Intelligent hospital guide method and system, intelligent terminal and storage medium

Abstract

The invention relates to an intelligent hospital guide method and system, an intelligent terminal and a storage medium, and relates to the field of intelligent medical treatment, and the method comprises the steps: collecting the case information of a patient, the case information comprising medical history information and symptom description; performing data processing on the case information by using an illness state analysis model to obtain illness state characteristics of the patient; potential diseases of the patient are retrieved in a disease database according to the illness state characteristics; determining a department and a hanging suggestion corresponding to the potential disease; screening out recommended doctors from a doctor qualification database according to the department; and pushing the patient hanging suggestion and registration information corresponding to the recommended doctor to a user terminal of the patient. The method has the effect of improving the accuracy and efficiency of medical triage.

Description

Technical Field

[0001] This application relates to the field of smart healthcare, and in particular to an intelligent triage method, system, intelligent terminal, and storage medium. Background Technology

[0002] In modern healthcare services, intelligent triage systems are a key technological support for improving medical efficiency and accurately matching medical resources.

[0003] Among related technologies, intelligent triage systems are based on large-scale language model technology. By intelligently analyzing patients' descriptions of their conditions, they automatically match historical case databases, recommend relevant departments and examinations, and finally guide patients to make payments.

[0004] Regarding the technologies mentioned above, there are issues with insufficient accuracy in triage recommendations and limited intelligence when matching departments, resulting in low patient registration efficiency and inaccurate department matching. Summary of the Invention

[0005] To improve the accuracy and efficiency of medical triage, this application provides an intelligent triage method, system, intelligent terminal, and storage medium.

[0006] Firstly, this application provides an intelligent triage method, which adopts the following technical solution: A smart triage method includes: Collect patient medical records, including medical history and symptom descriptions; Using a disease analysis model, the case information is processed to obtain the patient's disease characteristics; Based on the described disease characteristics, search the disease database for the patient's potential diseases; Determine the relevant department and recommended course of care for the potential disease; Recommended doctors will be selected from the physician qualification database based on the department mentioned. The suggested department and the registration information for the recommended doctor are pushed to the patient's user terminal.

[0007] By adopting the above technical solution, which analyzes patient medical records and combines them with a disease database to accurately match potential diseases and recommend corresponding departments, and intelligently selects suitable doctors from a physician qualification database and automatically pushes registration information, this solution significantly improves triage accuracy, optimizes the allocation of medical resources, reduces mis-registrations, and significantly improves medical efficiency and experience.

[0008] Optionally, medical entities and time features can be extracted from the case information using a named entity recognition model; The medical entities are mapped to their corresponding terms in a standard medical terminology database. The terms are structured and encoded using a clinical terminology coding system to obtain computer codes; The computer encoding and the time features are used to construct a symptom-time correlation data structure, which is used to characterize the correlation between symptom manifestation and time dimension; If the symptom-time association data structure corresponds to the emergency diagnosis type, then the department type is determined to be an emergency department; If the symptom-time association data structure corresponds to a non-emergency type, then the department type is obtained using the department determination method.

[0009] By adopting the above technical solution, and based on standardized terminology mapping and structured coding technology, combined with symptom-time dynamic correlation analysis, accurate digital representation of disease characteristics has been achieved. This solution significantly improves the scientific nature of triage decisions, especially the accuracy of emergency identification. Simultaneously, time-series feature analysis makes chronic disease triage more rational, optimizing the overall efficiency of medical resource allocation and reducing the risk of misdiagnosis.

[0010] Optionally, establish a symptom combination-department mapping knowledge base; If the symptom combination-department mapping knowledge base contains at least two symptom features that correspond to the term, then a joint department recommendation instruction is generated, which includes at least two candidate recommended departments; Real-time access to hospital department configuration database; Based on the priority ranking of the candidate recommended departments in the joint department recommendation instruction, the hospital department configuration database is checked sequentially to see if it is in a missing state; The candidate recommended departments that appear for the first time without being in a vacancy will be set as recommended departments.

[0011] By adopting the above technical solutions, and based on dynamic symptom combination matching and real-time departmental resource detection mechanisms, accurate triage of multiple overlapping cases and intelligent adaptation to the hospital's actual capacity to receive patients were achieved. This effectively solved the problem of difficulty in registering for departments for patients with complex symptoms. By prioritizing and real-time resource verification, the availability of recommended departments was ensured, significantly improving the triage success rate and medical resource utilization rate, while reducing the number of patients making duplicate registrations due to department closures.

[0012] Optionally, standardized symptom features can be extracted from the case information using natural language processing technology; Image features are extracted from the imaging examination data in the medical history information using computer vision algorithms; Establish a weighted correlation model between the standardized symptom features and the imaging features; A symptom-disease association map is constructed using image neural network technology. The nodes of the map include symptom nodes, disease nodes, and examination suggestion nodes. Match the symptom combinations and examination suggestion combinations corresponding to the weighted association model in the atlas; Obtain the edge weight combination between the symptom combination and the examination suggestion combination; Take the maximum value from the edge weight combination, activate the associated disease nodes and examination suggestion nodes in the graph, and obtain the disease characteristics.

[0013] By employing the aforementioned technical solution, and based on multimodal data fusion and knowledge graph reasoning technologies, intelligent correlation analysis between textual symptoms and imaging features is achieved. This enables the automatic identification of key disease characteristics and the generation of optimal examination recommendations. This solution overcomes the limitations of traditional single-data source analysis, significantly improving the early identification accuracy of complex diseases. Simultaneously, dynamic weight optimization ensures the scientific validity of diagnostic recommendations, providing more comprehensive and reliable intelligent support for clinical decision-making.

[0014] Optionally, by accessing the public health data stream in real time, a weight decay curve can be set. The weight decay curve is used to represent the relationship between the associated weight factor and the epidemic period. The associated weight factor is used to represent the degree of association between symptoms and the epidemic. The associated weight factors are dynamically updated based on the weight decay curve. The epidemic correlation degree is calculated based on the correlation weighting factor and the disease characteristics; When the epidemic correlation is greater than the baseline correlation, the potential disease is set as the epidemic.

[0015] By adopting the above technical solution and based on a dynamic epidemiological weight adjustment mechanism, real-time intelligent correlation analysis between symptom characteristics and epidemics was achieved. This significantly improved the early screening and warning capabilities for epidemics. By setting a weight decay curve to scientifically reflect the evolution of epidemics, it can quickly identify potential epidemic cases while ensuring the accuracy of routine disease triage, effectively enhancing the speed of public health emergency response and optimizing the efficiency of emergency allocation of medical resources.

[0016] Optionally, if at least two potential diseases exist, the potential diseases are defined as fuzzy diseases, and the disease matching degree corresponding to the fuzzy diseases is obtained; Set a disease matching confidence threshold; If the maximum value of the disease matching degree is higher than the threshold, then the potential disease corresponding to the maximum value is set as the potential disease; If the maximum value of the disease matching degree is lower than the threshold, a symptom-disease probability matrix is ​​constructed based on the disease characteristics and the fuzzy disease; the Top-N associated diseases are calculated based on the probability matrix; and the potential disease is determined from the Top-N associated diseases.

[0017] By adopting the above technical solution, and combining dynamic confidence threshold judgment with probability matrix analysis, intelligent triage decision-making for ambiguous diseases is achieved. This solution employs a dual judgment mechanism, ensuring rapid diagnosis of high-confidence diseases while providing Top-N accurate recommendations for low-matching cases through probability matrix calculation. This effectively solves the diagnostic ambiguity problem caused by complex symptoms, significantly improves the fault tolerance and diagnostic coverage of the triage system, and makes medical recommendations both efficient and robust.

[0018] Optionally, a multidimensional physician assessment model is constructed, which includes the following synergistic assessment dimensions; Based on the areas of expertise marked in the doctor qualification database and the characteristics of the patient's condition, the first candidate doctor is selected. The implicit areas of expertise of doctors are statistically analyzed from historical cases in the doctor qualification database, and candidate doctors corresponding to the disease characteristics are selected. The disease matching degree is calculated based on the historical medical records, and a second candidate doctor is selected from the candidate doctors based on the disease matching degree. Obtain the first satisfaction score and the first remaining number of appointment slots for the first candidate doctor; Obtain the second satisfaction score and the second remaining number of appointment slots for the second candidate doctor; The first recommendation score of the first candidate doctor is obtained by weighted calculation of the first satisfaction score and the first number of remaining appointment slots. The second satisfaction score and the second remaining number of appointment slots are weighted together to obtain the second recommendation score for the second candidate doctor. The recommended doctor is obtained based on the first recommendation score and the second recommendation score.

[0019] By adopting the above technical solution, and based on the construction of a multi-dimensional physician evaluation model, combined with explicit specialty labeling and implicit clinical data mining, dynamic intelligent matching of physician resources is achieved. This solution innovatively integrates multiple dimensions of factors, including professional field suitability, historical case matching rate, patient satisfaction, and real-time appointment availability. Through a weighted scoring mechanism, it scientifically balances medical quality and consultation efficiency, significantly improving the accuracy of physician recommendations. While ensuring professional diagnosis and treatment, it optimizes appointment utilization, effectively improving the patient experience and the rationality of medical resource allocation.

[0020] Secondly, this application provides an intelligent triage system, which adopts the following technical solution: An intelligent triage system includes: The acquisition module is used to acquire case information, disease analysis models, and disease databases. A memory for storing programs for intelligent triage methods as described above; The processor and the program in the memory can be loaded and executed by the processor to implement any of the intelligent triage methods mentioned above.

[0021] By adopting the above technical solution, which analyzes patient medical records and combines them with a disease database to accurately match potential diseases and recommend corresponding departments, and intelligently selects suitable doctors from a physician qualification database and automatically pushes registration information, this solution significantly improves triage accuracy, optimizes the allocation of medical resources, reduces mis-registrations, and significantly improves medical efficiency and experience.

[0022] Thirdly, this application provides a smart terminal, which adopts the following technical solution: A smart terminal includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed by any of the methods described above.

[0023] Fourthly, this application provides a computer storage medium capable of storing corresponding programs, which facilitates improving the accuracy and efficiency of medical triage, and adopts the following technical solution: A computer-readable storage medium storing a computer program that can be loaded by a processor and executed by any of the above-described intelligent triage methods.

[0024] In summary, this application includes at least one of the following beneficial technical effects: 1. By analyzing patient medical records and combining them with a disease database, the system accurately matches potential diseases and recommends corresponding departments. Simultaneously, it intelligently filters suitable doctors from a physician qualification database and automatically pushes registration information. This solution significantly improves triage accuracy, optimizes the allocation of medical resources, reduces mis-registrations, and substantially enhances medical efficiency and experience. 2. By employing multimodal data fusion and knowledge graph reasoning technologies, intelligent correlation analysis between textual symptoms and imaging features is achieved, enabling automatic identification of key disease characteristics and generation of optimal examination recommendations. This approach overcomes the limitations of traditional single-data source analysis, significantly improving the early identification accuracy of complex diseases. Simultaneously, dynamic weight optimization ensures the scientific validity of diagnostic recommendations, providing more comprehensive and reliable intelligent support for clinical decision-making.

[0025] 3. Through a dynamic epidemiological weight adjustment mechanism, real-time intelligent correlation analysis between symptom characteristics and epidemics is achieved. This significantly improves the early screening and warning capabilities for epidemics. By setting a weight decay curve to scientifically reflect the evolution of epidemics, it can quickly identify potential epidemic cases while ensuring the accuracy of routine disease triage, effectively enhancing the speed of public health emergency response and optimizing the efficiency of emergency allocation of medical resources. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating an intelligent triage method provided in an embodiment of this application.

[0027] Figure 2 This is a flowchart illustrating an emergency intelligent triage method provided in an embodiment of this application.

[0028] Figure 3 This is a flowchart illustrating a multi-symptom combined triage method provided in an embodiment of this application.

[0029] Figure 4 This is a flowchart illustrating a multimodal intelligent diagnostic method provided in an embodiment of this application.

[0030] Figure 5 This is a flowchart illustrating an epidemiological dynamic weighted triage method provided in an embodiment of this application.

[0031] Figure 6 This is a flowchart illustrating a fuzzy disease grading decision-making method provided in an embodiment of this application.

[0032] Figure 7 This is a flowchart illustrating a multi-dimensional intelligent doctor recommendation method provided in an embodiment of this application.

[0033] Figure 8 This is a schematic diagram of the structure of an intelligent triage system provided in an embodiment of this application. Detailed Implementation

[0034] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1 to 8 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0035] This application discloses an intelligent triage method. (Refer to...) Figure 1 The method includes: Step S101: Collect the patient's medical history information and symptom description.

[0036] Case information refers to disease-related data provided by the patient, including past medical history and current symptoms.

[0037] For example, this can be achieved through electronic forms filled out by patients or by retrieving historical information from the hospital information system.

[0038] Step S102: Use the disease analysis model to process the case information and obtain the patient's disease characteristics.

[0039] Disease analysis models refer to algorithmic models based on natural language processing (NLP) and machine learning, used to extract medical features of symptoms.

[0040] For example, entity recognition is performed on textual symptom descriptions, and non-standardized descriptions are mapped to a medical terminology database. Structured disease features can then be obtained through a disease analysis model.

[0041] Step S103: Search for the patient's potential diseases in the disease database based on the characteristics of the condition.

[0042] A disease database is a knowledge base containing information such as disease names, typical symptoms, and examination indicators. Disease feature retrieval refers to matching disease features with those in the database and calculating similarity.

[0043] For example, by inputting symptoms such as headache, nausea, and no fever, a disease database can be searched to match migraine or concussion.

[0044] Step S104: Determine the department corresponding to the potential disease and the recommended course of treatment.

[0045] For example, if the underlying disease is migraine, then the appropriate department should be neurology. The recommended department for initial consultation for migraine is neurology. If there are accompanying vision problems or acute symptoms, the patient can be referred to ophthalmology or the emergency department.

[0046] Step S105: Select recommended doctors from the doctor qualification database based on the department.

[0047] The doctor qualification database stores data such as doctors' areas of expertise, number of patients seen, and patient ratings.

[0048] For example, doctors corresponding to the disease are prioritized for matching, and recommendations are optimized by combining historical consultation data, while doctors without available appointments are excluded in real time.

[0049] Step S106: Push the suggested department and the corresponding registration information of the recommended doctor to the patient's user terminal.

[0050] For example, information such as suggested departments, recommended doctor's name, title, and available appointment time can be sent to the patient's mobile app or WeChat mini-program.

[0051] By adopting the above technical solution, which analyzes patient medical records and combines them with a disease database to accurately match potential diseases and recommend corresponding departments, and intelligently selects suitable doctors from a physician qualification database and automatically pushes registration information, this solution significantly improves triage accuracy, optimizes the allocation of medical resources, reduces mis-registrations, and significantly improves medical efficiency and experience.

[0052] This application discloses an intelligent triage method for emergency departments. (Refer to...) Figure 2 The method includes: Step S201: Extract medical entities and time features from case information using a named entity recognition model.

[0053] Named entity recognition models are deep learning-based natural language processing models used to identify medical-related entities in text.

[0054] Medical entities refer to independent elements with specific medical significance identified from medical texts, including key information such as symptoms, signs, disease names, and body parts.

[0055] Temporal characteristics refer to the time-related information of symptoms or diseases, including key parameters such as duration, onset time, and frequency.

[0056] For example, the input case information is persistent colic in the left chest for 4 hours, accompanied by cold sweats. The named entity recognition model extracts the symptoms as: chest pain, location: left chest, time: 4 hours, and accompanying symptoms: cold sweats.

[0057] Step S202: Map the medical entity to the corresponding term in the standard medical terminology database.

[0058] The standard medical terminology database refers to standardized terminology systems such as SNOMED CT and ICD-11, which can map medical entities to standardized medical terms.

[0059] For example, if the obtained medical entity is "heartache", then the mapping output is "chest pain".

[0060] Step S203: Use a clinical terminology coding system to perform structured coding of the terms to obtain computer codes.

[0061] Clinical terminology coding systems, such as LOINC and UMLS, assign unique machine-readable codes to terms.

[0062] Computer coding is the process of converting medical terms into machine-readable numbers or letters using a standardized coding system, which is then used by computer systems to process medical information.

[0063] For example, chest pain is coded as C0038031 in the UMLS terminology database, and the computer code obtained after conversion is 271807003.

[0064] Step S204: Construct a symptom-time correlation data structure by combining computer encoding with its temporal features. The symptom-time correlation data structure is used to characterize the relationship between symptom presentation and the time dimension.

[0065] The time dimension refers to the temporal manifestation of symptoms or disease characteristics, including temporal information such as duration, frequency of attacks, and time of onset.

[0066] For example, if a patient describes intermittent headaches for 3 days, each lasting 2 hours, the symptom-time association data structure can be represented as (symptom, number of days the symptom lasts, total duration of the symptom each day). For example, (headache, 3, 2), where "3" represents the number of days the symptom lasts and "2" represents the total duration of the symptom each day.

[0067] Step S205: If the symptom-time association data structure corresponds to the emergency diagnosis type, then the department type is determined as the emergency department.

[0068] For example, based on the symptom-time association data structure, the product of the number of days the symptom lasts and the total duration of symptom occurrence each day is calculated to obtain an initial classification score. The risk weight corresponding to the symptom is then obtained; this risk weight represents the degree of impact of the symptom on the user. The product of the initial classification score and the risk weight is calculated to obtain an emergency classification score. If the emergency classification score is greater than a preset score threshold, the symptom-time association data structure corresponds to the emergency category; if the emergency classification score is not greater than the preset score threshold, the symptom-time association data structure corresponds to the non-emergency category.

[0069] For example, the symptom-time association data structure corresponds to abdominal pain for two hours, which can recommend the emergency department.

[0070] Step S206: If the symptom-time association data structure corresponds to a non-emergency type, then use the department determination method to obtain the department type.

[0071] The method for determining the department can be found here. Figure 3 The embodiments shown are not described in detail here.

[0072] For example, the symptom-time association data structure corresponds to abdominal pain for 6 months, and the recommended department is non-emergency.

[0073] By adopting the above technical solution, and based on standardized terminology mapping and structured coding technology, combined with symptom-time dynamic correlation analysis, accurate digital representation of disease characteristics has been achieved. This solution significantly improves the scientific nature of triage decisions, especially the accuracy of emergency identification. Simultaneously, time-series feature analysis makes chronic disease triage more rational, optimizing the overall efficiency of medical resource allocation and reducing the risk of misdiagnosis.

[0074] This application discloses a multi-symptom combined triage method. (Refer to...) Figure 3 The method includes: Step S301: Establish a symptom combination-department mapping knowledge base.

[0075] The Symptom Combination-Department Mapping Knowledge Base is a database used to store the mapping relationship between symptom combinations and departments.

[0076] For example, if the combination of symptoms is chest pain and shortness of breath, it would be mapped to cardiology and pulmonology.

[0077] Step S302: If the symptom combination-department mapping knowledge base contains at least two symptom features that correspond to terms, then generate a joint department recommendation instruction, which includes at least two candidate recommended departments.

[0078] For example, if the input symptoms are abdominal pain and jaundice, the output candidate departments will be hepatobiliary surgery and gastroenterology.

[0079] Step 303: Real-time access to the hospital department configuration database.

[0080] The hospital department is configured with a database to store the data of the hospital department.

[0081] For example, data from departments can be obtained in real time through the API of a hospital's HIS system, with an update frequency of less than or equal to 1 minute.

[0082] Step 304: Based on the priority ranking of candidate recommended departments in the joint department recommendation instruction, check in turn whether the hospital department configuration database is in a missing state.

[0083] For example, if the candidate departments are hepatobiliary surgery (P1, 5 available appointments) and gastroenterology (P2, 0 available appointments), then hepatobiliary surgery is in a non-missing state, while gastroenterology is in a missing state.

[0084] Step 305: Set the candidate recommended departments that appear for the first time without being in a missing state as recommended departments.

[0085] For example, the priority order is hepatobiliary medicine, gastroenterology, and general internal medicine. Among them, hepatobiliary surgery (P1, 0 appointments), gastroenterology (P2, 2 appointments), and general internal medicine (P3, 4 appointments) are selected as the recommended departments.

[0086] By adopting the above technical solutions, and based on dynamic symptom combination matching and real-time departmental resource detection mechanisms, accurate triage of multiple overlapping cases and intelligent adaptation to the hospital's actual capacity to receive patients were achieved. This effectively solved the problem of difficulty in registering for departments for patients with complex symptoms. By prioritizing and real-time resource verification, the availability of recommended departments was ensured, significantly improving the triage success rate and medical resource utilization rate, while reducing the number of patients making duplicate registrations due to department closures.

[0087] This application discloses a multimodal intelligent diagnostic method. (Refer to...) Figure 4 The method includes: Step 401: Extract standardized symptom features based on case information using natural language processing technology.

[0088] Natural Language Processing (NLP) technology refers to the technology of understanding, analyzing, and generating human language through computer algorithms.

[0089] For example, the case information is intermittent pain in the upper left abdomen. The standardized symptom features extracted by natural language processing technology are: location is upper left abdomen, symptom is pain, and characteristic is intermittent.

[0090] Step 402: Extract image features from the imaging examination data in the medical history information using computer vision algorithms.

[0091] Computer vision algorithms refer to technologies that enable computers to automatically recognize and analyze image / video content through mathematical modeling and machine learning methods.

[0092] Imaging data includes, but is not limited to, at least one of X-ray images, CT images, magnetic resonance images, and ultrasound images.

[0093] For example, the imaging data in the medical history information includes a chest CT scan, and the feature extracted by the computer vision algorithm is ground-glass opacity.

[0094] Step 403: Establish a weighted association model between standardized symptom features and imaging features.

[0095] The weighted association model is built by automatically learning the weighted calculation results of symptom text features and image features through deep learning networks (such as attention mechanisms). This weighted calculation result can be used to represent the disease a user suffers from.

[0096] Step 404: Construct a symptom-disease association map using image neural network technology. The map nodes include symptom nodes, disease nodes, and examination suggestion nodes.

[0097] Image neural network technology refers to deep learning models (such as convolutional neural networks CNN) specifically designed for processing and analyzing medical image data. It automatically extracts multi-level features from images (such as lesion areas, tissue structures, etc.) and transforms them into numerical feature representations that can be processed by computers.

[0098] The symptom-disease association graph is constructed based on image features extracted by image neural networks and symptom features obtained by natural language processing. A multi-dimensional association network of three types of nodes, namely "symptom-disease-examination suggestion", is established through a deep learning model. The connection weights between nodes are obtained by training with massive clinical case data, and finally a medical knowledge graph with reasonable diagnostic paths is formed.

[0099] For example, there is a correlation strength of 0.8 between the symptom of "cough" and the disease of "pneumonia" (meaning that patients with coughs have a relatively high probability of having pneumonia), while there is a correlation strength of 0.9 between the disease of "pneumonia" and the recommendation of a "chest CT scan" (meaning that a chest CT scan is strongly recommended when diagnosing pneumonia). When a patient presents with cough symptoms, the system will consider the possibility of pneumonia based on a correlation weight of 0.8. If pneumonia is confirmed, a chest CT scan will be recommended for further diagnosis with a recommendation strength of 0.9.

[0100] Step 405: Match the symptom combinations and examination suggestion combinations in the atlas with the weighted association model.

[0101] For example, when inputting fever and moist rales in the lungs, the system will reason along the path of fever → pneumonia → moist rales in the lungs → chest CT scan, and finally output a diagnosis of pneumonia (confidence 0.88) and a recommendation for chest CT scan (recommendation 0.92).

[0102] Step 406: Obtain the edge weight combination between the symptom combination and the examination suggestion combination.

[0103] Edge weight combination is a combination of weights between symptoms and examination recommendations in the symptom-disease association graph.

[0104] For example, after analyzing a patient's symptoms of headache and blurred vision, the system identifies three associated pathways related to glaucoma and their weight values ​​(0.7, 0.8, and 0.85).

[0105] Step 407: Take the maximum value in the edge weight combination, activate the associated disease nodes and examination suggestion nodes in the graph, and obtain the disease characteristics.

[0106] For example, the intraocular pressure test recommendation corresponding to the highest weight of 0.85 in the previous step is selected, and a glaucoma diagnosis (75% confidence level) is given. Finally, it is recommended to prioritize the intraocular pressure test.

[0107] By employing the aforementioned technical solution, and based on multimodal data fusion and knowledge graph reasoning technologies, intelligent correlation analysis between textual symptoms and imaging features is achieved. This enables the automatic identification of key disease characteristics and the generation of optimal examination recommendations. This solution overcomes the limitations of traditional single-data source analysis, significantly improving the early identification accuracy of complex diseases. Simultaneously, dynamic weight optimization ensures the scientific validity of diagnostic recommendations, providing more comprehensive and reliable intelligent support for clinical decision-making.

[0108] This application discloses an epidemiological dynamic weighted triage method. (Refer to...) Figure 5 The method includes: Step 501: By accessing the public health data stream in real time, set up a weight decay curve. The weight decay curve is used to represent the relationship between the associated weight factor and the epidemic period. The associated weight factor is used to represent the degree of association between symptoms and the epidemic.

[0109] Public health data streams refer to collections of public health-related data that are updated in real time or near real time, including dynamic information such as infectious disease reports, emergency room visit trends, and drug sales monitoring.

[0110] For example, during the peak of the influenza epidemic (initial association weight w0=0.9), if a patient develops a fever, the system will assign an initial association weight of 0.9 to that symptom. As the epidemic enters the decline phase (the association weight decays by λ=0.01 per day for t=90 days), this weight decays, significantly reducing its contribution to influenza diagnosis.

[0111] Step 502: Dynamically update the associated weight factors based on the weight decay curve.

[0112] For example, during the decline of the influenza epidemic (60 days), the association weight between cough symptoms and influenza decreased from an initial 0.7 to approximately 0.38 using an exponential decay formula, reflecting the reduced diagnostic value of the symptom in the current epidemic phase.

[0113] Step 503: Calculate the epidemic correlation degree based on the correlation weighting factor and disease characteristics.

[0114] Epidemic correlation refers to the degree of matching between a patient's symptoms and a specific epidemic, quantified by an algorithm. The higher or lower the value, the greater the likelihood that the patient has contracted the epidemic.

[0115] For example, the epidemic correlation degree is obtained by weighting the disease characteristics using the correlation weight factor as the weight value.

[0116] For example, the weighted association between the patient's fever (weight 0.6) and muscle aches (weight 0.5) symptoms and influenza (fever had a match of 1.0 and muscle aches had a match of 0.8) was 1.0 (0.6 × 1.0 + 0.5 × 0.8).

[0117] Step 504: When the epidemic correlation is greater than the baseline correlation, the potential disease is set as an epidemic.

[0118] For example, when the epidemiological correlation of a patient’s symptom combination reaches 1.0 (exceeding the preset threshold of 0.8), the system automatically generates an intelligent triage suggestion of “suspected influenza, recommended to seek priority treatment in the infectious disease department”.

[0119] By adopting the above technical solution and based on a dynamic epidemiological weight adjustment mechanism, real-time intelligent correlation analysis between symptom characteristics and epidemics was achieved. This significantly improved the early screening and warning capabilities for epidemics. By setting a weight decay curve to scientifically reflect the evolution of epidemics, it can quickly identify potential epidemic cases while ensuring the accuracy of routine disease triage, effectively enhancing the speed of public health emergency response and optimizing the efficiency of emergency allocation of medical resources.

[0120] This application discloses a fuzzy disease grading decision-making method. (Refer to...) Figure 6 The method includes: Step 601: If there are at least two potential diseases, define the potential diseases as fuzzy diseases and obtain the disease matching degree corresponding to the fuzzy diseases.

[0121] Fuzzy disease refers to a situation where symptoms simultaneously match two or more diseases, but the matching degree of a single disease is not significantly dominant.

[0122] For example, when a patient's symptoms of chest pain and shortness of breath match both angina pectoris (disease match score of 0.72) and gastroesophageal reflux (disease match score of 0.68) and the difference in match score is less than 0.1, the system identifies the patient as a fuzzy disease case that requires further analysis.

[0123] Step 602: Set the disease matching confidence threshold.

[0124] The disease matching confidence threshold is the minimum matching score required for the system to determine a certain disease as a potential diagnosis. It is a dynamically adjustable critical value used to distinguish between confirmed diagnoses and ambiguous cases that require further evaluation.

[0125] It is feasible. If no disease matching confidence threshold is set, the system will be unable to distinguish between confirmed diagnoses and suspected cases, which may lead to low-match disease suggestions being misjudged as confirmed cases, increasing the risk of misdiagnosis.

[0126] Step 603: If the maximum value of the disease matching degree is higher than the threshold, then set the potential disease corresponding to the maximum value as a potential disease.

[0127] For example, when the pneumonia match score of 0.89 (exceeding the threshold of 0.8) is significantly higher than that of bronchitis (0.7), pneumonia is directly diagnosed without entering the fuzzy disease analysis process.

[0128] Step 604: If the maximum value of the disease matching degree is lower than the threshold, construct a symptom-disease probability matrix based on the disease characteristics and fuzzy diseases; calculate the Top-N associated diseases based on the probability matrix; identify potential diseases from the Top-N associated diseases.

[0129] The symptom-disease probability matrix refers to a quantitative model established by an algorithm to establish the relationship between symptom features and diseases, storing the contribution probability of each symptom to different diseases in matrix form.

[0130] Top-N related diseases refer to the N diseases (usually N=3) that are most closely matched with the current symptom combination extracted from the probability matrix and arranged in descending order of probability to form a graded diagnostic recommendation.

[0131] For example, when the disease match scores for a patient's chest pain and shortness of breath symptoms are both below the threshold, the system generates a graded diagnostic recommendation through probability matrix analysis: angina pectoris is the primary recommendation (disease match score of 0.72), and gastroesophageal reflux (disease match score of 0.68) and intercostal neuralgia (disease match score of 0.55) are listed as secondary alternatives.

[0132] By adopting the above technical solution, and combining dynamic confidence threshold judgment with probability matrix analysis, intelligent triage decision-making for ambiguous diseases is achieved. This solution employs a dual judgment mechanism, ensuring rapid diagnosis of high-confidence diseases while providing Top-N accurate recommendations for low-matching cases through probability matrix calculation. This effectively solves the diagnostic ambiguity problem caused by complex symptoms, significantly improves the fault tolerance and diagnostic coverage of the triage system, and makes medical recommendations both efficient and robust.

[0133] This application discloses a multi-dimensional intelligent doctor recommendation method. (Refer to...) Figure 7 The method includes: Step 701: Construct a multidimensional physician assessment model, which includes the following synergistic assessment dimensions.

[0134] The multidimensional doctor evaluation model refers to an intelligent comprehensive doctor evaluation system that integrates multidimensional data such as doctors' explicit qualifications (e.g., professional title, department), implicit expertise (e.g., historical case cure rate), and real-time status (e.g., appointment availability, satisfaction rate).

[0135] Optionally, without a multidimensional doctor evaluation model, recommendations will rely solely on basic tags (such as professional title and department), which may result in doctors with excellent actual clinical skills being overlooked (for example, a doctor may be an attending physician with the highest cure rate for a specific disease in the entire hospital, but may not be given priority due to a lower professional title). At the same time, it may not be able to dynamically respond to changes in appointment availability and patient satisfaction, thus reducing the accuracy of recommendations.

[0136] Step 702: Select the first candidate doctor based on the areas of expertise and disease characteristics marked in the doctor qualification database.

[0137] Specialty areas refer to the professional scope of diagnosis and treatment that a doctor has clearly marked in the qualification database (such as the diagnosis and treatment of arrhythmia). Specialty areas can be clinical directions recognized by department division, professional qualification certification or hospital official recognition.

[0138] For example, when a patient’s symptoms are palpitations and chest pain, Dr. Wang, whose specialty is the diagnosis and treatment of arrhythmias, is selected from the qualification database as the first candidate doctor.

[0139] Step 703: Analyze the implicit expertise areas of historical cases in the doctor qualification database and select candidate doctors that correspond to the characteristics of the patients' conditions.

[0140] Based on historical patient cases in the physician qualification database, identify physicians' implicit patient cases that do not match the aforementioned areas of expertise. Calculate the cure rate for these implicit patient cases. If the cure rate for an implicit patient case is greater than a threshold, then the medical field corresponding to that case is designated as an implicit area of ​​expertise.

[0141] For example, by analyzing 150 cases of abdominal pain and rectal bleeding treated by a gastroenterologist over the past 3 years (with a cure rate of 95%), the system discovers that the doctor's hidden expertise is in the diagnosis and treatment of inflammatory bowel disease, and automatically lists the doctor as a candidate doctor when patients have similar symptoms.

[0142] Step 704: Calculate the disease matching degree based on historical cases, and select a second candidate doctor from the candidate doctors based on the disease matching degree.

[0143] Disease matching rate is a quantitative indicator of the proportion of target diseases in a doctor's historical cases (calculated as: (number of target disease cases treated by the doctor / total number of cases treated) × 100%), used to assess the doctor's clinical experience with a specific disease.

[0144] For example, if a doctor has seen 300 cases in his / her history, and 180 of them are diabetic foot (60% match), the system will list him / her as the second candidate doctor, prioritizing other candidate doctors with a match of only 40%.

[0145] Step 705: Obtain the first satisfaction score and the first number of remaining appointment slots for the first candidate doctor.

[0146] The first satisfaction score is a quantitative indicator of physician service quality calculated based on patients' historical evaluation data.

[0147] The first remaining number of appointment slots refers to the number of real-time appointment slots available for that doctor in the hospital's registration system.

[0148] For example, the current satisfaction score of the first candidate, Dr. Zhang (Chief Physician of Cardiology), is 4.8 (based on nearly 100 patient reviews), and the real-time remaining appointment slots are 2 (from the hospital's registration system data).

[0149] Step 706: Obtain the second satisfaction score and the second remaining number of appointment slots for the second candidate doctor.

[0150] The second satisfaction score refers to the average patient evaluation of the second candidate doctor in recent times (e.g., within the last 3 months) obtained from the hospital information system (HIS) (e.g., Dr. Li's 4.6 / 5 score), used to quantify the quality of their service.

[0151] The second remaining number of appointment slots refers to the number of appointments that can be booked by the same doctor in real time on the same day (for example, Dr. Li has 2 remaining appointment slots), reflecting the accessibility of his services.

[0152] For example, real-time data of the second candidate, Dr. Li, is obtained from the HIS database, including a patient satisfaction score of 4.6 out of 5 over the past 3 months and 2 remaining appointment slots for the day, which are used for subsequent weighted calculation of the recommendation score.

[0153] Step 707: Calculate the first satisfaction score and the first remaining number of appointment slots using a weighted average to obtain the first recommendation score for the first candidate doctor.

[0154] For example, if the patient satisfaction score of the first candidate doctor is 4.8 (weight 60%) and there are 3 remaining appointment slots (with a weight of 0.45 calculated by the conversion function), the recommended score is 4.8 × 0.6 + 0.45 × 0.4 = 3.06, and the system generates a priority recommendation level accordingly.

[0155] Step 708: Calculate the second satisfaction score and the second remaining number of appointment slots using a weighted average to obtain the second recommendation score for the second candidate doctor.

[0156] For example, when the second candidate doctor has a satisfaction score of 4.5 (weight 60%) and there are 5 remaining appointment slots (weight 40%), the system calculates a recommendation score of 3.55 (4.5×0.6 + (1-e^(-5 / 5))×0.4×5≈3.55) using a weighted formula, and recommends this doctor first.

[0157] Step 709: Obtain the recommended doctor based on the first recommendation score and the second recommendation score.

[0158] The primary recommendation score is a comprehensive score calculated based on the match between the doctor's explicit qualifications (such as professional title and department) and the characteristics of the patient's condition.

[0159] The second recommendation score is a dynamically generated score based on the doctor's implicit expertise (such as historical case cure rate) and real-time patient reception ability (such as availability of appointments and patient satisfaction).

[0160] For example, when the scores of the first candidate doctor (Director Zhang of Cardiology, recommendation score 4.2) and the second candidate doctor (Dr. Li of Respiratory Medicine, recommendation score 4.5) are both higher than the benchmark, the system will prioritize Dr. Li, who has a higher score, as the final recommendation and mark it as "better match for respiratory symptoms and sufficient appointments available today".

[0161] By adopting the above technical solution, and based on the construction of a multi-dimensional physician evaluation model, combined with explicit specialty labeling and implicit clinical data mining, dynamic intelligent matching of physician resources is achieved. This solution innovatively integrates multiple dimensions of factors, including professional field suitability, historical case matching rate, patient satisfaction, and real-time appointment availability. Through a weighted scoring mechanism, it scientifically balances medical quality and consultation efficiency, significantly improving the accuracy of physician recommendations. While ensuring professional diagnosis and treatment, it optimizes appointment utilization, effectively improving the patient experience and the rationality of medical resource allocation.

[0162] Based on the same inventive concept, this application provides an intelligent triage system. Please refer to [link / reference]. Figure 8 The system includes: The acquisition module 801 is used to acquire case information, disease analysis models, and disease databases. Memory 802 is used to store a program for an intelligent triage method as described above; The processor 803 can load and execute programs in memory to implement any of the intelligent triage methods described above.

[0163] By adopting the above technical solution, which analyzes patient medical records and combines them with a disease database to accurately match potential diseases and recommend corresponding departments, and intelligently selects suitable doctors from a physician qualification database and automatically pushes registration information, this solution significantly improves triage accuracy, optimizes the allocation of medical resources, reduces mis-registrations, and significantly improves medical efficiency and experience.

[0164] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0165] This invention provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as an intelligent triage method.

[0166] Computer storage media include, for example, USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media that can store program code.

[0167] Based on the same inventive concept, embodiments of the present invention provide a smart terminal, including a memory and a processor, wherein the memory stores a computer program that can be loaded and executed by the processor to provide a smart triage method.

[0168] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. An intelligent triage method, characterized in that, include: Collect patient medical records, including medical history and symptom descriptions; Using a disease analysis model, the case information is processed to obtain the patient's disease characteristics; Based on the described disease characteristics, search the disease database for the patient's potential diseases; Determine the relevant department and recommended course of care for the potential disease; Recommended doctors will be selected from the physician qualification database based on the department mentioned. The suggested department and the registration information for the recommended doctor are pushed to the patient's user terminal.

2. The intelligent triage method according to claim 1, characterized in that, The method further includes: Medical entities and temporal features are extracted from the case information using a named entity recognition model. The medical entities are mapped to their corresponding terms in a standard medical terminology database. The terms are structured and encoded using a clinical terminology coding system to obtain computer codes; The computer encoding and the time features are used to construct a symptom-time correlation data structure, which is used to characterize the correlation between symptom manifestation and time dimension; If the symptom-time association data structure corresponds to the emergency diagnosis type, then the department type is determined to be an emergency department; If the symptom-time association data structure corresponds to a non-emergency type, then the department type is obtained using the department determination method.

3. The intelligent triage method according to claim 2, characterized in that, The methods for determining the department include: Establish a symptom combination-department mapping knowledge base; If the symptom combination-department mapping knowledge base contains at least two symptom features that correspond to the term, then a joint department recommendation instruction is generated, which includes at least two candidate recommended departments; Real-time access to hospital department configuration database; Based on the priority ranking of the candidate recommended departments in the joint department recommendation instruction, the hospital department configuration database is checked sequentially to see if it is in a missing state; The candidate recommended departments that appear for the first time without being in a vacancy will be set as recommended departments.

4. The intelligent triage method according to claim 1, characterized in that, The process of using a disease analysis model to process the case information and obtain the patient's disease characteristics includes: Standardized symptom features were extracted from the case information using natural language processing technology. Image features are extracted from the imaging examination data in the medical history information using computer vision algorithms; Establish a weighted correlation model between the standardized symptom features and the imaging features; A symptom-disease association map is constructed using image neural network technology. The nodes of the map include symptom nodes, disease nodes, and examination suggestion nodes. Match the symptom combinations and examination suggestion combinations corresponding to the weighted association model in the atlas; Obtain the edge weight combination between the symptom combination and the examination suggestion combination; Take the maximum value from the edge weight combination, activate the associated disease nodes and examination suggestion nodes in the graph, and obtain the disease characteristics.

5. The intelligent triage method according to claim 1, characterized in that, The disease database retrieval includes: By accessing public health data streams in real time, a weight decay curve is set up. The weight decay curve is used to represent the relationship between the associated weight factor and the epidemic period. The associated weight factor is used to represent the degree of association between symptoms and the epidemic. The associated weight factors are dynamically updated based on the weight decay curve. The epidemic correlation degree is calculated based on the correlation weighting factor and the disease characteristics; When the epidemic correlation is greater than the baseline correlation, the potential disease is set as the epidemic.

6. The intelligent triage method according to claim 1, characterized in that, The step of retrieving the patient's potential diseases from the disease database based on the described disease characteristics includes: If at least two potential diseases exist, the potential diseases are defined as fuzzy diseases, and the disease matching degree corresponding to the fuzzy diseases is obtained; Set a disease matching confidence threshold; If the maximum value of the disease matching degree is higher than the threshold, then the potential disease corresponding to the maximum value is set as the potential disease; If the maximum value of the disease matching degree is lower than the threshold, a symptom-disease probability matrix is ​​constructed based on the disease characteristics and the fuzzy disease; the Top-N associated diseases are calculated based on the probability matrix; and the potential disease is determined from the Top-N associated diseases.

7. The intelligent triage method according to claim 1, characterized in that, The process of selecting recommended doctors from the doctor qualification database based on the department includes: A multidimensional physician assessment model is constructed, which includes the following synergistic assessment dimensions; Based on the areas of expertise marked in the doctor qualification database and the characteristics of the patient's condition, the first candidate doctor is selected. The implicit areas of expertise of doctors are statistically analyzed from historical cases in the doctor qualification database, and candidate doctors corresponding to the disease characteristics are selected. The disease matching degree is calculated based on the historical medical records, and a second candidate doctor is selected from the candidate doctors based on the disease matching degree. Obtain the first satisfaction score and the first remaining number of appointment slots for the first candidate doctor; Obtain the second satisfaction score and the second remaining number of appointment slots for the second candidate doctor; The first recommendation score of the first candidate doctor is obtained by weighted calculation of the first satisfaction score and the first number of remaining appointment slots. The second satisfaction score and the second remaining number of appointment slots are weighted together to obtain the second recommendation score for the second candidate doctor. The recommended doctor is obtained based on the first recommendation score and the second recommendation score.

8. An intelligent triage system, characterized in that, include: The acquisition module is used to acquire case information, disease analysis models, and disease databases. A memory for storing the program of the intelligent triage method as described in any one of claims 1 to 7; The processor and the program in the memory can be loaded and executed by the processor to implement the intelligent triage method as described in any one of claims 1 to 7.

9. A smart terminal, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed according to any one of claims 1 to 7.

Images