Medical business system processing method based on multi-agent cooperation, medium and equipment

Abstract

The invention discloses a medical business system processing method based on multi-agent cooperation, a medium and equipment, and belongs to the technical field of medical informatization. In the method, a task analysis agent receives a medical service request, calls a large language model to analyze a user intention, and generates a standardized task description; the routing scheduling agent decomposes the task and distributes the task to the corresponding service execution agent according to the task type and the service rule; the plurality of service execution agents process respective responsible subtasks in parallel, and call corresponding medical information system interfaces to complete specific service operation; and the result summarizing agent collects the processing results of the execution agents, and the processing results are integrated to generate a final service response. According to the invention, through specialized division and cooperation of multiple agents, automatic processing of complex medical services is realized, and the processing efficiency and the intelligent level of a medical information system can be significantly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of medical informatization, and particularly relates to a medical business system processing method based on multi-agent cooperation, which is used for realizing intelligent integration and automatic processing of medical information systems. BACKGROUND

[0002] With the continuous deepening of medical informatization construction, modern hospitals have deployed a large number of information systems to support daily business operations. These systems include but are not limited to: hospital information system (HIS), laboratory information system (LIS), medical image archive and communication system (PACS), electronic medical record system (EMR), rational drug use system, surgical anesthesia system, intensive care system, etc. Each system manages medical data and business processes in a specific field, which together constitutes the information infrastructure of the hospital.

[0003] However, this coexistence of multiple systems brings many challenges. The problem of system silos is serious. Systems developed by different manufacturers use different technical architectures, data formats and interface standards, making it difficult to exchange data between systems. Medical staff need to frequently switch between different systems in their daily work, manually completing information queries, entries and transfers, which not only reduces efficiency, but also easily leads to medical errors due to human operation errors.

[0004] Business process fragmentation is another major problem. A complete medical business often needs to cross multiple systems to complete. For example, doctors need to operate in the HIS system to issue examination applications, view examination results in the PACS system, and access historical test reports in the LIS system. This fragmented operation mode seriously affects the continuity and efficiency of medical services.

[0005] More importantly, existing systems lack intelligent support. Traditional system integration solutions mainly solve the problem of data interconnection, but lack intelligent understanding and processing capabilities for complex business scenarios. Medical staff often describe business requirements in natural language, and existing systems cannot directly understand and respond, requiring users to learn complex operation processes. This not only increases the learning cost of medical staff, but also affects the efficiency of medical services.

[0006] High maintenance costs are also a problem that cannot be ignored. Traditional point-to-point integration or enterprise service bus (ESB) architecture requires modification of a large number of interface codes and business logic when facing business changes. With the continuous development of medical business and the continuous update of standards, the maintenance cost of the system grows exponentially. At the same time, when new medical information systems need to be connected or new business functions need to be added, the traditional architecture needs to be extensively modified, not only the development cycle is long, but also it may affect the stability of the existing system.

[0007] Although some solutions based on microservices, middleware and other technologies have emerged in recent years, these solutions mainly focus on technical-level system integration and lack deep understanding of medical business and intelligent processing capabilities. In particular, under the background of rapid development of large language model technology, how to deeply integrate artificial intelligence technology with medical information systems to realize truly intelligent medical services has become a problem to be solved. SUMMARY

[0008] In view of the above problems existing in the prior art, the purpose of the present application is to provide a medical business system processing method based on multi-agent cooperation, which realizes intelligent processing of medical business and seamless integration of the system by constructing a specialized agent team and combining the understanding ability of a large language model.

[0009] To achieve the above purpose, the technical solution adopted by the present application is as follows:

[0010] In the first aspect, the present application provides a medical business system processing method based on multi-agent cooperation, which is realized based on a medical business multi-agent cooperation framework. The medical business multi-agent cooperation framework includes a task analysis agent, a routing and scheduling agent, a plurality of business execution agents and a result aggregation agent. Each agent communicates through a standardized interface. The processing method includes the following steps:

[0011] S1, the task analysis agent receives a medical business request input by a user, calls a large language model for semantic analysis, extracts business key elements including business type, operation object and time range, and generates a standardized task description;

[0012] S2, the routing and scheduling agent receives the standardized task description, decomposes the complex task into a plurality of subtasks according to the predefined business rules and task types, and assigns the subtasks to the corresponding business execution agents;

[0013] S3, each business execution agent respectively calls the corresponding medical information system API interface according to the subtask received by itself, executes the specific business operation, and feeds back the execution status in real time;

[0014] S4, the result aggregation agent collects the processing results of all business execution agents, performs data integration, format conversion and quality checking, and calls a large language model to generate a final business response conforming to medical specifications.

[0015] As a preferred embodiment of the first aspect, in the task analysis agent, the medical business request text input by the user is first subjected to word segmentation, term standardization and entity recognition to extract key elements including operation objects, time information and business types; then the extracted key elements are combined with a medical knowledge base based on a pre-constructed prompt word template to form an input for a large language model, and the user intent is analyzed by the large language model to identify the business demand, and finally a standardized task description in a structured form is output, which includes the business type, operation parameters, execution conditions and task priority.

[0016] As a preferred embodiment of the first aspect, in the routing and scheduling agent, a business execution agent registration table is maintained, which records the business processing capabilities and state information of each agent; when the routing and scheduling agent receives the standardized task description, it first analyzes the task to identify the business type and complexity; for simple tasks, the corresponding business execution agent is directly matched; for complex tasks, the task is decomposed into multiple sub-tasks according to the business logic, and then appropriate business execution agents are assigned to each sub-task based on the business execution agent registration table to generate a task execution plan containing sub-tasks, business execution agents and execution sequence.

[0017] As a preferred embodiment of the first aspect, the routing and scheduling agent needs to follow the logical relationship of medical business when performing task decomposition, and the decomposition methods include serial decomposition, parallel decomposition and conditional decomposition; the serial decomposition means that sub-tasks with dependency relationship are executed in sequence; the parallel decomposition means that independent sub-tasks can be executed simultaneously; the conditional decomposition means that the execution path of the subsequent sub-task is determined according to the result of the previous sub-task.

[0018] As a preferred embodiment of the first aspect, the business execution agent includes:

[0019] A patient information management agent responsible for querying and updating patient basic information, medical records and medical history data;

[0020] A test and examination agent responsible for querying, booking and result acquisition of test and examination items;

[0021] A medical order management agent responsible for querying, executing, recording and tracking the status of medical orders;

[0022] A drug management agent responsible for drug information query, prescription review, inventory inspection and medication guidance;

[0023] A reservation and scheduling agent responsible for resource scheduling of outpatient reservation, examination reservation and surgery arrangement;

[0024] A medical insurance settlement agent responsible for cost calculation, medical insurance policy matching and settlement processing;

[0025] Each business execution intelligent agent realizes standardized service calling by encapsulating the API interface of the designated medical information system.

[0026] As a preferred embodiment of the first aspect, the execution process of each business execution intelligent agent is as follows: for the sub-task instruction received by the business execution intelligent agent itself, the task parameters are parsed, and after the permission verification and parameter checking, the API interface of the corresponding medical information system is called, the API return result is obtained and processed, then the task execution state is updated, and the result and the task execution state are sent to the result aggregation intelligent agent.

[0027] As a preferred embodiment of the first aspect, in the result aggregation intelligent agent, the returned results of each business execution intelligent agent are received asynchronously, and the collected results are data cleaned and format unified, then data fusion is performed to integrate multi-source data into complete business results, finally, after quality checking according to the medical data standard, a large language model is called to generate a final business response in natural language form.

[0028] As a preferred embodiment of the first aspect, the intelligent agents in the medical business multi-agent collaboration framework have one or more of the following collaboration mechanisms:

[0029] A communication mechanism based on a message queue, in which the intelligent agents communicate with each other through a pre-defined standard message format;

[0030] A task state synchronization mechanism, in which the intelligent agents notify other related intelligent agents through state update messages;

[0031] An exception handling collaboration mechanism, in which when a certain intelligent agent fails to execute, the task recovery or compensation is completed through collaboration.

[0032] In a second aspect, the present application provides a computer readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the medical business system processing method based on multi-agent collaboration according to any one of the first aspect can be realized.

[0033] In a third aspect, the present application provides a computer electronic device, which comprises a memory and a processor.

[0034] The memory is used to store a computer program.

[0035] The processor is used to realize the medical business system processing method based on multi-agent collaboration according to any one of the first aspect when the computer program is executed.

[0036] Compared with the prior art, the present application has the following advantages:

[0037] The present application realizes intelligent integration of the medical information system through the specialization and cooperation mechanism of multiple agents, realizes automatic processing of complex medical business, and can significantly improve the processing efficiency and intelligent level of the medical information system. Medical staff can use natural language to propose business requirements, and the system can automatically understand, decompose, execute and return results, greatly improving work efficiency. At the same time, since the intelligent agent encapsulation method is adopted, the access of the new system becomes very simple, and only the corresponding business execution agent needs to be developed, without changing the overall architecture. This flexible architecture design makes the system have good scalability and maintainability.

[0038] In addition, the present application makes full use of the capabilities of large language models, not only can accurately parse user intent in the task understanding stage, but also can generate professional and easy-to-understand responses in the result generation stage. This end-to-end intelligent processing truly realizes the intelligent upgrade of the medical information system and provides strong technical support for the construction of smart hospitals. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a schematic diagram of a multi-agent cooperation framework of the present application;

[0040] Figure 2 is a schematic diagram of the steps of the medical business system processing method based on multi-agent cooperation;

[0041] Figure 3 is a schematic diagram of the structure of a computer electronic device. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below in conjunction with the drawings. In the following description, a large number of specific details are set forth in order to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, so the present application is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the present application can be combined accordingly without conflict.

[0043] The present application provides a medical business system processing method based on multi-agent cooperation, which is realized by constructing a multi-agent cooperation framework. The core idea of the framework is to decompose complex medical business processing tasks to different functional agents, each agent focuses on a specific type of task, and the overall business process is completed through mutual cooperation. For example, Figure 1As shown, this multi-agent collaborative framework comprises four core agents: a task analysis agent responsible for understanding the user's natural language requests, a routing and scheduling agent responsible for task decomposition and allocation, multiple business execution agents responsible for specific business operations, and a result aggregation agent responsible for integrating processing results and generating a response. All agents communicate through standardized interfaces.

[0044] like Figure 2 As shown, in a preferred embodiment of the present invention, based on Figure 1 The multi-agent collaboration framework shown above, and the medical business system processing method based on multi-agent collaboration, can be implemented according to the following steps:

[0045] S1. The task analysis agent receives the medical service request input by the user, calls the large language model to perform semantic parsing, extracts key business elements including service type, operation object, and time range, and generates a standardized task description.

[0046] The task analysis agent serves as the system's entry point, receiving medical service requests expressed by users in natural language. These requests may range from simple queries, such as "view the patient's blood test results," to complex workflows, such as "schedule a comprehensive examination for the patient and schedule a specialist consultation." The task analysis agent leverages the powerful language understanding capabilities of a large language model to perform deep parsing of these natural language requests. The parsing process can be represented as follows:

[0047]

[0048] Where R represents the original user request text, P represents the prompt template designed for the medical field, K represents the medical knowledge base, and T represents the task understanding result output by the large language model. With the support of carefully designed prompt templates and the medical knowledge base, the large language model can accurately understand medical business requirements expressed in various ways and convert them into standardized task descriptions that the system can process.

[0049] In an embodiment of the present invention, the internal processing flow of the task analysis agent described above is as follows:

[0050] S11: Preprocess the input medical business request text. Preprocessing includes word segmentation, medical terminology standardization, entity recognition, and extraction of key elements such as the operation object (e.g., patient information), time information, and business type.

[0051] S12: Construct a prompt word template specifically for the medical field, and combine the preprocessed text (i.e., the extracted key elements) with the medical knowledge base to form the input of the large language model;

[0052] S13: Use the large language model to parse user intent and identify specific business requirements;

[0053] S14: Convert the output of the large language model into a standardized task description containing business type, operation parameters, execution conditions, and other structured information. The standardized task description TD is in the following form:

[0054]

[0055] wherein represents the business type, parameters is the operation parameter set, conditions is the execution condition, and priority is the task priority.

[0056] S2, the routing scheduling agent receives the standardized task description, decomposes the complex task into multiple sub-tasks according to the pre-defined business rules and task types, and assigns them to the corresponding business execution agent.

[0057] After the routing scheduling agent receives the standardized task description, intelligent task analysis and distribution are needed. For simple single operations such as querying the basic information of a patient, the routing scheduling agent will directly assign the task to the corresponding business execution agent. For complex business processes, the routing scheduling agent will decompose them into multiple sub-tasks according to the logical relationship of medical business. The task decomposition process can be described by the following formula:

[0058]

[0059] wherein represents the standardized task description, represents the task decomposition function, represents the set of decomposed sub-tasks, each sub-task belongs to the set of atomic tasks defined by the system .

[0060] In an embodiment of the present application, the task distribution process inside the routing scheduling agent is as follows:

[0061] S21: Maintain a business execution agent registration table, which records the business processing capabilities and state information of each agent;

[0062] S22: When the routing scheduling agent receives the standardized task description passed by the task analysis agent, it first parses the standardized task description to identify the business type and complexity of the task. The complexity is divided into simple tasks and complex tasks, and the business execution agent is assigned according to different strategies for tasks of different complexity types, with the following rules:

[0063] For simple tasks, no further sub-tasks are needed, and the corresponding business execution agent is directly matched by combining the business execution agent registration table;

[0064] For complex tasks, they are decomposed into multiple sub-tasks according to business logic:

[0065]

[0066] The intelligent agent registry combines business execution for each sub-task to assign appropriate business execution intelligent agent, and generates a task execution plan:

[0067]

[0068] Wherein is the i-th sub-task, is the business execution intelligent agent assigned for the i-th sub-task, is the execution order, and n is the total number of sub-tasks.

[0069] It should be noted that the routing and scheduling intelligent agent in the present application needs to follow the logical relationship of medical business when further decomposing sub-tasks of complex tasks, and the decomposition process needs to fully consider the dependency relationship between tasks. Some tasks must be executed in order, some can be processed in parallel, and some need to be executed according to the results of the pre-task to determine whether to execute, so as to avoid that the sub-tasks divided do not conform to the actual processing flow of medical business. Generally speaking, there are three ways of sub-task decomposition, including:

[0070] Serial decomposition, decomposing complex tasks into a series of sub-tasks with dependency relationship according to the dependency relationship, and the sub-tasks with dependency relationship need to be executed in order;

[0071] Parallel decomposition, decomposing complex tasks into a series of sub-tasks without dependency relationship, and the independent sub-tasks can be executed simultaneously;

[0072] Conditional decomposition, decomposing complex tasks into a series of sub-tasks that need to meet preconditions to be executed, and the execution path of the subsequent sub-tasks needs to be determined according to the results of the pre-sub-tasks.

[0073] S3, each business execution intelligent agent respectively receives the sub-tasks, calls the corresponding medical information system API interface, executes specific business operations, and feeds back the execution status in real time.

[0074] The business execution agent is the core component that actually completes the medical business operation. The above business execution agent can be set according to the actual medical task, and a large model or other tool class agent is preferably used to execute different tasks. Due to the complexity of medical business, a plurality of specialized business execution agents should be included in the system, each of which encapsulates the interface of a specific medical information system and can convert standardized task instructions into specific system operations. For example, the patient information management agent is responsible for interacting with the HIS system and processing the basic information and medical records of patients; the inspection and examination agent is responsible for interacting with the LIS and PACS systems to obtain inspection reports and image data; the medical order management agent is responsible for the creation, execution and management of medical orders; the drug management agent processes drug-related queries and management; the appointment and scheduling agent is responsible for the appointment arrangement of various medical resources; and the medical insurance settlement agent processes fee calculation and medical insurance reimbursement. The execution process of each agent can be represented as:

[0075]

[0076] wherein represents a subtask assigned to the agent, represents a target medical information system, represents a specific method called, represents a task parameter, represents an execution result. Each agent will perform necessary permission verification and parameter checking when executing a task to ensure the legality and security of the operation.

[0077] In an embodiment of the present application, the above business execution agent includes the following types:

[0078] The patient information management agent is responsible for querying and updating the basic information, medical records and medical record data of patients;

[0079] The inspection and examination agent is responsible for querying, booking and result obtaining of inspection (such as laboratory inspection) and examination (such as medical image) items;

[0080] The medical order management agent is responsible for querying, executing, recording and status tracking of medical orders;

[0081] The drug management agent is responsible for drug information query, prescription review, inventory inspection and medication guidance;

[0082] The appointment and scheduling agent is responsible for outpatient appointment, examination appointment, surgery arrangement and other resource scheduling;

[0083] The medical insurance settlement agent is responsible for fee calculation, medical insurance policy matching and settlement processing;

[0084] Each business execution agent encapsulates the API interface of the designated medical information system to achieve standardized service invocation. The execution process of each business execution agent can be represented as follows:

[0085] S31: Receive the sub-task instruction sent by the routing and scheduling agent, and parse the task parameters from it;

[0086] S32: Perform permission verification and parameter checking according to the settings of the corresponding medical information system;

[0087] S33: After verification, call the API interface of the corresponding medical information system to obtain the API return result :

[0088]

[0089] S34: Process the API return result , and perform necessary data conversion according to the form of the final result;

[0090] S35: Update the task execution status and send the result and task execution status to the result aggregation agent.

[0091] S4, the result aggregation agent collects the processing results of all business execution agents, performs data integration, format conversion and quality checking, and calls a large language model to generate a final business response that meets medical standards.

[0092] The result aggregation agent is responsible for collecting the processing results of all business execution agents, which is a complex data integration process. Since the data formats returned by different medical information systems are different, the result aggregation agent needs to perform deep data processing. The data fusion process can be represented by the following formula:

[0093]

[0094] where represents the return result of the i-th business execution agent, Transform represents the data format conversion function, Normalize represents the data standardization function, and Validate represents the verification function based on the medical standards Standards, represents the final fusion result. This process not only ensures the accuracy and completeness of the data, but also ensures that the data meets the relevant standards and specifications of the medical industry.

[0095] In an embodiment of the present application, the processing process of the above-mentioned result aggregation agent includes:

[0096] S41: Establish a result collection mechanism to asynchronously receive the return results of each business execution agent:

[0097]

[0098] S42: Data cleaning and format unification are performed on the collected results:

[0099]

[0100] S43: Data fusion is performed to integrate multi-source data into complete business results:

[0101]

[0102] The data fusion process can include the operations described in 1) ~ 4) below:

[0103] 1) Time alignment, organizing data at different time points in chronological order;

[0104] 2) Data deduplication, identifying and eliminating duplicate data;

[0105] 3) Conflict resolution, when there are conflicts between multiple data sources, select according to data source priority and timeliness;

[0106] 4) Association integration, establish the association between different data, and form a complete information view.

[0107] S44: Quality check according to medical data standards:

[0108]

[0109] S45: Call large language model to generate natural language response:

[0110]

[0111] In addition, the above-mentioned medical business multi-agent collaboration framework has the following A) ~ C) described collaboration mechanism between agents:

[0112] A) Establish a communication mechanism based on message queue, and each agent communicates with each other through a pre-defined standard message format, which is defined as follows:

[0113]

[0114] B) Implement a task state synchronization mechanism, and each agent notifies other related agents through a state update message;

[0115] C) Establish an exception handling collaboration mechanism, when a certain agent fails to execute, complete task recovery or compensation through collaboration.

[0116] The collaboration mechanism between agents is the key to the efficient operation of the system. All agents communicate through a standardized message format. The system uses an asynchronous communication mechanism based on message queues to ensure loose coupling between agents. When an agent completes a task or encounters an exception, it notifies the relevant agents through a message, triggering subsequent processing procedures.

[0117] In addition, in other embodiments of the present application, optimization mechanisms can also be provided in the entire system, such as by recording the execution log of each task, including execution time, success rate, abnormal situation, failure cause, etc. information, the system can continuously learn and optimize. At the same time, based on the analysis of historical data, the task decomposition strategy can be optimized, the agent allocation rules can be adjusted, and the exception handling process can be improved. The ability of continuous optimization enables the system to continuously adapt to the changes and development of medical business.

[0118] In order to better understand the specific implementation process of the present application, the following takes a comprehensive medical business scenario as an example to show how the present application handles complex medical processes involving multiple links. Suppose a hospital emergency department admitted a patient with suspected acute myocardial infarction, who needs to quickly complete a series of diagnosis and treatment processes, including admission registration, emergency examination, specialist consultation, drug treatment, etc.

[0119] The emergency physician inputs through the system interface: "Patient Wang Jianguo, male, 62 years old, chest pain for 2 hours, suspected acute myocardial infarction, needs emergency admission, complete electrocardiogram, myocardial enzyme spectrum, troponin examination, apply for cardiology consultation, prepare for thrombolysis treatment." This is a typical complex medical business request involving multiple systems and multiple operation steps.

[0120] After receiving this request, the task analysis agent starts in-depth semantic analysis. First, medical entity recognition is performed to extract key information such as patient name "Wang Jianguo", gender "male", age "62 years old", chief complaint "chest pain for 2 hours", preliminary diagnosis "suspected acute myocardial infarction", etc. Then identify the business operations to be performed, including "emergency admission", "electrocardiogram examination", "myocardial enzyme spectrum examination", "troponin examination", "cardiology consultation", "preparation for thrombolysis treatment", etc. The task analysis agent calls a large language model to understand that this is a time-sensitive emergency medical task that requires multiple departments to coordinate and respond quickly, combining the diagnosis and treatment guidelines for acute myocardial infarction and the hospital's standard operation procedures.

[0121] After analysis and processing, the task analysis agent generates a standardized task description, clearly defining the task type as "acute myocardial infarction emergency treatment" with a priority of "urgent" and listing all the operations to be performed and their parameters in detail. This standardized task description is sent to the routing and scheduling agent.

[0122] After receiving the task description, the routing and scheduling agent immediately starts the task decomposition process. According to the standard diagnosis and treatment path of acute myocardial infarction and the logical relationship between tasks, the entire process is decomposed into multiple sub-tasks. First, the patient's emergency record and admission registration need to be created in the HIS system, which is the prerequisite for all subsequent operations; Then multiple examination applications can be performed in parallel, including electrocardiogram, blood test, etc.; At the same time, a cardiology consultation application is initiated; Finally, the drugs and equipment required for thrombolytic therapy are prepared.

[0123] After task decomposition is complete, the routing and scheduling agent formulates a detailed execution plan based on the capabilities and current state of each business execution agent. The patient information management agent is responsible for creating the emergency record and handling the admission procedures; The examination and inspection agent is responsible for applying for electrocardiogram and various blood tests; The consultation management agent is responsible for sending an emergency consultation request to the cardiology department; The drug management agent is responsible for confirming the inventory of thrombolytic drugs and preparing related drugs.

[0124] Each business execution agent immediately begins execution after receiving the task. The patient information management agent first creates the patient's emergency visit record in the HIS system, generates a visit number, and then quickly completes the admission registration and assigns a bed. After the creation of these basic information, other agents are immediately notified that they can begin executing their respective tasks.

[0125] The examination and inspection agent simultaneously sends emergency examination applications to the electrocardiogram room and the laboratory. For electrocardiogram examination, the system automatically marks it as "emergency myocardial infarction" priority, and the electrocardiogram room prepares immediately after receiving the notification; For blood tests, the system not only applies for myocardial enzyme spectrum and troponin, but also automatically adds blood routine, coagulation function, liver and kidney function, etc. necessary examination items according to the examination package of acute myocardial infarction, all of which are marked as "emergency".

[0126] The consultation management agent sends an emergency consultation application to the on-duty doctor of the cardiology department, which automatically includes the patient's basic information, chief complaint, preliminary diagnosis, and other key information. The system also notifies the cardiology doctor through multiple ways, including system messages, mobile phone messages, etc. according to the urgency of the consultation, to ensure that the consultation doctor can respond in time.

[0127] The drug management agent checks the inventory of drugs required for thrombolytic therapy, confirms the availability of thrombolytic drugs such as alteplase, and sends an emergency drug preparation notice to the pharmacy. At the same time, the system also automatically prepares other emergency drugs that may be needed, such as nitroglycerin, aspirin, clopidogrel, etc.

[0128] In the process of each agent performing the task, they continuously report the progress and results to the result aggregation agent. The result aggregation agent collects this information in real time, integrates the data and tracks the status. After the electrocardiogram is completed, the system immediately obtains the electrocardiogram results and finds that there is indeed ST segment elevation, supporting the diagnosis of acute myocardial infarction; the blood test results show that troponin is significantly elevated, further confirming the diagnosis; the cardiologist has been notified and is on the way; thrombolytic drugs are ready.

[0129] The result aggregation agent intelligently integrates all this information, identifies the key diagnostic basis and the actions that need to be taken immediately. Then it calls the large language model to generate a clear and professional comprehensive report: "Patient Wang Jianguo emergency treatment progress: emergency admission registration has been completed, bed CCU-05. Electrocardiogram shows ST segment elevation in V1-V4 leads, consistent with acute anterior myocardial infarction. Troponin I is 8.5 ng / ml (normal value <0.04), and CK-MB is elevated. Dr. Li of the cardiology department has been notified for consultation and is expected to arrive within 5 minutes. Thrombolytic drug alteplase is ready, and it is recommended to start the thrombolytic treatment process immediately. Please note that the patient has been admitted for 2.5 hours since the onset, and is still within the thrombolytic time window."

[0130] The entire process from the doctor inputting the request to the system returning the comprehensive results takes less than 3 minutes. In the traditional mode, the doctor needs to log in to multiple systems, manually fill out various application forms, and make phone calls to coordinate various departments. The entire process may take 20-30 minutes. More importantly, the system automatically completes many operations that the doctor may miss during the execution process, such as automatically adding necessary examination items, notifying the consulting doctor through multiple ways, and reminding the thrombolytic time window. These intelligent processing greatly improves the quality of medical treatment and patient safety.

[0131] This embodiment fully demonstrates the core advantages of the present application. Through the collaborative work of multiple agents, complex medical processes are efficiently decomposed and executed; through the understanding and generation capabilities of the large language model, the system can accurately understand the doctor's intentions and generate professional responses; through intelligent process optimization, the system not only improves efficiency, but also improves the quality of medical services. This intelligent medical business processing method provides a new solution for the informatization construction of modern hospitals.

[0132] In addition, in order to verify the effectiveness of the method of the present application, a 2-month actual deployment test was conducted, covering emergency, outpatient, and inpatient scenarios. Table 1 shows the processing performance of tasks of different complexity during the actual deployment test, and Table 2 shows the efficiency of agent collaboration during the actual deployment test.

[0133] Table 1 Processing performance of tasks of different complexity

[0134]

[0135] Table 2 Intelligence body cooperation efficiency analysis

[0136]

[0137] The experimental results show that the method has significant advantages over the traditional integration method in various key indicators, especially in processing complex medical business processes, and the efficiency is improved by more than 75%.

[0138] It should be noted that the medical business system processing method based on multi-agent cooperation shown by S1-S4 can be essentially realized in the form of a computer program or a software function module.

[0139] In addition, based on the same inventive concept, as shown in Figure 3 The present application also provides a computer electronic device corresponding to the medical business system processing method based on multi-agent cooperation provided by the above-mentioned embodiment, which includes a memory and a processor.

[0140] The memory is used to store a computer program.

[0141] The processor is used to realize the medical business system processing method based on multi-agent cooperation as described above when the computer program is executed.

[0142] In addition, the logical instructions in the above-mentioned memory can be realized in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the various embodiments of the present application.

[0143] Therefore, based on the same inventive concept, the present application provides a computer readable storage medium corresponding to the medical business system processing method based on multi-agent cooperation, and the storage medium stores a computer program. When the computer program is executed by a processor, the medical business system processing method based on multi-agent cooperation as described above can be realized.

[0144] Therefore, based on the same inventive concept, the present application provides a computer program product, which includes a computer program / instruction that can be executed by a processor to realize the medical business system processing method based on multi-agent cooperation as described above. Therefore, based on the same inventive concept, the present application provides a computer program product, which includes a computer program / instruction that can be executed by a processor to realize the medical business system processing method based on multi-agent cooperation as described above.

[0145] Specifically, in the computer readable storage medium of the above three embodiments, the stored computer program is executed by the processor, and the steps S1-S4 can be executed.

[0146] It can be understood that the storage medium can include a random access memory (RAM) and a non-volatile memory (NVM), such as at least one disk memory. Meanwhile, the storage medium can also be a U disk, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0147] It can be understood that the processor can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; or a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0148] In addition, it should be noted that the skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here. In each embodiment provided in the present application, the division of steps or modules in the system and method is only a logical function division, and actual implementation can have another division mode, for example, multiple modules or steps can be combined or integrated together, or a module or step can be split.

[0149] The above-described embodiments are only preferred embodiments of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and supplements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A medical business system processing method based on multi-agent collaboration, characterized in that, The processing method is based on a multi-agent collaboration framework for medical services. This framework includes a task analysis agent, a routing and scheduling agent, multiple service execution agents, and a result aggregation agent. All agents communicate through a standardized interface. The processing method includes the following steps: S1. The task analysis agent receives the medical business request input by the user, calls the large language model to perform semantic parsing, extracts key business elements including business type, operation object, and time range, and generates a standardized task description. S2. The routing and scheduling agent receives a standardized task description, decomposes the complex task into multiple sub-tasks according to predefined business rules and task types, and assigns them to the corresponding business execution agents. S3. Each business execution agent calls the corresponding medical information system API interface to perform specific business operations based on the sub-tasks it receives, and provides real-time feedback on the execution status. S4. The result summarization agent collects the processing results of all business execution agents, integrates the data, performs format conversion and quality verification, and calls the large language model to generate the final business response that conforms to medical standards.

2. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, In the task analysis agent, the medical service request text input by the user is first segmented, standardized in terms of terminology, and identified as an entity to extract key elements including the operation object, time information, and service type. Then, based on a pre-built prompt word template, the extracted key elements are combined with a medical knowledge base to form the input of a large language model. The large language model then parses the user's intent, identifies the service requirements, and finally outputs a standardized task description in a structured form. The standardized task description includes the service type, operation parameters, execution conditions, and task priority.

3. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, The routing scheduling agent maintains a business execution agent registry, which records the business processing capabilities and status information of each agent. When the routing scheduling agent receives the standardized task description, it first parses it to identify the business type and complexity of the task. For simple tasks, it directly matches the corresponding business execution agent. For complex tasks, it decomposes them into multiple sub-tasks according to the business logic, and then assigns a suitable business execution agent to each sub-task based on the business execution agent registry, generating a task execution plan that includes sub-tasks, business execution agents, and execution order.

4. The medical business system processing method based on multi-agent collaboration according to claim 3, characterized in that, When the routing and scheduling agent performs task decomposition, it must follow the logical relationship of medical business. The decomposition methods are divided into serial decomposition, parallel decomposition, and conditional decomposition. The serial decomposition means that subtasks with dependencies are executed in sequence. The parallel decomposition means that independent subtasks can be executed simultaneously. The conditional decomposition means that the execution path of subsequent subtasks is determined based on the results of the preceding subtasks.

5. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, The business execution intelligent agent includes: The patient information management intelligent agent is responsible for querying and updating basic patient information, medical records, and medical data. The inspection and testing intelligent agent is responsible for querying, scheduling, and obtaining results for inspection and testing items. The medical order management intelligent agent is responsible for querying, executing, recording, and tracking the status of medical orders; The intelligent agent for drug management is responsible for drug information inquiry, prescription review, inventory check and medication guidance; The appointment scheduling intelligent agent is responsible for resource scheduling for outpatient appointments, examination appointments, and surgical arrangements; The intelligent medical insurance settlement agent is responsible for cost calculation, medical insurance policy matching, and settlement processing. Each business execution agent implements standardized service calls by encapsulating the API interface of the specified medical information system.

6. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, The execution process of each business execution agent is as follows: for the subtask instructions received by the business execution agent itself, it parses the task parameters, and after authorization verification and parameter validation, it calls the API interface of the corresponding medical information system, obtains and processes the API return results, updates the task execution status, and sends the results and task execution status to the result aggregation agent.

7. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, The result aggregation agent asynchronously receives the return results from each business execution agent, cleans and unifies the format of the collected results, then performs data fusion to integrate multi-source data into a complete business result. Finally, after quality verification according to medical data standards, it calls a large language model to generate the final business response in natural language form.

8. The medical business system processing method based on multi-agent collaboration according to claim 1, characterized in that, Within the multi-agent collaboration framework for medical services, the agents employ one or more of the following collaboration mechanisms: The communication mechanism based on message queues allows agents to communicate with each other through a predefined standard message format. The task state synchronization mechanism allows each agent to notify other relevant agents through state update messages. An exception handling collaboration mechanism is used to recover or compensate for task failures through collaboration when an agent fails to execute its task.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, can implement the medical business system processing method based on multi-agent collaboration as described in any one of claims 1 to 8.

10. A computer electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the medical business system processing method based on multi-agent collaboration as described in any one of claims 1 to 8.

Images