Intelligent hospital guide method and system and intelligent hospital guide robot

Abstract

The invention is suitable for the technical field of medical robots, and particularly relates to an intelligent hospital guide method and system and an intelligent hospital guide robot, and the method comprises the steps: obtaining department attribute data and a multi-mode operation record of a target patient, and integrating dynamic information; performing data feature fusion according to the multi-modal operation record and the department attribute data, and determining a triage adaptive feature group of the target patient; calling a historical triage database according to the triage adaptation feature group of the target patient to obtain the number of abnormal first triage instances in the historical triage database; screening in a historical triage database through the triage adaptation feature group of the target patient to obtain the number of second triage instances associated with the triage adaptation feature group; and according to the first triage instance number and the second triage instance number, the triage conflict coefficient is constructed, the hospital guide information of the target patient is determined according to the triage conflict coefficient, and special precise hospital guide based on the real-time features and historical data improves the doctor seeing experience of the patient.

Description

Technical Field

[0001] The present application belongs to the field of medical robot technology, and in particular relates to an intelligent medical guidance method, system and intelligent medical guidance robot. Background Art

[0002] Intelligent medical guidance is a key technology for clinical medical decision support. It aims to achieve accurate triage path recommendations by integrating patients' multimodal medical data. By analyzing dynamic indicators such as symptom descriptions, physical sign dimensions, and department resource allocation, it provides a decision-making basis for matching treatment departments. It is widely used in scenarios such as triage path optimization, medical resource scheduling, and improvement of treatment efficiency.

[0003] Traditional guidance methods have significant limitations, including insufficient data fusion capabilities, limited model processing capabilities, low department adaptability, and lack of effective triage decision-making methods, which lead to inefficient clinical triage and improper allocation of medical resources. At the same time, it restricts the optimization space for personalized medical treatment and affects patients' medical experience. Summary of the invention

[0004] The embodiments of the present application provide an intelligent medical guidance method, system and intelligent medical guidance robot, which can solve the problem that medical guidance lacks effective triage decision-making means, resulting in low clinical triage efficiency and improper allocation of medical resources, while restricting the optimization space for personalized medical treatment and affecting the patient's medical experience.

[0005] In a first aspect, an embodiment of the present application provides an intelligent diagnosis guidance method, including: Obtain department attribute data and multimodal operation records of target patients; Performing data feature fusion according to the multimodal operation record and the department attribute data to determine a triage adaptation feature group for the target patient; Acquire a historical triage database, retrieve the first triage instance with triage abnormality in the historical triage database, and obtain the number of first triage instances; wherein the historical triage database includes a plurality of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's department; Screening second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is triage instances with triage anomalies whose matching degree between the department attribute data of the visiting department and the physical sign data of the visiting patient and the triage adaptation feature group reaches a preset matching threshold; A triage conflict coefficient is constructed according to the number of the first triage instances and the number of the second triage instances, and the guidance information of the target patient is determined based on the triage conflict coefficient.

[0006] The above technical solutions in the embodiments of the present application have at least the following technical effects: The intelligent guidance method provided in the embodiment of the present application comprehensively integrates the dynamic information of patient signs, behaviors and department resource allocation by acquiring department attribute data and multimodal operation records of target patients. Data feature fusion is performed based on multimodal operation records and department attribute data to determine the triage adaptation feature group of the target patient, providing data support for subsequent analysis. The historical triage database is called based on the triage adaptation feature group of the target patient to obtain the number of first triage instances with abnormalities in the historical triage database; the historical triage database is screened through the triage adaptation feature group of the target patient to obtain the number of second triage instances associated with the triage adaptation feature group, thereby realizing guidance analysis based on historical data. According to the number of first triage instances and the number of second triage instances, a triage conflict coefficient is constructed, and the guidance information of the target patient is determined based on the triage conflict coefficient, so as to provide specialized and precise guidance based on real-time features and historical data and improve the patient's medical experience.

[0007] In a second aspect, an embodiment of the present application provides an intelligent medical guidance system, including: An acquisition unit, used to acquire department attribute data and multimodal operation records of target patients; A fusion unit, configured to perform data feature fusion according to the multimodal operation record and the department attribute data, and determine a triage adaptation feature group for the target patient; A retrieval unit is used to obtain a historical triage database, retrieve the first triage instance with triage abnormality in the historical triage database, and obtain the number of first triage instances; wherein the historical triage database includes a plurality of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's department; A screening unit is used to screen second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is triage instances with triage abnormalities in which the department attribute data of the visiting department and the patient's physical sign data match the triage adaptation feature group to a preset matching threshold; A guidance unit is used to construct a triage conflict coefficient according to the number of the first triage instances and the number of the second triage instances, and determine the guidance information of the target patient based on the triage conflict coefficient.

[0008] In a third aspect, an embodiment of the present application provides an intelligent medical guidance robot, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of the above aspects when executing the computer program.

[0009] In a fourth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on an intelligent medical guidance robot, the intelligent medical guidance robot executes the method of any one of the above aspects.

[0010] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the above aspects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 It is a flowchart of an intelligent diagnosis guidance method provided by an embodiment of the present application; Figure 2 It is a schematic diagram of the operation of the intelligent diagnosis guidance method provided in one embodiment of the present application; Figure 3 It is a structural diagram of an intelligent medical guidance system provided in one embodiment of the present application; Figure 4 It is a structural schematic diagram of an intelligent medical guidance robot provided in one embodiment of the present application. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0014] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0015] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] As used in the specification of this application and the appended claims, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if the described condition or event is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce the described condition or event is detected" or "in response to detecting the described condition or event" depending on the context.

[0017] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0019] Traditional medical guidance methods have significant limitations, including insufficient data fusion capabilities, limited model processing capabilities, low department matching adaptability, and lack of effective triage decision-making methods, which lead to inefficient clinical triage and improper allocation of medical resources. At the same time, it restricts the optimization space for personalized medical treatment and affects patients' medical experience.

[0020] To solve the above problems, the embodiments of the present application provide an intelligent guidance method, system and intelligent guidance robot. In this method, by acquiring department attribute data and multimodal operation records of target patients, the dynamic information of patient signs, behaviors and department resource configuration is fully integrated. Data feature fusion is performed based on multimodal operation records and department attribute data to determine the triage adaptation feature group of the target patient, providing data support for subsequent analysis. The historical triage database is called according to the triage adaptation feature group of the target patient to obtain the number of first triage instances with abnormalities in the historical triage database; the triage adaptation feature group of the target patient is screened in the historical triage database to obtain the number of second triage instances associated with the triage adaptation feature group, and the guidance analysis based on historical data is realized. According to the number of the first triage instances and the number of the second triage instances, the triage conflict coefficient is constructed, and the guidance information of the target patient is determined based on the triage conflict coefficient, so as to provide precise guidance based on the specialization of real-time features and historical data and improve the patient's medical experience.

[0021] The intelligent medical guidance method provided in the embodiment of the present application can be applied to an intelligent medical guidance robot. In this case, the intelligent medical guidance robot is the executor of the intelligent medical guidance method provided in the embodiment of the present application. The embodiment of the present application does not impose any restrictions on the specific type of the intelligent medical guidance robot.

[0022] For example, the intelligent medical guidance robot can be a variety of types of robots. For example, the intelligent medical guidance robot can be a humanoid robot, a counter robot, a wearable robot, a track robot, or a vehicle-mounted robot. The intelligent medical guidance robot can include, but is not limited to, mobile devices, communication devices, display devices, sensor devices, etc.

[0023] In order to better understand the intelligent medical guidance method provided in the embodiment of the present application, the specific implementation process of the intelligent medical guidance method provided in the embodiment of the present application is exemplarily introduced below.

[0024] Figure 1 A schematic flow chart of the intelligent diagnosis guidance method provided in an embodiment of the present application is shown. Figure 2 The operation flow chart of the intelligent diagnosis guidance method provided in the embodiment of the present application is shown, and the intelligent diagnosis guidance method includes: S100, acquiring department attribute data and multimodal operation records of target patients.

[0025] It can be understood that department attribute data refers to the characteristics and capabilities of each department in the hospital, which may include the department's professional field, medical equipment configuration, professional level of the doctor team, historical case handling success rate, types of diseases that the department is good at treating, etc. Department attribute data can be extracted from the hospital's management system or constructed in the form of a knowledge graph to describe the department's medical capabilities and resource distribution. The multimodal operation records of the target patient refer to the various types of data generated by the patient during the medical treatment process, which may include the target patient's physiological data, behavioral data, medical history data, and operation records generated by the patient during the interaction process. Multimodal operation records can come from multiple data sources such as the hospital's electronic medical record system, sensor devices, and mobile applications.

[0026] Exemplarily, when acquiring department attribute data, it can be directly extracted through the hospital's information system or queried from a pre-built knowledge base. When acquiring the multimodal operation records of the target patient, the patient's data in different time periods and scenarios can be collected and integrated. For example, the patient's physiological data and medical history can be extracted from the electronic medical record system, and real-time monitoring data and patient interaction records can be obtained from sensor devices. It can provide basic data support for subsequent data feature fusion and analysis, and provide an important basis for the decision-making of the intelligent guidance system.

[0027] S200, performing data feature fusion based on multimodal operation records and department attribute data to determine a triage adaptation feature group for the target patient.

[0028] It can be understood that multimodal operation records can include physiological data, behavioral data, and medical history of patients at different times and occasions. Multimodal operation records can be time-series aligned and semantically parsed to extract operation behavior feature vectors and mark abnormal operations. Time-series alignment ensures that data from different time periods are consistent on the same time scale, while semantic parsing converts these data into a format that can be calculated and analyzed. A departmental professional capability map can be constructed based on departmental attribute data. The process of constructing a departmental professional capability map involves encoding information such as the department's historical case processing success rate into three-dimensional features to form a network graph in which nodes represent the key capability features of the department. After feature extraction and departmental capability map construction are completed, the patient's behavioral feature vector can be cross-modally associated with the departmental capability map. By expanding the dimension of the behavioral feature vector and projecting it into the same feature space as the departmental capability map, the dynamic semantic similarity between each behavioral feature vector and the department node is calculated. After the dynamic semantic similarity is segmented and normalized, a standardized association weight is formed to construct the behavior-department association matrix. The dynamic attention mechanism can be used to assign weights to the three-dimensional feature encodings of abnormal operation marks and department capability maps, calculate the feature coupling degree, integrate the behavior-department association matrix and feature coupling degree, perform nonlinear fusion, and generate a triage adaptation feature group that reflects the compatibility between patients and departments. This is helpful in providing core data support for subsequent triage decisions and helping to determine the most suitable department for target patients.

[0029] In a possible implementation, S200, data feature fusion is performed according to the multimodal operation record and department attribute data to determine the triage adaptation feature group of the target patient, including: S210, performing time sequence alignment and semantic analysis on the multimodal operation records, and extracting operation behavior feature vectors and abnormal operation marks.

[0030] It can be understood that time series alignment refers to unifying data from different time points and different devices on the time axis to achieve consistency of data on the same time scale. For example, the blood pressure data and heart rate data of the target patient may be collected at different time points. They can be aligned to the same time axis through interpolation or other methods for subsequent analysis. Time series alignment can improve the availability of data and avoid analysis errors caused by time inconsistency.

[0031] Semantic parsing refers to the structural processing of unstructured or semi-structured data to extract key feature information. For example, the target patient's consultation content may be recorded in natural language, and it is necessary to extract key information such as symptom description and treatment recommendations through natural language processing technology; the patient's examination items may be recorded in the form of text or code, which needs to be parsed into specific examination types and result data. Semantic parsing can convert complex data into feature vectors that can be calculated and analyzed, facilitating subsequent feature fusion.

[0032] Based on time alignment and semantic parsing, the operation behavior feature vector can be extracted. The operation behavior feature vector is a numerical representation of the patient's operation record, which can reflect the target patient's behavior pattern during the medical treatment process. In addition, there may be abnormal operation marks in the operation record, which can be detected by predefined rules. Abnormal operation marks are indications of abnormal conditions in the patient's health status or behavior pattern, which can be used for subsequent weight allocation and feature fusion.

[0033] S220, constructing a department professional capability map based on department attribute data; wherein the department professional capability map includes a three-dimensional feature coding of the success rate of historical case processing.

[0034] It can be understood that the process of constructing a department's professional capability map includes feature extraction and encoding of department attribute data. Among them, the historical case handling success rate is one of the core features of department attribute data, reflecting the actual performance of the department in a specific disease or treatment field. The historical case handling success rate can be encoded in multiple dimensions, such as based on disease type, treatment method, patient age and other dimensions, to form a three-dimensional feature encoding, which more comprehensively describes the professional capabilities of the department and provides rich information for subsequent feature fusion.

[0035] Exemplarily, when constructing a department's professional capability map, the department's attribute data can be cleaned and preprocessed to remove redundant information and unify the data format. The key feature dimensions can be extracted by statistics and analysis of the success rate of historical case processing. For example, the success rate of different departments can be counted by disease type, the success rate can be counted by treatment methods (such as surgery, drug treatment), or the success rate can be counted by patient age group. These statistical results are encoded into three-dimensional feature vectors, each dimension corresponding to a specific feature, such as disease type dimension, treatment method dimension, patient age dimension, etc. The encoded feature vector can be integrated with other department attribute data (such as doctor's professional level, medical equipment configuration, etc.) to form a department's professional capability map. The department's professional capability map reflects the capabilities and resource distribution between departments in the form of a graph structure, providing basic data support for subsequent cross-modal association mapping.

[0036] S230, perform cross-modal correlation mapping between the operation behavior feature vector and the department professional capability map to generate a behavior-department correlation matrix.

[0037] It can be understood that cross-modal association mapping refers to associating data from different modalities (patient behavior and department capabilities) and calculating the similarity or correlation between them.

[0038] When performing cross-modal association mapping, the operation behavior feature vector and the department professional capability map can be encoded into the same feature space. The dynamic semantic similarity between each operation behavior feature vector and each node in the department professional capability map can be calculated. Dynamic semantic similarity refers to dynamically adjusting weights according to different feature dimensions to reflect the importance of different features to the association. The calculated dynamic semantic similarity can be segmented and normalized to convert it into a standardized association weight. Segmented normalization can eliminate the dimensional differences between different feature dimensions, making the similarity values ​​comparable within the same range. By arranging the association weights according to the node relationship between the patient behavior feature vector and the department professional capability map, a behavior-department association matrix is ​​generated. The behavior-department association matrix is ​​a two-dimensional matrix, in which the rows represent the patient's operation behavior feature vectors, the columns represent the nodes in the department professional capability map, and each element in the matrix represents the strength of the association between the patient's behavior and the department's capability. By generating the behavior-department association matrix, the patient's behavior characteristics can be quantitatively associated with the department's professional capabilities, providing an important basis for the subsequent determination of the triage adaptation feature group. The cross-modal association mapping method can comprehensively consider multiple dimensions of patient behavior and department capabilities, and improve the accuracy and scientificity of triage decisions.

[0039] Optionally, S230, by performing cross-modal correlation mapping between the operation behavior feature vector and the department professional capability map, a behavior-department correlation matrix is ​​generated, including: S231, expand the dimension of the operation behavior feature vector, and project the operation behavior feature vector to the same feature space as the department's professional capability map.

[0040] It can be understood that dimension expansion refers to increasing the dimension of the operation behavior feature vector to the same number of dimensions as the department's professional capability map through mathematical methods (such as zero filling, interpolation, or feature mapping) so that the two can be compared and analyzed in the same feature space. For example, if the feature space of the department's professional capability map includes three dimensions: disease type, treatment method, and patient age, and the operation behavior feature vector only contains two dimensions: registration frequency and examination item type, the dimension expansion can be completed by adding new feature dimensions (such as patient age) or expanding existing features (such as mapping registration frequency to disease type dimension). Dimension expansion can make the operation behavior feature vector comparable with the department's professional capability map in the same feature space, providing a basis for subsequent dynamic semantic similarity calculations.

[0041] S232, based on the mapping relationship in the preset symptom-department association rule library, calculate the dynamic semantic similarity between each operation behavior feature vector and the department professional capability map node.

[0042] It can be understood that the symptom-department association rule base is a pre-built knowledge base used to describe the association between different symptoms and departments. For example, the rule base may associate the symptom of "fever" with departments such as "internal medicine" and "respiratory medicine", or associate the symptom of "fracture" with departments such as "orthopedics" and "emergency department". The symptom-department association rule base can be constructed through expert experience and historical case analysis. When calculating the dynamic semantic similarity, the corresponding mapping relationship can be found in the symptom-department association rule base according to the features in the operation behavior feature vector (such as the patient's main complaint symptoms, examination results, etc.), and the dynamic semantic similarity between the operation behavior feature vector and the node of the department's professional ability map is calculated based on the mapping relationship and the node features in the department's professional ability map. Dynamic semantic similarity refers to dynamically adjusting weights according to different feature dimensions to reflect the importance of different features to the association in a specific context. For example, for fever symptoms, the weight of the success rate of historical case treatment in the internal medicine department may be higher, while for fracture symptoms, the weight of the success rate of historical case treatment in the orthopedics department may be higher. By calculating dynamic semantic similarity, the correlation strength between patient behavior characteristics and department capabilities can be quantified, providing a basis for the subsequent construction of the correlation matrix.

[0043] S233, performing segment normalization processing on the dynamic semantic similarity to generate a standardized association weight.

[0044] It can be understood that the dynamic semantic similarity may have a large difference in its numerical range because it may come from different feature dimensions or calculation methods. For example, the similarity values ​​of some features may be concentrated between 0 and 1, while the similarity values ​​of other features may be distributed between 0 and 100. In order to eliminate this dimensional difference, the dynamic semantic similarity can be segmented and normalized. Segmented normalization refers to standardizing the similarity values ​​according to different intervals (segments). For example, the similarity values ​​can be divided into three intervals: low, medium, and high, and normalization is performed in each interval so that all similarity values ​​are mapped to a unified range (such as between 0 and 1). Segmented normalization can retain the distribution characteristics of the similarity values ​​while eliminating the dimensional differences between different dimensions, making the similarity values ​​comparable and interpretable. The normalized value is the standardized association weight, which is used to represent the association strength between the target patient's operation behavior feature vector and the department's capabilities.

[0045] S234, construct an N×M sparse association matrix based on the standardized association weights, and perform time series noise reduction on the sparse association matrix to generate a behavior-department association matrix; wherein N is the total number of dimensions of the operation behavior feature vector, and M is the total number of nodes in the department's professional capability map.

[0046] It can be understood that the standardized association weight is a numerical representation of the strength of association between the operation behavior feature vector and the node of the department's professional capability map. Based on the standardized association weight, an N×M association matrix can be constructed, where N represents the total number of dimensions of the operation behavior feature vector (such as registration frequency, type of examination items, main complaint symptoms, etc.), and M represents the total number of nodes of the department's professional capability map (such as internal medicine, surgery, orthopedics, etc.). Since not all operation behavior features are strongly associated with all department nodes, the association matrix is ​​usually sparse, that is, most elements in the matrix are zero or close to zero. In order to reduce the influence of noise in the sparse matrix, it is necessary to perform time series denoising on the association matrix. Time series denoising refers to smoothing the weight values ​​in the matrix using time series analysis methods or filtering techniques (such as moving average filtering, low-pass filtering, etc.) to eliminate the influence of random noise or outliers. For example, if a weight value fluctuates violently in a short period of time, it may be regarded as noise, and it can be smoothed to a more stable value through filtering technology. The association matrix after denoising is the final behavior-department association matrix. The behavior-department association matrix can clearly quantify the correlation between the patient's operational behavior characteristics and the department's capabilities, providing an important basis for subsequent triage decisions.

[0047] S240, weights are assigned to abnormal operation marks and three-dimensional feature codes through a dynamic attention mechanism, and feature coupling is calculated.

[0048] It can be understood that abnormal operation marks are identifications of abnormal behaviors of patients during medical treatment, reflecting possible health risks or abnormal behaviors of patients. Three-dimensional feature coding is a multi-dimensional representation of the success rate of historical case handling in the department's professional capability map, such as coding based on disease type, treatment method, patient age and other dimensions. Feature coupling refers to the strength of association between abnormal operation marks and three-dimensional feature coding, which is used to quantify the degree of match between abnormal patient behavior and department capabilities.

[0049] Exemplarily, the dynamic attention mechanism is a model that can dynamically adjust the weight distribution according to the context of the input data. Its core idea is to adaptively weight the contribution of different features. When calculating the feature coupling, the abnormal operation mark and the three-dimensional feature code can be input into the dynamic attention mechanism. The dynamic attention mechanism determines the degree of association between each abnormal operation mark and each three-dimensional feature code by calculating the attention score between the two. Dynamic weight distribution is performed on the abnormal operation mark and the three-dimensional feature code based on the attention score, that is, different weight values ​​are assigned to different features to reflect their importance in the calculation of feature coupling. According to the weight distribution result, the feature coupling between the abnormal operation mark and the three-dimensional feature code is calculated. Feature coupling is a numerical indicator used to quantify the degree of match between the patient's abnormal behavior and the department's capabilities, providing an important basis for subsequent feature fusion.

[0050] S250, based on the behavior-department association matrix and feature coupling degree, nonlinear fusion of multimodal operation record features and department attribute features is performed to generate a triage adaptation feature group for the target patient.

[0051] It can be understood that nonlinear fusion refers to the use of nonlinear transformation methods to integrate features from different modalities to generate a feature group that can comprehensively reflect the compatibility of patients and departments.

[0052] Exemplarily, the behavior-department association matrix can be integrated with the feature coupling degree to generate a comprehensive association feature matrix. For example, the feature coupling degree can be used as a weight to adjust the behavior-department association matrix in a weighted manner to enhance the influence of features with strong correlation with abnormal behavior on the final feature group. The multimodal operation record features and department attribute features are input into a nonlinear fusion model, such as using models such as neural networks, decision trees, or support vector machines. The nonlinear fusion model model can capture complex nonlinear relationships between features, such as the interaction between patient behavior characteristics and department capability characteristics. Through the calculation of the nonlinear fusion model, the multimodal operation record features and department attribute features are integrated to generate a triage adaptation feature group for the target patient. The triage adaptation feature group is a comprehensive feature vector that contains information on patient behavior characteristics, abnormal behavior markers, department capability characteristics, and their correlations, and can provide comprehensive and accurate data support for subsequent triage decisions. By generating a triage adaptation feature group, the scientificity and accuracy of triage decisions can be improved to ensure that patients can be assigned to the most suitable department for treatment.

[0053] S300, obtaining a historical triage database, searching the historical triage database for the first triage instance with triage abnormality, and obtaining the number of first triage instances; wherein the historical triage database includes a number of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's visiting department.

[0054] It can be understood that the historical triage database is a data set that stores the triage records of past patients, where each record is a historical triage instance. The historical triage instance records the patient's physical sign data and the department attribute data of the assigned department. Physical sign data can include the patient's physiological indicators (such as blood pressure, heart rate, body temperature), medical history information (such as previous diseases, allergies) and behavioral data (such as the number of registrations, the number of examination items), etc. Department attribute data can include the department's professional field (such as internal medicine, surgery, pediatrics), equipment configuration (such as CT machine, ultrasound machine) and the level of the doctor team (such as doctor's qualifications, surgical success rate), etc. In the historical triage database, the first triage instance refers to the situation where the department to which the patient is assigned does not match his physical sign data, and there are historical triage instances with triage anomalies. For example, a patient has heart disease but is assigned to a non-cardiovascular department, or a patient needs a complex operation but is assigned to a department with insufficient equipment. The purpose of retrieving the first triage instance is to identify possible problems in historical triage, so as to optimize the triage process, improve triage accuracy and patient safety.

[0055] In a possible implementation, S300, obtaining a historical triage database, searching the historical triage database for first triage instances with triage exceptions, and obtaining the number of first triage instances includes: S310, obtaining a historical triage database.

[0056] It can be understood that obtaining the historical triage database refers to reading a data set containing historical triage instances from a storage system. The historical triage database can be stored in a relational database (such as MySQL, PostgreSQL), a non-relational database (such as MongoDB), or a distributed file system (such as HDFS). When obtaining the historical triage database, the historical triage instances can be loaded into memory or local storage through a database query interface (such as SQL query) or a data import tool (such as an ETL tool) for subsequent retrieval and analysis.

[0057] S320, searching the historical triage database according to a preset abnormality determination standard, obtaining the first triage instances in the historical triage database having triage abnormalities, and counting the number of the first triage instances to obtain the number of first triage instances.

[0058] It is understood that the preset abnormality determination criteria are rules or conditions for determining whether there are triage abnormalities in historical triage instances. The abnormality determination criteria can be formulated based on professional medical guidelines or historical triage experience. For example, the abnormality determination criteria can include the following conditions: The patient's vital signs indicated that she required priority treatment (such as high-risk pregnancy, acute myocardial infarction), but she was assigned to a non-priority department.

[0059] Patients require specific equipment or technology (such as heart stent surgery, tumor radiotherapy) but are assigned to a department that does not have the equipment or technology.

[0060] The patient's medical history data indicates that he or she has a higher need for a specific department (such as endocrinology for diabetes), but he or she is assigned to another department.

[0061] The patient's medical history data showed that the patient was reallocated to another department without any treatment after registration.

[0062] During the retrieval process, the historical triage instances can be traversed one by one, and each instance can be evaluated according to the abnormality judgment criteria. If an instance meets at least one condition in the abnormality judgment criteria, it is marked as the first triage instance with triage abnormality. For example, for a record, it can be checked whether the patient's physical sign data matches the department attribute data. If it matches, it is marked as normal, otherwise it is marked as abnormal. After the retrieval is completed, the number of all first triage instances marked as abnormal is counted to obtain the number of first triage instances. The number of first triage instances reflects the number of triage abnormalities in historical triages, providing data support for the subsequent optimization of triage processes. For example, if the number of first triage instances is high, it means that there may be major problems in the existing triage process, and it is necessary to further analyze the causes and optimize the triage rules; if the number of first triage instances is low, it means that the existing triage process is relatively reasonable. By counting the number of first triage instances, a scientific basis can be provided for triage dynamics.

[0063] S400, screening second triage instances associated with the triage adaptation feature group from all first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is the triage instances with triage abnormalities whose matching degree between the department attribute data of the visiting department and the physical sign data of the visiting patient and the triage adaptation feature group reaches a preset matching threshold.

[0064] It can be understood that the first triage instance is a triage anomaly instance that meets the anomaly judgment criteria in the historical triage database; the second triage instance is a triage anomaly instance that is further screened from the first triage instance and is associated with the triage adaptation feature group. The triage adaptation feature group is a feature vector used to reflect the adaptability of the patient to the department, including the patient's physical sign data, behavioral data, medical history data, and department attribute data. The preset matching threshold is a numerical standard used to measure the degree of matching between the department attribute data and the patient's physical sign data and the triage adaptation feature group. For example, instances with a matching degree greater than 0.8 are considered to be successfully matched. The purpose of screening the second triage instance is to identify situations in the historical triage anomaly instances that are highly similar to the current target patient's triage adaptation feature group, so as to discover potential triage rules or problems and optimize the triage process. The number of second triage instances reflects the degree of association between the historical anomaly instances and the current target patient's triage adaptation feature group, providing a reference for triage decisions.

[0065] In a possible implementation, S400, screening second triage instances associated with the triage adaptation feature group from all first triage instances to obtain the number of second triage instances includes: S410, constructing department attribute data matching conditions and physical sign data correlation constraint conditions based on the triage adaptation feature group of the target patient.

[0066] It can be understood that constructing department attribute data matching conditions means defining department matching rules for screening historical triage instances based on department attribute dimensions (such as department professional fields, equipment configuration, and doctor team level) in the triage adaptation feature group. For example, if the target patient's triage adaptation feature group contains the "cardiovascular disease" department attribute, the matching condition can be set to screen the records of the department's professional fields in the historical instances that include cardiovascular disease diagnosis and treatment capabilities. The physical sign data association constraint condition means defining the calculation method and threshold range of physical sign similarity based on the patient physical sign dimensions (such as blood pressure, heart rate, body temperature, and other physiological indicators) in the triage adaptation feature group. For example, the cosine similarity algorithm is used to calculate the similarity between the target patient's physical sign vector and the historical instance's physical sign vector, and the lower limit of the similarity is set to 0.7.

[0067] The matching conditions of department attribute data can be constructed through the vector space model. After encoding the department attribute characteristics into multidimensional vectors, the Euclidean distance is used to measure the matching degree. The constraint conditions of the correlation degree of physical sign data can be combined with the normalized physical sign data, and the correlation degree can be calculated by the Pearson correlation coefficient or the dynamic time warping algorithm (for time series physical sign data).

[0068] S420, based on the department attribute data matching condition and the physical sign data association constraint condition, all first triage instances are screened to obtain second triage instances, and the number of second triage instances is counted to obtain the number of second triage instances.

[0069] It can be understood that the department attribute data matching conditions and the sign data association constraint conditions can be applied to the data sets of all first triage instances through a database query engine or a data filtering algorithm. For each first triage instance, double verification can be performed: verify whether its department attributes meet the matching conditions, such as using Boolean logic to determine whether the department's professional field contains the target disease type; secondly, calculate the sign data association score, such as aligning the sign data of the historical instance with the target patient's sign data, and then using a predefined similarity algorithm to calculate the numerical result. If and only if both conditions are met at the same time (such as department matching ≥ 0.8 and sign association ≥ 0.7), the first triage instance is marked as a second triage instance.

[0070] For example, a distributed computing framework (such as Spark) can be used to perform parallel screening of large-scale historical data. The data set fragments of all first triage instances can be loaded into memory, and the filtering conditions can be applied to each data fragment. The condition matching and correlation calculation are completed through MapReduce operations, and finally the total number of second triage instances that meet the conditions is counted through aggregation operations. The number of second triage instances obtained will be used as one of the core parameters for the subsequent triage abnormal risk probability calculation.

[0071] S500: construct a triage conflict coefficient according to the number of first triage instances and the number of second triage instances, and determine the guidance information of the target patient according to the triage conflict coefficient.

[0072] It can be understood that the triage conflict coefficient is a quantitative indicator used to characterize the degree of decision difference between the triage adaptation feature group of the current target patient and the historical triage abnormal instances. The number of first triage instances reflects the overall scale of historical triage abnormal instances, and the number of second triage instances characterizes the number of abnormal instances that are highly associated with the characteristics of the current target patient. When constructing the triage conflict coefficient, the distribution difference between the first triage instance and the second triage instance in the multidimensional feature space can be modeled by the probability density distribution function, and the degree of conflict can be quantified using information theory methods. The process of determining the guidance information can be based on the correlation between the triage conflict coefficient and the risk of misdiagnosis in the department, and the department with the lowest triage conflict coefficient can be selected as the recommended target to reduce the risk of triage errors.

[0073] In a possible implementation, S500, constructing a triage conflict coefficient according to the number of first triage instances and the number of second triage instances, and determining the referral information of the target patient according to the triage conflict coefficient, includes: S510, obtaining the number of triage instances in the historical triage database.

[0074] It can be understood that the number of triage instances refers to the total number of all triage instance records in the historical triage database, including normal triage instances and abnormal triage instances. The number of triage instances can be directly obtained through database statistical functions (such as SQL's COUNT operation), or quickly read through a pre-generated metadata index. The number of triage instances provides a benchmark reference value for subsequent probability density distribution modeling, which is used to standardize the statistical ratio of the first triage instance and the second triage instance. For example, when the historical triage database contains 100,000 records, the number of triage instances is 100,000, and the number of triage instances will be used as the denominator to calculate the probability of occurrence of abnormal instances.

[0075] S520, mapping the first number of triage instances and the number of triage instances to a multidimensional feature space, and generating a first probability density distribution function corresponding to the first number of triage instances.

[0076] It can be understood that the multidimensional feature space is a mathematical modeling space composed of multiple feature dimensions of historical triage instances, such as department professional matching, physical sign data deviation, equipment requirement satisfaction, etc. The first probability density distribution function P(x) reflects the aggregation pattern of historical triage abnormal instances in the feature space. The distribution pattern of historical triage abnormal instances (first triage instances) in the multidimensional feature space can be quantified through probability density modeling.

[0077] Exemplarily, the first probability density distribution function P(x) can be calculated by the following formula, ,in: is the number of first triage instances, i.e., the total number of abnormal instances; Indicates The exception instance is Coordinates in the dimensional feature space; is the kernel bandwidth parameter, which controls the smoothness of the density estimation; is a kernel function (such as Gaussian kernel ; Any point in the feature space The probability density value at . The features of each first triage instance can be encoded: numerical features (such as blood pressure values) retain the original values ​​directly; categorical features (such as department types) are converted into binary vectors through One-Hot Encoding; text features (such as medical history descriptions) are mapped into low-dimensional dense vectors through word embedding (such as Word2Vec); and then standardized to eliminate dimensional differences. For example, for the j-th dimension feature x j , calculate its mean value as μ j , standard deviation is σ j ; The standardized value is The Gaussian kernel function is chosen to ensure the continuous differentiability of the probability density function. Bandwidth Through cross-validation optimization, the integrated mean square error (MISE) is minimized. dimensional space, the bandwidth matrix is ​​often simplified to a diagonal matrix ,in Initialization by Silverman criterion: ,in is the characteristic standard deviation. For any point in the feature space , its probability density It is obtained by weighted summation of kernel functions of all abnormal instances. For example, in three-dimensional space (department capability score ,body temperature , heart rate ), a certain point The density of is calculated as: ;in, For the The characteristic value of an abnormal instance.

[0078] S530, mapping the second number of triage instances and the number of triage instances to a multidimensional feature space, and generating a second probability density distribution function corresponding to the second number of triage instances.

[0079] It can be understood that the second probability density distribution function is used to quantify the distribution pattern of abnormal instances (second triage instances) that are highly correlated with the target patient's characteristics in the multidimensional feature space. The second triage instance is a subset of the first triage instance and must meet the department attribute matching conditions and the physical sign data correlation constraint conditions. The second probability density distribution function Q(x) focuses on historical abnormal instances that are highly correlated with the current target patient's characteristics, and its density peak reflects the "high-risk feature combination" that may cause triage conflicts. By comparing with the first probability density distribution function P(x) (such as calculating the KL divergence), the degree of deviation between the current patient's characteristics and the historical abnormal pattern can be accurately identified, providing a fine-grained risk assessment for triage decisions. For example, if Q(x) is significantly higher than P(x) in the "high medical history complexity + low department response time" area, it indicates that such feature combinations need to be avoided first.

[0080] Exemplarily, the mathematical form of the second probability density distribution function is similar to that of the first probability density distribution function. However, there are differences in data sources and parameter settings. The second probability density distribution function can be obtained by the following formula: ;in, is the number of second triage instances; Indicates feature vector of the second triage instance; is the kernel bandwidth of the second distribution, which is usually smaller than the bandwidth of the first distribution (Because the amount of data is smaller and the distribution is more concentrated).

[0081] S540, calculating the relative entropy ratio of the first probability density distribution function to the second probability density distribution function, and generating a feature sensitivity discriminant function.

[0082] It can be understood that the first probability density distribution function and the second probability density distribution function respectively represent the probability distribution of the characteristic variables under different conditions. For example, the first distribution may correspond to the distribution of physiological parameters (such as blood pressure and heart rate) of a patient group in a certain department, and the second distribution may correspond to the distribution of a patient group in another department. Relative entropy (Kullback-Leibler divergence) is an asymmetric indicator that measures the difference between two probability distributions, which is defined as the difference between the logarithmic expectations of the two distributions. The feature sensitivity discriminant function is a mathematical expression constructed based on this ratio, which is used to evaluate the discriminative ability of specific features in triage of different departments. For example, if the relative entropy ratio of a physiological feature deviates significantly from 1, it indicates that there is a significant distribution difference in the feature between patients in two types of departments, and it can be marked as a highly sensitive feature by the discriminant function.

[0083] Optionally, S540, calculating a relative entropy ratio between the first probability density distribution function and the second probability density distribution function to generate a feature sensitivity discriminant function includes: S541, determining a mixed distribution of the first probability density distribution and the second probability density distribution, and calculating a symmetric divergence and an asymmetric divergence between the first probability density distribution and the second probability density distribution according to the mixed distribution.

[0084] It can be understood that the mixed distribution is an intermediate distribution generated by weighted averaging of two original distributions, for example, it is defined as M(x)=1 / 2(P(x)+Q(x)), where P(x) and Q(x) are the first and second probability density distribution functions, respectively. Symmetric divergence usually refers to Jensen-Shannon divergence (JS divergence), which is calculated as JS(P||Q)=0.5KL(P||M)+0.5KL(Q||M), and the divergence is made symmetrical by introducing the mixed distribution M. Asymmetric divergence directly uses the original relative entropy KL(P||Q) or KL(Q||P).

[0085] The parameters of the mixed distribution M can be estimated based on the sample data of P and Q. For example, the expectation maximization algorithm can be used to optimize the weight coefficients for the Gaussian mixture model. Subsequently, the relative entropies between P and M and Q and M are calculated respectively, and the JS divergence value is obtained by weighted summation. At the same time, the original relative entropy of P to Q is independently calculated as the asymmetric divergence. These two divergence values ​​quantify the distribution differences from the symmetric and asymmetric perspectives, respectively, and provide a multi-dimensional measurement basis for the subsequent construction of the discriminant function.

[0086] S542, calculating the forward and reverse relative entropy differences between the first probability density distribution function and the second probability density distribution function, and dynamically generating a weight factor according to the relative entropy differences.

[0087] It can be understood that the positive relative entropy difference refers to the difference between KL(P||Q) and KL(Q||P), and the negative relative entropy difference is the difference between KL(Q||P) and KL(P||Q). For example, when KL(P||Q)=1.2 and KL(Q||P)=0.8, the positive difference is 0.4 and the negative difference is -0.4. Dynamic weight factor generation can be achieved by setting threshold conditions: if the positive difference exceeds the preset threshold (such as 0.5), the symmetric divergence is given a higher weight; if the negative difference is significant, the weight coefficient of the asymmetric divergence is increased.

[0088] For example, a sigmoid function can be used to map the difference value to the interval (0, 1) as a weight, for example, the weight α=1 / (1+exp(-k*(D f -D t ))), where k is the slope parameter, D f is the forward relative entropy difference, D tThis dynamic adjustment mechanism enables the discriminant function to adapt to the data distribution characteristics of different features. For example, for features with obvious directional differences (such as the significantly older age of patients in a certain department), the contribution of asymmetric divergence is strengthened, while for features with high distribution overlap, the contribution of symmetric divergence is relied on.

[0089] S543, weighting and fusing the weight factor with the symmetric divergence and the asymmetric divergence to generate a feature sensitivity discriminant function.

[0090] It can be understood that the symmetric divergence (JS) and asymmetric divergence (KL) of the weighted fusion process can be weighted by linear combination. For example, the discriminant function is set to F=α*JS(P||Q)+(1-α)*KL(P||Q), where α is the generated dynamic weight factor. When α approaches 1, the feature sensitivity discriminant function mainly reflects the symmetric characteristics of the distribution difference; when α approaches 0, it focuses on asymmetric differences. JS and KL can be normalized to avoid dimensional differences, for example, dividing them by their historical maximum values ​​or using z-score normalization. The final generated discriminant function value can be directly used as a feature sensitivity score for sorting or threshold screening.

[0091] S550, integrating the feature sensitivity discriminant function in the multidimensional feature space to obtain the triage conflict coefficient, and determining the guidance information of the target patient based on the triage conflict coefficient.

[0092] It can be understood that the triage conflict coefficient is an indicator obtained by quantifying the overall performance of the feature sensitivity discriminant function in the multidimensional feature space, which is used to measure the degree of conflict between the feature combination of the target patient in the triage decision of different departments. Each dimension in the multidimensional feature space corresponds to a patient feature (such as blood pressure, age, and medical history indicators), and the value of the feature sensitivity discriminant function represents the distinguishing ability of the feature in triage. By integrating the function value in the feature space, the triage uncertainty under the joint action of multiple features can be comprehensively evaluated. For example, if the discriminant function values ​​of multiple features of a patient (such as hypertension and hyperglycemia) in different departments are quite different, the integral result (triage conflict coefficient) is high, indicating that there is a large conflict in the triage decision. The generation of guidance information needs to combine the conflict coefficient with the department priority, and finally select the department with the least conflict as the recommendation target.

[0093] Optionally, S550, integrating the feature sensitivity discriminant function in the multidimensional feature space to obtain a triage conflict coefficient, and determining the guidance information of the target patient according to the triage conflict coefficient, including: S551, define the department feature distribution and integration domain range of each department in the multidimensional feature space; wherein the integration domain range is determined by the feature value distribution of historical triage instances.

[0094] It can be understood that the department characteristic distribution refers to the statistical distribution model of patient characteristics in each department in the historical triage instance. For example, the blood pressure and age of patients in the cardiology department may obey the Gaussian distribution with a specific mean and variance, while the vital capacity and cough frequency of patients in the respiratory department may conform to the Poisson distribution. The integral domain range is the characteristic value integral interval defined separately for each department, covering the characteristic value distribution range of the historical cases of the department. For each department, based on its historical case data, probability distribution fitting (such as Gaussian mixture model, kernel density estimation) or non-parametric statistical methods (such as histogram binning) can be used to construct the probability density function of multidimensional features. For example, the blood pressure feature of the cardiology department may be fitted as a normal distribution with a mean of 130 mmHg and a standard deviation of 15, while the cough frequency of the respiratory department may be modeled as a Poisson distribution with λ=3. For each characteristic dimension, the upper and lower limits of the integral are determined according to the statistical characteristics of the department's historical data (such as 5%-95% quantiles, 3σ principle). For example, if the 95% quantile of the age of cardiology patients is 65 years old, the integral domain can be set to [18,65] to exclude extreme outliers (such as centenarians). For discrete features (such as disease codes), the integral domain range is an enumeration of all possible values. The independence definition of department feature distribution avoids the dilution of department-specific data by the global integral domain. For example, the height range of pediatric patients is significantly different from that of orthopedics, and a separate integral domain can improve the accuracy of the integral. The integral domain range driven by historical data ensures that the calculation focuses on actual clinical scenarios and avoids invalid calculations for impossible feature combinations (such as neonatal prostate-specific antigen testing).

[0095] S552, performing multiple integral operations on the feature sensitivity discriminant function within the department integral domain according to the department feature distribution of each department, and generating a triage conflict coefficient between the target patient and each department.

[0096] It can be understood that the triage conflict coefficient is a scalar value calculated independently for each department, indicating the degree of conflict between the target patient characteristics and the historical case distribution of the department. For each department, within its exclusive integral domain, the feature sensitivity discriminant function Perform multiple integration, where Indicates the department number. is the patient characteristic. The triage conflict coefficient can be calculated by the following integral formula: ,in For the department No. The integration domain of the dimensional features. Randomly sample a large number of feature points in the department integration domain, calculate the mean of the discriminant function value and multiply it by the volume of the integration domain. Use polynomial approximation for continuous features and directly sum discrete features. For multi-department scenarios, a distributed computing framework (such as Spark) can be used to synchronously calculate the conflict coefficients of each department.

[0097] S553, according to the triage conflict coefficient between the target patient and each department, select the corresponding department with the lowest triage conflict coefficient as the target patient's target patient's guide information.

[0098] It can be understood that the triage conflict coefficient is a quantitative indicator of the degree of conflict between the target patient's feature combination and the historical case distribution of each department. The lower the value, the higher the match between the patient's features and the department, and the lower the risk of abnormal guidance. The triage conflict coefficients between the target patient and each department can be sorted in ascending order, and the department corresponding to the minimum value can be directly selected. If the conflict coefficients of multiple departments are the same (such as 0.3 for both the cardiology department and the neurology department), dynamic decisions can be made by reading and comparing the real-time load (number of waiting people) and resource availability (equipment idle rate) of the department from the department attribute data. For example, when the conflict coefficients are the same, departments with shorter waiting queues are recommended first to optimize resource allocation.

[0099] For example, assuming that the conflict coefficient of the target patient is mapped as {cardiology: 0.2, respiratory department: 0.5, orthopedics: 0.7}, the cardiology department can be directly output as the guidance information for the target patient; if it is mapped as {endocrinology department: 0.3, nephrology department: 0.3}, the real-time load of the department can be read and compared from the department attribute data. For example, if the current number of waiting patients in the endocrinology department is 10 and that in the nephrology department is 5, the nephrology department will be output as the guidance information for the target patient to balance the load. By combining the quantitative conflict intensity with dynamic rules, it is beneficial for the guidance results to take into account accuracy, efficiency and clinical safety, realize specialized and precise guidance based on real-time features and historical data, and improve the patient's medical experience.

[0100] Corresponding to the intelligent medical guidance method of the above embodiment, the embodiment of the present application also provides an intelligent medical guidance system, and each unit of the system can implement each step of the intelligent medical guidance method. Figure 3 A structural block diagram of the intelligent medical guidance system provided in an embodiment of the present application is shown. For the sake of convenience of explanation, only the parts related to the embodiment of the present application are shown.

[0101] Reference Figure 3 , the intelligent medical guidance system includes: An acquisition unit, used to acquire department attribute data and multimodal operation records of target patients; A fusion unit, configured to perform data feature fusion according to the multimodal operation record and the department attribute data, and determine a triage adaptation feature group for the target patient; A retrieval unit is used to obtain a historical triage database, retrieve the first triage instance with triage abnormality in the historical triage database, and obtain the number of first triage instances; wherein the historical triage database includes a plurality of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's department; A screening unit is used to screen second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is triage instances with triage abnormalities in which the department attribute data of the visiting department and the patient's physical sign data match the triage adaptation feature group to a preset matching threshold; A guidance unit is used to construct a triage conflict coefficient according to the number of the first triage instances and the number of the second triage instances, and determine the guidance information of the target patient based on the triage conflict coefficient.

[0102] It should be noted that the information interaction, execution process, etc. between the above-mentioned systems / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0103] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0104] The present application also provides an intelligent medical guidance robot. Figure 4 This is a schematic diagram of the structure of the intelligent medical guidance robot provided in one embodiment of the present application. Figure 4 As shown, the intelligent medical guidance robot 6 of this embodiment includes: at least one processor 60 ( Figure 4 Only one is shown), at least one memory 61 ( Figure 4 Only one is shown in the figure) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the intelligent medical guidance robot 6 implements the steps in any of the above-mentioned intelligent medical guidance method embodiments, or implements the functions of each unit in the above-mentioned system embodiments.

[0105] Exemplarily, the computer program 62 may be divided into one or more units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 62 in the intelligent medical guidance robot 6.

[0106] For example, the intelligent medical guidance robot can be a variety of types of robots. The intelligent medical guidance robot may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will appreciate that Figure 4 It is only an example of the intelligent medical guidance robot 6 and does not constitute a limitation of the intelligent medical guidance robot 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, buses, etc.

[0107] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0108] In some embodiments, the memory 61 may be an internal storage unit of the intelligent medical guidance robot 6, such as a hard disk or memory of the intelligent medical guidance robot 6. In other embodiments, the memory 61 may also be an external storage device of the intelligent medical guidance robot 6, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the intelligent medical guidance robot 6. Further, the memory 61 may also include both an internal storage unit and an external storage device of the intelligent medical guidance robot 6. The memory 61 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0109] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0110] An embodiment of the present application provides a computer program product. When the computer program product runs on an intelligent medical guidance robot, the intelligent medical guidance robot implements the steps in any of the above method embodiments.

[0111] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the intelligent guidance robot, a recording medium, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), an electric carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0112] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0113] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0114] In the embodiments provided in the present application, it should be understood that the disclosed intelligent medical guidance system / intelligent medical guidance robot and method can be implemented in other ways. For example, the intelligent medical guidance system / intelligent medical guidance robot embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0115] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent diagnosis guidance method, characterized in that: include: Obtain department attribute data and multimodal operation records of target patients; Performing data feature fusion according to the multimodal operation record and the department attribute data to determine a triage adaptation feature group for the target patient; Acquire a historical triage database, retrieve the first triage instance with triage abnormality in the historical triage database, and obtain the number of first triage instances; wherein the historical triage database includes a plurality of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's department; Screening second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is triage instances with triage anomalies whose matching degree between the department attribute data of the visiting department and the physical sign data of the visiting patient and the triage adaptation feature group reaches a preset matching threshold; A triage conflict coefficient is constructed according to the number of the first triage instances and the number of the second triage instances, and the guidance information of the target patient is determined based on the triage conflict coefficient.

2. The intelligent diagnosis guidance method according to claim 1, characterized in that: The step of fusing data features according to the multimodal operation record and the department attribute data to determine the triage adaptation feature group for the target patient includes: Performing time sequence alignment and semantic analysis on the multimodal operation records, and extracting operation behavior feature vectors and abnormal operation marks; Constructing a department professional capability map based on the department attribute data; wherein the department professional capability map includes a three-dimensional feature code of the success rate of historical case processing; Performing cross-modal correlation mapping between the operation behavior feature vector and the department professional capability map to generate a behavior-department correlation matrix; Weights are assigned to the abnormal operation mark and the three-dimensional feature code through a dynamic attention mechanism to calculate the feature coupling degree; According to the behavior-department association matrix and the feature coupling degree, the multimodal operation record features and the department attribute features are nonlinearly fused to generate the triage adaptation feature group of the target patient.

3. The intelligent diagnosis guidance method according to claim 2, characterized in that: By performing cross-modal correlation mapping between the operation behavior feature vector and the department professional capability map, a behavior-department correlation matrix is ​​generated, including: Expanding the dimension of the operation behavior feature vector, and projecting the operation behavior feature vector to the same feature space as the department professional capability map; Based on the mapping relationship in the preset symptom-department association rule library, the dynamic semantic similarity between each of the operation behavior feature vectors and the department professional capability map node is calculated; Performing segment-by-segment normalization processing on the dynamic semantic similarity to generate a standardized association weight; An N×M sparse association matrix is ​​constructed based on the standardized association weights, and the sparse association matrix is ​​subjected to time series noise reduction to generate a behavior-department association matrix; wherein N is the total number of dimensions of the operation behavior feature vector, and M is the total number of nodes in the department's professional capability map.

4. The intelligent diagnosis guidance method according to claim 1, characterized in that: The acquiring of the historical triage database, retrieving the first triage instance with triage abnormality in the historical triage database, and obtaining the number of the first triage instances, comprises: Access to historical triage databases; The historical triage database is searched according to a preset abnormality determination standard to obtain first triage instances in the historical triage database with triage abnormalities, and the number of the first triage instances is counted to obtain the number of first triage instances.

5. The intelligent diagnosis guidance method according to claim 1, characterized in that: The step of screening the second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances includes: According to the triage adaptation feature group of the target patient, construct department attribute data matching conditions and physical sign data correlation constraint conditions; Based on the department attribute data matching condition and the physical sign data association constraint condition, all the first triage instances are screened to obtain second triage instances, and the number of the second triage instances is counted to obtain the number of second triage instances.

6. The intelligent diagnosis guidance method according to claim 1, characterized in that: The constructing a triage conflict coefficient according to the first triage instance number and the second triage instance number, and determining the referral information of the target patient according to the triage conflict coefficient, includes: Obtaining the number of triage instances in the historical triage database; Mapping the first number of triage instances and the number of triage instances to a multidimensional feature space to generate a first probability density distribution function corresponding to the first number of triage instances; Mapping the second number of triage instances and the number of triage instances to a multidimensional feature space to generate a second probability density distribution function corresponding to the second number of triage instances; Calculating a relative entropy ratio of the first probability density distribution function to the second probability density distribution function to generate a feature sensitivity discriminant function; The feature sensitivity discriminant function is integrated in the multidimensional feature space to obtain a triage conflict coefficient, and the guidance information of the target patient is determined based on the triage conflict coefficient.

7. The intelligent diagnosis guidance method according to claim 6, characterized in that: The calculating the relative entropy ratio of the first probability density distribution function to the second probability density distribution function to generate a feature sensitivity discriminant function includes: Determine a mixed distribution of the first probability density distribution and the second probability density distribution, and calculate a symmetric divergence and an asymmetric divergence between the first probability density distribution and the second probability density distribution according to the mixed distribution; Calculating the forward and reverse relative entropy differences between the first probability density distribution function and the second probability density distribution function, and dynamically generating a weight factor according to the relative entropy difference; The weight factor is weighted and fused with the symmetric divergence and the asymmetric divergence to generate a feature sensitivity discriminant function.

8. The intelligent diagnosis guidance method according to claim 6, characterized in that: The step of integrating the feature sensitivity discriminant function in the multidimensional feature space to obtain a triage conflict coefficient, and determining the guidance information of the target patient according to the triage conflict coefficient includes: Define the department feature distribution and integral domain range of each department in the multidimensional feature space; wherein the integral domain range is determined by the feature value distribution of the historical triage instance; According to the department feature distribution of each department, the feature sensitivity discriminant function is multi-integrated within the department integral domain to generate a triage conflict coefficient between the target patient and each department; According to the triage conflict coefficient between the target patient and each department, the corresponding department with the lowest triage conflict coefficient is selected as the guidance information for the target patient.

9. An intelligent medical guidance system, characterized in that: include: An acquisition unit, used to acquire department attribute data and multimodal operation records of target patients; A fusion unit, configured to perform data feature fusion according to the multimodal operation record and the department attribute data, and determine a triage adaptation feature group for the target patient; A retrieval unit is used to obtain a historical triage database, retrieve the first triage instance with triage abnormality in the historical triage database, and obtain the number of first triage instances; wherein the historical triage database includes a plurality of historical triage instances, each of which records the patient's physical sign data and the department attribute data of the patient's department; A screening unit is used to screen second triage instances associated with the triage adaptation feature group from all the first triage instances to obtain the number of second triage instances; wherein the number of second triage instances is triage instances with triage abnormalities in which the department attribute data of the visiting department and the patient's physical sign data match the triage adaptation feature group to a preset matching threshold; A guidance unit is used to construct a triage conflict coefficient according to the number of the first triage instances and the number of the second triage instances, and determine the guidance information of the target patient based on the triage conflict coefficient.

10. An intelligent medical guidance robot, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images