Online medicine process management method and system

Abstract

The invention relates to the technical field of Internet, in particular to an online medicine process management method and system, and the method comprises the following steps: S1, disassembling an online medicine process into independent industrial micro-services, and carrying out the high-cohesion and low-coupling of functions; according to the invention, through multi-mode demand accurate analysis, micro-service chain dynamic combination and priority intelligent sorting, high-efficiency adaptation of patient demands and service resources is realized, the urgent and severe order response timeliness is improved, the demand analysis accuracy is also greatly improved, real-time data monitoring and resource self-adaptive adjustment based on reinforcement learning are realized, and the system has good application prospects. On the premise that compliance is guaranteed, optimal allocation of physicians, medicines and distribution resources is achieved, the requirement satisfaction rate of high-priority patients is increased, the resource utilization rate is optimized, and the response speed and resource allocation efficiency of online medical services are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, in particular to an online medicine process management method and system. BACKGROUND

[0002] Online medicine refers to a service mode that traditional medicine and health services (covering disease consultation, diagnosis and treatment guidance, prescription issuance, medicine purchase, logistics distribution, health management and other links) are completed online relying on Internet technology, mobile communication technology and digital platform. It breaks through the time and space limit, enables patients to obtain medical services without going out, greatly improves the service convenience, and plays an important role in remote medical resource coverage, chronic disease long-term management and other scenarios. However, in the prior art, the traditional online medicine service process is rigid, the demand matching accuracy is low, the response to urgent and severe orders is lagging, the resource scheduling is inefficient, and the compliance risk is difficult to control.

[0003] Based on this, the present application provides an online medicine process management method and system to solve the above technical problems. SUMMARY

[0004] The present application aims to provide an online medicine process management method and system. The present application accurately analyzes multi-modal demands, dynamically combines microservices and intelligently sorts priorities, efficiently adapts patient demands to service resources, improves the response time of urgent and severe orders, greatly improves the accuracy of demand analysis, and based on real-time data monitoring and resource self-adaptive adjustment of reinforcement learning, realizes the optimal allocation of doctors, medicines and distribution resources under the premise of ensuring compliance, improves the satisfaction rate of high-priority patient demands, optimizes resource utilization, and significantly improves the response speed and resource allocation efficiency of online medicine services.

[0005] To achieve the above purpose, the present application provides the following technical scheme:

[0006] The present application provides an online medicine process management method, comprising the following steps:

[0007] S1: The online medicine process is disassembled into independent industrial microservices, with high cohesion and low coupling of functions;

[0008] S2: The AI algorithm analyzes the patient demand characteristics and real-time resource status, dynamically generates a microservice combination scheme, optimizes the service priority based on an adaptive weight model, and converts unstructured patient input into standardized microservice call parameters through an interface adaptation layer;

[0009] S3: A cross-domain collaboration layer is constructed based on microservice bus technology, integrating microservices of each link of the medicine supply chain, diagnosis and treatment services and logistics distribution, and realizing end-to-end visual collaboration;

[0010] S4: Construct a "scene awareness-intelligent decision-making-execution feedback" closed-loop network based on real-time data flow between microservices, dynamically monitor the state of the link through multi-dimensional indicators, and generate resource adjustment strategies combined with reinforcement learning algorithms to adaptively optimize the configuration of service resources;

[0011] S5: Embed each microservice interaction node with a microservice security middleware to perform medical data compliance and system-level risk prevention and control.

[0012] Based on the above method, the application also provides an online medical process management system, which comprises a microservice decoupling module, an intelligent matching engine module, a cross-domain collaborative bus module, a dynamic optimization control module, and a security compliance middleware module.

[0013] The microservice decoupling module is used to decompose the online medical process into independent microservice units with high cohesion and low coupling, and define the functional boundaries and interaction standards of each service.

[0014] The intelligent matching engine module is based on AI algorithms to dynamically combine microservices, optimize service priorities, and adapt to the standardized conversion of unstructured patient inputs.

[0015] The cross-domain collaborative bus module is used to integrate drug supply chain, medical treatment services, and logistics distribution microservices through microservice bus technology for end-to-end visual collaborative management.

[0016] The dynamic optimization control module is based on real-time data flow and reinforcement learning algorithms to monitor process status and adaptively adjust resource configuration.

[0017] The security compliance middleware module is used to embed microservice interaction nodes to perform medical data compliance and system-level risk prevention and control.

[0018] The microservice decoupling module comprises a process decomposition unit, a service definition unit, and an interface standardization unit.

[0019] The process decomposition unit is used to split the process into independent microservices according to online medical business links, and contains redundant microservices for improving system fault tolerance, such as backup distribution node microservices and emergency physician resource microservices.

[0020] The service definition unit is used to clearly define the functional scope, input and output parameters, and performance indicators of each microservice, and the redundant microservices need to define additional trigger activation conditions, such as automatically activating the backup distribution node microservice when the main distribution node fails.

[0021] The interface standardization unit is used to develop interface specifications and communication protocols for data interaction between microservices.

[0022] The intelligent matching engine module comprises a demand analysis unit, a service combination unit, a priority sorting unit and an input adaptation unit, wherein:

[0023] The demand analysis unit is configured to analyze patient demand characteristics and real-time resource states by using an AI algorithm.

[0024] The service combination unit is configured to dynamically combine and adapt a micro-service chain according to the demand and generate an optimal service path.

[0025] The priority sorting unit is configured to sort the service execution order based on an adaptive weight model.

[0026] The input adaptation unit is configured to convert patient unstructured input into standardized micro-service calling parameters.

[0027] The service combination unit dynamically combines and adapts a micro-service chain according to the demand and generates an optimal service path, and the specific operation is as follows:

[0028] A1: Based on the patient demand characteristics output by the demand analysis unit, a candidate micro-service set that meets the functional demand is screened out, such as a “physician matching micro-service” and a “prescription generation micro-service” corresponding to the inquiry demand;

[0029] A2: Compatibility verification is performed on the candidate micro-service set, and micro-services with interface conflicts or compliance conflicts are removed, such as prohibiting the combination of a physician micro-service that has not passed a qualification audit and a prescription micro-service;

[0030] A3: Based on preset evaluation indexes including service response time, resource consumption cost and historical execution success rate, the compatible micro-service chain is scored, and the chain with the highest comprehensive score is selected as the optimal service path;

[0031] A4: If there are multiple paths with similar scores, the path with a backup distribution node micro-service is preferentially selected.

[0032] The priority sorting unit sorts the service execution order based on an adaptive weight model, and the specific operation is as follows:

[0033] B1: Real-time resource state indexes of each micro-service are obtained, including the number of online doctors, the inventory level of medicines and the logistics distribution timeliness;

[0034] B2: A weight configuration template is dynamically loaded according to the patient prescription type, wherein the logistics timeliness weight is preferentially configured for emergency prescriptions;

[0035] B3: A comprehensive priority score is calculated based on multi-dimensional indexes, and the specific formula is as follows:

[0036]

[0037] wherein, , , are weight coefficients respectively, is the disease emergency degree, which is generated by the demand analysis unit according to the patient demand, is the resource availability, is the average response time;

[0038] B4: generating a service execution queue in descending order of scores.

[0039] The cross-domain collaborative bus module comprises a service registration and discovery unit, a data integration unit, a collaborative scheduling unit and a visualization unit, wherein:

[0040] The service registration and discovery unit is configured to manage the registration information of each microservice and support automatic discovery and calling of cross-domain services.

[0041] The data integration unit is configured to integrate the microservice data of the drug supply chain, diagnosis and treatment, and distribution links and eliminate information silos.

[0042] The collaborative scheduling unit is configured to coordinate the execution timing of cross-link microservices through the bus.

[0043] The visualization unit is configured to display the end-to-end process status in real time through a graphical interface.

[0044] The dynamic optimization control module comprises a data acquisition unit, a state monitoring unit, an algorithm decision unit and an execution feedback unit, wherein:

[0045] The data acquisition unit is configured to acquire the running data of each microservice in real time.

[0046] The state monitoring unit is configured to monitor the running state of the process node based on multi-dimensional indexes.

[0047] The algorithm decision unit is configured to analyze the monitoring data through a reinforcement learning algorithm and generate a resource adjustment strategy.

[0048] The execution feedback unit is configured to issue the adjustment strategy to the corresponding microservice and track the optimization effect to form a closed loop.

[0049] The algorithm decision unit analyzes the monitoring data through a reinforcement learning algorithm to generate a resource adjustment strategy, and the specific operation is as follows:

[0050] C1: State space construction:

[0051] Define the state set of reinforcement learning as:

[0052]

[0053] wherein, Resource state vector, including physician load rate, drug inventory level, real-time location of delivery vehicle, Patient demand vector, including urgency of illness in B3 , service time requirement, Compliance status indicators, such as prescription review completion rate;

[0054] C2: Action space definition:

[0055] Set the executable resource adjustment action set , including:

[0056] I. Physician resource scheduling, such as temporary allocation of physicians from low-load departments to high-load departments;

[0057] II. Drug inventory allocation, such as emergency replenishment from regional warehouses to short supply pharmacies;

[0058] III. Delivery path optimization, such as switching direct delivery routes for high-priority orders;

[0059] C3: Reward function design:

[0060] Construct a reward function that takes into account efficiency, safety and compliance :

[0061]

[0062] Among them, Efficiency reward, positively related to the reduction in process time after resource adjustment; Safety reward, positively related to the satisfaction rate of high-priority patient demand, with double weight for critical patients; Compliance reward, positively related to compliance indicators such as prescription review and privacy protection; , and when the compliance indicators are not met, , Penalty coefficient;

[0063] C4: Strategy learning and execution:

[0064] ① Use DQN algorithm to pre-train the model through offline historical data to get the initial strategy ;

[0065] ② In online operation, according to the real-time state , the optimal adjustment action is output through the strategy ;

[0066] ③ After executing the action, collect the reward value of the environmental feedback ​, for model parameter iterative optimization, so that the strategy gradually converges to the "optimal allocation of resources under compliance premise";

[0067] C5: policy output:

[0068] The converged optimal action Converts into executable resource adjustment instructions, such as "allocate 5 boxes of a certain targeted drug to A pharmacy", and issues to the corresponding microservice.

[0069] The security compliance middleware module comprises a data encryption unit, a compliance verification unit, an authority management unit, and a risk interception unit, wherein:

[0070] The data encryption unit is used for transmitting and storing medical privacy data.

[0071] The compliance verification unit is used for embedding medical industry regulations and rules to verify the compliance of the process in real time.

[0072] The authority management unit assigns microservice access authority based on roles and limits the data operation range.

[0073] The risk interception unit is used for monitoring abnormal interactions and triggering system-level risk warning and interception.

[0074] Compared with the prior art, the beneficial effects of the present application are:

[0075] The present application realizes efficient adaptation of patient demand and service resources through multi-modal demand precise analysis, microservice chain dynamic combination and priority intelligent sorting, improves the response timeliness of critical order and greatly improves the demand analysis accuracy, and based on real-time data monitoring and resource self-adaptive adjustment of reinforcement learning, realizes the optimal allocation of doctors, drugs and distribution resources under the premise of guaranteeing compliance, improves the satisfaction rate of high-priority patient demand, optimizes resource utilization, and significantly improves the response speed and resource allocation efficiency of online medical services. BRIEF DESCRIPTION OF DRAWINGS

[0076] Figure 1 The present application is a flow chart of an online medical process management method.

[0077] Figure 2 The present application is a system diagram of an online medical process management system.

[0078] Explanation of reference numerals:

[0079] 1. Microservice Decoupling Module; 11. Process Decomposition Unit; 12. Service Definition Unit; 13. Interface Standardization Unit; 2. Intelligent Matching Engine Module; 21. Requirements Analysis Unit; 22. Service Composition Unit; 23. Priority Sorting Unit; 24. Input Adaptation Unit; 3. Cross-Domain Collaboration Bus Module; 31. Service Registration and Discovery Unit; 32. Data Integration Unit; 33. Collaborative Scheduling Unit; 34. Visualization Unit; 4. Dynamic Optimization Control Module; 41. Data Acquisition Unit; 42. Status Monitoring Unit; 43. Algorithm Decision-Making Unit; 44. Execution Feedback Unit; 5. Security and Compliance Middleware Module; 51. Data Encryption Unit; 52. Compliance Verification Unit; 53. Access Control Unit; 54. Risk Interception Unit. Detailed Implementation

[0080] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0081] Example 1:

[0082] like Figure 1 As shown, this embodiment provides an online medical process management system, including a microservice decoupling module 1, an intelligent matching engine module 2, a cross-domain collaboration bus module 3, a dynamic optimization control module 4, and a security and compliance middleware module 5. Specifically: the microservice decoupling module 1 decomposes the online medical process into highly cohesive, loosely coupled independent microservice units, defining the functional boundaries and interaction standards of each service; the intelligent matching engine module 2 dynamically combines microservices based on AI algorithms, optimizes service priorities, and adapts to the standardized conversion of unstructured patient input; the cross-domain collaboration bus module 3 integrates drug supply chain, medical services, and logistics distribution microservices through microservice bus technology, enabling end-to-end visualized collaborative management; the dynamic optimization control module 4 monitors process status and adaptively adjusts resource configuration based on real-time microservice data flow and reinforcement learning algorithms; and the security and compliance middleware module 5 embeds microservice interaction nodes to perform medical data compliance and system-level risk prevention and control.

[0083] It should be noted that the microservice decoupling module 1 builds basic services, the intelligent matching engine module 2 realizes demand-driven dynamic service orchestration, the cross-domain collaborative bus module 3 connects the entire pharmaceutical data interaction chain, the dynamic optimization control module 4 optimizes resource configuration based on real-time feedback closed loop, and the security and compliance middleware module 5 ensures system security throughout all aspects.

[0084] In the present embodiment, it is also necessary to point out that the micro-service decoupling module 1 comprises a process disassembling unit 11, a service defining unit 12, and an interface standardization unit 13, wherein: the process disassembling unit 11 is used for splitting the process according to the online medical business link into independent micro-services, and contains redundant micro-services for improving system fault tolerance, such as a backup distribution node micro-service and an emergency physician resource micro-service; the service defining unit 12 is used for specifying the function range, input and output parameters, and performance indicators of each micro-service, wherein the redundant micro-service needs to additionally define the triggering activation condition, such as automatically activating the backup distribution node micro-service when the main distribution node fails; and the interface standardization unit 13 is used for formulating the interface specification and communication protocol for data interaction between micro-services.

[0085] It is necessary to point out that the process disassembling unit 11 realizes the modularization segmentation and redundancy design of the business link, the service defining unit 12 specifies the function boundary and activation condition of each micro-service, and the interface standardization unit 13 unifies the interaction protocol.

[0086] Further, it is necessary to point out that the HTTP / 2+gRPC hybrid protocol is adopted in the interface standardization unit 13, gRPC (binary transmission, reducing protocol overhead) is used for high-frequency and low-latency interaction (such as “real-time synchronization of order status”), and HTTP / 2 (compatible with existing gateways, supporting Header extension) is used for low-frequency and cross-domain interaction (such as “cross-regional medicine inventory query”); the interface defaults to enable TLS1.3 encryption, and through the interface specification, it is forcibly required that “medical sensitive data (such as prescriptions and medical records) must carry an encrypted identification header (X-Encrypt-Type: SM4)”, so as to ensure the security of the data link.

[0087] In the present embodiment, it is also necessary to point out that the intelligent matching engine module 2 comprises a demand analysis unit 21, a service combination unit 22, a priority sorting unit 23, an input adaptation unit 24, wherein: the demand analysis unit 21 is used for analyzing patient demand characteristics and resource real-time state through an AI algorithm; the service combination unit 22 is used for dynamically combining an adaptive microservice chain according to the demand to generate an optimal service path; the specific operation is as follows: A1: based on the patient demand characteristics output by the demand analysis unit 21, a candidate microservice set meeting the functional demand is screened out, such as the “physician matching microservice” and “prescription generation microservice” corresponding to the inquiry demand; A2: compatibility checking is performed on the candidate microservice set, and microservices with interface conflicts or compliance conflicts are excluded, such as prohibiting the combination of unqualified physician microservices and prescription microservices; A3: based on preset evaluation indexes including service response time, resource consumption cost and historical execution success rate, the compatible microservice chain is scored, and the chain with the highest comprehensive score is selected as the optimal service path; A4: if there are multiple paths with similar scores, the path with the standby distribution node microservice is preferentially selected. The priority sorting unit 23 sorts the service execution order based on an adaptive weight model; the specific operation is as follows: B1: real-time acquisition of resource state indexes of each microservice, including the number of online doctors, drug inventory level and logistics distribution timeliness; B2: dynamically loading a weight configuration template according to the patient prescription type, wherein the logistics timeliness weight is preferentially configured for emergency prescriptions; B3: calculating a comprehensive priority score based on multi-dimensional indexes, and the specific formula is:

[0088]

[0089] wherein, , , are weight coefficients, is the severity of the illness, which is generated by the demand analysis unit 21 analyzing the patient demand, is the resource availability, is the average response time; B4: generating a service execution queue in descending order of score. The input adaptation unit 24 is used for converting patient unstructured input into standardized microservice calling parameters.

[0090] It is necessary to point out that the demand analysis unit 21 analyzes patient demand characteristics, the service combination unit 22 dynamically generates an optimal microservice chain based on the analysis result, the priority sorting unit 23 adjusts the execution order according to the resource state and medical emergency, and the input adaptation unit 24 realizes the standardized conversion of unstructured data.

[0091] Further, it is necessary to point out that the service path scoring formula is modified as:

[0092]

[0093] wherein, is the business importance weight of the ith microservice (medical specific parameter); is a vector containing response time, availability, compliance score; is the weight coefficient of the jth resource (CPU / memory / bandwidth); is the resource occupancy rate. The text input of the input adaptation unit 24 is: using the "BERT pre-training model" for intent recognition, the patient inputs "open antihypertensive drugs", and it is parsed as "service type = prescription issuance, drug type = antihypertensive drugs, urgency = routine"; the voice input: through the "medical field ASR model" conversion, the recognition rate of medical terms such as "chest pain, shortness of breath" is optimized (compared with the general ASR, the accuracy rate is improved by 15%), and the "urgency label" is automatically supplemented; image input (such as a handwritten prescription photo): call "OCR + prescription recognition model" to extract drug name, dose, physician signature, etc. Information is converted into "prescription upload + manual review" microservice call parameters (automatically marked "compliance verification required").

[0094] In the present embodiment, it also needs to be explained that the cross-domain collaborative bus module 3 includes a service registration and discovery unit 31, a data integration unit 32, a collaborative scheduling unit 33, and a visualization unit 34, wherein: the service registration and discovery unit 31 is used to manage the registration information of each microservice, and supports automatic discovery and calling of cross-domain services; the data integration unit 32 is used to integrate microservice data of the drug supply chain, diagnosis and treatment, and distribution links, and eliminate information silos; the collaborative scheduling unit 33 is used to coordinate the execution timing of cross-link microservices through the bus; the visualization unit 34 is used to display the end-to-end process status in a graphical interface in real time.

[0095] It needs to be explained that the service registration and discovery unit 31 realizes unified management and intelligent discovery of microservices, the data integration unit 32 breaks down the data barriers of the medical and pharmaceutical whole chain, the collaborative scheduling unit 33 optimizes the execution timing of cross-domain services, and the visualization unit 34 provides whole process state monitoring.

[0096] Further, it needs to be explained that the service registration and discovery unit 31 builds a "medical and pharmaceutical industry data platform" to solve the problem of "same object different name" data silos: drug data standardization: mapping the "commodity code" of the supply chain, the "drug generic name" of the diagnosis and treatment system, and the "package specification code" of the distribution system to the unified "national drug standard code", to ensure that "aspirin" is uniquely identified in the whole chain; diagnosis and treatment data association: through the "patient unique identifier (such as the electronic certificate number of medical insurance)", the inquiry record, prescription information, and distribution track are connected to form a complete data chain of "patient-diagnosis-drug-logistics", supporting cross-domain query of "patient's drug purchase record within 3 months" (access permission needs to be verified by the permission management unit 53).

[0097] In this embodiment, it also needs to be explained that the dynamic optimization control module 4 comprises a data acquisition unit 41, a state monitoring unit 42, an algorithm decision unit 43, and an execution feedback unit 44, wherein: the data acquisition unit 41 is used for acquiring the running data of each microservice in real time; the state monitoring unit 42 is used for monitoring the running state of the process node based on multi-dimensional indexes; the algorithm decision unit 43 is used for analyzing the monitoring data through a reinforcement learning algorithm to generate a resource adjustment strategy; and the specific operation is as follows: C1: state space construction: defining the state set of reinforcement learning For:

[0098]

[0099] Among them, is a resource state vector, including the physician load rate, the drug inventory balance, and the real-time position of the delivery vehicle, is a patient demand vector, including the disease emergency degree in B3 , the service timeliness requirement, is a compliance state index, such as the prescription audit completion rate; C2: action space definition: setting an executable resource adjustment action set , including: I, physician resource scheduling, such as temporarily deploying physicians from low-load departments to high-load departments; II, drug inventory allocation, such as urgently replenishing medicines from regional warehouses to short-of-supply pharmacies; III, delivery path optimization, such as switching the direct delivery route for high-priority orders; C3: reward function design: constructing a reward function considering efficiency, safety, and compliance :

[0100]

[0101] Among them, is an efficiency reward, which is positively correlated with the time reduction of the process after resource adjustment; is a safety reward, which is positively correlated with the satisfaction rate of high-priority patient demand, and the weight of critical patients is doubled; is a compliance reward, which is positively correlated with the compliance index of prescription audit and privacy protection; , and when the compliance index is not up to standard, , is a penalty coefficient; C4: strategy learning and execution: ①using the DQN algorithm, pre-training the model through offline historical data to obtain an initial strategy ; ②in online operation, according to the real-time state , the optimal adjustment action is output through the strategy ; ③after executing the action, the reward value of the environmental feedback is collected , for model parameter iterative optimization, so that the strategy gradually converges to the "resource optimal configuration under compliance premise"; C5: policy output: the optimal action after convergence into executable resource adjustment instructions such as "allocate 5 boxes of a certain targeted drug to pharmacy A", and issue them to the corresponding microservices. Execution feedback unit 44: for issuing adjustment strategies to corresponding microservices and tracking optimization effects to form a closed loop.

[0102] It should be noted that the data acquisition unit 41 acquires microservice running data in real time, the state monitoring unit 42 performs multi-dimensional health assessment, the algorithm decision unit 43 generates the optimal resource allocation strategy based on reinforcement learning (including state space modeling, action space definition, composite reward function design, and DQN algorithm implementation), and the execution feedback unit 44 implements the strategy and forms a closed loop optimization.

[0103] Further, it should be noted that a "three-level health assessment model" is constructed in the state monitoring unit 42, covering efficiency, safety, and compliance dimensions: ① primary indicators (efficiency layer): service response timeliness: if the first reply delay of the inquiry microservice is greater than 5 minutes, an early warning is triggered; resource utilization rate: if the physician load rate is greater than 80%, it is marked as "high load" and automatically pushed to the algorithm decision unit 43; ② secondary indicators (safety layer): drug inventory early warning: if the inventory of emergency drugs is less than 3 days of use, an "orange warning" (priority allocation) is triggered; distribution trajectory deviation: if the actual distribution route deviates from the planned route by more than 2 kilometers, it is automatically checked whether it is "detouring" or "abnormal stopping"; ③ tertiary indicators (compliance layer): prescription review coverage rate: if the proportion of unreviewed prescriptions is greater than 1%, the out-of-stock permission of the corresponding pharmacy is immediately frozen; data desensitization rate: if the desensitization failure rate of sensitive information such as patient name and medical record is greater than 0, a "red warning" is triggered and synchronized to the safety compliance module 5. The "graded response mechanism" of the warning: ① mild warning (such as "certain pharmacy inventory is slightly low"): only notify the operation and maintenance personnel of the corresponding microservice to supplement through automatic allocation of the system; ② moderate warning (such as "cardiovascular physician load rate > 90%"): trigger the "resource scheduling strategy generation" of the algorithm decision unit 43, and push it to the medical operation team; ③ severe warning (such as "prescription non-compliance rate suddenly rises to 5%"): suspend the relevant service chain, and the execution feedback unit 44 automatically sends a "service adjustment notice" to the user end (such as "current prescription review delay, expected to resume in 10 minutes").

[0104] In this embodiment, it also needs to be explained that the security compliance middleware module 5 includes a data encryption unit 51, a compliance verification unit 52, a permission management unit 53, and a risk interception unit 54, wherein: the data encryption unit 51 is used for encrypting the transmission and storage of medical privacy data; the compliance verification unit 52 is used for embedding medical industry regulations and rules to verify the compliance of the process in real time; the permission management unit 53 is used for assigning microservice access permissions based on roles and limiting the scope of data operations; and the risk interception unit 54 is used for monitoring abnormal interactions and triggering system-level risk warning and interception.

[0105] It needs to be explained that the data encryption unit 51 ensures the safety of medical data throughout the life cycle, the compliance verification unit 52 verifies the legality of the business process in real time, the permission management unit 53 implements fine-grained access control, and the risk interception unit 54 actively defends against abnormal behavior.

[0106] Further, it needs to be explained that the verification process includes: pre-verification: before generating a prescription, verify whether the attending physician has the corresponding prescription right for the drug (e.g., a psychiatrist cannot prescribe narcotic drugs); in-process verification: when the drug is delivered, real-time check whether the delivery address is consistent with the address of the prescription visit (to prevent the transfer of prescription drugs in different places); and post-verification: a compliance audit task is performed every morning, 10% of the orders are randomly selected, and the data encryption integrity and permission access compliance are verified (e.g., if a pharmacist is found to have accessed patient records without authorization, the risk interception unit 54 is automatically associated to trace the responsibility). The risk interception unit 54 has multiple dimensions of abnormal behavior identification: ① static identification based on rules: intercepting obvious abnormal operations such as “10 failed login attempts within 1 hour from the same IP” and “exporting more than 100 patient data at a time”; and ② dynamic identification based on AI: identifying hidden abnormalities through a “user behavior baseline model”: the normal operation habit of a pharmacist is “viewing a prescription → auditing → confirming” (with an average time of 3 minutes), if a pharmacist suddenly audits 20 prescriptions within 10 seconds, it is marked as “suspected bulk review” and intercepted; the normal trajectory of a delivery person is “pharmacy → straight line to the delivery address”, if a delivery person frequently “takes a detour to non-delivery areas”, it triggers an “abnormal drug flow” warning. The risk disposal has a “graded response and traceability”: ① mild risk (e.g., “incomplete permission application materials”): automatically rejected and prompted to correct the materials; ② moderate risk (e.g., “suspected omission in prescription auditing”): suspend the current operation and push it to the “two-person review queue” (another experienced pharmacist is required to review it again); and ③ severe risk (e.g., “hacker attempts to download patient records in bulk”): immediately cut off the network connection of the related microservice, freeze the account involved, and synchronize the logs (including attack IP, operation trajectory, and encrypted evidence) to the network security department.

[0107] Embodiment 2:

[0108] As Figure 2As shown, in this embodiment, an online medical process management method, specifically, includes the following steps:

[0109] S1. Microservice deconstruction phase

[0110] 1. Business link modularization split:

[0111] According to the core business domains of "consultation, prescription issuance, drug supply chain, logistics distribution, compliance verification", it is disassembled into atomized microservices (such as "physician matching microservice", "prescription review microservice", "drug inventory query microservice");

[0112] At the same time, redundant microservices (such as backup distribution node microservice, emergency physician resource microservice) are pre-installed to improve system fault tolerance (such as automatically switching to a backup node when the main distribution node fails);

[0113] 2. Service boundary and interaction specification definition:

[0114] The functional scope (such as "prescription generation microservice" is only responsible for generating electronic prescriptions according to diagnostic results, and does not participate in drug inventory queries), input and output parameters (input "patient ID + diagnosis result", output "prescription number + drug list") and performance indicators (such as response time ≤ 3 seconds) of each microservice are clear;

[0115] Develop a unified interface protocol: high-frequency interactions (such as order status synchronization) use gRPC protocol, cross-domain low-frequency interactions (such as cross-region inventory queries) use HTTP / 2 protocol, all interfaces default to enable TLS 1.3 encryption, and medical sensitive data must carry an encrypted identification header (X-Encrypt-Type:SM4);

[0116] S2: Intelligent service matching phase

[0117] 1. Multi-modal demand analysis:

[0118] Receive unstructured patient input (text, voice, image) and convert it to standardized demand features through AI algorithms:

[0119] I. Text input (such as "open a depressant"): use the BERT model to identify intent and output "service type = prescription issuance, drug type = depressant, urgency = routine";

[0120] II. Voice input (such as "chest pain, difficulty breathing"): convert text using a medical ASR model, extract keywords from a knowledge graph, and label "urgency = 9 (on a 10-point scale), suspected acute chest pain";

[0121] III. Image input (such as a handwritten prescription photo): extract drug information through OCR + prescription recognition model and automatically label "compliance verification required";

[0122] 2. Microservice chain dynamic combination:

[0123] ①Screening candidate services: match microservices based on demand characteristics (such as emergency demand matching "emergency physician microservice + emergency medicine allocation microservice");

[0124] ②Compatibility check: exclude services with interface conflicts (such as "pharmacy not connected to prescription audit microservice cannot be combined with prescription generation microservice") or compliance conflicts (such as physician microservice without anesthesia prescription rights cannot be combined with anesthesia drug distribution microservice);

[0125] ③Optimal path selection: according to "service response time, resource consumption cost, historical execution success rate, business importance weight" comprehensive score, specific formula:

[0126]

[0127] Where, is the business importance weight of the ith microservice (medical specific parameter); is the vector containing response time, availability, and compliance score; is the weight coefficient of the jth resource (CPU / memory / bandwidth); is the resource occupancy rate; if the scores are similar, prefer the path with standby distribution nodes;

[0128] 3. Service priority ranking:

[0129] ①Calculate priority score based on adaptive weight model:

[0130]

[0131] Where, , , are weight coefficients, is the severity of the disease, generated by the demand analysis unit 21, is the resource availability, is the average response time;

[0132] ②Dynamic adjustment of weight template: emergency prescription increases logistics timeliness weight ( ), chronic disease prescription increases resource cost weight ( ), generate execution queue in descending order of score;

[0133] S3: Cross-domain collaborative execution phase

[0134] 1. Service registration and intelligent discovery:

[0135] ①Microservices online in the registration center before the completion of the qualification record (such as pharmacy microservices need to upload "drug operation license"), run through the service gateway to automatically discover available nodes (prefer to choose health degree > 90 points, network delay < 50 ms node);

[0136] ②Support cross-domain service call (such as Beijing inquiry service call Shanghai drug inventory service), through the gateway to hide the underlying IP, enhance security;

[0137] 2. Multi-domain data fusion and collaborative scheduling:

[0138] Building a data center in the pharmaceutical industry, unified data standards:

[0139] I, drug data: mapping supply chain "commodity code" and diagnosis and treatment system "generic name" to national drug standard code (such as "aspirin" full link unique identification);

[0140] II, patient data: through the medical insurance electronic certificate number to link the inquiry record, prescription, distribution track, form a complete data chain (access needs to be verified by permission);

[0141] Based on the directed acyclic graph to coordinate the service sequence: such as "prescription generation → prescription review → drug out of warehouse → distribution" dependence, ensure that the previous service is completed before triggering the subsequent service (such as prescription not passed, drug out of warehouse service does not execute);

[0142] 3. Full process visualization monitoring:

[0143] Graphical interface to display microservice running state (node color marking: green = normal, yellow = warning, red = fault), order flow track (such as "inquiry → prescription → A pharmacy → delivery man Zhang San") and key indicators (real-time order quantity, prescription compliance rate, resource utilization rate), support abnormal warning (such as distribution delay rate > 10% automatically pop-up);

[0144] S4: dynamic optimization control phase

[0145] 1. Multi-dimensional data collection and state evaluation:

[0146] ① Real-time collection of microservice running data:

[0147] I, resource status (physician load rate, drug inventory, delivery vehicle location);

[0148] II, patient demand (urgency of illness, service time requirement);

[0149] III, compliance indicators (prescription review completion rate, privacy data desensitization rate);

[0150] ② Based on "three-level health model" evaluation: efficiency layer (e.g. physician load rate > 80% triggers early warning), safety layer (e.g. emergency drug inventory < 3 days of use triggers orange warning), compliance layer (e.g. prescription non-compliance rate > 1% triggers red warning);

[0151] 2. Reinforcement learning decision and resource adjustment:

[0152] ① State space construction:

[0153]

[0154] Where, is the resource state vector, including physician load rate, drug inventory, and real-time location of delivery vehicles, is the patient demand vector, including the severity of illness in B3 , service time requirement, is the compliance status indicator;

[0155] ② Action space definition: including physician scheduling (allocating low-load departments to high-load departments), drug allocation (replenishing short supply pharmacies from regional warehouses), and delivery route optimization (switching high-priority orders to direct routes);

[0156] ③ Reward function design:

[0157]

[0158] Where, is the efficiency reward, positively related to the reduction in process time after resource adjustment; is the safety reward, positively related to the satisfaction rate of high-priority patient demand, with the weight of critical patients doubled; is the compliance reward, positively related to the compliance indicators of prescription review and privacy protection, and , is the penalty coefficient;

[0159] ④ Strategy learning and execution: pre-training the model using DQN algorithm, and outputting the optimal action (e.g. "allocate 5 boxes of thrombolytic drugs to B pharmacy") based on real-time state during online operation. After execution, collect reward values to iteratively optimize the model;

[0160] 3. Closed-loop feedback and policy iteration: track the effects of resource adjustment (e.g. whether drug allocation meets inventory requirements, whether physician scheduling reduces load rate), short-term feedback is used to correct the current strategy (e.g. secondary allocation), and long-term data is used for model retraining (e.g. update DQN parameters every week to adapt to seasonal or policy changes);

[0161] S5: Safety and compliance guarantee phase

[0162] 1. Data encryption throughout the life cycle:

[0163] ①Transmission encryption: medical privacy data (medical records, prescriptions) are encrypted using SM4 algorithm, and non-sensitive data are encrypted using TLS 1.3;

[0164] ②Storage encryption: homomorphic encryption technology is used to support data analysis in encrypted state (such as statistical disease incidence);

[0165] ③Key management: root key is stored offline in HSM, data key is automatically rotated regularly, and emergency recovery requires multi-signature approval;

[0166] 2. Full-process compliance verification:

[0167] ①Pre-verification: check the qualifications of doctors before generating prescriptions (e.g. psychiatrists cannot prescribe narcotic drugs);

[0168] ②In-process verification: check the consistency of "delivery address and prescription visit address" during drug delivery to prevent the transfer of prescription drugs to other places;

[0169] ③Post-verification: randomly check 10% of orders every morning, audit data encryption integrity and access compliance (e.g. if a pharmacist exceeds his authority to view medical records, trace the responsibility);

[0170] 3. Abnormal risk identification and interception:

[0171] ①Static rule interception: such as "10 failed login attempts within 1 hour from the same IP" and "exporting more than 100 patient data at a time";

[0172] ②AI dynamic identification: identify hidden abnormalities through user behavior baseline model (e.g. a pharmacist reviews 20 prescriptions in 10 seconds, suspected bulk review);

[0173] ③Hierarchical disposal: automatically reject for light risk, trigger double review for moderate risk, and cut off connection and report to relevant departments for severe risk.

[0174] In the description of the present specification, the description of the reference terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0175] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments described. Obviously, many modifications and variations can be made in light of the teachings above. The description is chosen and described in order to best explain the principles of the application and its practical application to thereby enable others skilled in the art to best utilize the application and get the best results from the application. The application is only limited by the claims and their full scope and equivalents.

Claims

1. An online medical procedure management method, characterized by, Comprise the following steps: S1: the online medical process is disassembled into independent industrial microservices, and the functions are highly cohesive and low-coupled; S2: Through AI algorithm analysis of patient demand characteristics and real-time resource state, dynamically generate microservice combination scheme, and optimize service priority based on adaptive weight model, for unstructured patient input, through interface adaptation layer conversion to standardized microservice call parameters; S3: Based on the microservice bus technology, a cross-domain collaboration layer is constructed, integrating the microservices of drug supply chain, diagnosis and treatment services, and logistics distribution, and performing end-to-end visual collaboration; S4: Based on the real-time data flow between microservices, a "scene awareness-intelligent decision-making-execution feedback" closed-loop network is constructed, and the state of the link is dynamically monitored through multi-dimensional indicators, and the resource adjustment strategy is generated by combining the reinforcement learning algorithm, and the service resources are adaptively optimized and configured; S5: Through the microservice security middleware embedded in each microservice interaction node, the medical data compliance and system-level risk prevention and control are performed.

2. A process management system for online medicine according to the process management method for online medicine of claim 1, characterized by, It comprises a microservice decoupling module (1), an intelligent matching engine module (2), a cross-domain collaboration bus module (3), a dynamic optimization control module (4), and a security compliance middleware module (5), wherein: The microservice decoupling module (1) is used to disassemble the online medical process into independent microservice units with high cohesion and low coupling, and define the function boundaries and interaction standards of each service; The intelligent matching engine module (2) dynamically combines microservices based on AI algorithms, optimizes service priority, and adapts the standardized conversion of unstructured patient input; The cross-domain collaboration bus module (3) is used to integrate drug supply chain, diagnosis and treatment services, and logistics distribution microservices through microservice bus technology, and perform end-to-end visual collaboration management; The dynamic optimization control module (4) monitors the process state and adaptively adjusts resource configuration based on real-time data flow and reinforcement learning algorithm of microservices; The security compliance middleware module (5) is used to embed microservice interaction nodes to perform medical data compliance and system-level risk prevention and control.

3. The system according to claim 2, wherein, The microservice decoupling module (1) comprises a process disassembly unit (11), a service definition unit (12), and an interface standardization unit (13), wherein: The process disassembly unit (11) is used to split the process into independent microservices according to the online medical business links, and contains redundant microservices for improving system fault tolerance, such as backup distribution node microservices and emergency physician resource microservices; The service definition unit (12) is used to clearly define the function scope, input and output parameters, and performance indicators of each microservice, wherein the redundant microservices need to additionally define the trigger activation conditions, such as automatically activating the backup distribution node microservice when the main distribution node fails; The interface standardization unit (13) is used to formulate the interface specification and communication protocol for data interaction between microservices.

4. The system according to claim 2, wherein, The intelligent matching engine module (2) comprises a demand analysis unit (21), a service combination unit (22), a priority sorting unit (23), and an input adaptation unit (24), wherein: The demand analysis unit (21) is used to analyze patient demand characteristics and real-time resource state through AI algorithm; The service combination unit (22) is configured to dynamically combine the adaptive microservice chain according to the demand, and generate an optimal service path. The priority sorting unit (23) is configured to sort the service execution order based on an adaptive weight model. The input adaptation unit (24) is configured to convert the patient unstructured input into standardized microservice call parameters.

5. The system according to claim 4, wherein, The service combination unit (22) dynamically combines the adaptive microservice chain according to the demand, and generates an optimal service path, which specifically operates as follows: A1: Based on the patient demand characteristics output by the demand analysis unit (21), a candidate microservice set that meets the functional demand is selected, such as the "physician matching microservice" and "prescription generation microservice" corresponding to the inquiry demand; A2: The candidate microservice set is subjected to compatibility verification, and microservices with interface conflicts or compliance conflicts are removed, such as prohibiting the combination of unqualified physician microservices and prescription microservices; A3: Based on the preset evaluation indexes including service response time, resource consumption cost, and historical execution success rate, the compatible microservice chain is scored, and the chain with the highest comprehensive score is selected as the optimal service path; A4: If there are multiple paths with similar scores, the path with the standby distribution node microservice is preferentially selected.

6. The system according to claim 4, wherein, The priority sorting unit (23) sorts the service execution order based on an adaptive weight model, which specifically operates as follows: B1: Real-time acquisition of resource state indexes of each microservice, including the number of online doctors, drug inventory level, and logistics distribution timeliness; B2: Dynamically loading a weight configuration template according to the patient prescription type, wherein the logistics timeliness weight is preferentially configured for emergency prescriptions; B3: Calculating a comprehensive priority score based on multi-dimensional indexes, and the specific formula is: wherein, , , are weight coefficients, respectively, is the urgency of the condition, generated by the demand analysis unit (21) from the patient demand, is the resource availability, is the average response time; B4: Generating a service execution queue in descending order of the score.

7. The system according to claim 2, wherein, The cross-domain collaborative bus module (3) includes a service registration and discovery unit (31), a data integration unit (32), a collaborative scheduling unit (33), and a visualization unit (34), wherein: The service registration and discovery unit (31) is configured to manage the registration information of each microservice, and support automatic discovery and calling of cross-domain services; The data integration unit (32) is configured to integrate microservice data of the drug supply chain, diagnosis and treatment, and distribution links, and eliminate information silos; The collaborative scheduling unit (33) is configured to coordinate the execution timing of cross-link microservices through the bus; The visualization unit (34) is configured to display the end-to-end process state in a graphical interface in real time.

8. The system according to claim 2, wherein, The dynamic optimization control module (4) includes a data acquisition unit (41), a state monitoring unit (42), an algorithm decision unit (43), and an execution feedback unit (44), wherein: The data acquisition unit (41) is configured to acquire the running data of each microservice in real time; The state monitoring unit (42) is configured to monitor the running state of the process node based on multi-dimensional indexes; The algorithm decision unit (43) is configured to analyze the monitoring data through a reinforcement learning algorithm, and generate a resource adjustment strategy; The execution feedback unit (44) is configured to issue the adjustment strategy to the corresponding microservice, and track the optimization effect to form a closed loop.

9. The system according to claim 8, wherein, The algorithm decision unit (43) analyzes the monitoring data by a reinforcement learning algorithm to generate a resource adjustment strategy, and the specific operations are as follows: C1: State space construction: Defining a set of states for reinforcement learning is: wherein, is a resource status vector, containing physician load rate, drug inventory level, delivery vehicle real-time location, is a patient demand vector, containing the urgency of illness in B3 , service time requirement, is a compliance status indicator, such as prescription review completion rate; C2: Action space definition: Setting a set of executable resource adjustment actions , comprising: I. Physician resource scheduling, such as temporary allocation of physicians from low-load departments to high-load departments; II. Drug inventory allocation, such as emergency replenishment from regional warehouses to short supply pharmacies; III. Distribution path optimization, such as switching direct distribution routes for high-priority orders; C3: Reward function design: Constructing a reward function that balances efficiency, safety, and compliance : Wherein: is an efficiency reward, positively related to the time reduction of the process after resource adjustment; is a safety reward, positively related to the satisfaction rate of high-priority patient demand, with the weight of critical patients doubled; is a compliance reward, positively related to the compliance index of prescription review and privacy protection; , and when the compliance index is not up to standard, , is a penalty coefficient; C4: Strategy learning and execution: ①Using DQN algorithm, pre-training model through offline historical data, get the initial strategy ; ② When online, output optimal adjustment action according to real-time state through strategy ;​ ③ After performing the action, collect the reward value of the environmental feedback , for iterative optimization of model parameters, so that the strategy gradually converges to "optimal allocation of resources under compliance premise"; C5: Strategy output: the converged optimal action converted into executable resource adjustment instructions, such as "allocate 5 boxes of a certain targeted drug to A pharmacy", and issued to the corresponding microservice.

10. The online medical procedure management system of claim 2, wherein, The security compliance middleware module (5) includes a data encryption unit (51), a compliance verification unit (52), a permission management unit (53), and a risk interception unit (54), wherein: The data encryption unit (51) is used for encrypting the transmission and storage of medical privacy data; The compliance verification unit (52) is used for embedding medical industry regulations and rules to verify the compliance of the process in real time; The permission management unit (53) assigns microservice access permissions based on roles to limit the scope of data operations; The risk interception unit (54) is used for monitoring abnormal interactions and triggering system-level risk warning and interception.

Images