A hospital system interaction method based on data association matching

Abstract

The present application discloses a hospital system interaction method based on data association matching, which relates to the field of data association matching, and the method includes: an online hospital system obtains a business request through a patient client; the online hospital system converts the business request into a consultation request information through a front-end network and transmits it to an offline hospital system; the offline hospital system parses the consultation request information to obtain the patient's disease information and the patient's basic information; the offline hospital system recommends multiple departments to the patient based on the patient's disease information and the patient's basic information, matches the patient with a suitable doctor based on the recommended departments, and obtains a doctor-patient matching result; the offline hospital system transmits the doctor-patient matching result to the online hospital system through a front-end gateway. The present invention has the effect of realizing business interaction and information synchronization between the online hospital system and the offline hospital system.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of hospital system interaction, and in particular, to a hospital system interaction method based on data association matching. Background Art

[0002] With the rapid development of 4G and 5G mobile communication technologies and mobile Internet technologies, the informatization process of the medical system has also progressed rapidly. Hospitals of all levels and types have successively established online hospital systems that cover the entire process of patients seeking medical treatment in hospitals. Online payment, online appointments, and online viewing of electronic medical records can be realized through the online hospital system. Patients can make online appointments with relevant departments or doctors in advance through the online hospital system, and then go to the hospital where they have made an appointment for offline consultation.

[0003] However, for patients who have difficulty traveling, it is often difficult to go to the hospital independently for offline consultations, so the above basic functions cannot meet their needs. At present, hospitals on the market have also begun to provide online telemedicine services to patients through online hospital systems. However, since the online hospital system is independent of the traditional offline hospital management system, there is no business and information interaction between the two systems, resulting in information asynchrony between the online hospital system and the offline hospital system, which is prone to duplicate appointments and duplicate consultations, resulting in a waste of medical resources and increased medical costs. Summary of the invention

[0004] The embodiment of the present application provides a hospital system interaction method based on data association matching, which is used to solve the problem of information asynchrony between online hospital systems and offline hospital systems in the prior art.

[0005] To achieve the above objectives, the present application provides a hospital system interaction method based on data association matching, the method comprising:

[0006] The online hospital system obtains a service request through a patient client held by the target patient, and sends the service request to the front-end gateway;

[0007] The front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system;

[0008] The offline hospital system analyzes the consultation request information to obtain the patient disease information and patient basic information of the target patient;

[0009] The offline hospital system selects a plurality of recommended departments that match the patient's disease information and the patient's basic information;

[0010] The offline hospital system retrieves corresponding target doctor information from a preset hospital database based on the plurality of recommended departments;

[0011] The offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information to obtain a doctor-patient matching result;

[0012] The offline hospital system transmits the doctor-patient matching result to the online hospital system through the front-end gateway.

[0013] Optionally, the offline hospital system selects multiple recommended departments that match the patient's disease information and the patient's basic information, including the following steps:

[0014] The offline hospital system extracts a plurality of disease symptom keywords from the patient's disease information using a TF-IDF algorithm, and integrates the plurality of disease symptom keywords into a disease keyword set;

[0015] The offline hospital system retrieves the patient historical medical record information of the target patient from a preset hospital database based on the patient basic information, wherein the patient historical medical record information includes the patient's disease history information;

[0016] The offline hospital system extracts a plurality of disease history symptom keywords from the patient's disease history information using the TF-IDF algorithm, and integrates the plurality of disease history symptom keywords into a disease history keyword set;

[0017] The offline hospital system calculates the word frequency vectors of the disease keyword set and the disease history keyword set to obtain a disease word frequency vector and a disease history word frequency vector;

[0018] The offline hospital system calculates the similarity between the disease word frequency vector and the disease history word frequency vector using a similarity formula to obtain a first similarity;

[0019] If the first similarity is greater than a preset first similarity threshold, the offline hospital system determines the suspected disease type of the target patient based on the patient's historical disease types included in the patient's medical history information;

[0020] The offline hospital system selects all departments associated with the suspected disease type as recommended departments.

[0021] Optionally, the patient's historical medical record information also includes the patient's historical medical consultation information, and the method further includes the following steps:

[0022] If the first similarity is less than or equal to the first similarity threshold, the offline hospital system retrieves the comprehensive historical medical record information of all patients from the hospital database;

[0023] The offline hospital system analyzes and calculates the heterogeneous information complexity of the comprehensive historical medical record information;

[0024] If the complexity of the heterogeneous information is less than or equal to a preset complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by constructing a disease association recognition model;

[0025] If the complexity of the heterogeneous information is greater than the complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information to construct a disease association graph network, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network;

[0026] The offline hospital system extracts a plurality of related symptom keywords from the suspected related disease information using the TF-IDF algorithm, and integrates the plurality of related symptom keywords into a related symptom keyword set;

[0027] The offline hospital system calculates the word frequency vector of the associated symptom keyword set to obtain the associated disease word frequency vector;

[0028] The offline hospital system calculates the similarity between the disease word frequency vector and the associated disease word frequency vector using a similarity formula to obtain a second similarity;

[0029] If the second similarity is greater than a preset second similarity threshold, the offline hospital system selects all departments associated with the suspected related disease types included in the suspected related disease information as recommended departments;

[0030] If the second similarity is less than or equal to the second similarity threshold, the offline hospital system uses the historical medical department in the patient's historical medical information as the recommended department.

[0031] Optionally, the offline hospital system constructs a disease association graph network in combination with the comprehensive historical medical record information and the patient's historical medical record information, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network, including the following steps:

[0032] The offline hospital system generates a target patient node and multiple target historical disease nodes corresponding to the target patient based on the patient's historical medical record information, and generates first node edges between all the target historical disease nodes and the target patient nodes to construct an initial disease association graph network, wherein the first node attribute and the first node edge of the target historical disease node are both generated according to the patient's disease history information;

[0033] In the initial disease association graph network, the offline hospital system generates a plurality of other patient nodes and a plurality of other disease nodes based on the comprehensive historical medical record information, and respectively generates a second node edge between each of the other patient nodes and the corresponding plurality of other disease nodes, so as to construct a disease association graph network, wherein the second node attribute of the other disease node is generated according to the comprehensive disease history information in the comprehensive historical medical record information;

[0034] The offline hospital system uses each of the target historical disease nodes as a clustering center, and performs node clustering on all the other disease nodes in combination with the first node attribute and the second node attribute to obtain a plurality of disease node clusters, wherein the number of the disease node clusters is the same as the number of the target historical disease nodes;

[0035] For each of the disease node clusters, the offline hospital system performs node link prediction on the disease node clusters based on node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster;

[0036] The offline hospital system uses the target patient node corresponding to the disease node cluster as the path starting point, the first node edge as the initial path segment, and any other disease node in the disease node cluster as the path end point, and selects multiple disease meta-paths in the disease node cluster, wherein the disease meta-path has at least one node prediction edge as the other path segment;

[0037] The offline hospital system constructs a plurality of disease meta-path neighbor graphs according to the plurality of disease meta-paths, and repeatedly performs the reconstruction and generalization step on all the disease meta-path neighbor graphs using a pre-configured classification mixed membership random block model, and screens out uncertain disease meta-paths in the disease meta-path neighbor graphs until the number of executions of the reconstruction and generalization step reaches a preset number threshold;

[0038] When the offline hospital system completes the step of screening out the uncertain disease meta-paths for all the disease node clusters, the offline hospital system aggregates and generates a node adjacency matrix of the n-th order neighbor nodes of the target patient node based on all the remaining disease meta-paths in the disease association graph network, where n≥2;

[0039] The offline hospital system extracts disease-related fusion features from the node adjacency matrix, and determines suspected associated disease information of the target patient based on the disease-related fusion features.

[0040] Optionally, the offline hospital system performs node link prediction on the disease node cluster based on node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster, including the following steps:

[0041] The offline hospital system calculates the first node degrees of all the target historical disease nodes in the disease node cluster, and calculates the second node degrees of all the other disease nodes;

[0042] The offline hospital system calculates the node similarity between the target historical disease node and the other disease nodes by combining the first node degree and the second node degree;

[0043] If the node similarity between any target disease node among all the other disease nodes and any target historical disease node exceeds a preset similarity threshold, the offline hospital system generates a first node prediction edge between the target disease node and the target historical disease node;

[0044] The offline hospital system inputs the disease node cluster containing the first node prediction edge as a disease-related subgraph network into a preset graph convolutional neural network model, and extracts the node association feature vector between the target historical disease node and the other disease nodes through the graph convolutional neural network model;

[0045] The offline hospital system converts the node-associated feature vector into a predicted node edge vector by calculating the Hadamard distance, and generates a second node predicted edge between the target historical disease node and all the other disease nodes except the target disease node based on the predicted node edge vector.

[0046] Optionally, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by constructing a disease association recognition model as follows:

[0047] The offline hospital system builds a disease association recognition model based on a neural network model;

[0048] The offline hospital system initializes the model parameters of the disease association identification model;

[0049] The offline hospital system inputs the comprehensive historical medical record information into the model as a training set, and updates the model parameters to the optimal model parameters through repeated iterative training;

[0050] The offline hospital system inputs the patient's historical medical record information into the disease-related identification model, outputs an identification result through the disease-related identification model, and determines the suspected associated disease information of the target patient based on the identification result.

[0051] Optionally, the offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information to obtain a doctor-patient matching result, including the following steps:

[0052] The offline hospital system uses a convolutional neural network model to extract features of the doctor information and doctor consultation case features;

[0053] The offline hospital system calculates the similarity between the patient's disease information and the doctor's case characteristics using a similarity formula to obtain the doctor-patient similarity;

[0054] The offline hospital system performs real number encoding on the doctor information, the patient basic information, and the doctor-patient similarity, performs initialization processing, and obtains a doctor-patient matching association set;

[0055] The offline hospital system pre-constructs a fitness function, and uses the fitness function to evaluate the fitness of each individual in the doctor-patient matching association set to obtain the fitness function value of each individual;

[0056] If the fitness function value is less than or equal to a preset fitness function value threshold, the primary doctor-patient matching result is subjected to crossover operation and mutation operation to obtain a target doctor-patient matching association set, and the fitness of the target doctor-patient matching association set is re-evaluated;

[0057] If the fitness function value is greater than a preset fitness function value threshold, the result with the largest fitness function value is selected as the doctor-patient matching result, and the doctor-patient matching result is output.

[0058] Optionally, the offline hospital system uses a similarity formula to calculate the similarity between the patient's disease information and the doctor's case characteristics, and obtaining the doctor-patient similarity includes the following steps:

[0059] The offline hospital system vectorizes the patient disease information and the doctor's case features to obtain a vectorized result of the patient disease information and a vectorized result of the doctor's case features;

[0060] The offline hospital system uses a similarity formula to calculate the similarity between the vectorized results of the patient's disease information and the vectorized results of the doctor's consultation cases to obtain the doctor-patient similarity.

[0061] Optionally, the front-end gateway includes a first front-end gateway configured by the online hospital system and a second front-end gateway configured by the offline hospital system, and the front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system, including the following steps:

[0062] The first front-end gateway calculates the request address according to the routing addressing algorithm to obtain the address parameter;

[0063] The first front-end gateway encapsulates the address parameter into the service request to obtain a request parameter set;

[0064] The first front-end gateway performs protocol conversion on the request parameter set to obtain consultation request information;

[0065] The first front-end gateway sends an interface call instruction to the offline hospital system to transmit the consultation request information to the offline hospital system.

[0066] Optionally, the offline hospital system transmits the doctor-patient matching result to the online hospital system through the front-end gateway, including the following steps:

[0067] The offline hospital system transmits the doctor-patient matching result to the second front-end gateway;

[0068] The second front-end gateway calculates the doctor-patient matching result request address according to the routing addressing algorithm to obtain the doctor-patient matching result address parameter;

[0069] The second front-end gateway encapsulates the doctor-patient matching result address parameter into the doctor-patient matching result to obtain a doctor-patient matching result parameter set;

[0070] The second front-end gateway performs protocol conversion on the doctor-patient matching result parameter set to obtain doctor-patient matching result information;

[0071] The second front-end gateway sends an interface call instruction to the online hospital system to transmit the doctor-patient matching result information to the offline hospital system.

[0072] Through the above technical solution, the online hospital system in this application obtains the patient's business request through the patient client held by the patient, and the online hospital system sends the obtained business request to the front-end gateway. The front-end gateway converts the business request into a consultation request information, and sends the consultation request information to the offline hospital system. The offline hospital system parses the consultation information to obtain the patient's disease information and the patient's basic information. The offline hospital system recommends multiple suitable departments for the patient based on the patient's disease information and the patient's basic information. In this process, the offline hospital system first compares the patient's disease information with the patient's disease history information in the hospital database, and calculates the first similarity. If the first similarity is sufficient, the department can be recommended based on the patient's disease history information. If the first similarity is not enough, the second similarity between the patient's disease information and the patient's suspected related disease information is continued to be calculated. If the second similarity is sufficient, the department can be recommended based on the department associated with the patient's suspected related disease type. If the second similarity is not enough, the historical consultation department in the patient's historical consultation information is output as the recommended department. The offline hospital system retrieves the doctor information of the corresponding department from the preset hospital database based on the recommended multiple departments. The offline hospital system intelligently matches the patient's disease information and doctor's information, assigns the corresponding doctor to the patient, and then transmits the assignment result as the doctor-patient matching result to the online hospital system through the front-end gateway, realizing business interaction and information synchronization between the online and offline hospital systems.

[0073] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 A schematic diagram of the system structure of an online and offline integrated hospital system and hospital system associated modules provided in an embodiment of the present application;

[0075] Figure 2 A flowchart of a hospital system interaction method based on data association matching provided in an embodiment of the present application. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0077] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), such directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0078] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0079] Figure 1 , applied to online hospital systems and offline hospital systems, which interact with each other through the front-end gateway. The front-end gateway includes the first front-end gateway configured by the online hospital system and the second front-end gateway configured by the offline hospital system. Both the first front-end gateway and the second front-end gateway include a routing addressing module, a request forwarding module, a protocol conversion module and an interface call module. The offline hospital system includes eight modules: department recommendation module, doctor-patient matching module, HIS&EMR (hospitalization management system & electronic medical record system), integration platform, unified payment platform, appointment registration platform, appointment examination platform and report center. The online hospital system includes an account management module, a permission management module, a medical directory module, a data dictionary module, a message push module, a resource management module and a data synchronization module.

[0080] Figure 2 The flowchart of a hospital system interaction method based on data association matching according to an embodiment of the present application is schematically shown. Figure 2 As shown, the embodiment of the present application provides a method for hospital system interaction based on data association matching, which may include the following steps:

[0081] S101. The online hospital system obtains a service request through a patient client held by a target patient, and sends the service request to a front-end gateway.

[0082] In this embodiment, the patient client held by the patient mainly includes the patient APP, the patient applet, etc. The patient can describe his or her symptoms and fill in personal information on the patient client. The online hospital system can integrate the information of the patient client into a business request and send it to the offline hospital system through the front gateway to achieve online and offline integrated management. The medical staff of the hospital can provide medical services offline or use the online hospital system to provide telemedicine services to patients. At the same time, online consultation data and offline consultation data will also be synchronized. Regardless of whether the patient chooses online or offline consultation, the patient's disease information and basic patient information can be interconnected. The doctor can also retrieve the patient's previous medical records, electronic medical records and historical medical orders, whether online or offline, which improves the efficiency of consultation. Patients can also view their own diagnosis results online, which greatly saves the time of doctors and patients.

[0083] S102: The front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system.

[0084] In this embodiment, the front gateway includes a first front gateway and a second front gateway. The first front gateway converts the service request into consultation request information through four modules: routing addressing, request forwarding, protocol conversion, and interface call, and then sends it to the offline hospital system. The first front gateway calculates the request address of the service request through the routing addressing algorithm to obtain the address parameters, and then encapsulates the address parameters into the service request to obtain the request parameter set, and obtains the consultation request information after performing protocol conversion on the request parameter set. After the first front gateway sends the interface call instruction to the offline hospital system, the consultation request information is transmitted to the offline hospital system.

[0085] S103. The offline hospital system analyzes the consultation request information to obtain the target patient's disease information and basic patient information.

[0086] In this embodiment, the offline hospital system classifies and extracts the consultation request information to obtain the patient's disease information and basic patient information. The patient's disease information mainly includes the patient's description of the disease he or she suffers from, and the basic patient information mainly includes the patient's age and gender.

[0087] S104. The offline hospital system selects multiple recommended departments that match the patient's disease information and basic information.

[0088] In this embodiment, the offline hospital system uses the TF-IDF algorithm to extract multiple disease symptom keywords from the patient's disease information, and integrates the multiple disease symptom keywords into a disease keyword set, and then calculates the word frequency vector of the disease keyword set to obtain the disease word frequency vector. The offline hospital system retrieves the historical disease information of the target patient from the preset hospital database according to the patient's basic information, and then uses the TF-IDF algorithm to extract multiple disease history symptom keywords from the patient's historical disease information, and then integrates the multiple disease history symptom keywords into a disease history keyword set, and then calculates the word frequency vector of the disease history keyword set to obtain the disease history word frequency vector. After obtaining the disease word frequency vector and the disease history word frequency vector, the cosine similarity formula is used to calculate the similarity between the disease word frequency vector and the disease history word frequency vector to obtain the first similarity. If the first similarity is greater than the preset first similarity threshold, the offline hospital system can determine the suspected disease type of the target patient according to the patient's historical disease type contained in the patient's disease history information. Specifically, the large first similarity indicates that the patient's disease this time is very likely to be the same as the disease type the patient suffered in the past. The offline hospital system can recommend multiple related departments based on the patient's historical disease type.

[0089] Specifically, the patient's disease information mainly includes the patient's description of the disease. The main function of the TF-IDF algorithm is to extract multiple disease symptom keywords, such as "cough" and "headache", from the patient's description of the disease, and then integrate them into a disease keyword set and calculate the word frequency vector of the disease history keyword set. Similarly, the keyword extraction principle for the target patient's historical disease information is the same.

[0090] Specifically, the preset hospital database stores the information of all patients who see a doctor in the hospital and the information of doctors working in the hospital. Patient information refers to the patient's online consultation information and offline consultation information. The patient information mainly includes the patient's disease information, patient basic information, doctor's diagnosis information, doctor's medical advice and other information.

[0091] S105. The offline hospital system retrieves corresponding target doctor information from a preset hospital database based on multiple recommended departments.

[0092] In this embodiment, the offline hospital system retrieves all the doctor information corresponding to the multiple departments recommended in step S104 from the hospital database, and the doctor information includes the doctor's case characteristics, doctor's medical experience, and doctor's resume. The hospital database is the same as the hospital database in step S104.

[0093] S106. The offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information to obtain a doctor-patient matching result.

[0094] In this embodiment, the offline hospital system matches doctors who are good at treating similar diseases according to the characteristics of the patient's disease. The offline hospital system first extracts the characteristics of the doctor's case, obtains the doctor's medical experience, doctor's resume and other information, and then uses genetic algorithms to intelligently match the patient's disease information and doctor's information to select the best doctor-patient matching results. This process can match the patient with a suitable doctor, can cure the patient more efficiently, save medical resources, and improve medical efficiency.

[0095] S107. The offline hospital system transmits the doctor-patient matching results to the online hospital system through the front-end gateway.

[0096] In this embodiment, the offline hospital system transmits the doctor-patient matching result to the second front-end gateway, and the second front-end gateway performs protocol conversion on the doctor-patient matching result and transmits it to the online hospital system.

[0097] In one embodiment, the offline hospital system selects multiple recommended departments that match the patient's disease information and the patient's basic information, including the following steps:

[0098] The offline hospital system uses the TF-IDF algorithm to extract multiple disease symptom keywords from the patient's disease information and integrates the multiple disease symptom keywords into a disease keyword set;

[0099] The offline hospital system retrieves the historical medical record information of the target patient from the preset hospital database based on the basic information of the patient, and the historical medical record information includes the patient's disease history information;

[0100] The offline hospital system uses the TF-IDF algorithm to extract multiple disease history symptom keywords from the patient's disease history information, and integrates the multiple disease history symptom keywords into a disease history keyword set;

[0101] The offline hospital system calculates the word frequency vectors of the disease keyword set and the disease history keyword set to obtain the disease word frequency vector and the disease history word frequency vector;

[0102] The offline hospital system uses the similarity formula to calculate the similarity between the disease word frequency vector and the disease history word frequency vector to obtain the first similarity;

[0103] If the first similarity is greater than a preset first similarity threshold, the offline hospital system determines the suspected disease type of the target patient based on the patient's historical disease types included in the patient's medical history information;

[0104] The offline hospital system selects all departments associated with the suspected disease types as recommended departments.

[0105] In this embodiment, TF-IDF (term frequency–inverse document frequency) is a weighting technique used in information retrieval and text mining. TF-IDF can be used to evaluate the importance of a word to a document set or one of the documents in a corpus. The importance of a word increases in direct proportion to the number of times it appears in a document, but decreases in inverse proportion to the frequency of its appearance in the corpus.

[0106] Specifically, the TF-IDF algorithm is first used to extract multiple disease symptom keywords from the patient's disease information, and then these keywords are integrated into a disease symptom keyword set. At the same time, if the patient has also been to this hospital before, the offline hospital system will retrieve the patient's historical medical record information from the preset hospital database. Continue to use the TF-IDF algorithm to extract disease history symptom keywords from the patient's disease history information in the patient's historical medical record information. For example, if the patient has suffered from hypertension in the past, the symptoms of hypertension can be feature extracted, and disease history symptom keywords such as "headache", "fatigue", "arrhythmia", "palpitations", and "tinnitus" can be extracted. These disease history symptom keywords are integrated into a disease history keyword set. After that, the word frequency of these disease history symptom keywords is calculated. The higher the word frequency, the more important the keyword. For example, when describing their symptoms, patients may repeatedly mention keywords such as "chest" and "vomiting". When calculating the word frequency, the word frequency will also be higher, which means that these two disease keywords are very important. The steps to calculate the word frequency of patient disease information are as follows:

[0107] Step 1.1: Count the number of occurrences of each disease keyword;

[0108] Step 1.2: Calculate the primary word frequency of each disease keyword, that is, divide the number of occurrences of each disease keyword by the total number of words in the patient's disease information;

[0109] Step 1.3: Calculate the inverse document frequency. Create a corpus containing all patient disease information in advance to simulate the language environment. The method for calculating the inverse document frequency is as follows:

[0110] Inverse Document Frequency (IDF) = log(the total number of patient disease information in the corpus / (the number of patient disease information containing the keyword + 1))

[0111] Step 1.4: Multiply the keyword's word frequency by the inverse document frequency to get the final word frequency of the document.

[0112] Similarly, calculate the keyword frequency in the patient's historical medical record information, and then add the keyword frequency of the patient's disease information to

[0113] The keyword frequency in the patient's historical medical record information is converted into a word frequency vector. For example, the keyword frequencies of the patient's disease information keywords "chest", "stuffy", "nausea", and "vomiting" are "3", "2", "1", and "1" respectively. Then the keyword frequency vector of the patient's disease information is [3, 2, 1, 1]. The same is true for the keyword frequency vector of the patient's historical medical record information.

[0114] After the calculation is completed, the cosine similarity formula is used to calculate the similarity between the disease word frequency vector and the disease history word frequency vector. The calculation formula is as follows:

[0115] ;

[0116] Among them, Ai represents the i-th dimension vector in the n-dimensional vector of the disease word frequency vector, and Bi represents the i-th dimension vector in the n-dimensional vector of the disease history word frequency vector.

[0117] The closer the cosine value calculated by the above formula is to 1, the closer the angle between the two vectors is to 0 degrees, which means that the two vectors are more similar, that is, the disease and the disease history are highly similar.

[0118] The calculated cosine value is the first similarity. If the first similarity is greater than the preset first similarity threshold (for example, the first similarity threshold is set to 0.8), it means that the disease the patient suffers this time is very likely to be the same as the type of disease the patient suffered in the past. The offline hospital system can recommend multiple related departments based on the patient's historical disease types.

[0119] In one embodiment, the historical medical record information also includes the patient's historical medical visit information, and the method further includes the following steps:

[0120] If the first similarity is less than or equal to the first similarity threshold, the offline hospital system retrieves the comprehensive historical medical record information of all patients from the hospital database;

[0121] Offline hospital system analyzes and calculates the complexity of heterogeneous information of comprehensive historical medical records;

[0122] If the complexity of the heterogeneous information is less than or equal to the preset complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by building a disease association recognition model;

[0123] If the complexity of heterogeneous information is greater than the complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information to construct a disease association graph network, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network;

[0124] The offline hospital system uses the TF-IDF algorithm to extract multiple related symptom keywords from the suspected related disease information, and integrates the multiple related symptom keywords into a related symptom keyword set;

[0125] The offline hospital system calculates the word frequency vector of the associated symptom keyword set to obtain the associated disease word frequency vector;

[0126] The offline hospital system uses the similarity formula to calculate the similarity between the disease word frequency vector and the associated disease word frequency vector to obtain the second similarity;

[0127] If the second similarity is greater than a preset second similarity threshold, the offline hospital system selects all departments associated with the suspected related disease types included in the suspected related disease information as recommended departments;

[0128] If the second similarity is less than or equal to the second similarity threshold, the offline hospital system uses the historical medical department in the patient's historical medical information as the recommended department.

[0129] In this embodiment, if the first similarity is less than or equal to the first similarity threshold, the offline hospital system retrieves the comprehensive historical medical record information of all patients from the hospital database. The comprehensive historical medical record information of patients refers to the medical record information of all patients who have visited this hospital, and the medical record information of these patients mainly includes the historical patient disease types, historical patient disease information, historical patient basic information, etc.

[0130] Next, the offline hospital system analyzes and calculates the heterogeneous information complexity of the comprehensive historical medical record information. Specifically, the comprehensive historical medical record information is first divided into different information categories according to the information type of the heterogeneous information, and the information in each information category is standardized. After the standardization process, the marginal entropy, joint entropy and conditional entropy of all information in each information category are calculated respectively, and then the marginal entropy, joint entropy and conditional entropy are calculated by weighted average to obtain the information complexity of all information in the information category. Finally, the information complexity of all information categories is weighted averaged again to obtain the heterogeneous information complexity of the comprehensive historical medical record information.

[0131] If the complexity of heterogeneous information is less than or equal to the preset complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by constructing a disease association recognition model. The offline hospital system uses a neural network model (for example, the ALBERT model can be used) to construct a disease association recognition model, and inputs the patient's historical medical record information into the trained disease association recognition model to obtain the target patient's suspected associated disease information. The patient's historical medical record information contains the patient's disease history information, and the associated disease mainly refers to the complications of the patient's disease history. If the complexity of heterogeneous information is greater than the complexity threshold, it means that it is difficult to directly use the neural network model to extract the information association features between the comprehensive historical medical record information and the patient's historical medical record information. Therefore, it is necessary to construct a disease association graph network by combining the comprehensive historical medical record information and the patient's historical medical record information through the offline hospital system, and determine the target patient's suspected associated disease information based on the disease association fusion features extracted from the disease association graph network.

[0132] The offline hospital system uses the TF-IDF algorithm to extract multiple related symptom keywords from the suspected related disease information, and integrates the multiple related symptom keywords into a related symptom keyword set, and then calculates the related word frequency vector of the related symptom keyword set. The offline hospital system uses the cosine similarity formula to calculate the similarity between the disease word frequency vector and the related disease word frequency vector of the related symptom keyword set to obtain the second similarity.

[0133] If the second similarity is greater than the preset second similarity threshold (for example, the second similarity threshold is set to 0.8), the offline hospital system selects all departments associated with the suspected associated disease types contained in the suspected associated disease information as recommended departments. In this step, by comparing the similarity between the patient's disease information and the patient's associated disease information, it is determined whether the patient's disease type and the associated disease type are likely to be the same or similar. If the second similarity is greater than the preset second similarity threshold, it means that the patient's disease type is likely to be the same or similar to the associated disease type, and the relevant department is recommended according to the associated disease type. If the second similarity is less than or equal to the second similarity threshold, it means that the patient's disease type may not be similar to the associated disease type, and the offline hospital system will use the historical treatment department in the patient's historical treatment information as the recommended department.

[0134] In one embodiment, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information to construct a disease association graph network, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network, including the following steps:

[0135] The offline hospital system generates a target patient node and multiple target historical disease nodes corresponding to the target patient based on the patient's historical medical record information, and generates the first node edges between all target historical disease nodes and the target patient nodes to construct an initial disease association graph network. The first node attributes and first node edges of the target historical disease nodes are generated based on the patient's disease history information.

[0136] In the initial disease association graph network, the offline hospital system generates multiple other patient nodes and multiple other disease nodes based on the comprehensive historical medical record information, and respectively generates second node edges between each other patient node and the corresponding multiple other disease nodes to construct a disease association graph network. The second node attributes of the other disease nodes are generated according to the comprehensive disease history information in the comprehensive historical medical record information.

[0137] The offline hospital system uses each target historical disease node as the cluster center, and clusters all other disease nodes by combining the first node attribute and the second node attribute to obtain multiple disease node clusters. The number of disease node clusters is the same as the number of target historical disease nodes.

[0138] For each disease node cluster, the offline hospital system predicts the node links of the disease node cluster based on the node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster;

[0139] The offline hospital system takes the target patient node corresponding to the disease node cluster as the path starting point, the first node edge as the initial path segment, and any other disease node in the disease node cluster as the path end point. Multiple disease meta-paths are selected in the disease node cluster, and the disease meta-path has at least one node prediction edge as the other path segment;

[0140] The offline hospital system constructs multiple disease meta-path neighbor graphs based on multiple disease meta-paths, and uses a pre-configured classification mixed membership random block model to repeatedly perform the reconstruction generalization step on all disease meta-path neighbor graphs to screen out uncertain disease meta-paths in the disease meta-path neighbor graphs until the number of executions of the reconstruction generalization step reaches a preset threshold number of times;

[0141] When the offline hospital system completes the step of screening out uncertain disease meta-paths for all disease node clusters, the offline hospital system aggregates and generates a node adjacency matrix of the n-order neighbor nodes of the target patient node based on all remaining disease meta-paths in the disease association graph network, where n≥2;

[0142] The offline hospital system extracts disease-related fusion features from the node adjacency matrix and determines the suspected associated disease information of the target patient based on the disease-related fusion features.

[0143] In this embodiment, the offline hospital system first constructs an initial disease association graph network based on the patient's historical medical record information. The system creates a target patient node for the target patient and creates corresponding target historical disease nodes for each historical disease of the patient. For example, if the patient has suffered from hypertension, diabetes, and coronary heart disease, these three disease nodes will be created. The system also creates a first node edge between the target patient node and each target historical disease node to represent the association between the patient and the disease. The attributes of these nodes and edges are generated based on the patient's specific medical history information. For example, the attributes of the disease node may include onset time, duration, severity, etc., while the attributes of the edge may include diagnosis date, treatment method, etc. In this way, the system converts the patient's medical history information into a structured graph network, laying the foundation for subsequent analysis.

[0144] After constructing the initial disease association graph network, the offline hospital system further expands the network to include more patient and disease information. Based on the data in the comprehensive historical medical record information base, the system creates patient nodes for other patients and disease nodes for the diseases of these patients. For example, if there are 10,000 patients in the database, and each has an average of 5 diseases, the system will create 10,000 other patient nodes and about 50,000 other disease nodes. The system will also create a second node edge between each other patient node and its corresponding disease node to represent the association between these patients and the disease. The attributes of these newly added nodes and edges are generated based on the comprehensive historical medical record information and contain rich clinical data. Through this step, the initial simple network is expanded into a large and complex disease association graph network, which contains a large amount of patient and disease information. This expanded network not only contains the information of the target patient, but also covers a large amount of disease information of other patients, providing a rich data basis for subsequent analysis. This network structure enables the system to analyze the association between diseases at the group level.

[0145] Next, the offline hospital system clusters the nodes in the disease association graph network. The system uses each historical disease node of the target patient as the cluster center and clusters all other disease nodes in the network. The clustering process takes into account the attributes of the nodes (such as disease type, symptoms, severity, etc.) and the network structure characteristics. For example, the system can use the K-means algorithm, take the attributes of the target historical disease node as the initial center point, calculate the similarity of other disease nodes with these center points, and assign them to the most similar categories. The clustering process may require multiple iterations until the clustering results are stable. In the end, the system obtains the same number of clusters as the number of historical diseases of the target patient. For example, if the target patient has 3 historical diseases, then 3 disease node clusters will eventually be formed. Each cluster contains other disease nodes that are similar or related to a target historical disease.

[0146] After completing the clustering of disease nodes, the offline hospital system performs node link prediction for each cluster. Based on node similarity, the system predicts the possible association between the target historical disease node and other disease nodes in the same cluster. The node similarity here may consider multiple factors, such as the similarity of disease symptoms, similarity of treatment methods, overlap of patient groups, etc. The system can use a variety of link prediction algorithms, such as common neighbor-based methods, path-based methods, or machine learning methods. For example, the Jaccard coefficient is used to calculate the common neighbor similarity of two nodes. If the similarity exceeds the preset threshold, the system generates a node prediction edge between the two nodes. These predicted edges represent potential disease associations, which reflect common risk factors, similar pathogenesis, or treatment methods. In this way, the system not only utilizes known disease association information, but also predicts possible disease associations that have not been directly observed, enriching the structure of the disease network and providing more possibilities for subsequent path analysis.

[0147] After completing the node link prediction, the offline hospital system begins to select disease meta-paths in each disease node cluster. The system starts with the target patient node, connects to the target historical disease node through the first node edge, and then uses the predicted node edges and existing edges to find paths leading to other disease nodes in the cluster. For example, a disease meta-path may be: target patient → hypertension (target historical disease) → hyperlipidemia (predicted association) → atherosclerosis (other diseases). The system will select multiple such paths, each containing at least one predicted edge. The selection of paths may be based on a variety of strategies, such as the shortest path, the maximum flow path, or a random walk. These disease meta-paths represent possible development pathways from the known diseases of the target patient to other potentially related diseases. By analyzing these paths, the system can explore the potential associations and evolution patterns between diseases, providing a basis for predicting the health risks that patients may face in the future. This graph network-based path analysis method can capture complex disease association patterns, surpassing simple dual disease association analysis.

[0148] Based on the selected disease meta-paths, the offline hospital system constructs multiple disease meta-path neighbor graphs. Each neighbor graph is centered on a disease meta-path and contains all nodes on the path and their directly connected nodes. For example, for the path A→B→C, the neighbor graph will contain three nodes A, B, and C, as well as other nodes directly connected to them. The system then reconstructs and generalizes these neighbor graphs using a pre-configured categorical mixed membership stochastic block model (MMSBM). MMSBM is a probabilistic graph model that assumes that nodes in a network can belong to multiple potential groups, with different connection probabilities between each group. The goal of the model is to learn the group membership of nodes and the connection probability between groups. For example, the model may find that some disease nodes belong to both the "cardiovascular disease" and "metabolic disease" groups, and there is a high connection probability between these two groups. The system repeatedly performs the reconstruction and generalization steps, and each iteration updates the model parameters until the preset number of iterations is reached. In this process, the system will screen out those disease meta-paths that are unstable or have low probabilities in multiple reconstructions, and these paths are considered uncertain. In this way, the system can identify more reliable and universal disease association patterns, reduce the impact of noise and accidental factors, and improve the accuracy and reliability of subsequent analysis.

[0149] When the offline hospital system has completed the screening of uncertain disease meta-paths for all disease node clusters, it will construct an adjacency matrix of the n-th order neighbor nodes of the target patient node based on the remaining disease meta-paths. n ≥ 2 means that the system considers not only directly connected diseases but also indirectly related diseases. The adjacency matrix is ​​a two-dimensional matrix, in which the matrix element a_ij represents the strength of the relationship between nodes i and j. For example, for the third-order neighbors, the system first finds all disease nodes that are at most 3 steps away from the target patient node, and then calculates the strength of the relationship between these nodes. The strength of the relationship may be based on multiple factors, such as the number of paths, path length, edge weight, etc. This adjacency matrix combines direct and indirect disease association information, providing rich structured data for subsequent feature extraction. By considering high-order neighbors, the system is able to capture more complex and deeper disease association patterns that may be overlooked when only direct associations are considered.

[0150] Finally, the offline hospital system extracts disease-associated fusion features from the constructed node adjacency matrix. This step aims to convert complex network structure information into feature vectors that can be used for analysis and prediction. The system can use a variety of feature extraction methods, such as spectral clustering, matrix decomposition, or graph neural networks. For example, the adjacency matrix A is decomposed using singular value decomposition (SVD), and then the left singular vectors corresponding to the first k singular values ​​are selected as features. Alternatively, a graph convolutional network (GCN) is used to learn the embedded representation of the nodes. The extracted features fuse the structural information of the network and the node attribute information, and can comprehensively represent the association pattern between diseases. Based on these features, the system can use support vector machines for further feature classification to determine the suspected associated diseases of the target patients.

[0151] In one embodiment, the offline hospital system performs node link prediction on the disease node cluster based on node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster, including the following steps:

[0152] The offline hospital system calculates the first node degree of all target historical disease nodes in the disease node cluster, and calculates the second node degree of all other disease nodes;

[0153] The offline hospital system combines the first node degree and the second node degree to calculate the node similarity between the target historical disease node and other disease nodes;

[0154] If the node similarity between any target disease node and any target historical disease node among all other disease nodes exceeds a preset similarity threshold, the offline hospital system generates a first node prediction edge between the target disease node and the target historical disease node;

[0155] The offline hospital system inputs the disease node clusters containing the first node prediction edge as the disease association subgraph network into the preset graph convolutional neural network model, and extracts the node association feature vectors between the target historical disease node and other disease nodes through the graph convolutional neural network model;

[0156] The offline hospital system converts the node association feature vector into a predicted node edge vector by calculating the Hadamard distance, and generates a second node prediction edge between the target historical disease node and all other disease nodes except the target disease node based on the predicted node edge vector.

[0157] In this embodiment, first, the offline hospital system calculates the first node degree of all target historical disease nodes in the disease node clustering cluster, and the second node degree of all other disease nodes. The node degree refers to the number of edges directly connected to the node, reflecting the importance and degree of connection of the node in the network. For the target historical disease node, the system counts the number of existing connection edges between each node and other nodes to obtain the first node degree. Similarly, for other disease nodes, the system also counts the number of connection edges of each node to obtain the second node degree. The purpose of this step is to provide basic data for subsequent calculation of node similarity. By comparing the degrees of different nodes, the similarity between nodes can be preliminarily judged. In specific implementation, the system can traverse each node in the clustering cluster, count the number of its connected edges, and store the results in the corresponding data structure. Next, the offline hospital system evaluates the similarity between nodes by comparing the node degrees. There are many ways to calculate similarity, for example, algorithms such as cosine similarity or Jaccard similarity can be used.

[0158] The system then determines whether there is any pair of target disease nodes and target historical disease nodes whose node similarity exceeds the preset similarity threshold. If such a node pair exists, the system generates a first node prediction edge between the two nodes. The purpose of this step is to identify highly correlated disease node pairs and establish a direct connection between them. The setting of the similarity threshold is critical, as it determines the system's criteria for judging relevance. Setting the threshold too low may result in too many false positive connections, while setting it too high may ignore some potentially important associations.

[0159] Next, the offline hospital system will input the disease node clusters containing the first node prediction edge as the disease-related subgraph network into the preset graph convolutional neural network model. Graph convolutional neural network (GCN) is a deep learning model specifically designed to process graph-structured data, which can effectively capture local and global relationships between nodes. In this step, the GCN model performs multi-layer convolution operations on the input disease-related subgraph network to extract the node association feature vectors between the target historical disease node and other disease nodes. Through the multi-layer convolution operation of GCN, GCN can gradually aggregate the information of neighboring nodes, thereby learning richer and more abstract node representations. The effect of this step is to convert the original disease node relationship into a high-dimensional feature vector, which not only contains the information of the node itself, but also contains the structural information and semantic information of the node in the entire network.

[0160] Finally, the offline hospital system converts the node association feature vector into a predicted node edge vector by calculating the Hadamard distance, and generates a second node prediction edge between the target historical disease node and other disease nodes except the target disease node based on this predicted node edge vector. Hadamard distance is a method for calculating the similarity between vectors. It is obtained by multiplying the corresponding elements of two vectors bit by bit and then summing them. The advantage of using Hadamard distance is that it can capture fine-grained differences between vectors and is particularly suitable for processing high-dimensional feature vectors. In this step, the system calculates the Hadamard distance between the feature vector of the target historical disease node and the feature vector of each other disease node (except the target disease node), and the result is the predicted node edge vector. If an element of this predicted edge vector exceeds the preset threshold, the system generates a second node prediction edge between the corresponding two nodes. The effect of implementing this step is to further improve the disease association network, considering not only the direct high-similarity association, but also the indirect association implied by the network structure, thereby constructing a more comprehensive and accurate disease association network.

[0161] In one embodiment, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by building a disease association recognition model as follows:

[0162] The offline hospital system builds a disease association recognition model based on the neural network model;

[0163] The offline hospital system initializes the model parameters of the disease association recognition model;

[0164] The offline hospital system inputs the comprehensive historical medical record information into the model as a training set, and updates the model parameters to the optimal model parameters through repeated iterative training;

[0165] The offline hospital system inputs the patient's historical medical record information into the disease association recognition model, outputs the recognition results through the disease association recognition model, and determines the suspected related disease information of the target patient based on the recognition results.

[0166] In this embodiment, the offline hospital system uses a neural network model (such as the ALBERT model) to build a disease association recognition model. After initializing the model parameters of the disease association recognition model, the comprehensive historical medical record information is input into the model as a training set. Through repeated iterative training, until the model parameters are updated to the optimal model parameters, it is indicated that the disease association recognition model training is completed.

[0167] The function of the disease association identification model is to identify the patient's historical medical record information and obtain information about suspected related diseases based on the patient's historical medical record information. Related diseases refer to complications caused by the patient's historical disease information. For example, hypertension may cause heart failure, cerebral thrombosis, cerebral hemorrhage and other diseases. Then, heart failure, cerebral thrombosis, cerebral hemorrhage and other diseases are related diseases of hypertension.

[0168] The disease association recognition model constructed using the ALBERT model includes an information input layer, an information feature extraction layer, an attention mechanism layer, and an output layer. The text information contained in the comprehensive historical medical record information is used as a data set for training the disease association recognition model. The data set is input into the information input layer of the disease association recognition model. The information input layer vectorizes the data and converts the text information contained in the comprehensive historical medical record information into text vector information. Then, the text vector information is input into the information feature extraction layer. The information feature extracts the contextual text semantics to obtain semantic feature information. Next, the semantic feature information obtained by the information feature extraction layer is input into the attention mechanism layer. This step can assign a higher weight value to the key feature information in the semantic feature information. Finally, the activation function (such as the sigmoid function) is used to update the model parameters to the optimal model parameters according to the semantic feature information and the corresponding weight value, and the disease association recognition model training is completed. The offline hospital system inputs the patient's historical medical record information into the trained disease association recognition model, outputs the recognition result through the disease association recognition model, and determines the suspected associated disease information of the target patient based on the recognition result.

[0169] In one embodiment, the offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information, and obtaining a doctor-patient matching result includes the following steps:

[0170] The offline hospital system uses a convolutional neural network model to extract features of doctor information and doctor consultation case characteristics;

[0171] The offline hospital system uses the similarity formula to calculate the similarity between the patient's disease information and the doctor's case characteristics to obtain the doctor-patient similarity;

[0172] The offline hospital system performs real number encoding on the doctor information, patient basic information, and doctor-patient similarity, performs initialization processing, and obtains a doctor-patient matching association set;

[0173] The offline hospital system pre-constructs a fitness function, uses the fitness function to evaluate the fitness of each individual in the doctor-patient matching association set, and obtains the fitness function value of each individual;

[0174] If the fitness function value is less than or equal to the preset fitness function value threshold, the primary doctor-patient matching result is subjected to crossover and mutation operations to obtain a target doctor-patient matching association set, and the fitness of the target doctor-patient matching association set is re-evaluated;

[0175] If the fitness function value is greater than a preset fitness function value threshold, the result with the largest fitness function value is selected as the doctor-patient matching result, and the doctor-patient matching result is output.

[0176] In this embodiment, the Convolutional Neural Networks (ConvNets, CNN) model is a type of neural network. The offline hospital system uses the Convolutional Neural Network model to classify and extract features from the doctor information retrieved from the hospital database, that is, the doctor information is convoluted and then reduced in dimension, and then feature mapping is performed on the reduced-dimensional doctor information to obtain four different features: doctor's medical age, doctor's resume features, doctor's professional title level, and doctor's case features. The offline hospital system performs word segmentation on the patient's disease features and the doctor's case features, and then vectorizes the patient's disease feature word set and the doctor's case feature word set obtained after word segmentation, and then uses the cosine similarity formula to calculate the doctor-patient similarity.

[0177] Specifically, the offline hospital system uses genetic algorithms to match patient disease information with doctor information, and obtaining the optimal doctor-patient matching result includes the following steps:

[0178] Step 2.1: The offline hospital system takes each doctor-patient matching result as a chromosome, and the patient information, patient disease information, doctor information and doctor-patient similarity in each chromosome as chromosome genes. Each doctor-patient matching result is encoded to obtain a complete chromosome code, which is the doctor-patient matching association set.

[0179] Step 2.2: The offline hospital system pre-constructs a fitness function based on the constraints of doctor-patient matching. The doctor-patient matching optimization conditions required to construct the fitness function mainly include two aspects. On the one hand, the doctor-patient similarity is greater than the preset doctor-patient similarity threshold. On the other hand, for special groups with poor physical fitness such as the elderly, young children, and pregnant women, doctors with longer medical experience, rich resumes, and higher professional titles should be given priority. After the fitness function is constructed, the offline hospital system uses the fitness function to calculate the fitness of each doctor-patient matching result in the doctor-patient matching association set to evaluate the fitness and obtain the fitness function value of each doctor-patient matching result;

[0180] Step 2.3: If the fitness function value is less than or equal to the preset fitness function value threshold, the primary doctor-patient matching result is subjected to crossover and mutation operations to obtain a target doctor-patient matching association set, and the target doctor-patient matching association set is re-evaluated for fitness. The process of performing crossover and mutation operations on the primary doctor-patient matching result can re-match the patient information, patient disease information, doctor information, and doctor-patient similarity information in the doctor-patient matching association set to obtain a re-matched target doctor-patient matching association set;

[0181] Step 2.4: If the fitness function value is greater than the preset fitness function value threshold, the result with the largest fitness function value is selected as the doctor-patient matching result, and the offline hospital system outputs the doctor-patient matching result. The maximum fitness function value means that the corresponding doctor-patient matching result is the optimal doctor-patient matching result. Through this doctor-patient matching method, the most suitable doctor can be matched for the patient, which improves medical efficiency and saves medical resources.

[0182] In one embodiment, the offline hospital system uses a similarity formula to calculate the similarity between the patient's disease information and the doctor's case characteristics, and obtaining the doctor-patient similarity includes the following steps:

[0183] The offline hospital system vectorizes the patient's disease information and the doctor's case features to obtain the vectorized results of the patient's disease information and the doctor's case features;

[0184] The offline hospital system uses a similarity formula to calculate the similarity between the quantized results of the patient's disease information and the quantized results of the doctor's consultation cases to obtain the doctor-patient similarity.

[0185] In this embodiment, the offline hospital system performs word segmentation processing on the patient's disease characteristics and the doctor's treatment case characteristics to obtain the patient's disease characteristic word set and the doctor's treatment case characteristic word set, and then uses the TF-IDF algorithm to calculate the word frequency of the patient's disease characteristic word set and the doctor's treatment case characteristic word set, and vectorizes the calculated word frequency, and then uses the cosine similarity formula to calculate the doctor-patient similarity.

[0186] In one embodiment, the front-end gateway includes a first front-end gateway configured by the online hospital system and a second front-end gateway configured by the offline hospital system. The front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system, including the following steps:

[0187] The first front-end gateway calculates the request address according to the routing addressing algorithm to obtain the address parameters;

[0188] The first front-end gateway encapsulates the address parameters into the service request to obtain a request parameter set;

[0189] The first front-end gateway performs protocol conversion on the request parameter set to obtain consultation request information;

[0190] The first front-end gateway sends an interface call instruction to the offline hospital system to transmit the consultation request information to the offline hospital system.

[0191] In this embodiment, the main function of the first front gateway is to convert the request information of the online hospital system into a format and send it to the offline hospital system. The first front gateway mainly includes four modules: routing addressing, request forwarding, protocol conversion and interface calling.

[0192] Specifically, the first front-end gateway calculates the request address according to the routing addressing algorithm to obtain the address parameters, and then encapsulates the address parameters into the business request to obtain a request parameter set, performs protocol conversion on the request parameter set to obtain the consultation information, and the first front-end gateway sends an interface call instruction to the offline hospital system to transmit the obtained consultation request information to the offline hospital system.

[0193] Specifically, the function of the routing addressing module is to determine the request address of the peer gateway through a certain addressing algorithm according to the call parameters. Generally, the function call of the online hospital system to the offline hospital system can be expressed as a triple (x, y, z), where x represents the business module, y represents the interface name, and z represents the parameter set. Therefore, the function of the routing addressing module can be formally expressed as url = f (x, y, z). In the formula, url is the request address for forwarding the request to the peer gateway. For example, the business request sent by the online hospital system to the first front-end gateway is to perform doctor-patient matching, then the triple can be expressed as url = f (doctor-patient matching module, Web Service interface, patient disease information and patient information). The function of the request forwarding module is to forward the request to the peer gateway using the request address calculated by the routing addressing module, and encapsulate the parameter set z in the caller request triple into the request and pass it to the peer gateway. The function of the protocol conversion module is to convert the parameter set format of the request forwarded by the peer gateway according to the interface requirements of the docking system on this side; at the same time, convert the result set returned by the docking system on this side into the result set required by the peer gateway. The function of the interface call module is to respond to the request of the peer gateway for a certain address path, issue an interface call to the local docking system with a parameter set converted by the protocol, and obtain the call result. After obtaining the call result, the first front-end gateway transmits the business request information to the offline hospital system.

[0194] In one embodiment, the offline hospital system transmits the doctor-patient matching result to the online hospital system through the front-end gateway, including the following steps:

[0195] The offline hospital system transmits the doctor-patient matching results to the second front-end gateway;

[0196] The second front-end gateway calculates the doctor-patient matching result request address according to the routing addressing algorithm to obtain the doctor-patient matching result address parameter;

[0197] The second front-end gateway encapsulates the doctor-patient matching result address parameter into the doctor-patient matching result to obtain a doctor-patient matching result parameter set;

[0198] The second front-end gateway performs protocol conversion on the doctor-patient matching result parameter set to obtain doctor-patient matching result information;

[0199] The second front-end gateway sends an interface call instruction to the online hospital system to transmit the doctor-patient matching result information to the offline hospital system.

[0200] In this embodiment, the internal logical structure of the second front gateway is the same as that of the first front gateway, and also includes four modules: routing addressing, request forwarding, protocol conversion, and interface calling. The main function of the second front gateway is to help the offline hospital system send business requests and business processing results to the online hospital system, and ensure information exchange and information security between the offline hospital system and the online hospital system. The second front gateway calculates the doctor-patient matching result request address according to the routing addressing algorithm, obtains the doctor-patient matching result address parameter, encapsulates the doctor-patient matching result address parameter into the doctor-patient matching result, obtains the doctor-patient matching result parameter set, performs protocol conversion on the obtained doctor-patient matching result parameter set, obtains the doctor-patient matching result information, and then the second front gateway sends an interface call instruction to the online hospital system to transmit the doctor-patient matching result information to the offline hospital system.

[0201] An embodiment of the present application also provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned hospital system interaction method based on data association matching.

[0202] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0203] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0204] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0205] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0206] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0207] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0208] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0209] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0210] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A hospital system interaction method based on data association matching, applied to an online hospital system and an offline hospital system, wherein the online hospital system and the offline hospital system interact with each other through a front-end gateway, characterized in that: The method comprises the following steps: The online hospital system obtains a service request through a patient client held by the target patient, and sends the service request to the front-end gateway; The front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system; The offline hospital system analyzes the consultation request information to obtain the patient disease information and patient basic information of the target patient; The offline hospital system extracts a plurality of disease symptom keywords from the patient's disease information using a TF-IDF algorithm, and integrates the plurality of disease symptom keywords into a disease keyword set; The offline hospital system retrieves the patient's historical medical record information of the target patient from a preset hospital database based on the patient's basic information, wherein the patient's historical medical record information includes the patient's disease history information and the patient's historical medical visit information; The offline hospital system extracts a plurality of disease history symptom keywords from the patient's disease history information using the TF-IDF algorithm, and integrates the plurality of disease history symptom keywords into a disease history keyword set; The offline hospital system calculates the word frequency vectors of the disease keyword set and the disease history keyword set to obtain a disease word frequency vector and a disease history word frequency vector; The offline hospital system calculates the similarity between the disease word frequency vector and the disease history word frequency vector using a similarity formula to obtain a first similarity; If the first similarity is greater than a preset first similarity threshold, the offline hospital system determines the suspected disease type of the target patient based on the patient's historical disease types included in the patient's medical history information; The offline hospital system selects all departments associated with the suspected disease type as recommended departments; If the first similarity is less than or equal to the first similarity threshold, the offline hospital system retrieves the comprehensive historical medical record information of all patients from the hospital database; The offline hospital system analyzes and calculates the heterogeneous information complexity of the comprehensive historical medical record information; If the complexity of the heterogeneous information is less than or equal to a preset complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by constructing a disease association recognition model; If the complexity of the heterogeneous information is greater than the complexity threshold, the offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information to construct a disease association graph network, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network; The offline hospital system extracts a plurality of related symptom keywords from the suspected related disease information using the TF-IDF algorithm, and integrates the plurality of related symptom keywords into a related symptom keyword set; The offline hospital system calculates the word frequency vector of the associated symptom keyword set to obtain the associated disease word frequency vector; The offline hospital system calculates the similarity between the disease word frequency vector and the associated disease word frequency vector using a similarity formula to obtain a second similarity; If the second similarity is greater than a preset second similarity threshold, the offline hospital system selects all departments associated with the suspected related disease types included in the suspected related disease information as recommended departments; If the second similarity is less than or equal to the second similarity threshold, the offline hospital system uses the historical medical department in the patient's historical medical information as the recommended department; The offline hospital system retrieves corresponding target doctor information from a preset hospital database based on the plurality of recommended departments; The offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information to obtain a doctor-patient matching result; The offline hospital system transmits the doctor-patient matching result to the online hospital system through the front-end gateway.

2. The method according to claim 1, characterized in that The offline hospital system constructs a disease association graph network by combining the comprehensive historical medical record information and the patient's historical medical record information, and determines the suspected associated disease information of the target patient based on the disease association fusion features extracted from the disease association graph network, including the following steps: The offline hospital system generates a target patient node and multiple target historical disease nodes corresponding to the target patient based on the patient's historical medical record information, and generates first node edges between all the target historical disease nodes and the target patient nodes to construct an initial disease association graph network, wherein the first node attribute and the first node edge of the target historical disease node are both generated according to the patient's disease history information; In the initial disease association graph network, the offline hospital system generates a plurality of other patient nodes and a plurality of other disease nodes based on the comprehensive historical medical record information, and respectively generates a second node edge between each of the other patient nodes and the corresponding plurality of other disease nodes, so as to construct a disease association graph network, wherein the second node attribute of the other disease node is generated according to the comprehensive disease history information in the comprehensive historical medical record information; The offline hospital system uses each of the target historical disease nodes as a clustering center, and performs node clustering on all the other disease nodes in combination with the first node attribute and the second node attribute to obtain a plurality of disease node clusters, wherein the number of the disease node clusters is the same as the number of the target historical disease nodes; For each of the disease node clusters, the offline hospital system performs node link prediction on the disease node clusters based on node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster; The offline hospital system uses the target patient node corresponding to the disease node cluster as the path starting point, the first node edge as the initial path segment, and any other disease node in the disease node cluster as the path end point, and selects multiple disease meta-paths in the disease node cluster, wherein the disease meta-path has at least one node prediction edge as the other path segment; The offline hospital system constructs a plurality of disease meta-path neighbor graphs according to the plurality of disease meta-paths, and repeatedly performs the reconstruction and generalization step on all the disease meta-path neighbor graphs using a pre-configured classification mixed membership random block model, and screens out uncertain disease meta-paths in the disease meta-path neighbor graphs until the number of executions of the reconstruction and generalization step reaches a preset number threshold; When the offline hospital system completes the step of screening out the uncertain disease meta-paths for all the disease node clusters, the offline hospital system aggregates and generates a node adjacency matrix of the n-th order neighbor nodes of the target patient node based on all the remaining disease meta-paths in the disease association graph network, where n≥2; The offline hospital system extracts disease-related fusion features from the node adjacency matrix, and determines suspected associated disease information of the target patient based on the disease-related fusion features.

3. The method according to claim 2, characterized in that The offline hospital system performs node link prediction on the disease node cluster based on node similarity, and generates node prediction edges between the target historical disease node and other disease nodes in the same disease node cluster, including the following steps: The offline hospital system calculates the first node degrees of all the target historical disease nodes in the disease node cluster, and calculates the second node degrees of all the other disease nodes; The offline hospital system calculates the node similarity between the target historical disease node and the other disease nodes by combining the first node degree and the second node degree; If the node similarity between any target disease node among all the other disease nodes and any target historical disease node exceeds a preset similarity threshold, the offline hospital system generates a first node prediction edge between the target disease node and the target historical disease node; The offline hospital system inputs the disease node cluster containing the first node prediction edge as a disease-related subgraph network into a preset graph convolutional neural network model, and extracts the node association feature vector between the target historical disease node and the other disease nodes through the graph convolutional neural network model; The offline hospital system converts the node-associated feature vector into a predicted node edge vector by calculating the Hadamard distance, and generates a second node predicted edge between the target historical disease node and all the other disease nodes except the target disease node based on the predicted node edge vector.

4. The method according to claim 1, characterized in that: The offline hospital system combines the comprehensive historical medical record information and the patient's historical medical record information and determines the suspected associated disease information of the target patient by constructing a disease association recognition model as follows: The offline hospital system builds a disease association recognition model based on a neural network model; The offline hospital system initializes the model parameters of the disease association identification model; The offline hospital system inputs the comprehensive historical medical record information into the model as a training set, and updates the model parameters to the optimal model parameters through repeated iterative training; The offline hospital system inputs the patient's historical medical record information into the disease-related identification model, outputs an identification result through the disease-related identification model, and determines the suspected associated disease information of the target patient based on the identification result.

5. The method according to claim 1, characterized in that The offline hospital system uses a genetic algorithm to intelligently match patient disease information and doctor information, and obtains a doctor-patient matching result, including the following steps: The offline hospital system uses a convolutional neural network model to extract features of the doctor information and doctor consultation case features; The offline hospital system calculates the similarity between the patient's disease information and the doctor's case characteristics using a similarity formula to obtain the doctor-patient similarity; The offline hospital system performs real number encoding on the doctor information, the patient basic information, and the doctor-patient similarity, performs initialization processing, and obtains a doctor-patient matching association set; The offline hospital system pre-constructs a fitness function, and uses the fitness function to evaluate the fitness of each individual in the doctor-patient matching association set to obtain the fitness function value of each individual; If the fitness function value is less than or equal to a preset fitness function value threshold, the primary doctor-patient matching result is subjected to crossover operation and mutation operation to obtain a target doctor-patient matching association set, and the fitness of the target doctor-patient matching association set is re-evaluated; If the fitness function value is greater than a preset fitness function value threshold, the result with the largest fitness function value is selected as the doctor-patient matching result, and the doctor-patient matching result is output.

6. The method according to claim 5, characterized in that The offline hospital system uses a similarity formula to calculate the similarity between the patient's disease information and the doctor's case characteristics, and obtaining the doctor-patient similarity includes the following steps: The offline hospital system vectorizes the patient disease information and the doctor's case features to obtain a vectorized result of the patient disease information and a vectorized result of the doctor's case features; The offline hospital system uses a similarity formula to calculate the similarity between the vectorized results of the patient's disease information and the vectorized results of the doctor's consultation cases to obtain the doctor-patient similarity.

7. The method according to claim 1, characterized in that The front-end gateway includes a first front-end gateway configured by the online hospital system and a second front-end gateway configured by the offline hospital system. The front-end gateway performs protocol conversion on the service request to obtain consultation request information, and sends the consultation request information to the offline hospital system, including the following steps: The first front-end gateway calculates the request address according to the routing addressing algorithm to obtain the address parameter; The first front-end gateway encapsulates the address parameter into the service request to obtain a request parameter set; The first front-end gateway performs protocol conversion on the request parameter set to obtain consultation request information; The first front-end gateway sends an interface call instruction to the offline hospital system to transmit the consultation request information to the offline hospital system.

8. The method according to claim 7, characterized in that The offline hospital system transmits the doctor-patient matching result to the online hospital system through the front gateway, including the following steps: The offline hospital system transmits the doctor-patient matching result to the second front-end gateway; The second front-end gateway calculates the doctor-patient matching result request address according to the routing addressing algorithm to obtain the doctor-patient matching result address parameter; The second front-end gateway encapsulates the doctor-patient matching result address parameter into the doctor-patient matching result to obtain a doctor-patient matching result parameter set; The second front-end gateway performs protocol conversion on the doctor-patient matching result parameter set to obtain doctor-patient matching result information; The second front-end gateway sends an interface call instruction to the online hospital system to transmit the doctor-patient matching result information to the offline hospital system.

Images