Clinical database full-chain common intelligence system

Abstract

The invention discloses a clinical database full-chain common intelligence system, relates to the technical field of medical data, and is technically characterized in that an intelligent design platform of a queue research scheme and automatic configuration is established; the intelligent design platform comprises a knowledge base, a queue research scheme intelligent recommendation algorithm and a Qwen2.5-72B-Instruct intelligent dialogue system based on an RAGflow framework, and further comprises a data acquisition platform with a standardized data acquisition process, a data security management authorization process and a data intelligent quality control process, and the intelligent design platform covers the knowledge base, the queue research scheme intelligent recommendation algorithm and the Qwen2.5-72B-Instruct intelligent dialogue system based on the RAGflow framework. Through standardized data acquisition, safety management authorization and intelligent quality control processes, a data acquisition platform and an intelligent design platform are fused, the requirements of clinical doctors, departments, hospitals and scientific research institutions for developing high-quality queue research scheme design can be met, efficient implementation of the scheme is achieved, and the working efficiency is improved. And the scientificity of a research scheme and the accuracy, standardization, timeliness and safety of data acquisition are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data, in particular to a clinical database full-chain co-intelligence system. BACKGROUND

[0002] With the continuous deepening and expansion of biomedical research projects, the number of research projects continues to increase, and the use demand of cohort data also increases. In the construction of clinical cohort database, the design of research plan and data collection are crucial front-end links. Whether the research plan is scientific and professional, and whether the data can be accurately, timely and normatively collected will directly affect the quality and value of the cohort database, and effectively shorten the research cycle.

[0003] At present, the design of clinical cohort database plan and data collection lacks professional research design plan guidance, and lacks intelligent professional knowledge base platform and technical support platform, which is difficult to meet the needs of clinical doctors, departments, hospitals and scientific research units to carry out high-quality cohort research; Lack of professional data collection platform and management process. At the same time, for the long-term follow-up data collection needs, there is no automatic and intelligent dynamic follow-up management process. In addition, there is a lack of standardized data authority intelligent management and authorization mechanism, and the data security is difficult to guarantee. At present, the quality control of clinical cohort data mostly depends on simple system logic rule configuration and manual checking, which not only makes it difficult to ensure the accuracy of data quality control, but also consumes a lot of manpower and time cost.

[0004] Therefore, the present application aims to provide a clinical database full-chain co-intelligence system to solve the above problems. SUMMARY

[0005] The purpose of the present application is to solve the above problems, and to provide a clinical database full-chain co-intelligence system. The present application builds a professional high-quality clinical cohort data research plan knowledge base, a professional knowledge base platform and a technical platform to meet the needs of clinical doctors, departments, hospitals and scientific research units to carry out high-quality cohort research.

[0006] In order to achieve the above purpose, the technical scheme of the present application is as follows:

[0007] The present application provides a clinical database full-chain co-intelligence system, which comprises the following steps:

[0008] S1, building an intelligent design platform with automatic configuration of cohort study plan;

[0009] S2, building a data collection platform with standardized data collection process, data security management authorization process and data intelligent quality control process;

[0010] S3, the intelligent design platform includes a knowledge base, a cohort study protocol intelligent recommendation algorithm, and an intelligent dialogue agent system based on the RAGflow framework, the intelligent dialogue agent system uses the Qwen2.5-72B-Instruct model as the dialogue base, and realizes document embedding and efficient retrieval by combining the BGE embedding model. The knowledge base contains knowledge, experience and supporting data of multiple disciplines such as epidemiology, clinical medicine, medical statistics and economics. The content of the protocol is labeled with its applicable scope and application scenario after being evaluated by experts. The intelligent recommendation algorithm realizes protocol recommendation, and meets the needs of clinical doctors, departments, hospitals and research institutions for high-quality cohort study protocol design at the present stage;

[0011] An intelligent recommendation algorithm for cohort study protocols is formed to realize the project configuration assistant function based on the question and answer mode. A large language model is used, which is a natural language generation model extended from the transformer architecture and can be fine-tuned to complete various natural language processing tasks such as text generation, code generation, video generation, text question and answer, image generation, paper writing, film and television creation, and scientific experiment design.

[0012] The intelligent dialogue agent system collects user preferences and choices by parameter adaptability training of the Qwen2.5-72B-Instruct model for specific tasks and combining pre-set logical rules in the dialogue process, obtains the collected cohort and its construction scheme demand information, makes personalized research protocol recommendation and question answering based on the knowledge base, and uses visual configuration to deploy and modify the scheme. If the question in the collected cohort is not matched in the knowledge base, access the external API to use the online GPT large model to retrieve answers for recommendation, and realize intelligent retrieval assistance outside the knowledge base.

[0013] S4, the intelligent design platform includes a knowledge base and a project configuration function, and realizes full-process intelligent recommendation according to the process of establishing a project-determining disease selection and inclusion and exclusion scheme-selecting scheme and project cycle configuration-CRF form visual configuration-follow-up point event configuration-project personnel setting-project research process summary display and accounting estimation of economic cost-saving and establishing a complete cohort study protocol, to assist researchers to complete high-quality cohorts;

[0014] S5, the standardized data collection process includes establishing a data collection platform suitable for multiple terminals, integrating biological sample information management, designing multi-modal data API interface and docking process, constructing follow-up intelligent automatic management function module, customizing data export function and electronic signature and document collection function;

[0015] S5 Establishing a multi-terminal data collection platform developed in Java language, using B / S architecture, using J2EE open technology system to establish a multi-terminal data collection platform software, adapting to computers, mobile phones, embedded small programs, smart bracelets and other customized electronic information products and hardware facilities. The multi-terminal data collection platform software can conveniently and accurately collect clinical cohort questionnaire data, biological sample collection and management data, and dynamically interface with multiple device measurement data, realize unified data collection platform for professional standardized data collection, and complete high-quality data collection of cohort study data through the establishment of standardized data collection process, data security management authorization process and intelligent data quality control process. The specific functions of the platform are as follows:

[0016] Integrating OCR-image recognition function, extracting file image content through image recognition, and automatically filling the extracted content into the system specified business field, supporting subject registration, form automatic filling and other scenarios; conveniently and accurately collecting clinical cohort questionnaire data, biological sample collection and management data, and dynamically interfacing with multiple device measurement data, realizing unified and standardized data collection; completing high-quality data collection through standardized data collection process, data security management authorization process and intelligent data quality control process;

[0017] Integrating biological sample information management to realize whole process management from sample warehousing, processing, sub-packaging, identification, freezing, warehousing, and re-warehousing, supporting sample type definition, annotation, frozen space allocation, expiration date management, dynamic inventory statistics, warehousing management, quality control record management, sample pedigree traceability management, achieving sample management standardization, warehousing automation, inventory operation visualization, and sample research management digitization;

[0018] Designing multi-modal data API interface and interfacing process According to the characteristics of multi-modal data (including text, voice, image, video and other modal information), design API interface and standardized data collection, interfacing process, establish cross-index for data correlation, realize real-time collection and integration of multi-modal data, dynamic interfacing of image, electroencephalogram, gait and other multi-device measurement data;

[0019] Building follow-up intelligent automatic management function module Through follow-up priority sorting algorithm to dynamically arrange follow-up tasks, the following functions are realized:

[0020] The system pushes the task to the investigator in real time, or pushes the online follow-up related information to the subject through WeChat and SMS;

[0021] Integrating WeChat customized tweet function, according to the subject's form filling situation, regularly pushing personalized image and text content such as medical treatment precautions and follow-up report, enhancing the interaction between the platform and the subjects, and improving the possibility of cloud data collection;

[0022] Real-time statistical summary of queue progress indicators (master queue progress) and investigator work indicators (baseline completion rate, follow-up completion rate, etc.), to achieve daily dynamic scheduling and real-time message pushing of follow-up tasks, and to achieve automatic and intelligent dynamic follow-up management process;

[0023] Customized data export function Based on the design of DBMS, customized data export function can realize research plan export; the collected electronic data can be exported after screening and de-privatization processing, and business data export is also supported;

[0024] Electronic signature and document collection function provides diversified document data collection methods for different scenarios:

[0025] Electronic informed consent provides electronic signature function;

[0026] Paper version of informed consent supports mobile phone shooting and uploading;

[0027] It can be extended to other document information collection, and provides one-key export of original file function, which is convenient for transmission to third-party system.

[0028] S6, the data security management authorization process includes user security management, identity authentication strategy, data permission hierarchical access management strategy, audit and monitoring strategy, data storage and transmission encryption control and other general security strategies to ensure data security;

[0029] S6, the user security management optimizes and integrates the platform user creation process, realizes one-key association of queue, and sends user information to user configuration mailbox, supporting contactless and secure management of platform user login;

[0030] Identity authentication strategy ensures the security of the authentication process by enforcing strong password policy, introducing mobile phone verification code authentication, and using OAuth2.0, OpenIDConnect and other mature identity verification protocols;

[0031] Data permission hierarchical access management strategy adopts multi-tenant data isolation mode, and converts department control architecture into cloud platform (SAAS mode) architecture;

[0032] Configure independent queue permissions for each tenant, configure queue data permissions and business application rules (such as inclusion and exclusion criteria, forms, and cycles) for each sub-center of the queue, and configure the number of queues, the number of centers and new business configuration rules for tenants;

[0033] Design a standardized data access rights intelligent management and authorization process: user access rights are opened by application, and after submission, they are authorized by managers; authorized roles include project managers, data collectors, data auditors, center managers, etc. Managers assign appropriate access rights to different users or roles to achieve role-based access control;

[0034] Audit and monitor system access logs in real time, and alarm and audit abnormal behavior; limit session life cycle to avoid security risks, while supporting session recovery function to improve user experience;

[0035] Data storage and transmission encryption control uses a unique and innovative internal and external network data control strategy: complete physical isolation of data input and output is achieved from the network architecture; data collection is transmitted to the data center through the data center, and after standardization with data governance tools, it is encrypted in a symmetric and asymmetric manner, and transmitted to the data center through text or audio and video files; decrypted by the data center and distributed to different business servers to ensure data security;

[0036] Other general security strategies include real-time backup and recovery functions, establishment of network boundary firewalls to limit external network services and management port openings, provision of intrusion detection systems to monitor network traffic in real time, identify and report potential attack behavior, and regular penetration testing, irregular security vulnerability scanning and repair.

[0037] S7, the data intelligent quality control process realizes efficient and intelligent data quality control through a queue follow-up data automatic correction model based on voice recognition; mature and open source voice recognition algorithms are used to build acoustic models for audio data in the data collection system, and semantic extraction algorithms are used to develop related tool sets to realize automatic correction, improve the accuracy of medical interview data collection, save data quality control costs, and improve the automation of audio data quality control functions;

[0038] S701: noise reduction of original audio data signal;

[0039] S702: automatic recognition of voice information as text;

[0040] S703: natural language recognition algorithm to extract keywords;

[0041] S704: fuzzy matching algorithm for automatic correction of voice information and original questionnaire data;

[0042] S705: data inspection and traceability: set up data quality control function modules, establish a data initiation query-reply query-data modification-correction confirmation quality control business process, and at the same time, after the first submission of the collected data, any data modification will be logged and can be customized and exported.

[0043] The voice recognition algorithm in S7 is one or more integrated models in a voice recognition algorithm based on a hidden Markov model, a voice recognition algorithm based on a Gaussian mixture model, a voice recognition algorithm based on a neural network, and a voice recognition algorithm based on deep learning, and two types of models are accessed in parallel: based on the local deployment of medical data corpus and voice recognition and natural language recognition algorithm in the system, automatic proofreading is performed; the API accesses the third-party large language online model to perform automatic proofreading.

[0044] Compared with the prior art, the beneficial effects of the present scheme are:

[0045] The present application meets the needs of clinical doctors, departments, hospitals and scientific research units to carry out high-quality cohort study by constructing a professional high-quality clinical cohort data research scheme knowledge base, a professional knowledge base platform and a technical platform; the system provides an intelligent research scheme recommendation algorithm according to diversified collection scenarios, and collects demand information in combination with a GPT question and answer mode; a multi-terminal data collection platform is developed by using J2EE technology, an API interface is designed and connected to a process, real-time collection and integration of data are realized, and efficient data quality control is ensured through a voice recognition and automatic proofreading model; the system also realizes dynamic scheduling and message pushing of follow-up tasks through an automatic management module, and guarantees the accuracy, timeliness, security and standardization of data collection; the system uses mature open source voice recognition algorithms to save quality control costs, and realizes standardized data collection through a unified platform, further improves the accuracy and efficiency of medical interview data, optimizes the data quality control process, and ensures data security through standardized data permission management. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a flowchart of the system in the embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to enable personnel in the art to better understand the present application scheme, the technical solutions of the present application will be further described in detail below in combination with the embodiments of the present application and the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0048] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below in combination with the embodiments.

[0049] Embodiment: The system comprises the following steps:

[0050] S1, construct a queue study scheme and an intelligent design platform for automatic configuration, which combines CRF form visualization and componentized configuration functions, integrates form attributes, association relationships, rules, etc. in the knowledge base, realizes form topic management in the form of componentization and drag-and-drop, can simplify the knowledge base form entry process, reduce system operation steps, and significantly improve the form configuration efficiency in the scheme design stage.

[0051] S2, construct a data collection platform with standardized data collection process, data security management authorization process and data intelligent quality control process.

[0052] S3, the intelligent design platform includes a knowledge base, a queue study scheme intelligent recommendation algorithm and an intelligent dialogue agent system based on the RAGflow framework; the knowledge base contains knowledge, experience and supporting data of multiple disciplines such as epidemiology, clinical medicine, medical statistics and economics, the scheme content is labeled with its applicable range and application scenario after expert evaluation, and the intelligent recommendation algorithm is used to realize scheme recommendation, which meets the needs of clinical doctors, departments, hospitals and research institutions to design high-quality queue study schemes at present;

[0053] An intelligent queue study scheme recommendation algorithm is formed to realize the project configuration assistant function based on the question and answer mode, a large language model is used, and the large language model is a natural language generation model extended from the transformer architecture, which can be fine-tuned to complete various natural language processing tasks such as text generation, code generation, video generation, text question and answer, image generation, paper writing, film and television creation, and scientific experiment design;

[0054] The intelligent dialogue agent system uses Qwen2.5-72B-Instruct model as the dialogue base, combines BGEembedding model to realize document embedding and efficient retrieval, reasons and answers based on the existing knowledge base in the dialogue process, so as to avoid the generation of false information; the system further collects user preferences and queue construction scheme demand information through the question and answer mode and the preset logic, provides personalized research scheme recommendation and problem solving based on the knowledge base, and realizes the deployment and modification of the scheme through visual configuration, so as to ensure the controllability and stability of the interaction process, effectively support the intelligent generation and optimization of queue study scheme; if the problem collected in the queue is not matched in the knowledge base, the online GPT large model is called through external API to realize supplementary retrieval and answer recommendation, so as to realize intelligent retrieval assistance outside the knowledge base.

[0055] S4, the intelligent design platform includes a knowledge base and project configuration function, according to the establishment of project-determine disease selection and screening scheme-select scheme and project cycle configuration-CRF form visualization configuration-follow-up point event configuration-project personnel setting-project research process summary display and accounting estimated economic cost-save and establish the complete cohort study scheme process to realize the whole process intelligent recommendation, auxiliary researchers to complete high-quality cohort. Among them, in the form configuration link, in addition to componentized drag management, it also supports form topic drawing function: form topic can realize manual drawing input, and supports the configuration of option as picture type, which can meet the form demand of special scene such as describing symptoms and signs through graphics in medical research, and improve the flexibility and adaptability of form design.

[0056] S5, the standardized data collection process includes establishing a data collection platform adapting to multiple terminals, integrating biological sample information management, designing multi-modal data API interface and docking process, building follow-up intelligent automatic management function module, customizing data export function and electronic signature and document collection function.

[0057] S5, the data collection platform software is developed by using java language, using B / S architecture, using J2EE open technology system to establish the data collection platform software adapting to multiple terminals, adapting to computer, mobile phone, embedded applet, smart bracelet and other customized electronic information products and hardware facilities. The data collection platform software can conveniently and accurately collect clinical cohort questionnaire data, biological sample collection and management data, and dynamically dock multiple device measurement data, realize unified data collection platform for professional standardized data collection, and complete high-quality cohort study data collection through standardized data collection process, data security management authorization process and intelligent data quality control process. The specific functions of the platform are as follows:

[0058] Integrating OCR-text recognition function, through image recognition to extract file text content, and automatically filling the extracted content into the system specified business field, supporting subject registration, form automatic filling and other scenes; conveniently and accurately collect clinical cohort questionnaire data, biological sample collection and management data, and dynamically dock multiple device measurement data, realize unified and standardized data collection; complete high-quality data collection through standardized data collection process, data security management authorization process and intelligent data quality control process.

[0059] Integrating biological sample information management realizes the whole process management from sample warehousing, processing, sub-packaging, identification, freezing, warehousing, re-warehousing, supports sample type definition, annotation, freezing space allocation, shelf life management, dynamic inventory statistics, warehousing management, quality control record management, sample pedigree traceability management, achieves sample management standardization, warehousing automation, inventory operation visualization, and sample research management digitization.

[0060] Based on the characteristics of multimodal data (including text, voice, image, video and other modal information), we designed API interfaces and standardized data collection and integration processes, established cross-indexes for data association, and realized real-time collection and integration of multimodal data, dynamically integrating measurement data from multiple devices such as imaging, EEG, and gait.

[0061] A smart, automated follow-up management module is constructed, dynamically scheduling follow-up tasks through a follow-up priority ranking algorithm. By employing this algorithm, the system dynamically schedules follow-up tasks, pushing tasks to investigators in real time or sending notifications to subjects via WeChat or SMS for online follow-up. It also provides real-time statistics and displays of queue progress indicators, allowing for monitoring of queue progress and summarizing research data. Furthermore, it provides real-time statistics and summaries of investigators' performance indicators, including baseline completion rate and follow-up completion rate, for queue work management and business data aggregation. This enables daily dynamic scheduling of follow-up tasks and real-time message pushes, thereby achieving an automated and intelligent dynamic follow-up management process.

[0062] This module integrates WeChat's customized article function, which can push personalized text and image content such as medical treatment precautions and follow-up reports to subjects on a regular basis according to the subjects' form completion. This can increase the frequency of interaction between the platform and subjects, improve subjects' follow-up compliance, and thus improve the completeness and timeliness of cloud data collection.

[0063] The customized data export function is designed based on DBMS and can realize the export of research plans; the collected electronic data can be exported after being screened and de-anonymized, and business data export is also supported.

[0064] The electronic signature and document collection functions offer diverse methods for collecting document data in different scenarios. Electronic informed consent forms support electronic signatures, while paper informed consent forms support uploading via mobile phone photos. This function can be extended to the collection of other document information and provides the ability to export the original file with one click, making it convenient for users to transfer to third-party systems and achieve standardized archiving and cross-system transfer of document data.

[0065] S6. The data security management authorization process includes user security management, identity authentication strategy, data permission hierarchical access management strategy, audit and monitoring strategy, data storage and transmission encryption control, and other general security strategies to ensure data security.

[0066] In S6, the user security management system optimizes and integrates the platform's user creation process, enabling one-click queue association and sending user information to the user's configured email address, supporting contactless and secure management of platform user login.

[0067] Identity authentication strategy ensures the security of the authentication process by enforcing strong password policies, introducing mobile phone verification authentication, and using mature identity verification protocols such as OAuth2.0 and OpenID Connect.

[0068] The data permission level access management strategy adopts a multi-tenant data isolation mode, converting the department control architecture into a cloud platform (SAAS mode) architecture. Each tenant is configured with independent queue permissions, and each sub-center of the queue is configured with data permissions and business application rules (such as inclusion and exclusion criteria, forms, and cycles). The number of queues, the number of centers, and the addition of business configuration rules for each tenant are configured to ensure that different tenant data is isolated and to meet the needs of multi-agency collaborative research permission management.

[0069] Standardized data permission intelligent management and authorization processes are designed: user access permissions are opened through an application process, and after submission, management personnel authorize them. Authorized roles include project managers, data collectors, data auditors, and center administrators. Management personnel assign appropriate access permissions to different users or roles to achieve role-based access control.

[0070] Audit and monitoring strategies monitor system access logs in real time, alert and audit abnormal behavior, limit session lifecycles to avoid security risks, and support session recovery functions to improve user experience.

[0071] Data storage and transmission encryption management uses a unique and innovative internal and external network data management strategy: data input and output are completely physically isolated from the network architecture. Data collection is transmitted to data management through data centers. After standardizing data using data governance tools, data is encrypted using symmetric and asymmetric encryption methods, and transmitted as text or audio / video files to data centers for decryption and distribution to different business servers, ensuring data security.

[0072] Other general security strategies include real-time data backup and recovery functions, network boundary firewall establishment to limit external network services and management port openings, intrusion detection system installation for real-time network traffic monitoring, identification and reporting of potential attack behavior, and regular penetration testing, as well as periodic security vulnerability scanning and repair.

[0073] S7、Data intelligent quality control processes use a queue follow-up data automatic correction model based on voice recognition to achieve efficient and intelligent data quality control. Mature and open-source voice recognition algorithms are used to build acoustic models for audio data in the data collection system, and semantic extraction algorithms are used to develop related tool sets to achieve automatic correction, improving the accuracy of medical interview data collection, saving data quality control costs, and improving the automation of audio data quality control functions.

[0074] S701: Noise reduction of raw audio source data signal;

[0075] S702: Voice information is automatically recognized as text;

[0076] S703: Natural Language Recognition Algorithm for Keyword Extraction;

[0077] S704: Fuzzy matching algorithm is used to automatically verify voice information and original questionnaire data;

[0078] S705: Data Auditing and Traceability: Set up a data quality control function module to establish a quality control business process of data inquiry - response to inquiry - data modification - correction confirmation. At the same time, after the initial submission of collected data, any data modification will be logged and can be customized for export.

[0079] The speech recognition algorithm in S7 is one or more integrated models among Hidden Markov Model-based speech recognition algorithms, Gaussian Mixture Model-based speech recognition algorithms, Neural Network-based speech recognition algorithms, and Deep Learning-based speech recognition algorithms. It integrates two types of models in parallel: it automatically verifies the speech recognition and natural language recognition algorithms based on the medical data corpus deployed locally in this system; and it automatically verifies the speech recognition by integrating third-party large-scale online language models through API.

[0080] In summary, this solution establishes a professional, high-quality clinical cohort study protocol knowledge base and technology platform, combining intelligent recommendation algorithms, multi-terminal data collection, end-to-end security management and intelligent quality control functions, and integrating business capabilities adapted to clinical scenarios such as OCR recognition, electronic signatures, and WeChat push notifications, to meet the needs of clinicians, departments, hospitals, and research institutions for the design and implementation of high-quality cohort study protocols.

[0081] The above specific embodiments are merely explanations of the present invention and are not intended to limit the present invention. After reading this specification, those skilled in the art can make modifications to these embodiments without contributing any inventive step, but as long as they are within the scope of the claims of the present invention, they are protected by patent law.

Claims

1. A clinical database end-to-end collaborative intelligence system, characterized by: The system includes the following steps: S1. Construct an intelligent design platform for cohort research schemes and automated configuration; S2. Build a data acquisition platform with standardized data acquisition processes, data security management and authorization processes, and intelligent data quality control processes; S3. The intelligent design platform includes a knowledge base, an intelligent recommendation algorithm for cohort study schemes, and an intelligent dialogue agent system built on the RAGflow framework. The knowledge base contains professional knowledge in epidemiology, clinical medicine, medical statistics, and economics. The intelligent recommendation algorithm for cohort study schemes uses a large language model to generate project configurations based on a question-and-answer pattern. The intelligent dialogue agent system uses the Qwen2.5-72B-Instruct model as the dialogue base model, combined with the BGEembedding model to achieve document embedding and efficient retrieval. During the dialogue, it performs reasoning and responses based on the existing knowledge base, thereby avoiding the generation of false information. The system collects user preferences and queue construction plan requirements through question-and-answer patterns and logical rules. Based on the knowledge base, it recommends personalized research plans and answers questions. It also uses visual configuration to allocate and modify plans. The system constrains the dialogue process through preset logical rules, ensuring the controllability and stability of the interaction. It effectively supports the intelligent generation and optimization of queue research plans. If the questions in the queue are not matched in the knowledge base, it connects to an external API to use the online GPT model to retrieve answers and make recommendations, realizing intelligent retrieval assistance outside the knowledge base. S4. The intelligent design platform includes knowledge base and project configuration functions. It realizes intelligent recommendation throughout the entire process of establishing a project, determining disease selection and inclusion / exclusion schemes, selecting schemes and configuring project cycle, visual configuration of CRF forms, configuration of follow-up point events, setting up project personnel, displaying a summary of the project research process and calculating estimated economic costs, and saving and establishing a complete cohort study scheme. This helps researchers complete the design of high-quality cohort study schemes. S5. The standardized data collection process includes establishing a data collection platform adapted to multiple terminals, integrating biological sample information management, designing multimodal data API interfaces and docking processes, constructing a follow-up intelligent automatic management function module, customizing data export functions, and electronic signature and document collection functions. S6. The data security management authorization process includes user security management, identity authentication strategy, data permission hierarchical access management strategy, audit and monitoring strategy, data storage and transmission encryption control and other general security strategies to ensure data security. S7. The data intelligent quality control process achieves efficient and intelligent data quality control through an automatic verification model of queue follow-up data based on speech recognition; it adopts mature and open-source speech recognition algorithms to build acoustic models of the recording data in the data acquisition system, and uses semantic crawling algorithms to develop related toolsets to achieve automatic verification, thereby improving the accuracy of medical interview data collection, saving data quality control costs, and improving the automated quality control function of recording data.

2. The clinical database end-to-end collaborative intelligence system as described in claim 1, characterized in that: The S5-based multi-terminal data acquisition platform is developed using Java, employs a B / S architecture, and leverages the J2EE open technology system. This platform has the following functions: Integrated with OCR-text recognition functionality, it extracts text and image content from documents through image recognition and automatically fills the extracted content into designated business fields in the system; it facilitates, speeds up, and accurately collects clinical cohort questionnaire data, biological sample collection and management data, and dynamically connects to data measured by multiple devices; it achieves high-quality data collection through standardized data collection processes, data security management authorization processes, and intelligent data quality control processes; The integrated biological sample information management system enables full-process management of samples from receipt, processing, dispensing, labeling, cryopreservation, outbound, and re-entry. It supports sample type definition, annotation, cryopreservation space allocation, expiration date management, dynamic inventory statistics, inbound and outbound management, quality control record management, and sample lineage traceability management, achieving standardized sample management, automated inbound and outbound operations, visualized inventory operations, and digitalized sample scientific research management. Based on the characteristics of multimodal data, we designed API interfaces and standardized data collection and integration processes, established cross-indexes for data association, and realized real-time collection and integration of multimodal data, as well as dynamic multi-device measurement data. A smart, automated follow-up management module is built to dynamically schedule follow-up tasks using a follow-up priority ranking algorithm, achieving the following functions: The system will push tasks to investigators in real time, or push online follow-up information to subjects via WeChat or SMS; It integrates WeChat's customized article function, and pushes personalized graphic and text content such as medical treatment precautions and follow-up reports on a regular basis according to the subjects' form filling, which enhances the interaction between the platform and the subjects and increases the possibility of cloud data collection; Real-time statistics and summaries of queue progress indicators and investigator work indicators enable daily dynamic scheduling of follow-up tasks and real-time message push, achieving an automated, intelligent, and dynamic follow-up management process. The customized data export function is designed based on DBMS and can realize the export of research plans; the collected electronic data can be exported after being screened and de-anonymized, and business data export is also supported. The electronic signature and document collection functions provide diverse document data collection methods for different scenarios. Electronic informed consent forms can be signed through the electronic signature function, while paper informed consent forms can be uploaded by taking photos with a mobile device. This feature can also be extended to other types of document information collection and provides a one-click export function for original files, making it easy to transfer data to third-party systems.

3. The clinical database end-to-end collaborative intelligence system as described in claim 1, characterized in that: The S6 optimizes and integrates the platform user creation process for user security management, enabling one-click queue association and sending user information to the user's configured email address, supporting contactless and secure management of platform user login; The identity authentication policy ensures the security of the authentication process by enforcing strong password policies and identity verification protocols; The data access control strategy adopts a multi-tenant data isolation model, transforming the departmental management architecture into a cloud platform architecture. Configure independent queue permissions for each tenant, configure queue data permissions and business application rules for each sub-center of the queue, and configure the number of queues, number of centers, and new business configuration rules for each tenant. Design a standardized intelligent data access management and authorization process: User access permissions are granted through an application process, and administrators authorize access after the application is submitted; administrators assign appropriate access permissions to different users or roles to achieve role-based access control; The auditing and monitoring strategy monitors system access logs in real time, alerts and audits abnormal behavior; limits session lifecycle to avoid security risks, and supports session recovery to improve user experience; The data storage and transmission encryption management adopts a unique and innovative internal and external network data management strategy: achieving complete physical isolation between data input and output from the network architecture. Data is collected and transmitted to data management via the data center. After standardization using data governance tools, it is encrypted using symmetric and asymmetric encryption methods and transmitted to the data center as text or audio / video files. The data center then decrypts the data and distributes it to different business servers to ensure data security. Other general security strategies include: designing real-time data backup and recovery functions; establishing network boundary firewalls to restrict the opening of external network services and management ports; equipping intrusion detection systems to monitor network traffic in real time, identify and report potential attack behaviors; conducting penetration tests regularly, and performing security vulnerability scans and remediation periodically.

4. The clinical database end-to-end collaborative intelligence system as described in claim 1, characterized in that: The specific steps of S7 are as follows: S701: Noise reduction of raw audio source data signal; S702: Voice information is automatically recognized as text; S703: Natural Language Recognition Algorithm for Keyword Extraction; S704: Fuzzy matching algorithm is used to automatically verify voice information and original questionnaire data; S705: Data Auditing and Traceability: Set up a data quality control function module to establish a quality control business process of data inquiry - response to inquiry - data modification - correction confirmation. At the same time, after the initial submission of collected data, any data modification will be logged and can be customized for export.

5. A clinical database end-to-end collaborative intelligence system as described in claim 1, characterized in that: The terminal includes computers, mobile phones, embedded mini-programs, smart bracelets, other electronic information products and hardware facilities.

6. A clinical database end-to-end collaborative intelligence system as described in claim 4, characterized in that: The speech recognition algorithm in S7 is one or more integrated models among Hidden Markov Model-based speech recognition algorithms, Gaussian Mixture Model-based speech recognition algorithms, Neural Network-based speech recognition algorithms, and Deep Learning-based speech recognition algorithms. Two types of models are accessed in parallel: automatic proofreading is performed based on the medical data corpus deployed locally in this system and speech recognition and natural language recognition algorithms; automatic proofreading is performed by accessing third-party large language online models through API.

7. A clinical database end-to-end collaborative intelligence system as described in claim 2, characterized in that: The multimodal data feature in S5 refers to the input of multiple modal information, including text, voice, image and video information.

Images