An ai-based standard management method and system for traditional chinese medicine prescription chain diagnosis and treatment

Abstract

The application discloses an AI-based standardized management method and system for traditional Chinese medicine (TCM) prescription chain diagnosis and treatment, and belongs to the technical field of artificial intelligence and TCM informatization technology. The application builds a cloud-based standardized prescription knowledge base, and internally builds a priority three-fold drug use risk checking rule. A store collects multi-modal diagnosis and treatment data and uploads the data to the cloud, and AI completes auxiliary syndrome differentiation and prescription recommendation. A prescription can be issued only after being reviewed on site by a certified physician and after identity verification. The cloud realizes full-chain grading quality control and full-process traceability, and the efficacy data backflow drives AI model incremental iteration to form a self-optimizing closed loop. The application can unify the diagnosis and treatment standards of prescriptions, reduce the risk of drug use, improve the efficiency of quality control, and support the scaled, standardized and intelligent operation of TCM prescription chains.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, big data informatization, and clinical diagnosis and treatment management of TCM classic prescriptions. It is particularly applicable to the clustered operation scenario of multi-store TCM chain institutions. It can realize TCM classic prescription-assisted diagnosis in chain stores, intelligent prescription recommendation, on-site prescription review in stores, headquarters-wide quality control, digital accumulation of famous doctors' classic prescription experience, and closed-loop iteration of AI models. Specifically, it is an AI-based standardized management method and system for TCM classic prescription chain diagnosis and treatment. Background Technology

[0002] While TCM chain institutions are rapidly expanding in scale, the industry's standardization of diagnosis and treatment and risk management capabilities are severely lagging behind. Existing technologies suffer from four core deficiencies: First, the standards for classical prescription diagnosis and treatment are inconsistent. The skill levels of physicians vary across branches, and there is a lack of unified digital execution standards for the compatibility, dosage, and applicable populations of classic prescriptions. The experience of renowned doctors is difficult to replicate in a standardized manner, leading to significant differences in treatment quality between branches. Second, medication risk control is weak and singular. Risk management of classical prescription medications relies heavily on manual control by physicians, lacking automated pre-verification mechanisms. Furthermore, existing verification tools are mostly parallel, general-purpose rule-based audits, lacking priority risk control logic specific to classical prescription scenarios. This results in a high risk of missed detection of high-risk contraindications, highlighting significant medication safety hazards. Third, the management capabilities of chain headquarters are insufficient. The dispersed operation of branches makes it impossible for headquarters to collect real-time data on the entire treatment process, making it difficult to trace treatment behavior and resulting in high post-treatment quality control costs. Moreover, high-risk prescriptions only use a single early warning mode without a tiered management mechanism, leading to low management efficiency and accuracy. Fourth, the AI ​​diagnosis and treatment models are static and fixed. Most existing TCM AI syndrome differentiation models are trained offline and then put online. Clinical efficacy data cannot drive model iteration, and the model accuracy cannot be continuously improved with the accumulation of clinical data. Existing general TCM AI systems are mostly geared towards single stores or internet diagnosis and treatment scenarios, and are disconnected from the operation and management system of chain stores. There is currently no integrated solution for the entire process that deeply integrates the five major links of standardized knowledge base, AI-assisted syndrome differentiation, offline mandatory prescription review, full-domain quality control and traceability, and efficacy-driven model iteration for the classical prescription chain scenario. Summary of the Invention

[0003] 1. Technical problems to be solved To address the aforementioned shortcomings of existing technologies, this invention provides an AI-based standardized management method and system for chain TCM formula diagnosis and treatment. It aims to resolve industry pain points such as inconsistent standards for formula diagnosis and treatment in chain institutions, lack of proactive medication risk control, difficulties in headquarters control and traceability, static AI model optimization limitations, and fragmented functions of existing systems. The invention unifies the standards for formula diagnosis and treatment in chain institutions, proactively controls medication risks, enables headquarters-wide quality control and full-process traceability, and constructs a self-optimizing closed loop for efficacy-driven AI models, supporting the large-scale, standardized, and intelligent operation of TCM formula chain institutions. Technical solution

[0004] 2.1 A Standardized Management Method for Chain Treatment of Traditional Chinese Medicine Formulas Based on AI The method includes the following steps: S1. A standardized classical formula knowledge base is built in the cloud, which uniformly encodes and stores symptom terms, TCM syndrome types, classic formulas, compatibility rules, dosage ranges, contraindications, and standardized treatment pathways. The knowledge base has built-in three-layer verification rules for compatibility, dosage, and contraindications, and supports unified updates from headquarters and real-time synchronization to all store terminals. S2. The store terminal collects multimodal diagnostic and treatment data such as patient identity information, medical history, physical signs, tongue appearance, and pulse appearance in a structured manner, and uploads them to the cloud database through an encrypted channel; S3: The cloud calls a multimodal AI diagnostic model to integrate multi-dimensional patient data to complete the auxiliary diagnostic classification and match the corresponding classical prescription treatment path in the knowledge base; S4. The system automatically recommends suitable classical prescriptions and compliant addition / subtraction schemes based on the built-in verification rules in the knowledge base. It performs three-fold medication compliance verification in sequence according to preset priority, including special population contraindication verification, medicinal material dosage threshold verification, and classical prescription compatibility contraindication verification. If any step fails, the subsequent verification process is terminated and the corresponding level of risk labeling is directly triggered. When abnormal items are identified during verification, the risk content is highlighted on the store terminal and adjustment suggestions are pushed. S5. The verified recommended treatment plan and complete medical records are sent to the store terminal for on-site review by licensed physicians. Physicians can adjust the recommended prescriptions and issue electronic prescriptions after identity verification. Valid prescriptions cannot be generated without physician review and confirmation. S6: Real-time cloud-based collection of data from the entire chain of stores' diagnosis and treatment processes; automatic triggering of graded warnings for high-risk prescriptions based on risk level; binding of store number, operator ID, and timestamp to the entire process operation and encrypting and storing it; support for headquarters to conduct multi-dimensional traceability queries and export of quality control reports by store, doctor, and time period; high-risk prescriptions are divided into three levels according to risk level, with different levels corresponding to different warning intensities and review requirements. S7. Collect patient follow-up efficacy data in the cloud, screen the collected follow-up efficacy data for effectiveness, and only include valid samples that have completed the entire treatment cycle and have complete follow-up records. Generate standardized training samples, and periodically use incremental methods to iteratively optimize the AI ​​diagnosis model to form a self-optimizing closed loop of diagnosis and treatment.

[0005] Furthermore, the model iteration adopts an incremental training mechanism and is automatically updated monthly / quarterly; the entire chain of diagnosis and treatment data is encrypted and transmitted and stored using national cryptographic algorithms; data access adopts a hierarchical permission control mechanism, with data access scopes isolated between headquarters and stores.

[0006] 2.2 A Standardized Management System for Chain Treatment of Traditional Chinese Medicine Formulas Based on AI This system is used to implement the aforementioned management methods. It connects a cloud-based central control platform with several offline store terminals via encrypted network communication. Each functional unit interacts and operates in tandem via a cloud data bus. Five functional units are deployed on the cloud-based central control platform, and two functional units are distributed across the various store terminals. Specifically, these include: (1) Cloud-based Classical Formula Knowledge Base: Responsible for the encoding and storage of all elements of classic classical formulas, the operation of the verification rule engine, the unified updates by headquarters and the synchronous distribution to all stores; (2) Store information collection terminal: Distributed deployment in each store terminal to complete the standardized collection and encrypted uploading of patients' multimodal four diagnostic methods and medical record information; (3) AI Classical Formula Diagnosis and Reasoning Module: Connects to the cloud knowledge base, integrates multimodal patient data to complete reasoning calculations, assist in syndrome differentiation and classification, and match classical formula treatment paths; (4) Classical prescription recommended medication verification module: Link the syndrome differentiation results to adapt to the classical prescription plan, automatically complete the triple medication risk verification according to the preset priority and mark the risk warning; (5) On-site review terminal for store physicians: Deployed on the local terminal of the store, it allows licensed physicians to review prescriptions, adjust contents, verify identity and issue / reject electronic prescriptions on-site; (6) Chain Cloud Quality Control Platform: Collects diagnosis and treatment data from all stores, realizes high-risk prescription classification and early warning, encrypted storage of full-process operation logs, multi-dimensional traceability query and quality control report export; (7) Efficacy feedback and model iteration module: Summarize clinical efficacy follow-up data from stores, generate a standardized training sample set after effectiveness screening, and incrementally update the weight parameters of the AI ​​diagnostic model.

[0007] Preferably, the system sets hierarchical data permissions, isolating data access between headquarters and stores; the entire chain uses national cryptographic algorithms for encrypted transmission and storage; the model adopts an incremental training mechanism and is automatically updated monthly / quarterly to reduce computing power consumption. Beneficial effects

[0008] (1) Achieve standardization and unification of classical prescription diagnosis and treatment: Through the cloud-based standardized classical prescription knowledge base, unify the syndrome types, classical prescriptions, compatibility, dosage and contraindication rules of the entire chain, transform the experience of famous doctors into replicable digital execution standards, and fundamentally solve the problem of uneven diagnosis and treatment levels among stores; (2) Construct an efficient two-tier medication risk control system: The system implements a three-level automatic verification based on priority to achieve pre-control, and combines it with the mandatory final review by licensed physicians offline to form a dual defense line of "system initial screening + manual backup"; the priority verification mechanism can prioritize the interception of the most dangerous contraindicated drugs, significantly improve the efficiency of risk interception, and greatly reduce the safety risks of clinical medication of classical prescriptions. (3) Achieve refined quality control across the entire chain: The entire process of diagnosis and treatment data is collected and recorded in the cloud, supporting multi-dimensional traceability and query; at the same time, a high-risk prescription graded early warning mechanism is adopted, and corresponding control processes are matched for different risk levels. Compared with the single early warning mode, it significantly improves the efficiency of headquarters quality control and reduces the cost of chain operation and management. (4) Forming a self-optimization closed loop in diagnosis and treatment: After the efficacy follow-up data is automatically collected and screened for effectiveness, the AI ​​diagnostic model is continuously optimized through incremental training. The model accuracy is dynamically improved with the accumulation of clinical data, which solves the industry pain point that traditional static AI models cannot continuously evolve. (5) Integrated architecture reduces implementation costs: It is customized for the traditional Chinese medicine chain store scenario. The knowledge base, diagnosis, verification, prescription review, quality control and iteration are integrated in one process. There is no need to connect to multiple heterogeneous third-party systems. The deployment and maintenance costs are low and it is suitable for the implementation needs of small and medium-sized chain stores. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the overall workflow of the present invention.

[0010] Attached labels: 1-Cloud-based classical formula knowledge base; 2-Store information collection terminal; 3-AI classical formula syndrome differentiation and reasoning module; 4-Classical formula recommended medication verification module; 5-Store physician on-site review terminal; 6-Chain cloud-based quality control platform; 7-Efficacy feedback and model iteration module; 8-Offline chain store cluster. Detailed Implementation

[0011] like Figure 1 As shown, the overall data flow logic of this invention is as follows: the cloud-based classical formula knowledge base 1 provides rule support for the AI ​​classical formula syndrome differentiation and reasoning module 3 and the classical formula recommended medication verification module 4; the data collected by the store collection terminal 2 is sequentially transmitted to the AI ​​classical formula syndrome differentiation and reasoning module 3, the classical formula recommended medication verification module 4, and the store physician on-site review terminal 5; the full-process data of the store collection terminal 2 and the store physician on-site review terminal 5 are synchronously uploaded to the chain cloud quality control platform 6; the efficacy data collected by the chain cloud quality control platform 6 flows back to the efficacy feedback and model iteration module 7, and the efficacy feedback and model iteration module 7 optimizes the model parameters of the AI ​​classical formula syndrome differentiation and reasoning module 3, forming a complete optimization closed loop.

[0012] The technical solution of the present invention will be further described in a complete and clear manner below with reference to the accompanying drawings and two embodiments. Those skilled in the art can completely reproduce the entire technical solution based on these embodiments.

[0013] Example 1: Routine Complete Diagnosis and Treatment Process in Chain Stores The headquarters pre-deployed and configured the system: setting hierarchical data permissions, with headquarters administrators having access to view all store data and edit the knowledge base, while store physicians can only access their own store's patient treatment data; a national cryptographic encryption channel was configured throughout the entire process to ensure data transmission and storage security. Simultaneously, classic prescriptions from *Shanghan Lun* and *Jinkui Yaolue*, along with corresponding syndrome differentiation criteria, compatibility rules, tiered dosage ranges, contraindications, and standard additions and subtractions, were entered into the cloud-based classical prescription knowledge base, completing standardized coding for all elements; PC data acquisition terminals, tongue imaging equipment, and pulse diagnosis digital acquisition terminals were deployed in each chain TCM store, and on-site physician review interfaces were configured.

[0014] When a patient visits the clinic, staff use a terminal to input the patient's identity information, chief complaint, past medical history, and allergy history in a structured format. They also take standardized tongue photos and collect digitized pulse data. After confirmation, all treatment data is uploaded to a cloud database via an encrypted channel. The cloud-based multimodal AI diagnostic model integrates textual medical records, tongue images, and pulse values ​​to perform reasoning, outputting auxiliary diagnostic classification results and matching corresponding classical prescription treatment paths. For example, if the patient is diagnosed with Taiyang stroke syndrome, the system automatically matches the Guizhi Tang treatment path and recommends the original prescription.

[0015] The prescription recommendation and medication verification module performs a three-tiered verification process based on priority: First, the system verifies whether the patient has contraindications such as pregnancy or abnormal liver and kidney function. If the verification passes, it then verifies whether the dosage of medicinal materials is within the range of the pharmacopoeia and prescription standards. Finally, it verifies whether the prescription compatibility conforms to the principles of classical prescription formulation. If the contraindication verification fails, a Level 1 high-risk warning is directly issued, and no further verification is performed. After confirming no risk, the complete medical record and recommended treatment plan are simultaneously sent to the corresponding store's review terminal.

[0016] At the in-store clinic, licensed physicians verify the patient's actual condition on-site. After confirming the recommended treatment plan is correct, they complete identity verification via facial recognition and click to confirm and issue the electronic prescription. If the physician needs to adjust the prescription based on the patient's individual condition, they can manually modify it on the terminal, and the system will automatically perform a second verification. Prescriptions with Level 1 risk are forcibly locked and require a second review by a pharmacist at headquarters before issuance; prescriptions with Level 2 risk are highlighted and require manual confirmation by the physician before issuance; prescriptions with Level 3 risk are marked with a reminder but are not forcibly blocked.

[0017] The entire process of this diagnosis and treatment is automatically synchronized to the chain's cloud-based quality control platform. The system automatically binds the store number, the operating physician ID, the precise operation time, and permanently encrypts and stores the data. Headquarters management personnel can retrieve traceability records and export quality control reports in the background by store, physician, and time period. If the prescription has high-risk situations such as exceeding the dosage threshold, the cloud platform will automatically push the corresponding level of warning notification to the headquarters administrator.

[0018] When patients return for follow-up visits, store staff enter efficacy feedback data, and the medical records and efficacy results are linked to generate training samples. When the cloud collects follow-up data every month, it automatically removes invalid samples with insufficient follow-up periods or missing medical records, and only includes valid samples that meet the standards into the monthly incremental training set. The AI ​​syndrome differentiation model weights are updated using incremental training to continuously optimize the accuracy of syndrome differentiation and classical prescription recommendations.

[0019] Example 2: Risk Prevention and Interception Scenarios in the Use of Classical Formulas When a pregnant woman with spleen and stomach deficiency-cold visits the clinic, the clinic's terminal collects her four diagnostic methods (inspection, auscultation and olfaction, palpation, and olfaction) information and uploads it to the cloud via an encrypted channel. The AI ​​diagnostic model determines that the patient has external cold and internal phlegm retention syndrome, matches the Xiao Qinglong Tang treatment path, and outputs a recommended prescription. The prescription recommendation and medication verification module performs priority verification. The first step, special population contraindication verification, determines that the prescription contains herbs contraindicated during pregnancy, directly terminating subsequent verification, marking a level one high-risk warning, and pushing appropriate alternative prescription reference suggestions.

[0020] After reviewing the risk warnings on-site, the on-duty physician at the store abandoned the original recommended treatment and manually replaced it with a mild traditional Chinese medicine formula suitable for pregnant women. After the system performed a second verification and found no risks, the prescription was issued, thus achieving proactive interception of medication risks and avoiding the risk of medication contraindicated during pregnancy.

[0021] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention; all equivalent changes and modifications made in accordance with the claims of the present invention should fall within the protection scope of the present invention.

Claims

1. A standardized management method for chain diagnosis and treatment of TCM classic prescriptions based on AI, characterized in that, Includes the following steps: S1. A standardized classical formula knowledge base is built in the cloud, which uniformly encodes and stores symptom terms, TCM syndrome types, classic formulas, compatibility rules, dosage ranges, contraindications, and standardized treatment pathways. The knowledge base has built-in three-layer verification rules for compatibility, dosage, and contraindications, and supports unified updates from headquarters and real-time synchronization to all store terminals. S2. The store terminal collects multimodal diagnostic and treatment data, including patient identity information, medical history, physical signs, tongue appearance, and pulse appearance, in a structured manner, and uploads it to the cloud database through an encrypted channel; S3. The cloud calls a multimodal AI diagnostic model to integrate multi-dimensional patient data to complete auxiliary diagnostic classification and match the corresponding classical prescription treatment path in the knowledge base; S4. Based on the built-in verification rules of the knowledge base, the system automatically recommends suitable classical prescriptions and compliant addition and subtraction schemes, and performs three-fold medication compliance verification in sequence according to preset priority, including screening for incompatibilities of classical prescriptions, verification of drug dosage thresholds, and determination of medication contraindications for special populations; when the system verifies and identifies abnormal items, it highlights the risk content through the store terminal and pushes adjustment suggestions. S5. The verified recommended treatment plan and complete medical records are sent to the store terminal for on-site review by licensed physicians. Physicians can adjust the recommended prescriptions and issue electronic prescriptions after identity verification. Valid prescriptions cannot be generated without physician review and confirmation. S6: Real-time cloud-based collection of data from the entire chain of stores' diagnosis and treatment processes; automatic triggering of graded warnings for high-risk prescriptions based on risk level; binding of store number, operator ID, and timestamp to the entire process operation and encrypted storage; support for multi-dimensional traceability query and quality control report export from headquarters. S7 collects patient follow-up efficacy data in the cloud, generates standardized training samples, and periodically iterates and optimizes the AI ​​diagnostic model to form a self-optimizing closed loop for diagnosis and treatment.

2. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S7, the model iteration adopts an incremental training mechanism, which automatically triggers updates on a monthly or quarterly basis.

3. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, The entire process of diagnosis and treatment data is encrypted and transmitted and stored using national cryptographic algorithms.

4. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, The method also includes a hierarchical access control step: data access permissions are assigned to system accounts, with headquarters administrators having the right to view all store data, export quality control reports, and edit the knowledge base; store physicians can only view the medical records and prescription data of patients treated at their own stores.

5. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S5, the identity verification is performed using facial recognition.

6. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S6, the multi-dimensional traceability query supports three dimensions: store, doctor, and time period.

7. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S4, the triple medication compliance verification is performed in sequence according to the priority of "contraindication verification for special populations - dosage threshold verification for medicinal materials - contraindication verification for classical prescriptions"; if any verification step fails, the subsequent verification is terminated and the corresponding risk label is triggered.

8. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S6, the high-risk prescriptions are divided into three levels according to risk level: Level 1 risk is the use of the drug in cases of contraindications such as pregnancy or abnormal liver and kidney function; Level 2 risk is the use of the drug in cases of exceeding the standard dosage range of the pharmacopoeia; and Level 3 risk is the use of the prescription in cases of non-compliance with the principles of classical prescription. Different risk levels correspond to different warning prompt intensities and mandatory review levels.

9. The AI-based standardized management method for chain diagnosis and treatment of TCM classic prescriptions according to claim 1, characterized in that, In step S7, before generating standardized training samples, the follow-up efficacy data is first screened for validity, and only valid samples with complete treatment cycles and complete follow-up records are included for incremental training.

10. A standardized management system for chain diagnosis and treatment of TCM formulas based on AI, used to implement the management method described in any one of claims 1 to 9, characterized in that, The system includes a cloud-based central control platform and several offline store terminals, which are connected via encrypted network communication. The system is configured with seven functional units, and each functional unit realizes data interaction and coupling linkage through the cloud data bus; among them, the cloud-based classical prescription knowledge base (1), the AI ​​classical prescription diagnosis and reasoning module (3), the classical prescription recommended medication verification module (4), the chain cloud quality control platform (6), and the efficacy feedback and model iteration module (7) are deployed on the cloud-based central control platform; the store information collection terminal (2) and the store physician on-site review terminal (5) are respectively set in each of the chain stores; The cloud-based classical prescription knowledge base (1) is responsible for the encoding and storage of all elements of classical prescriptions, the operation of the verification rule engine, the unified updates by the headquarters and the synchronous distribution to all stores; The store information collection terminal (2) completes the standardized collection of patients' multimodal four diagnostic methods and medical record information, and uploads it to the cloud control platform through an encrypted network; The AI ​​classical prescription syndrome differentiation and reasoning module (3) connects to the cloud knowledge base and integrates multimodal patient data to complete reasoning calculations, assist in syndrome differentiation and classification, and match classical prescription treatment paths; The classical prescription recommended medication verification module (4) links the syndrome differentiation results to adapt the classical prescription scheme, automatically completes the triple medication risk verification and marks the risk warning according to the preset priority; The store physician on-site review terminal (5) is used by licensed physicians to review prescriptions, adjust contents, verify identity, and issue or reject electronic prescriptions on-site. The chain cloud quality control platform (6) collects all store diagnosis and treatment data, automatically triggers graded early warning for high-risk prescriptions according to risk level, encrypts and stores the full-process operation log, and provides multi-dimensional traceability query and quality control report export. The efficacy feedback and model iteration module (7) summarizes the clinical efficacy follow-up data of the stores, generates a standardized training sample set after effectiveness screening, and iteratively updates the weight parameters of the AI ​​diagnostic model.

11. The AI-based standardized management system for chain diagnosis and treatment of TCM classic prescriptions according to claim 10, characterized in that, The therapeutic effect feedback and model iteration module (7) uses an incremental training mechanism to update the AI ​​diagnostic model, and supports automatic updates triggered monthly or quarterly.

12. The AI-based standardized management system for chain diagnosis and treatment of TCM classic prescriptions according to claim 10, characterized in that, The system is equipped with a hierarchical data access system, and the data access scope of the headquarters administrator and the store physician account are isolated from each other; the entire data chain adopts national cryptographic algorithms for encrypted transmission and partitioned storage.

13. The AI-based standardized management system for chain diagnosis and treatment of TCM classic prescriptions according to claim 10, characterized in that, The on-site physician verification terminal (5) at the store has a built-in facial recognition verification unit for verifying the physician's identity before issuing a prescription.

Images