Intelligent triage method and system

Abstract

The present invention relates to the field of medical technology, and specifically discloses an intelligent triage method and system, wherein the method comprises: S1, obtaining the patient's condition information and identifying key information in the condition information; S2, constructing a dynamic condition model according to the patient's historical data, medical knowledge and key information; S3, reasoning the key information according to a large model and a knowledge base, generating multiple condition hypotheses of the patient and the credibility of each condition hypothesis; the large model is used to identify and predict the possibility of disease progression; the knowledge base is used to provide relevant medical rules and facts; S4, according to the condition model and the condition hypothesis, using the large model to analyze the patient's urgency and triage needs, and generating one or more triage plans; S5, collating the patient's triage report according to the triage plan; the triage report includes at least multiple items of the following information: the patient's basic information, condition description, triage suggestion, urgency, treatment suggestion, required medical resources and precautions.

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to an intelligent triage method and system. Background Technology

[0002] The purpose of triage is to improve the efficiency and quality of medical care, reduce patient waiting time and mortality risk, and optimize resource utilization. However, traditional triage methods often rely on human experience and judgment, which leads to problems such as human resource shortages, subjectivity, inconsistency, errors, and delays.

[0003] Therefore, how to utilize artificial intelligence technology to combine large models and knowledge bases to automatically generate reasonable, effective, and interpretable triage strategies is a research question of great significance and challenge. Summary of the Invention

[0004] To address the aforementioned problems, the purpose of this invention is to provide an intelligent triage method and system that, by combining a large model and a knowledge base, automatically generates a reasonable and effective triage plan based on the patient's condition information, thereby reducing patient waiting time and improving medical efficiency.

[0005] The intelligent triage method provided by this invention includes:

[0006] S1, Obtain the patient's medical information and identify key information in the medical information;

[0007] S2, Based on the patient's historical data, medical knowledge, and key information, construct a dynamic disease model;

[0008] S3, reasoning based on the key information using the large model and knowledge base to generate multiple hypotheses about the patient's condition and the credibility of each hypothesis; the large model is used to identify and predict the likelihood of disease progression; the knowledge base is used to provide relevant medical rules and facts;

[0009] S4. Based on the disease model and the disease assumptions, use the large model to analyze the patient's urgency and triage needs, and generate one or more triage plans.

[0010] S5. Based on the triage plan, a triage report for the patient is compiled; the triage report includes at least several of the following: the patient's basic information, description of the condition, triage suggestions, urgency level, treatment suggestions, required medical resources, and precautions.

[0011] In one possible implementation, S1 includes:

[0012] The medical information was analyzed using natural language processing technology to obtain the first information;

[0013] The second information was obtained by using a BERT pre-trained language model to identify the patient's semantics and context.

[0014] Keywords are extracted from the described medical condition information to obtain third-party information;

[0015] The key information is obtained based on the first information and / or the second information and / or the third information; the key information includes at least: symptoms, duration, and severity.

[0016] In one possible implementation, S1 further includes:

[0017] The patient's speech-based medical information was converted into text-based medical information using a deep learning algorithm called a convolutional neural network (CNN).

[0018] In one possible implementation, the disease model includes at least the patient's basic information, symptoms, signs, past medical history, and family medical history; S2 includes:

[0019] By combining the key information and historical data through the disease model, a disease view is obtained;

[0020] The patient's current health status is assessed using the disease model; the health status includes the severity of symptoms and possible diseases.

[0021] The disease model identifies potential health risks and situations requiring urgent intervention in the patient's current health status.

[0022] The disease model is updated in real time based on new disease information.

[0023] In one possible implementation, S3 includes:

[0024] The knowledge base was constructed using ontology modeling tools.

[0025] The key information is analyzed using the large model, and possible hypotheses about the disease are extracted from the knowledge base;

[0026] The credibility of each disease hypothesis is evaluated using an evidence-based reasoning algorithm.

[0027] In one possible implementation, S4 includes:

[0028] An objective function is constructed to maximize the cure rate and minimize the waiting time for patients; the constraints of the objective function are the availability of medical resources and patient priority.

[0029] The optimal or suboptimal triage scheme is found in the feasible solution space of the objective function using a heuristic search algorithm, and the triage level is determined by combining relevant knowledge in the medical knowledge graph.

[0030] In one possible implementation, the use of a heuristic search algorithm to find the optimal or suboptimal triage scheme among the feasible solutions of the objective function includes:

[0031] Multiple feasible solutions are randomly generated based on the objective function to form an initial population;

[0032] The selection probability is calculated based on the fitness of each feasible solution, and multiple feasible solutions are selected from the initial population based on the selection probability to generate a temporary population.

[0033] The first and second chromosomes are randomly selected as parents from the temporary population;

[0034] Randomly select a crossover point, and exchange the genes of the first chromosome and the second chromosome after the crossover point to generate a first new chromosome and a second new chromosome;

[0035] With a mutation rate as a probability, a gene is randomly selected on the first and second new chromosomes to perform a mutation operation, resulting in a first mutated chromosome and a second mutated chromosome; the mutation operation includes any of the following: randomly changing the gene value and exchanging the gene position.

[0036] The first mutated chromosome and the second mutated chromosome are added to the new generation population;

[0037] When the maximum number of iterations is reached or the rate of change of fitness is less than a preset value, the optimal solution, i.e., the triage scheme, is output; otherwise, the new generation population is used as the initial population.

[0038] In one possible implementation, calculating the selection probability based on the fitness of each individual includes:

[0039] Based on the fitness function F(c) Calculate the probability of selection Pselect(c) The formula is as follows:

[0040] ;

[0041] in, N For the number of triage plans, c The chromosomes encoded for each triage protocol, c j The representative of the triage plan j Each decision-making gene, for each chromosome. c Calculate fitness F(c) , j It is a positive integer;

[0042] fitness function F(c) The following formula:

[0043] ;

[0044] in, Oi(c) It is the first i An evaluation function for each optimization objective. wi It is the first i The weights of each optimization objective, i It is a positive integer.

[0045] In one possible implementation, S5 includes:

[0046] A structured triage report is generated based on the template engine, natural language generation technology, and the triage scheme described above;

[0047] The triage report is reviewed through manual and / or machine review.

[0048] The present invention also provides an intelligent triage system for implementing any of the intelligent triage methods described above, comprising:

[0049] The information acquisition module is used to acquire the patient's medical condition information and identify key information in the medical condition information;

[0050] The model building module is used to build a dynamic disease model based on the patient's historical data, medical knowledge, and key information.

[0051] The information reasoning module is used to reason about the key information based on a large model and a knowledge base, generating multiple hypotheses about the patient's condition and the credibility of each hypothesis; the large model is used to identify and predict the probability of disease progression; the knowledge base is used to provide relevant medical rules and facts.

[0052] The disease analysis module is used to analyze the patient's urgency and triage needs using a large model based on the disease model and the disease assumptions, and generate one or more triage plans.

[0053] The report generation module is used to compile a triage report for the patient based on the triage plan; the triage report includes at least several of the following information: the patient's basic information, description of the condition, triage suggestions, urgency level, treatment suggestions, required medical resources, and precautions.

[0054] The purpose of this invention is to provide an intelligent triage method and system that constructs a disease model based on each patient's condition information. By combining a large model with a knowledge base, it automatically generates a reasonable and effective triage plan based on the patient's condition information, reducing patient waiting time and improving medical efficiency. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the intelligent triage method provided in an embodiment of the present invention. Detailed Implementation

[0056] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following detailed description of the embodiments and the accompanying drawings are used to illustrate the principles of the present invention by way of example, but should not be used to limit the scope of the present invention. That is, the present invention is not limited to the described preferred embodiments, and the scope of the present invention is defined by the claims.

[0057] In the description of this invention, it should be noted that, unless otherwise stated, "a plurality of" means two or more; the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance; those skilled in the art can understand the specific meaning of the above terms in this invention as appropriate.

[0058] Figure 1 This is a flowchart illustrating the intelligent triage method provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the intelligent triage method provided by the present invention includes:

[0059] S1, obtain the patient's medical information and identify key information within the medical information;

[0060] In one example, patient information includes basic information such as age, gender, and past medical history; chief complaint such as fever and chest tightness; physical signs such as blood pressure and heart rate; and auxiliary examinations such as complete blood count and electrocardiogram.

[0061] In one possible implementation, S1 includes: parsing medical condition information using natural language processing technology to obtain first information, such as medical terminology and symptoms; using a BERT pre-trained language model to identify the patient's semantics and context to obtain second information, such as the severity and urgency of symptoms; extracting keywords from the medical condition information to obtain third information, such as symptoms, duration, and past medical history; and obtaining key information based on the first and / or second and / or third information; the key information includes at least: symptoms, duration, and severity.

[0062] In one example, after using a deep learning algorithm convolutional neural network for speech-to-text conversion, the output of the RNN or CNN is reordered using an N-gram language model or a deep language model to improve recognition accuracy. The performance of the model is evaluated on an independent test set, and metrics such as word error rate are used to measure recognition accuracy.

[0063] In one possible implementation, since patients or callers may communicate via various means such as telephone, gestures, or online text, when the medical information is in voice format, it needs to be converted into text format first; the same applies to gesture information. This conversion ensures the accuracy of the medical information and avoids misunderstandings. S1 also includes: using a deep learning algorithm, a convolutional neural network (CNN), to convert the patient's voice-formatted medical information into text-formatted medical information.

[0064] S2, based on the patient's historical data, medical knowledge, and key information, constructs a dynamic disease model;

[0065] In one possible implementation, S2 includes: combining key information and historical data through a disease condition model to obtain a disease condition view; the disease condition view helps to understand the patient's health status. The disease condition model assesses the patient's current health status; the health status includes the severity of symptoms and possible diseases; the disease condition model identifies potential health risks and situations requiring urgent intervention in the patient's current health status; and the disease condition model is updated in real time based on new disease condition information.

[0066] The disease model includes at least the patient's basic information, symptoms, signs, past medical history, and family medical history; the disease model can identify potential health risks and their urgency, providing support for triage programs.

[0067] Therefore, by analyzing patient data, disease models can assist doctors in quickly identifying potential diseases and health risks. They can also recommend personalized treatment plans based on each patient's specific situation, providing a basis for preventative medicine. By prioritizing high-risk patients, disease models help optimize the allocation of medical resources, automate the disease assessment process, reduce doctors' workload, and improve the efficiency of medical services. By building patient disease models, medical professionals can more accurately assess patient conditions, make more informed clinical decisions, and provide more personalized medical services. Simultaneously, this also contributes to improving the overall quality and efficiency of medical services.

[0068] S3, based on the large model and knowledge base, reason about key information to generate multiple hypotheses about the patient's condition and the credibility of each hypothesis;

[0069] Among them, large models are used to identify and predict the likelihood of disease progression. Large models refer to machine learning models with large-scale parameters and complex computational structures, such as ChatGPT; knowledge bases are used to provide relevant medical rules and facts, such as medical book databases and case databases.

[0070] In one possible implementation, S3 includes constructing a knowledge base using ontology modeling tools, such as OWL or RDF formats; analyzing key information using a large model to extract possible disease hypotheses from the knowledge base; and evaluating the credibility of each disease hypothesis using an evidence-based reasoning algorithm.

[0071] S4. Based on the disease model and disease assumptions, use a large model to analyze the patient's urgency and triage needs, and generate one or more triage plans. The triage plan needs to consider a variety of factors, such as the severity of the disease, the availability of medical resources, and the time it takes for the patient to arrive at the hospital.

[0072] In one possible implementation, S4 includes: constructing an objective function to maximize the patient's cure rate and minimize the patient's waiting time; the objective function is constrained by the availability of medical resources and the patient's priority; using a heuristic search algorithm, such as a multi-objective genetic algorithm or a multi-objective particle swarm optimization algorithm, to find the optimal or suboptimal triage scheme in the feasible solution space of the objective function, and combining relevant knowledge in the medical knowledge graph to determine the triage level.

[0073] Among them, heuristic search algorithms are algorithms based on intuition or experience, providing a feasible solution for each instance of the combinatorial optimization problem under acceptable cost; the deviation of this feasible solution from the optimal solution is generally not estimable. Multi-objective genetic algorithms improve upon genetic algorithms by stratifying individuals according to their dominance and non-dominance relationships before performing selection operations. Multi-objective particle swarm optimization algorithms obtain a series of optimal solutions by ensuring that ions maintain a set of non-dominated solutions during the process, thus meeting the needs of different fields.

[0074] In one possible implementation, using heuristic search algorithms to find the optimal or suboptimal triage scheme among feasible solutions to the objective function includes:

[0075] Multiple feasible solutions are randomly generated based on the objective function to form the initial population;

[0076] The selection probability is calculated based on the fitness of each feasible solution, and multiple feasible solutions are selected from the initial population based on the selection probability to generate a temporary population.

[0077] The first and second chromosomes are randomly selected as parents from the temporary population;

[0078] Randomly select a crossover point, and exchange the genes of the first chromosome and the second chromosome after the crossover point to generate a first new chromosome and a second new chromosome;

[0079] Using the mutation rate as a probability, a gene is randomly selected on the first and second new chromosomes to perform mutation operations, resulting in the first mutated chromosome and the second mutated chromosome; the mutation operations include any of the following: randomly changing the gene value and exchanging the gene position.

[0080] Add the first and second variant chromosomes to the new generation population;

[0081] When the maximum number of iterations is reached or the rate of change of fitness is less than the preset value, the optimal solution, i.e., the triage scheme, is output; otherwise, the new generation of population is used as the initial population.

[0082] In one possible implementation, calculating the selection probability based on each individual's fitness includes:

[0083] Based on the fitness function F(c) Calculate the probability of selection Pselect(c) The formula is as follows:

[0084] ;

[0085] in, N For the number of triage plans, c The chromosomes encoded for each triage protocol, c j The representative of the triage plan j Each decision-making gene, for each chromosome. c Calculate fitness F(c) , j It is a positive integer;

[0086] fitness function F(c) The following formula:

[0087] ;

[0088] in, Oi(c) It is the first i An evaluation function for each optimization objective. wi It is the first i The weights of each optimization objective are used to balance the importance of different objectives. i It is a positive integer.

[0089] The purpose of using large-scale models for triage level decisions is to improve the efficiency and accuracy of the triage process through automation and intelligent means, ensuring that patients receive appropriate medical services in a timely manner based on the severity of their condition. By optimizing the allocation of medical resources, patient wait times are reduced, the overall quality of medical services is improved, and strong decision support is provided to healthcare professionals, enhancing the transparency and explainability of triage plans. Furthermore, this invention also emphasizes fairness, reducing human bias, improving patient safety, and enabling rapid and effective responses in emergency situations, thereby comprehensively enhancing the responsiveness and service quality of the healthcare system.

[0090] S5. Compile the patient's triage report according to the triage plan; the triage report shall include at least several of the following information: the patient's basic information, description of the condition, triage suggestions, urgency level, treatment suggestions, required medical resources and precautions.

[0091] In one possible implementation, S5 includes: generating a structured triage report based on a template engine, natural language generation technology, and triage scheme; and reviewing the triage report through human review and / or machine review.

[0092] The template engine aims to separate the business logic layer from the presentation layer, converting template code in a specified format into business data for online algorithms. Natural language generation technology is a technique that converts structured information understood by computers into natural language text, resulting in text with natural, coherent, and accurate expression.

[0093] In one possible implementation, data fusion technology can be used to integrate information from different sources to ensure the comprehensiveness of triage reports. Personalized treatment recommendations can be provided using clinical decision support systems.

[0094] The role of triage reports is to integrate key medical information, serving as a bridge between doctors, patients, and other healthcare professionals, while providing decision support to help medical teams make evidence-based diagnostic and treatment choices. Automated triage report generation not only improves the quality and efficiency of medical records, ensuring the accuracy and completeness of information, but also protects patient privacy, enhancing patient trust in healthcare services. Furthermore, it promotes patient education, enabling patients to better understand their health status and necessary medical procedures, thus actively participating in their own health management.

[0095] The present invention also provides an intelligent triage system for implementing any of the intelligent triage methods described above, comprising:

[0096] The information acquisition module is used to acquire the patient's medical information and identify key information within that information.

[0097] The model building module is used to construct dynamic disease models based on patients' historical data, medical knowledge, and key information.

[0098] The information reasoning module is used to reason about key information based on a large model and a knowledge base, generating multiple hypotheses about the patient's condition and the credibility of each hypothesis; the large model is used to identify and predict the probability of disease progression; the knowledge base is used to provide relevant medical rules and facts.

[0099] The disease analysis module is used to analyze the patient's urgency and triage needs based on the disease model and disease assumptions, and generate one or more triage plans.

[0100] The report generation module is used to generate triage reports for patients based on the triage plan. The triage report includes at least several of the following information: basic patient information, description of the condition, triage suggestions, urgency level, treatment suggestions, required medical resources, and precautions.

[0101] The purpose of this invention is to provide an intelligent triage method and system that constructs a disease model based on each patient's condition information. By combining a large model with a knowledge base, it automatically generates a reasonable and effective triage plan based on the patient's condition information, reducing patient waiting time and improving medical efficiency.

[0102] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent triage method, characterized in that: include: S1, obtaining the patient's condition information and identifying key information in the condition information; S2, constructing a dynamic disease model based on the patient's historical data, medical knowledge and the key information; S3, reasoning the key information based on the big model and the knowledge base to generate multiple hypotheses of the patient's condition and the credibility of each of the hypotheses; the big model is used to identify and predict the possibility of disease progression; The knowledge base is used to provide relevant medical rules and facts; S4, analyzing the urgency and triage needs of the patient using a large model according to the disease model and the disease hypothesis, and generating one or more triage plans; S5, obtaining a triage report for the patient according to the triage plan; the triage report includes at least multiple items of the following information: basic information of the patient, description of the condition, triage recommendations, urgency, treatment recommendations, required medical resources, and precautions; The S4 includes: An objective function is constructed to maximize the cure rate of patients and minimize the waiting time of patients; the constraints of the objective function are the availability of medical resources and the priority of patients; Using a heuristic search algorithm to find the optimal or suboptimal triage solution in the feasible solution space of the objective function, and determining the triage level in combination with relevant knowledge in the medical knowledge graph; The method of using a heuristic search algorithm to find an optimal or suboptimal triage solution in feasible solutions of the objective function includes: Randomly generate multiple feasible solutions according to the objective function to form an initial population; The selection probability is calculated according to the fitness of each feasible solution, and multiple feasible solutions are selected from the initial population according to the selection probability to generate a temporary population; randomly selecting a first chromosome and a second chromosome from the temporary population as parents; Randomly selecting a crossover point, exchanging genes of the first chromosome and the second chromosome after the crossover point, to generate a first new chromosome and a second new chromosome; A gene is randomly selected in the first new chromosome and the second new chromosome according to the probability of mutation rate to perform mutation operation to obtain a first mutation chromosome and a second mutation chromosome; the mutation operation includes any one of the following: randomly changing the gene value and exchanging the gene position; adding the first variant chromosome and the second variant chromosome to a new generation population; When the maximum number of iterations is reached or the rate of change of fitness is less than a preset value, the optimal solution, i.e., the triage plan, is output; otherwise, the new generation population is used as the initial population.

2. The intelligent triage method according to claim 1, characterized in that: The S1 includes: Analyze the medical condition information using natural language processing technology to obtain first information; Using the BERT pre-trained language model to identify the semantics and context of the patient to obtain second information; Extracting keywords from the condition information to obtain third information; The key information is obtained according to the first information and / or the second information and / or the third information; the key information at least includes: symptoms, duration, and severity.

3. The intelligent triage method according to claim 1, characterized in that: The S1 further comprises: The patient's medical condition information in voice format is converted into medical condition information in text format through a deep learning algorithm, a convolutional neural network (CNN).

4. The intelligent triage method according to claim 1, characterized in that: The disease model at least includes the patient's basic information, symptoms, signs, past medical history and family medical history; S2 includes: Combining the key information with the historical data through the disease model to obtain a disease view; Assessing the patient's current health status through the disease model; the health status includes the severity of symptoms and possible diseases; Identifying potential health risks and situations requiring urgent intervention in the patient's current health condition through the disease model; The condition model is updated in real time according to new condition information.

5. The intelligent triage method according to claim 1, characterized in that: The S3 includes Using ontology modeling tools to construct the knowledge base; Analyzing the key information using the large model and extracting possible disease hypotheses from the knowledge base; Through the evidence reasoning algorithm, the credibility of each disease hypothesis is evaluated.

6. The intelligent triage method according to claim 1, characterized in that: The calculation of the selection probability according to the fitness of each individual includes: The selection probability Pselect(c) is calculated according to the fitness function F(c), and the formula is as follows: Where N is the number of triage schemes, c is the chromosome that encodes each triage scheme, and c j Represents the gene of the jth decision in the triage scheme, and calculates the fitness F(c) for each chromosome c, where j is a positive integer; The fitness function F(c) is as follows: F(c)=wi·Oi(c) Among them, Oi(c) is the evaluation function of the i-th optimization objective, wi is the weight of the i-th optimization objective, and i is a positive integer.

7. The intelligent triage method according to claim 1, characterized in that: The S5 includes: Generate a structured triage report based on the template engine, natural language generation technology and the triage scheme; The triage report is reviewed by manual review and / or machine review.

8. An intelligent triage system, used to implement the intelligent triage method according to any one of claims 1 to 7, characterized in that: include: An information acquisition module is used to acquire the patient's condition information and identify key information in the condition information; A model building module, used to build a dynamic disease model based on the patient's historical data, medical knowledge and the key information; An information reasoning module, used to reason the key information based on the large model and the knowledge base, and generate multiple hypotheses of the patient's condition and the credibility of each of the hypotheses; The macro model is used to identify and predict the likelihood of disease progression; The knowledge base is used to provide relevant medical rules and facts; A condition analysis module, used to analyze the urgency and triage needs of the patient using a large model according to the condition model and the condition hypothesis, and generate one or more triage plans; A report generation module, used to obtain a triage report for the patient according to the triage plan; the triage report includes at least multiple items of the following information: basic information of the patient, description of the condition, triage suggestions, urgency, treatment suggestions, required medical resources and precautions; The condition analysis module is also used to: construct an objective function to maximize the patient's cure rate and minimize the patient's waiting time; the constraint conditions of the objective function are the availability of medical resources and the priority of the patient; Using a heuristic search algorithm to find the optimal or suboptimal triage solution in the feasible solution space of the objective function, and determining the triage level in combination with relevant knowledge in the medical knowledge graph; The method of using a heuristic search algorithm to find an optimal or suboptimal triage solution in feasible solutions of the objective function includes: Randomly generate multiple feasible solutions according to the objective function to form an initial population; The selection probability is calculated according to the fitness of each feasible solution, and multiple feasible solutions are selected from the initial population according to the selection probability to generate a temporary population; randomly selecting a first chromosome and a second chromosome from the temporary population as parents; Randomly selecting a crossover point, exchanging genes of the first chromosome and the second chromosome after the crossover point, to generate a first new chromosome and a second new chromosome; A gene is randomly selected in the first new chromosome and the second new chromosome according to the probability of mutation rate to perform mutation operation to obtain a first mutation chromosome and a second mutation chromosome; the mutation operation includes any one of the following: randomly changing the gene value and exchanging the gene position; adding the first variant chromosome and the second variant chromosome to a new generation population; When the maximum number of iterations is reached or the rate of change of fitness is less than the preset value, the optimal solution, i.e. the triage plan, is output; Otherwise, the new generation population is used as the initial population.

Images