Doctor recommendation method and device, electronic equipment and storage medium

Abstract

The invention provides a doctor recommendation method and device, electronic equipment and a computer readable storage medium, and relates to the technical field of data mining. The doctor recommendation method comprises the steps that a heterogeneous information network model and medical history information of a target patient are acquired, and the medical history information comprises at least two target diseases and corresponding disease weights; according to a heterogeneous information network model and the medical history information, the matching degree between the target patient and each doctor is evaluated, and the heterogeneous information network model is used for representing the multi-dimensional connection relation between each disease and each doctor; the matching degrees are ranked from high to low, doctors corresponding to the first N matching degrees are selected and recommended to the target patient, and N refers to a preset number. The problem that in the prior art, only a single disease factor is considered, and similar doctors meeting multi-aspect treatment requirements of patients with multiple medical histories and complicated illness conditions cannot be accurately recommended is solved. The method is suitable for doctor matching and recommendation scenes.

Description

Technical Field

[0001] This invention relates to the field of data mining technology, and in particular to a doctor recommendation method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] In multi-hospital unified registration platforms, patients often face the situation where their target doctor is unavailable. Existing doctor recommendation methods mainly include the following: ① Constructing a binary relationship between disease and doctor, but this has a narrow scope of application and is only suitable for patients with simple conditions.

[0003] ② The assessment of doctors' abilities is based on the similarity between doctors and diseases; however, since it only considers data from the patient's visit, the recommendations may deviate from the patient's actual needs. ③

[0004] Acquiring a large amount of patient data, but with limited doctor information, still affects the accuracy of recommendations. ④ Patients fill out a lot of consultation information to build a hypergraph for doctor recommendations, which is a heavy burden for patients and only supports recommendations for a single disease.

[0005] In summary, most existing doctor recommendation methods only consider a single disease factor, and cannot accurately recommend similar doctors that meet the diverse treatment needs of patients with multiple medical histories and complex conditions. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the above-mentioned shortcomings of the prior art by providing a doctor recommendation method, device, electronic device and computer-readable storage medium, which can enable doctors to make accurate and effective recommendations.

[0007] In a first aspect, the present invention provides a doctor recommendation method, comprising: acquiring a heterogeneous information network model and medical history information of a target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights; evaluating the matching degree between the target patient and each doctor based on the heterogeneous information network model and the medical history information, wherein the heterogeneous information network model is used to characterize the multi-dimensional connection relationship between each disease and each doctor; sorting the matching degrees from high to low, and selecting the doctors corresponding to the top N matching degrees to recommend to the target patient, wherein N refers to a preset number.

[0008] Preferably, obtaining a heterogeneous information network model specifically includes: acquiring doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data; performing data governance on the doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data to obtain a multi-dimensional diagnosis and treatment data matrix, wherein the multi-dimensional diagnosis and treatment data matrix includes each doctor's department, specialty diseases, published / participated projects, and diseases previously treated by the doctor at the same doctor's department; constructing a heterogeneous information network model based on the multi-dimensional diagnosis and treatment data matrix and the preset relationships between departments and diseases, wherein the heterogeneous information network model includes entities and relationships between entities, entities include doctor entities, department entities, disease entities, and project entities, and relationships between entities include the employment relationship between doctor entities and department entities, the specialty relationship between doctor entities and disease entities, the participation relationship between doctor entities and project entities, the diagnosis and treatment relationship between department entities and disease entities, and the inclusion relationship between project entities and disease entities.

[0009] Preferably, the entity also includes a cross-identifier entity, which is used to characterize the cross relationships between departmental entities.

[0010] Preferably, based on the heterogeneous information network model and medical history information, the matching degree between the target patient and each doctor is evaluated. Specifically, this includes: associating the disease entity corresponding to the target disease in the heterogeneous information network model with each doctor entity to obtain the target meta-path between the target disease and each doctor; determining the candidate meta-path between each disease and each doctor, wherein the candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model; evaluating the similarity score between the target meta-path and the candidate meta-path; and evaluating the matching degree between the target patient and each doctor based on the similarity score and the corresponding disease weight of the target disease.

[0011] Preferably, evaluating the similarity score between the target meta-path and the candidate meta-path specifically includes: evaluating the similarity score between the target meta-path and the candidate meta-path according to formula (1):

[0012]

[0013] Among them, Sim(P j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c d represents the candidate metapaths between each disease and doctor c. j Let P represent the disease entity corresponding to the j-th target disease, C represent the disease entity corresponding to at least two target diseases, and P represent the disease entity corresponding to at least two target diseases. i P represents cThe i-th candidate metapath, where ‖C‖ represents the number of diseases related to the target disease C. Indicates in d j ∈P i The time indicator is 1, in The first indicator function is 0 when the time indicator is set, λ represents the preset enhancement factor, and Cross1(e∈P) is the first indicator function. i ) indicates that in P i When including the cross-identifier entity e, the indication is the first preset value, in P i The second indicator function, which indicates the second preset value when the cross-identifier entity e is not included. P represents i With P j→q 'Shared entities, P represents j→q The entity of ′ P represents c The entity.

[0014] Preferably, the matching degree between the target patient and each doctor is assessed based on the similarity score and the corresponding disease weight of the target disease. Specifically, this includes: assessing the matching degree between the target patient and each doctor according to formula (2):

[0015]

[0016] Among them, Match final (q) represents the matching degree between the target patient and doctor q, d j , represents the disease entity corresponding to the j-th target disease, C represents the disease entity corresponding to at least two target diseases, β j Sim(P) represents the disease weight corresponding to the j-th target disease. j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c Let P represent the candidate meta-paths between each disease and doctor c, and let P represent all candidate meta-paths. i P represents c The i-th meta-path in the middle, Cross2(e∈P) i ) indicates that in P i When including the cross-identifier entity e, the indicator is set to the third preset value, in P i The second indicator function, which indicates the fourth preset value when the cross-identifier entity e is not included.

[0017] Secondly, the present invention also provides a doctor recommendation device, comprising an acquisition module, an evaluation module, and a recommendation module. The acquisition module is used to acquire a heterogeneous information network model and the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights. The evaluation module, connected to the acquisition module, is used to evaluate the matching degree between the target patient and each doctor based on the heterogeneous information network model and the medical history information, wherein the heterogeneous information network model is used to characterize the multi-dimensional connection relationship between each disease and each doctor. The recommendation module, connected to the evaluation module, is used to sort the matching degree from high to low and select the doctors corresponding to the top N matching degrees to recommend to the target patient, wherein N refers to a preset number.

[0018] Preferably, the evaluation module includes an association unit, a determination unit, a first evaluation unit, and a second evaluation unit. The association unit is used to associate the disease entity corresponding to the target disease with each doctor entity in the heterogeneous information network model to obtain the target meta-path between the target disease and each doctor. The determination unit is used to determine the candidate meta-path between each disease and each doctor, wherein the candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model. The first evaluation unit is used to evaluate the similarity score between the target meta-path and the candidate meta-path. The second evaluation unit is used to evaluate the matching degree between the target patient and each doctor based on the similarity score and the corresponding disease weight of the target disease.

[0019] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to implement the doctor recommendation method provided in the first aspect above.

[0020] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the doctor recommendation method provided in the first aspect above.

[0021] This invention provides a doctor recommendation method, device, electronic device, and computer-readable storage medium. By utilizing the multi-dimensional connections between diseases and doctors in a heterogeneous information network model, it accurately recommends doctors who are truly skilled in managing the overall combination of their conditions for patients with multiple related or concurrent diseases. This reduces the time cost for patients who have to make multiple appointments with multiple departments due to incomplete recommendations. Therefore, this invention enables accurate and effective doctor recommendations. Attached Figure Description

[0022] Figure 1 This is a flowchart of a doctor recommendation method according to Embodiment 1 of the present invention;

[0023] Figure 2 This is an example diagram of the heterogeneous information network model in Embodiment 1 of the present invention;

[0024] Figure 3 This is an example diagram of the target meta-path and candidate meta-path in Embodiment 1 of the present invention;

[0025] Figure 4 This is a flowchart of a doctor recommendation method according to Embodiment 2 of the present invention;

[0026] Figure 5 This is a schematic diagram of a doctor recommendation device according to Embodiment 3 of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0028] It is understood that the specific embodiments and accompanying drawings described herein are merely for explaining the invention and are not intended to limit the invention.

[0029] It is understood that, without conflict, the various embodiments and features in the embodiments of the present invention can be combined with each other.

[0030] It is understood that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, while the parts unrelated to the present invention are not shown in the drawings.

[0031] It is understood that each unit or module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units or modules may be integrated into one entity structure.

[0032] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of this invention may occur in a different order than that marked in the accompanying drawings.

[0033] It is understood that the flowcharts and block diagrams of this invention illustrate the possible architecture, functions, and operations of systems, apparatuses, devices, and methods according to various embodiments of this invention. Each block in the flowchart or block diagram may represent a unit, module, program segment, or code, containing executable instructions for implementing the specified function. Furthermore, each block or combination of blocks in the block diagram and flowchart can be implemented using a hardware-based system to achieve the specified function, or using a combination of hardware and computer instructions.

[0034] It is understood that the units and modules involved in the embodiments of the present invention can be implemented by software or by hardware. For example, the units and modules can be located in a processor.

[0035] Example 1:

[0036] like Figure 1 As shown, this embodiment provides a doctor recommendation method.

[0037] In this embodiment, the doctor recommendation method is applied to the registration platform, which includes, but is not limited to: registration platforms corresponding to various hospitals, and multiple hospitals on the same registration platform.

[0038] Doctor-recommended methods include:

[0039] S101, Obtain the heterogeneous information network model and the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights.

[0040] In this embodiment, a heterogeneous information network refers to a directed graph structure containing multiple types of entities and multiple types of relationships, used to model the interaction relationships of complex systems. Therefore, the heterogeneous information network model used to characterize the multi-dimensional connection relationships between various diseases and doctors has entity types including but not limited to: doctors and diseases, and relationship types including but not limited to: multi-dimensional connection relationships between various diseases and doctors.

[0041] Obtain disease weights for at least two target diseases, specifically including: determining the severity and urgency of the target diseases based on a pre-defined relationship between the disease and its severity and urgency. According to the formula... Calculate the disease weight corresponding to the target disease, where β j D represents the disease weight corresponding to the j-th target disease, Sick represents at least two target diseases, and D j D represents the j-th target disease. k Severity(D) represents the k-th target disease. j ) represents D j The severity of Urgency (D) j ) represents D j The degree of urgency, Urgency (D k ) represents D k The severity of the disease, Severity (D k ) represents D k The urgency level is assessed. This embodiment assigns corresponding disease weights to different diseases to comprehensively capture the patient's overall diagnosis and treatment needs, thereby providing data support for subsequently matching doctors who cover treatment needs across multiple medical histories.

[0042] Specifically, obtaining the heterogeneous information network model includes steps S1011-S1013:

[0043] S1011, obtain relevant medical data from doctors and patients.

[0044] In this embodiment, doctor-related medical data refers to the multi-dimensional data of doctors stored in the registration platform, and patient-related medical data refers to the multi-dimensional data of patients stored in the registration platform. Doctor-related medical data includes, but is not limited to: basic information, professional skills information, academic achievements information, and patient evaluation information of each doctor; patient-related medical data includes, but is not limited to: basic information, historical registration information, and disease information of each patient.

[0045] Basic information is used to locate the fundamental attributes of doctors and patients. Therefore, the basic information for each doctor includes, but is not limited to: name, gender, age, professional title, hospital, and department. The basic information for each patient includes, but is not limited to: name, gender, and age. Professional skills information is used to locate the doctor's skill attributes. Professional skills information includes, but is not limited to: diseases of expertise, treatment techniques, and surgical proficiency. Academic achievements information is used to quantify the doctor's professional depth. Academic achievements information includes, but is not limited to: journals / fields of publications, research project topics, and roles. Patient evaluation information is used to supplement the doctor's comprehensive ability attributes. Patient evaluation information includes, but is not limited to: treatment effectiveness satisfaction ratings and service attitude feedback texts. Historical registration information is used to analyze patient treatment habits and, combined with the basic information of each patient, to assist in assessing the doctor's patient preferences. Historical registration information includes, but is not limited to: doctors previously consulted, departments, and consultation times. Disease information refers to the patient's description and requests regarding their own disease. Disease information includes, but is not limited to: detailed symptoms entered by the patient, a list of past medical history, and treatment expectations. This embodiment obtains heterogeneous information such as patients' historical registration information, disease information, doctors' professional skills information, academic achievements information, and patient evaluation information as the basis for doctor recommendations. It breaks through the limitations of existing technologies that make doctor recommendations based on medical records, improves the accuracy of doctor recommendations, and realizes end-to-end processing from raw data to accurate doctor recommendations.

[0046] S1012. Data governance is performed on doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data to obtain a multi-dimensional diagnosis and treatment data matrix. The multi-dimensional diagnosis and treatment data matrix includes each doctor's department, diseases they specialize in, projects they have published / participated in, and diseases that the patients they treat have previously sought treatment for at their clinic.

[0047] In this embodiment, data governance is performed on doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data to obtain a multi-dimensional diagnosis and treatment data matrix, which specifically includes: performing data cleaning, data standardization, and data structuring on doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data to obtain structured data corresponding to doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data. Extract the names of each doctor, the hospitals where they work, the departments where they work, the diseases they are good at, the projects they have published / participated in, the names of the patients they have treated, and the evaluations of the treated patients on themselves from the structured data corresponding to doctor-related diagnosis and treatment data; extract the names of each patient, at least one disease they have ever seen, and the names of the doctors they have seen from the structured data corresponding to patient-related diagnosis and treatment data. According to the names of each doctor, the names of the patients they have treated, the names of each patient, and the names of the doctors they have seen, determine and associate the hospitals where each doctor works, the departments where they work, the diseases they are good at, the projects they have published / participated in, the evaluations of the treated patients on themselves, and the diseases that the treated patients have ever seen at themselves, to obtain a multi-dimensional diagnosis and treatment data matrix. In this embodiment, by performing data governance on doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data, the dimensional difference between doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data is eliminated, data accuracy is ensured, noise is filtered, the consistency of cross-source data is achieved, which is convenient for subsequent unified analysis, and provides high-quality and standardized data for the construction of subsequent heterogeneous information network models.

[0048] Data cleaning includes, but is not limited to: removing irrelevant characters (such as garbled characters, special symbols), stop words (such as "de", "le"), extracting core semantics, and numerical normalization (such as mapping the surgical proficiency level to the 0-1 interval). Data standardization includes, but is not limited to: unifying the formats of the same type of data. For example, unify "Department of Cardiology", "Cardiovascular Medicine Department" into "Department of Cardiovascular Medicine", and unify the "paper publication time" into the "YYYY-MM-DD" format. Data structuring includes, but is not limited to: converting unstructured data (such as disease information) and semi-structured data (such as professional skill information) into structured tables (clearly defining fields such as "disease name", "onset time"), and establishing data association indexes (such as "disease ID", "doctor ID" for subsequent association of their professional skill information and academic achievement information).

[0049] S1013, construct a heterogeneous information network model according to the multi-dimensional diagnosis and treatment data matrix and the preset relationship between departments and diseases. Among them, the heterogeneous information network model includes entities and the relationships between entities. The entities include doctor entities, department entities, disease entities, and project entities. The relationships between entities include the employment relationship between doctor entities and department entities, the expertise relationship between doctor entities and disease entities, the participation relationship between doctor entities and project entities, the diagnosis and treatment relationship between department entities and disease entities, and the inclusion relationship between project entities and disease entities.

[0050] Optionally, the entity also includes a cross-identifier entity, which is used to characterize the cross relationships between departmental entities.

[0051] In this embodiment, the employment relationship describes a doctor's professional affiliation and scope of responsibilities within a specific department; the expertise relationship quantifies a doctor's diagnostic and treatment capabilities for specific diseases; the participation relationship describes a doctor's role in a paper / project; the diagnosis and treatment relationship describes the department's standardized treatment capabilities for diseases; the inclusion relationship describes the research correlation between papers / projects and diseases; and the cross-relationship describes multi-departmental collaborative diagnosis and treatment capabilities. The cross-relationship indirectly reflects a doctor's cross-departmental diagnostic and treatment capabilities. The heterogeneous information network model is as follows: Figure 2 As shown, the doctor entities include Doctor 1, Doctor 2, and Doctor 3; the department entities include Department a, Department b, and Department c; the project entities include Project 1, Project 2, and Project 3; the disease entities include Disease a1, Disease a2, Disease b1, Disease b2, and Disease c1; and the cross-identifier entity is... Figure 2 The cross-identifiers in the model. This embodiment uses a heterogeneous information network model to intuitively present the multi-dimensional connection between doctors and diseases, providing a path foundation for subsequent doctor recommendations.

[0052] Based on the multi-dimensional diagnosis and treatment data matrix and the pre-defined relationships between departments and diseases, a heterogeneous information network model is constructed. Specifically, this includes: constructing entities corresponding to each doctor and their department, their areas of expertise, diseases that patients treated by the doctor have previously treated at the doctor's department, and projects published / participated in; the employment relationship between doctor entities and department entities; the expertise relationship between doctor entities and disease entities; the participation relationship between doctor entities and project entities; the diagnosis and treatment relationship between department entities and disease entities; and the inclusion relationship between project entities and disease entities. Here, doctor entities are the entities corresponding to each doctor, department entities are the entities corresponding to the doctor's department, disease entities are the entities corresponding to the diseases the doctor is skilled in or the diseases that patients treated by the doctor have previously treated at the doctor's department, and project entities are the entities corresponding to the projects published / participated in. Determine whether the entity corresponding to the disease the doctor specializes in or the disease the patient treated has previously been treated by the doctor has a treatment relationship with the entity corresponding to the department of each doctor; in response to whether the entity corresponding to the disease the doctor specializes in or the disease the patient treated has previously been treated by the doctor has a treatment relationship with the entity corresponding to the department of each doctor, if no treatment relationship exists, construct a cross-identifier entity between the entity corresponding to the disease the doctor specializes in or the disease the patient treated has previously been treated by the doctor and the entity corresponding to the department of each doctor. For example: Department a is responsible for treating diseases a1 and a2. If patient A has been treated for disease b1 by doctor 1, and disease b1 is also a disease that doctor 1 specializes in, but the department responsible for treating disease b1 is department b, then doctor 1 has cross-departmental treatment capabilities for both departments a and b, and a cross-identifier needs to be constructed between departments a and b.

[0053] It should be noted that the heterogeneous information network model also includes evaluation entities (such as...) Figure 2 Evaluation 1, Evaluation 2 and Evaluation 3), hospital entity (e.g. Figure 3 This embodiment establishes relationships between hospitals, hospital entities, and doctor entities, as well as between evaluation entities and doctor entities. The receiving relationship describes the comprehensive evaluations received by each doctor, while the affiliation relationship describes the employment relationship between doctors and hospitals. If the evaluations of treated patients involve doctors, their hospitals, departments, published / participated projects, areas of expertise, and diseases previously treated by the doctor, this embodiment can also construct relationships between evaluation entities and hospital entities, department entities, project entities, and disease entities. Furthermore, this embodiment can construct relationships between hospital entities and department entities, disease entities, and project entities based on preset relationships between hospitals and departments, hospitals and diseases, and hospitals and projects.

[0054] After constructing the heterogeneous information network model, this embodiment can also obtain the severity and urgency of each disease and the professional title of each doctor. The severity and urgency of each disease are used as attributes of the disease entity corresponding to each disease, and the professional title of each doctor is used as attributes of the doctor entity corresponding to each doctor.

[0055] S102, based on the heterogeneous information network model and medical history information, assess the matching degree between the target patient and each doctor. The heterogeneous information network model is used to characterize the multi-dimensional connection relationship between each disease and each doctor.

[0056] Specifically, S102: Based on the heterogeneous information network model and medical history information, assess the matching degree between the target patient and each doctor, including steps S1021-S1024:

[0057] S1021, in the heterogeneous information network model, the disease entity corresponding to the target disease and each doctor entity are associated to obtain the target meta-path between the target disease and each doctor.

[0058] In this embodiment, taking the medical history information including coronary heart disease and diabetes, and the heterogeneous information network model including disease entity 1 corresponding to coronary heart disease, disease entity 1 corresponding to diabetes, disease entity 3 corresponding to tenosynovitis, doctor entity 1 corresponding to Dr. Zhang, and doctor entity 2 corresponding to Dr. Luo as an example, then the target meta-path includes: disease entity 1 → doctor entity 1, disease entity 1 → doctor entity 2, disease entity 2 → doctor entity 1, disease entity 2 → doctor entity 2, that is, P′={P chd→zhang ′,P chd→luo ′,P dm→zhang ′,P dm→luo ′}.

[0059] S1022, determine the candidate meta-paths between each disease and each doctor, where the candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model.

[0060] In this embodiment, by determining the connection relationship between each disease entity and each doctor entity as candidate meta-paths, the similarity score between the target meta-path and the candidate meta-path can still be evaluated even when there is no disease entity corresponding to the target disease in the heterogeneous information network model. This score serves as the theoretical basis for subsequent doctor recommendations. In other words, when there is no disease entity corresponding to the target disease in the heterogeneous information network model, doctors with cross-departmental diagnosis and treatment capabilities can be recommended to treat the target patient.

[0061] It should be noted that if the heterogeneous information network model covers all diseases, this embodiment can also determine the connection relationships between the disease entity corresponding to the target disease and the doctor entities as candidate meta-paths, such as... Figure 3As shown, the disease entity corresponding to the target disease is... Figure 3 The disease in the target metapath is Figure 3 In the diagram, P1' (disease → doctor), P2' (disease → doctor), and P3' (disease → doctor) represent the individual doctor entities. Figure 3 In the context of doctors, candidate metapaths are... Figure 3 In the diagram, P1 (Doctor → Disease), P2 (Doctor → Academic Achievement → Disease), and P3 (Doctor → Department → Disease) represent project entities. Figure 3 The academic achievements in the department are the entity of the department. Figure 3 The departments in the evaluation entity are... Figure 3 Patient reviews in China.

[0062] S1023, Evaluate the similarity score between the target metapath and the candidate metapath.

[0063] Specifically, S1023: Evaluate the similarity score between the target meta-path and the candidate meta-path, including: evaluating the similarity score between the target meta-path and the candidate meta-path according to formula (1):

[0064]

[0065] Among them, Sim(P j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c d represents the candidate metapaths between each disease and doctor c. j Let P represent the disease entity corresponding to the j-th target disease, C represent the disease entity corresponding to at least two target diseases, and P represent the disease entity corresponding to at least two target diseases. i P represents c The i-th candidate metapath, where ‖C‖ represents the number of diseases related to the target disease C. Indicates in d j ∈P i The time indicator is 1, in The first indicator function is 0 when the time indicator is set, λ represents the preset enhancement factor, and Cross1(e∈P) is the first indicator function. i ) indicates that in P i When including the cross-identifier entity e, the indication is the first preset value, in P i The second indicator function, which indicates the second preset value when the cross-identifier entity e is not included. P represents i With P j→q 'Shared entities, P represents j→q The entity of ′ P representsc The entity.

[0066] In this embodiment, taking candidate meta-paths including Doctor Entity 1 → Disease Entity 1, Doctor Entity 1 → Department Entity 1 → Disease Entity 1, Doctor Entity 1 → Department Entity 1 → Disease Entity 2, Doctor Entity 2 → Department Entity 2 → Disease Entity 3, Doctor Entity 2 → Project Entity 1 → Disease Entity 2, and Doctor Entity 2 → Disease Entity 1 as an example, with λ = 0.5, a first preset value of 0.8, and a second preset value of 0.2, the similarity score between the target meta-path and the candidate meta-path includes Sim(P) chd→zhang ′,P zhang Sim(P) chd→zhang ′,P luo Sim(P) chd→luo ′,P zhang Sim(P) chd→luo ′,P luo Sim(P) dm→zhang ′,P zhang Sim(P) dm→zhang ′,P luo Sim(P) dm→luo ′,P zhang Sim(P) dm→luo ′,P luo ), where P zhang P represents the candidate metapaths between each disease and Dr. Zhang. luo P represents the candidate metapaths between each disease and Dr. Luo. zhang Includes: Doctor Entity 1 → Disease Entity 1, Doctor Entity 1 → Department Entity 1 → Disease Entity 1, Doctor Entity 1 → Department Entity 1 → Disease Entity 2, P luo This includes: Doctor Entity 2 → Department Entity 2 → Disease Entity 3, Doctor Entity 2 → Project Entity 1 → Disease Entity 2, Doctor Entity 2 → Disease Entity 1.

[0067]

[0068] S1023, assess the matching degree between target patients and doctors based on similarity scores and the corresponding disease weights of the target disease.

[0069] Specifically, S1023: Based on the similarity score and the corresponding disease weight of the target disease, assess the matching degree between the target patient and each doctor, including: assessing the matching degree between the target patient and each doctor according to formula (2):

[0070]

[0071] Among them, Match final (q) represents the matching degree between the target patient and doctor q, dj , represents the disease entity corresponding to the j-th target disease, C represents the disease entity corresponding to at least two target diseases, β j Sim(P) represents the disease weight corresponding to the j-th target disease. j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c Let P represent the candidate meta-paths between each disease and doctor c, and let P represent all candidate meta-paths. i P represents c The i-th meta-path in the middle, Cross2(e∈P) i ) indicates that in P i When including the cross-identifier entity e, the indicator is set to the third preset value, in P i The second indicator function, which indicates the fourth preset value when the cross-identifier entity e is not included.

[0072] In this embodiment, taking β1 as 0.7, β2 as 0.3, the third preset value as 1.2, and the fourth preset value as 1 as an example, Match final (zhang)=0.7×(1.12+0.3)+0.3×(0.57+0.24)=1.237, Match final (luo)=0.7×(0.54+0.68)+0.3×(0.24+0.57)=1.097, meaning the matching degree between the target patient and Dr. Zhang is 1.237, and the matching degree between the target patient and Dr. Zhang is 1.097. This embodiment uses meta-path similarity measurement, multi-medical history-doctor association, and cross-disciplinary priority adjustment (i.e., Cross1(e∈P)). i Cross2(e∈P) i This improves the accuracy of recommendations by 20%.

[0073] S103: Sort the matching degree from high to low, and select the doctors corresponding to the top N matching degrees to recommend to the target patient, where N refers to the preset number.

[0074] In this embodiment, the display content for the top N matching doctors is generated and shown to the target patient. The displayed content includes, but is not limited to: doctor's name, department, matching reason (e.g., "specializes in treating coronary heart disease and diabetes"), and appointment information. The target patient can select a doctor for appointment based on the displayed content, or trigger a new doctor recommendation request if dissatisfied with the recommendation results.

[0075] The doctor recommendation method provided in this embodiment utilizes the multi-dimensional connection relationships between diseases and doctors in a heterogeneous information network model to accurately recommend doctors who are truly skilled in handling the overall combination of their conditions for patients with multiple related or concurrent diseases. This reduces the time cost for patients to make multiple appointments with multiple departments due to incomplete recommendations, and achieves accurate and effective doctor recommendations.

[0076] Example 2:

[0077] like Figure 4 As shown, this embodiment provides a doctor recommendation method. The doctor recommendation method includes:

[0078] S201, Obtain relevant medical data from doctors and patients.

[0079] In this embodiment, the doctor's relevant diagnosis and treatment data is... Figure 4 The doctor data and patient-related medical data in the middle Figure 4 Patient data in the database.

[0080] S202 involves data governance of doctor-related and patient-related medical data to obtain a multi-dimensional medical data matrix. This matrix includes each doctor's department, areas of expertise, published / participated projects, and the diseases that patients treated by the doctor had previously sought treatment for.

[0081] In this embodiment, data governance is... Figure 4 Data cleaning, data standardization, and data structuring are involved.

[0082] S203. Based on the multi-dimensional diagnosis and treatment data matrix and the preset relationship between departments and diseases, a heterogeneous information network model is constructed. The heterogeneous information network model includes entities and the relationships between entities. Entities include doctor entities, department entities, disease entities, and project entities. The relationships between entities include the employment relationship between doctor entities and department entities, the expertise relationship between doctor entities and disease entities, the participation relationship between doctor entities and project entities, the diagnosis and treatment relationship between department entities and disease entities, and the inclusion relationship between project entities and disease entities.

[0083] In this embodiment, the heterogeneous information network model is... Figure 4 The entity types, relation types, and network structures within it.

[0084] S204, Obtain the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights; Associate the disease entities corresponding to the target diseases with each doctor entity in the heterogeneous information network model to obtain the target meta-path between the target diseases and each doctor; Determine the candidate meta-path between each disease and each doctor, wherein the candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model.

[0085] In this embodiment, the disease weights corresponding to at least two target diseases are obtained. Figure 4 In the context of medical history weighting, the target meta-path between the target disease and each doctor, and the candidate meta-path between each disease and each doctor, are... Figure 4 The multi-dimensional meta-path set in the middle.

[0086] S205, assess the similarity score between the target metapath and candidate metapaths; based on the similarity score and the corresponding disease weight of the target disease, assess the matching degree between the target patient and each doctor.

[0087] In this embodiment, the similarity score between the target meta-path and the candidate meta-path is evaluated. Figure 4 The process involves dynamically calculating meta-path weights, path similarity calculations, and multi-path weighted fusion. It also assesses the matching degree between the target patient and each doctor. Figure 4 Doctor-medical history matching calculation

[0088] S206 Sort the matching degree from high to low, and select the doctors corresponding to the top N matching degrees to recommend to the target patient, where N refers to the preset number.

[0089] In this embodiment, the doctors corresponding to the first N matching degrees are... Figure 4 Similar doctors in the text.

[0090] The doctor recommendation method provided in this embodiment utilizes the multi-dimensional connection relationships between diseases and doctors in a heterogeneous information network model to accurately recommend doctors who are truly skilled in handling the overall combination of their conditions for patients with multiple related or concurrent diseases. This reduces the time cost for patients to make multiple appointments with multiple departments due to incomplete recommendations, and achieves accurate and effective doctor recommendations.

[0091] Example 3:

[0092] like Figure 5 As shown, this embodiment also provides a doctor recommendation device, including an acquisition module 31, an evaluation module 32, and a recommendation module 33. The acquisition module 31 is used to acquire a heterogeneous information network model and the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights. The evaluation module 32, connected to the acquisition module 31, is used to evaluate the matching degree between the target patient and each doctor based on the heterogeneous information network model and the medical history information, wherein the heterogeneous information network model is used to characterize the multi-dimensional connection relationship between each disease and each doctor. The recommendation module 33, connected to the evaluation module 32, is used to sort the matching degree from high to low and select the doctors corresponding to the top N matching degrees to recommend to the target patient, wherein N refers to a preset number.

[0093] Specifically, the acquisition module 31 includes: an acquisition unit 311, a data governance unit 312, and a construction unit 313. The acquisition unit 311 is used to acquire doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data. The data governance unit 312 is used to perform data governance on the doctor-related diagnosis and treatment data and patient-related diagnosis and treatment data to obtain a multi-dimensional diagnosis and treatment data matrix. The multi-dimensional diagnosis and treatment data matrix includes each doctor's department, specialty diseases, published / participated projects, and diseases that the patients treated have previously sought treatment at the doctor's department. The construction unit 313 is used to construct a heterogeneous information network model based on the multi-dimensional diagnosis and treatment data matrix and the preset relationship between departments and diseases. The heterogeneous information network model includes entities and relationships between entities. Entities include doctor entities, department entities, disease entities, and project entities. Relationships between entities include the employment relationship between doctor entities and department entities, the specialty relationship between doctor entities and disease entities, the participation relationship between doctor entities and project entities, the diagnosis and treatment relationship between department entities and disease entities, and the inclusion relationship between project entities and disease entities.

[0094] Specifically, the evaluation module 32 includes: an association unit 321, a determination unit 322, a first evaluation unit 323, and a second evaluation unit 324. The association unit 321 is used to associate the disease entity corresponding to the target disease in the heterogeneous information network model with each doctor entity to obtain the target meta-path between the target disease and each doctor. The determination unit 322 is used to determine the candidate meta-path between each disease and each doctor, wherein the candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model. The first evaluation unit 323 is used to evaluate the similarity score between the target meta-path and the candidate meta-path. The second evaluation unit 324 is used to evaluate the matching degree between the target patient and each doctor based on the similarity score and the corresponding disease weight of the target disease.

[0095] Specifically, the first evaluation unit 323 includes: a first evaluation subunit, used to evaluate the similarity score between the target meta-path and the candidate meta-path according to formula (1):

[0096] Among them, Sim(P j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c d represents the candidate metapaths between each disease and doctor c. j Let P represent the disease entity corresponding to the j-th target disease, C represent the disease entity corresponding to at least two target diseases, and P represent the disease entity corresponding to at least two target diseases. i P representsc The i-th candidate metapath, where ‖C‖ represents the number of diseases related to the target disease C. Indicates in d j ∈P i The time indicator is 1, in The first indicator function is 0 when the time indicator is set, λ represents the preset enhancement factor, and Cross1(e∈P) is the first indicator function. i ) indicates that in P i When including the cross-identifier entity e, the indication is the first preset value, in P i The second indicator function, which indicates the second preset value when the cross-identifier entity e is not included. P represents i With P j→q 'Shared entities, P represents j→q The entity of ′ P represents c The entity.

[0097] Specifically, the second assessment unit 324 includes: a second assessment subunit for assessing the matching degree between the target patient and each physician according to formula (2):

[0098]

[0099] Among them, Match final (q) represents the matching degree between the target patient and doctor q, d j , represents the disease entity corresponding to the j-th target disease, C represents the disease entity corresponding to at least two target diseases, β j Sim(P) represents the disease weight corresponding to the j-th target disease. j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c Let P represent the candidate meta-paths between each disease and doctor c, and let P represent all candidate meta-paths. i P represents c The i-th meta-path in the middle, Cross2(e∈P) i ) indicates that in P i When including the cross-identifier entity e, the indicator is set to the third preset value, in P i The second indicator function, which indicates the fourth preset value when the cross-identifier entity e is not included.

[0100] Understandably, the doctor recommendation device provided above performs the doctor recommendation method corresponding to Embodiment 1 above. Therefore, the beneficial effects it can achieve can be referred to the beneficial effects of the scheme corresponding to the doctor recommendation method in Embodiment 1 above, which will not be repeated here.

[0101] Example 4:

[0102] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the doctor-recommended method in Embodiment 1 or Embodiment 2 above.

[0103] Example 5:

[0104] This embodiment also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, it implements the doctor recommendation method in Embodiment 1 or Embodiment 2 above.

[0105] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A doctor recommendation method, characterized in that, include: Obtain a heterogeneous information network model and the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights; Based on the heterogeneous information network model and the medical history information, the matching degree between the target patient and each doctor is evaluated. The heterogeneous information network model is used to characterize the multi-dimensional connection relationship between each disease and each doctor. Sort the matching scores from high to low, and select the doctors corresponding to the top N matching scores to recommend to the target patients, where N refers to the preset number.

2. The doctor recommendation method according to claim 1, characterized in that, The acquisition of the heterogeneous information network model specifically includes: Obtain relevant medical data from doctors and patients; Data governance was performed on doctor-related and patient-related diagnosis and treatment data to obtain a multi-dimensional diagnosis and treatment data matrix. The multi-dimensional diagnosis and treatment data matrix includes each doctor's department, diseases they specialize in, projects they have published / participated in, and diseases that their patients have previously sought treatment for at their clinic. Based on the multi-dimensional diagnosis and treatment data matrix and the pre-defined relationships between departments and diseases, a heterogeneous information network model is constructed. The heterogeneous information network model includes entities and the relationships between entities. Entities include doctor entities, department entities, disease entities, and project entities. The relationships between entities include the employment relationship between doctor entities and department entities, the expertise relationship between doctor entities and disease entities, the participation relationship between doctor entities and project entities, the diagnosis and treatment relationship between department entities and disease entities, and the inclusion relationship between project entities and disease entities.

3. The doctor recommendation method according to claim 2, characterized in that, Entities also include cross-identifier entities, which are used to represent the cross relationships between departmental entities.

4. The doctor recommendation method according to claim 3, characterized in that, The process of assessing the matching degree between the target patient and each doctor based on the heterogeneous information network model and the medical history information specifically includes: By associating the disease entity corresponding to the target disease with each doctor entity in the heterogeneous information network model, the target meta-path between the target disease and each doctor is obtained. Determine candidate meta-paths between each disease and each doctor, where candidate meta-paths refer to the connection relationships between each disease entity and each doctor entity in the heterogeneous information network model; Evaluate the similarity score between the target meta-path and the candidate meta-path; The matching degree between the target patient and each doctor is assessed based on the similarity score and the corresponding disease weight of the target disease.

5. The doctor recommendation method according to claim 4, characterized in that, The evaluation of the similarity score between the target meta-path and the candidate meta-path specifically includes: According to formula (1), the similarity score between the target meta-path and the candidate meta-path is evaluated: Among them, Sim(P j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c d represents the candidate metapaths between each disease and doctor c. j Let P represent the disease entity corresponding to the j-th target disease, C represent the disease entity corresponding to at least two target diseases, and P represent the disease entity corresponding to at least two target diseases. i P represents c The i-th candidate metapath, where ‖C‖ represents the number of diseases related to the target disease C. Indicates in d j ∈P i The time indicator is 1, in The first indicator function is 0 when the time indicator is set, λ represents the preset enhancement factor, and Cross1(e∈P) is the first indicator function. i ) indicates that in P i When including the cross-identifier entity e, the indication is the first preset value, in P i The second indicator function, which indicates the second preset value when the cross-identifier entity e is not included. P represents i With P j→q 'Shared entities,' P represents j→q The entity of ′ P represents c The entity.

6. The doctor recommendation method according to claim 5, characterized in that, The process of assessing the matching degree between the target patient and each doctor based on the similarity score and the corresponding disease weight of the target disease specifically includes: According to formula (2), assess the matching degree between the target patient and each doctor: Among them, Match final (q) represents the matching degree between the target patient and doctor q, d j , represents the disease entity corresponding to the j-th target disease, C represents the disease entity corresponding to at least two target diseases, β j Sim(P) represents the disease weight corresponding to the j-th target disease. j→q ′,P c ) represents P j→q ′ and P c Similarity score between them, P j→q ' represents the target meta-path between the j-th target disease and doctor q, P c Let P represent the candidate meta-paths between each disease and doctor c, and let P represent all candidate meta-paths. i P represents c The i-th meta-path in the middle, Cross2(e∈P) i ) indicates that in P i When including the cross-identifier entity e, the indicator is set to the third preset value, in P i The second indicator function, which indicates the fourth preset value when the cross-identifier entity e is not included.

7. A doctor recommendation device, characterized in that, It includes an acquisition module, an evaluation module, and a recommendation module. The acquisition module is used to acquire a heterogeneous information network model and the medical history information of the target patient, wherein the medical history information includes at least two target diseases and their corresponding disease weights. The evaluation module, connected to the acquisition module, is used to assess the matching degree between the target patient and each doctor based on the heterogeneous information network model and the medical history information. The heterogeneous information network model is used to characterize the multi-dimensional connections between each disease and each doctor. The recommendation module, connected to the evaluation module, is used to sort the matching degree from high to low and select the doctors corresponding to the top N matching degrees to recommend to the target patient, where N refers to the preset number.

8. The doctor recommendation device according to claim 7, characterized in that, The evaluation module includes an association unit, a determination unit, a first evaluation unit, and a second evaluation unit. The association unit is used to associate the disease entity corresponding to the target disease with each doctor entity in the heterogeneous information network model, thereby obtaining the target meta-path between the target disease and each doctor. The determining unit is used to determine candidate meta-paths between each disease and each doctor. Here, a candidate meta-path refers to the connection relationship between each disease entity and each doctor entity in the heterogeneous information network model. The first evaluation unit is used to evaluate the similarity score between the target meta-path and the candidate meta-path. The second assessment unit is used to assess the matching degree between the target patient and each doctor based on the similarity score and the corresponding disease weight of the target disease.

9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to implement a physician recommendation method as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements a doctor recommendation method as described in any one of claims 1 to 6.

Images