divided into multiple levels. After obtaining medical master data with a standard level, the method also includes the following steps:

[0103] If it is detected that the medical resources corresponding to the medical master data are changed, the fields in the medical master data are updated and maintained so that the fields in the medical master data comply with the latest data specifications and value ranges.

[0104] In this embodiment, some important medical master data, such as clinical diagnosis and treatment guidelines, medical insurance policy-related data, etc., may change with time and knowledge update. Version management can be used to record different versions of data and their applicable time ranges. For example, when the medical insurance reimbursement policy is adjusted, the old reimbursement policy version data is saved and the effective date and expiration date are marked so that historical data can be queried when needed. Specifically, before updating and maintaining the fields in the medical master data, it is necessary to clarify the update mechanism of the medical master data to detect whether the medical resources corresponding to the medical master data have changed. For example, when the patient's basic information changes (such as contact information changes, medical insurance type changes, etc.), the patient or medical staff should modify it through the update interface provided by the system; after each visit or treatment, the medical service information is updated by the medical staff in a timely manner. Related records, such as updating outpatient diagnosis results, hospitalization course records, etc., if changes occur, the fields in the medical master data are updated and maintained so that the fields in the medical master data meet the latest data specifications and value ranges.

[0105] Accordingly, after establishing the medical master data of at least one medical service system, the method further includes the following steps:

[0106] The medical master data of the at least one medical service system are merged according to the consistency of the field descriptions in the medical master data, so that the same medical master data in different medical service systems are combined and stored and different medical master data are stored in parallel.

[0107] In this embodiment, the medical master data of at least one medical service system can be integrated by seeking common ground while reserving differences. That is, the medical master data that can be merged in different medical service systems are merged and stored in the same project. For example, the medical master data with exactly the same name and meaning are merged and stored. The medical master data that cannot be merged in different medical service systems are stored in different projects. For example, medical resource data with different names are stored in parallel. This can ensure that the subsequent medical master data will not cause any information loss or distortion during the integration process.

[0108] It should be noted that, considering that there will be a certain redundancy in the fusion process of medical master data, for example, in the medical master data where "fever" and "fever" may exist at the same time, professionals with medical knowledge can merge and adjust the fused medical master data to reduce the redundancy of the medical master data.

[0109] Considering the data, further, in this embodiment, if Figure 3 As shown, after step 101, the method further includes the following steps:

[0110] 301. Perform standard cleaning on each medical field in the medical resource data.

[0111] 302. According to the blood relationship recorded in the standard cleaning process of the medical field, the medical resource data is tracked and traced.

[0112] Correspondingly, in step 102, on the basis that the medical resource data has traceability value, a pre-trained correlation model is used to perform correlation analysis on the medical resource data and the subjective medical record data, so as to convert the subjective medical record data into a standardized expression.

[0113] In this embodiment, after the medical master data is fused, the platform can clean and normalize the fused medical master data. The cleaning process can find unreasonable data based on the data specification of each field in the medical master data, and set modification rules. For medical master data that cannot be directly normalized, the platform can mark this part of the medical master data, and use the model to find similar medical master data, and use the modification rules corresponding to similar medical master data for normalization. It can be understood that since the medical master data all come from a data unit in the original data, the corresponding medical master data will record the blood relationship of the data during the cleaning and normalization process, so that each data unit can be traced during cleaning and normalization.

[0114] Specifically, in the process of tracing and tracing the medical resource data according to the blood relationship recorded by the medical field in the standard cleaning process, the data records of the medical resource data on different processing links can be determined according to the blood relationship recorded by the medical field in the standard cleaning process. If the data records of the medical resource data on different processing links pass the traceability verification, the medical resource data is determined to have traceability value. For example, when it is necessary to trace the source and change process of a certain medical master data, the corresponding medical master data is audited and monitored by querying the data records of the medical resource data on different processing links. If the audited and monitored medical master data meets the security compliance, the medical resource data is determined to have traceability value.

[0115] In the above embodiment, by tracking and tracing the medical resource data, it is helpful to locate the abnormal steps in the medical master data processing process to discover the corresponding abnormal data source, thereby facilitating the authorized use of the medical master data and the backtracking of the data value.

[0116] In the above embodiment, specifically, Figure 4 As shown, step 102 includes the following steps:

[0117] 401. Extract structured medical record information for auxiliary diagnosis from the subjective medical record data through medical record structuring technology.

[0118] 402. Use a pre-trained correlation model to map the structured medical record information to a standard vector having a corresponding characteristic dimension with the medical resource data, so as to obtain subjective case data with standardized expression.

[0119] In this embodiment, for some medical fields described in natural language (such as chief complaint, current medical history, physical examination, etc.), the corresponding key information such as symptoms, examinations, medication, and treatment can be extracted and normalized from these fields through natural language processing (NLP) technology. Accordingly, the extracted key information is used to annotate the sample data to be learned, so that the model can be trained through the annotated sample data, the versatility of the model can be expanded, and the accuracy of the model can be improved.

[0120] Specifically, in the process of extracting structured medical record information for auxiliary diagnosis from subjective medical record data through structured technology, considering that the characteristics of subjective medical record data are relatively simple language structure, a large number of professional vocabulary, and many long words, a deep learning model can be used here to form an understanding of the basic meaning and medical knowledge, and then the subjective medical record data is input into the trained information extraction model, and all possible word segmentation sequence probabilities are calculated, and the sequence with the highest probability is found as the result of word segmentation of the subjective medical data, and then medical entities are extracted based on the word segmentation results. A deep learning model similar to word segmentation can be used here. On the basis of manual annotation, important medical information in the medical records written by doctors can be extracted, and the corresponding mapping relationship between medical entity information can be achieved to realize the structuring of subjective medical record data.

[0121] Furthermore, a correlation model based on a deep learning language expression model can be used to establish a correlation model. During the model training process, each pair of structured mapping relationships annotated with structured medical record information is used for training. Through the trained correlation model, medical resource data can be mapped to a standard vector with a corresponding characteristic dimension to the medical resource data, and subjective case data with standardized expression can be obtained. Taking drugs as an example, when the medical service system needs to find the medical master data closest to the platform, the medical master data of the medical service structure can be mapped to a 256-dimensional vector v, and then the vector closest to this vector is found from all vectors {wi} of the medical master data corresponding to the platform drug name, that is, wi*v is closest to 1, and the similarity is calculated. If the similarity exceeds 50% (or a higher set value), it is considered that the two medical master data have a corresponding standardized association, and the subjective medical record data is further converted into the standardized expression of the medical master data.

[0122] In the above embodiment, specifically, Figure 5 As shown, step 103 includes the following steps:

[0123] 501. Use the pre-trained diagnosis and treatment model to perform association query on the standardized subjective medical record data to obtain the association relationship between entities in the subjective medical record data.

[0124] 502. Determine the association relationship between the disease and the disease diagnosis result in the subjective medical record data according to the association relationship between the entities in the subjective medical record data.

[0125] 503. Obtain the associated data of the subjective medical record according to the associated relationship between the disease and the disease diagnosis result in the subjective medical record data.

[0126] In this embodiment, the pre-trained diagnosis and treatment model is a knowledge graph model constructed through the relationship between entities. After describing the relationship between the medical concept ontology and the ontology with the knowledge graph, the association between the disease and the disease diagnosis result in the subjective medical record data can be clearly described. The disease diagnosis result includes at least one of the disease type, symptoms, drugs and test results. The corresponding association between the disease and the disease diagnosis result includes the association between diseases (for example, lung disease and bronchitis are both respiratory diseases), the connection between diseases and symptoms, the association between diseases and drugs, the association between diseases and test results, etc. The diagnosis and treatment model established through the knowledge graph can not only provide medical workers with better correlation query and search, but also provide reasoning basis for disease diagnosis and treatment, help clinicians assist in decision-making, and can also be used for clinical quality monitoring to find unreasonable aspects in diagnosis and treatment.

[0127] In this embodiment, based on structured medical record information and combined with objective medical record data such as tests and examinations, a patient's condition can be expressed using a high-dimensional (tens of thousands of dimensions) vector. Here, each dimension of the vector corresponds to the presence or absence of a symptom, physical examination results, test results, medical history records, treatments already taken, etc. Based on this high-dimensional vector and the doctor's actual diagnosis, prescription, and treatment, a corresponding auxiliary diagnosis model is established to learn from the doctor's diagnosis and treatment experience.

[0128] Specifically, what is learned through the model is actually the probability of giving a diagnosis based on all clinical observations. Common technologies often use a two-layer architecture. The first layer consists of multiple sub-model systems, which process a single type of information and make a preliminary diagnosis; the second layer integrates the preliminary diagnostic information of each sub-model system and makes a second judgment. However, this type of structure has the following problems: first, the judgment of each subsystem depends only on the internal information of the subsystem and cannot be assisted by other information; second, the system is difficult to expand linearly. To solve the above problems, this embodiment adopts a hierarchical high-dimensional vector prediction model, which can effectively reduce the dimensionality of multimodal data through a deep neural network, and ultimately achieve effective diagnosis of the disease. The entire system has a three-layer architecture:

[0129] The first layer is the embedding layer, which completes the dimensionality reduction and projection of high-dimensional data into the "low-dimensional space", realizing data dimensionality reduction while maintaining the inline relationship of information. Different neural networks are used to reduce the dimensionality according to different types of information. For example, CNN can be used to reduce the dimensionality of image data.

[0130] The second layer is the deep fusion layer, which combines various information from different types into a unified "super long vector" as input, completes information fusion through a multi-layer neural network, and generates high-order features;

[0131] The third layer is the prediction layer, which uses high-order features as input and uses the commonly used softmax or multiple independent logistic regressions to make the final judgment.

[0132] It is understandable that compared with many existing models, the multimodal hierarchical high-dimensional vector prediction model fully integrates different types of heterogeneous information, utilizes the implicit information between various types of information, and uses the embedding method to provide a unified architecture and cooperate with deep learning methods, which can easily achieve linear expansion of the system.

[0133] Through the above process, different types of information (structured chief complaint, current medical history, test results, imaging results, etc.) can be combined into a unified "very long vector" as input, and information fusion is completed through multiple embedding and dimensionality reduction to generate lower-dimensional high-order features. These high-order features can be used as a guarantee of the number of medical records for disease auxiliary diagnosis, and high-quality auxiliary diagnosis effects can be achieved.

[0134] Further, as Figure 1-5 The specific implementation of the method, the embodiment of the present application provides a device for generating auxiliary diagnosis information, such as Figure 6 As shown, the device includes: an acquisition unit 61, an analysis unit 62, a query unit 63, and a diagnosis unit 64.

[0135] An acquisition unit 61 is used to acquire medical record data to be diagnosed and medical resource data, wherein the medical record data to be diagnosed includes subjective medical record data and objective medical record data;

[0136] An analysis unit 62, configured to perform correlation analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model, so as to convert the subjective medical record data into a standardized expression;

[0137] A query unit 63 is used to perform an association query on the standardized expressed subjective medical record data using a pre-trained diagnosis and treatment model to obtain the association data of the subjective medical record, wherein the diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the diagnosis and treatment model is used to describe the association relationship between the disease and the disease diagnosis result in the subjective medical record data;

[0138] The diagnosis unit 64 is used to fuse the associated data of the subjective medical record, the objective medical record data and the medical resource data, and use a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

[0139] The auxiliary diagnosis information generation device provided by the embodiment of the present invention is compared with the current prior art method of generating auxiliary diagnosis information by establishing a series of rule combinations based on the expert system of the knowledge base or knowledge graph. The present application obtains the medical record data to be diagnosed and the medical resource data, and the medical record data to be diagnosed includes subjective medical record data and objective medical record data; uses a pre-trained correlation model to perform association analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression; uses a pre-trained diagnosis and treatment model to perform association query on the standardized expression of the subjective medical record data to obtain the associated data of the subjective medical record; fuses the associated data of the subjective medical record, the objective medical record data and the medical resource data, and uses a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information. The entire process uses medical resource data to express the subjective data of medical records in a structured and standardized manner to obtain the associated data of subjective cases, describe the relationship between disease and symptoms through the associated data of subjective medical records, and further integrate the associated data of subjective medical records, objective medical record data and medical resource data. It can effectively reduce the dimension of multimodal medical data to achieve correct auxiliary judgment of diseases and improve the accuracy of auxiliary diagnosis information.

[0140] In a specific application scenario, the device further includes:

[0141] An establishing unit, used for establishing medical master data of at least one medical service system before acquiring the medical record data to be diagnosed and the medical resource data, wherein the medical data includes a plurality of medical fields describing disease symptoms;

[0142] A division unit, for dividing the medical master data into multiple levels based on the medical resource category system to obtain medical master data with a standard level;

[0143] A construction unit is used to construct medical resource data according to the medical master data with a standard hierarchy.

[0144] In a specific application scenario, the device further includes:

[0145] The update unit is used to divide the medical master data into multiple levels based on the medical resource category system. After obtaining the medical master data with a standard level, if it is detected that the medical resources corresponding to the medical master data have changed, the fields in the medical master data are updated and maintained to make the fields in the medical master data comply with the latest data specifications and value ranges.

[0146] In a specific application scenario, the device further includes:

[0147] A fusion unit is used to fuse the medical master data of at least one medical service system after the medical master data of at least one medical service system is established, based on the consistency of the field descriptions in the medical master data, so that the same medical master data in different medical service systems can be merged and stored and different medical master data can be stored in parallel.

[0148] In a specific application scenario, the device further includes:

[0149] A cleaning unit, used for, after obtaining the medical record data to be diagnosed and the medical resource data, performing standardized cleaning on each medical field in the medical resource data;

[0150] A traceability unit, used to trace the medical resource data according to the blood relationship recorded in the standard cleaning process of the medical field;

[0151] Correspondingly, the analysis unit is further used to perform association analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model on the basis that the medical resource data has traceability value, so as to convert the subjective medical record data into a standardized expression;

[0152] The traceability unit is specifically used for:

[0153] Determine the data records of the medical resource data on different processing links according to the blood relationship recorded in the standard cleaning process of the medical field;

[0154] If the data records of the medical resource data on different processing links pass the traceability verification, it is determined that the medical resource data has traceability value.

[0155] In a specific application scenario, the analysis unit is specifically used to:

[0156] Extracting structured medical record information for auxiliary diagnosis from the subjective medical record data by using medical record structuring technology;

[0157] The structured medical record information is mapped to a standard vector having a corresponding characteristic dimension with the medical resource data using a pre-trained correlation model, thereby obtaining subjective case data with standardized expression.

[0158] In a specific application scenario, the query unit is specifically used to:

[0159] Use the pre-trained diagnosis and treatment model to perform association queries on the standardized subjective medical record data to obtain the association relationship between entities in the subjective medical record data;

[0160] Determine, based on the association relationship between entities in the subjective medical record data, an association relationship between a disease in the subjective medical record data and a disease diagnosis result, wherein the disease diagnosis result includes at least one of a disease type, a symptom, a drug, and an examination result;

[0161] According to the association relationship between the disease and the disease diagnosis result in the subjective medical record data, the association data of the subjective medical record is obtained.

[0162] It should be noted that for other corresponding descriptions of the functional units involved in the device for generating auxiliary diagnostic information provided in this embodiment, reference can be made to Figure 1-Figure 5 The corresponding description in will not be repeated here.

[0163] Based on the above Figure 1-Figure 5 The method shown in the embodiment of the present application accordingly provides a storage medium on which a computer program is stored, and when the program is executed by a processor, the above-mentioned Figure 1-Figure 5 The method for generating auxiliary diagnostic information shown.

[0164] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0165] Based on the above Figure 1-Figure 5 The method shown, and Figure 6 In order to achieve the above-mentioned purpose, the embodiment of the present application also provides a physical device for assisting in the generation of diagnostic information, which can be a computer, a smart phone, a tablet computer, a smart watch, a server, or a network device, etc. The physical device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to achieve the above-mentioned Figure 1-Figure 5 The method for generating auxiliary diagnostic information shown.

[0166] Optionally, the physical device may also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0167] In an exemplary embodiment, see Figure 7The physical device includes a communication bus, a processor, a memory and a communication interface, and may also include an input / output interface and a display device, wherein each functional unit can communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory and execute the method for generating auxiliary diagnostic information in the above embodiment.

[0168] Those skilled in the art will appreciate that the physical device structure for generating auxiliary diagnostic information provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or a combination of certain components, or different component arrangements.

[0169] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the physical device for generating the auxiliary diagnostic information, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing physical device.

[0170] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. By applying the technical solution of the present application, compared with the current existing methods, the present application uses medical resource data to express the subjective data of medical records in a structured and standardized manner to obtain the associated data of subjective cases, and describes the relationship between diseases and symptoms through the associated data of subjective medical records. The associated data of subjective medical records, objective medical record data and medical resource data are further integrated, which can effectively reduce the dimensionality of multimodal medical data to achieve correct auxiliary judgment of diseases and improve the accuracy of auxiliary diagnosis information.

[0171] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0172] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.

Claims

1. A method for generating auxiliary diagnosis information, characterized in that: include: Acquire medical record data to be diagnosed and medical resource data, wherein the medical record data to be diagnosed includes subjective medical record data and objective medical record data; Performing correlation analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model to convert the subjective medical record data into a standardized expression; Using a pre-trained diagnosis and treatment model to perform an association query on the standardized subjective medical record data to obtain the association data of the subjective medical record, the diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the diagnosis and treatment model is used to describe the association relationship between the disease and the disease diagnosis result in the subjective medical record data; The associated data of the subjective medical record, the objective medical record data and the medical resource data are fused, and a pre-trained auxiliary diagnosis model is used to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

2. The method according to claim 1, characterized in that Before obtaining the medical record data to be diagnosed and the medical resource data, the method further includes: Establishing medical master data of at least one medical service system, wherein the medical data includes a plurality of medical fields describing disease symptoms; Based on the medical resource category system, the medical master data is divided into multiple levels to obtain medical master data with a standard level; Medical resource data is constructed based on the medical master data with a standard hierarchy.

3. The method according to claim 2, characterized in that After the medical master data is divided into multiple levels based on the medical resource category system to obtain medical master data with a standard level, the method further includes: If it is detected that the medical resources corresponding to the medical master data are changed, the fields in the medical master data are updated and maintained so that the fields in the medical master data comply with the latest data specifications and value ranges.

4. The method according to claim 2, characterized in that: After establishing the medical master data of at least one medical service system, the method further includes: The medical master data of the at least one medical service system are merged according to the consistency of the field descriptions in the medical master data, so that the same medical master data in different medical service systems are combined and stored and different medical master data are stored in parallel.

5. The method according to any one of claims 1 to 4, characterized in that After obtaining the medical record data to be diagnosed and the medical resource data, the method further includes: Performing standardized cleaning on each medical field in the medical resource data; According to the blood relationship recorded in the standard cleaning process of the medical field, the medical resource data is tracked and traced; Accordingly, the use of a pre-trained correlation model to perform correlation analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression includes: On the basis that the medical resource data has traceability value, a pre-trained correlation model is used to perform correlation analysis on the medical resource data and the subjective medical record data, so as to convert the subjective medical record data into a standardized expression; The medical resource data is tracked and traced according to the blood relationship recorded in the standard cleaning process of the medical field, including: Determine the data records of the medical resource data on different processing links according to the blood relationship recorded in the standard cleaning process of the medical field; If the data records of the medical resource data on different processing links pass the traceability verification, it is determined that the medical resource data has traceability value.

6. The method according to any one of claims 1 to 4, characterized in that The using of a pre-trained correlation model to perform correlation analysis on the medical resource data and the subjective medical record data to convert the subjective medical record data into a standardized expression includes: Extracting structured medical record information for auxiliary diagnosis from the subjective medical record data by using medical record structuring technology; The structured medical record information is mapped to a standard vector having a corresponding characteristic dimension with the medical resource data using a pre-trained correlation model, thereby obtaining subjective case data with standardized expression.

7. The method according to any one of claims 1 to 4, characterized in that The pre-trained diagnosis and treatment model is used to perform an association query on the standardized subjective medical record data to obtain the association data of the subjective medical record, including: Use the pre-trained diagnosis and treatment model to perform association queries on the standardized subjective medical record data to obtain the association relationship between entities in the subjective medical record data; Determine, based on the association relationship between entities in the subjective medical record data, an association relationship between a disease in the subjective medical record data and a disease diagnosis result, wherein the disease diagnosis result includes at least one of a disease type, a symptom, a drug, and an examination result; According to the association relationship between the disease and the disease diagnosis result in the subjective medical record data, the association data of the subjective medical record is obtained.

8. A device for generating auxiliary diagnostic information, characterized in that: include: An acquisition unit, used to acquire medical record data to be diagnosed and medical resource data, wherein the medical record data to be diagnosed includes subjective medical record data and objective medical record data; An analysis unit, configured to perform association analysis on the medical resource data and the subjective medical record data using a pre-trained correlation model, so as to convert the subjective medical record data into a standardized expression; A query unit, used to perform an association query on the subjective medical record data in a standardized expression using a pre-trained diagnosis and treatment model to obtain the association data of the subjective medical record, wherein the diagnosis and treatment model is a knowledge graph model constructed by the relationship between entities, and the diagnosis and treatment model is used to describe the association relationship between the disease and the disease diagnosis result in the subjective medical record data; The diagnosis unit is used to fuse the associated data of the subjective medical record, the objective medical record data and the medical resource data, and use a pre-trained auxiliary diagnosis model to perform auxiliary diagnosis on the fused multimodal features to obtain auxiliary diagnosis information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for generating auxiliary diagnosis information according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for generating auxiliary diagnosis information according to any one of claims 1 to 7 are implemented.

Images