Intelligent ward AI assistant collaborative view docking data interaction system

Abstract

The invention discloses a data interaction system for cooperative view docking of an intelligent ward AI assistant. The data interaction system comprises an AI intelligent data processing module, an AI identification unified adaptation module, an AI medical intelligent analysis and model training module, a remote intelligent monitoring and control center module and an AI enhanced interaction visualization platform and intelligent report module. Through the AI identification unified adaptation module and the AI intelligent data processing module, manual intervention is reduced, demand identification, template matching, data cleaning and docking logic generation are automatically completed, the docking efficiency is improved, and the period is shortened; the AI medical intelligent analysis and model training module constructs a fault prediction model, combines real-time monitoring and multistage early warning functions of the remote intelligent monitoring and management and control center module, identifies potential problems in advance, pushes a repair scheme, and enhances system stability to reduce business delay.

Description

Technical Field

[0001] This invention relates to the field of smart ward technology, and in particular to a data interaction system for connecting a smart ward AI assistant with a collaborative view. Background Technology

[0002] In current enterprise-level data integration scenarios, cross-system data interaction has become a core requirement for business collaboration. However, existing data integration solutions have significant limitations in terms of adaptability, intelligence, user experience, and stability, failing to meet the needs of rapid business expansion and operation by non-technical personnel. Specific drawbacks include: 1. Template integration relies on manual intervention: Existing solutions require technical personnel to manually select and match templates and manually configure field mappings; this repeated manual selection and configuration leads to low integration efficiency. 2. Insufficient intelligent data processing: Existing data integration relies on fixed interface protocols. For interface development, technical personnel must manually write parsing logic and debug code according to the data format each time, extending the integration cycle. 3. Limited user interaction experience: Existing data integration requirements generally rely on backend development... For engineers, the challenges are multifaceted: Firstly, non-technical personnel may lack programming knowledge and be unable to solve problems independently, leading to lengthy processing times from request submission to backend response. Secondly, data integration failures typically require checking log files for diagnosis, which non-technical personnel cannot quickly pinpoint, resulting in reduced work efficiency. Thirdly, the system suffers from poor stability: traditional systems lack real-time monitoring capabilities. This results in weak risk assessment and an inability to monitor server CPU usage, memory utilization, and other resource loads in real time. Furthermore, the absence of an early warning mechanism for interface failures necessitates business personnel to detect data loss and report it to backend staff, leading to long response times and business delays. In light of these factors, this application proposes a data interaction system for intelligent ward AI assistant collaborative view integration. Summary of the Invention

[0003] Based on the technical problems existing in the background technology, the present invention proposes a data interaction system for intelligent ward AI assistant collaborative view docking.

[0004] The present invention proposes a data interaction system for AI assistant collaborative view docking in smart wards, including an AI intelligent data processing module, an AI recognition unified adaptation module, an AI medical intelligent analysis and model training module, a remote intelligent monitoring and control center module, and an AI-enhanced interactive visualization platform and intelligent reporting module; The AI ​​intelligent data processing module is used for data cleaning and standardization. The AI ​​recognition and unified adaptation module is used to automatically retrieve the smart ward data docking document view library and view docking based on the smart ward data docking document docking specification, automatically identify the protocol type of external devices and systems, and generate adaptation logic. The AI ​​medical intelligent analysis and model training module is used to receive standardized data and generate equipment fault prediction models through data preprocessing, feature extraction, AI algorithm integration and model training verification. The remote intelligent monitoring and control central module is equipped with a fault prediction model to obtain prediction results in real time, enabling remote monitoring, control, alarm, and log recording of equipment. The AI-enhanced interactive visualization platform and intelligent reporting module provide managers and medical staff with a data visualization interface and maintenance suggestions, supporting fault warning viewing and maintenance plan formulation.

[0005] Preferably, the specific operational logic steps of the AI ​​intelligent data processing module are as follows: S101: Obtain third-party interface data through AI recognition unified adaptation module, and load smart ward interface documents as cleaning rule base; S102: AI automatically executes cleaning data and determines the cleaning pass rate; S103: If the cleaning pass rate is >95%, a report will be generated and the data will enter the docking process; S104: If the cleaning pass rate is less than 95%, then unrepaired data will be filtered out, and the AI ​​assistant will push the solution in the docking document. After the user repairs the data, the AI ​​cleaning will be triggered again.

[0006] Preferably, in step S103, the generated report includes the amount of cleaned data, the distribution of error types, the repair rate, and a list of unrepaired data. The cleaning rule base in S101 includes cleaning types and AI automatic cleaning logic. The cleaning types include: missing value identification and processing, outlier identification and processing, format error identification and processing, and correlation verification. The corresponding AI automatic cleaning logic is as follows: (1) Missing value identification and processing: Mark missing data in required fields, mark and log them for feedback, use non-required fields to fill and record in the log for feedback, and trigger "backtracking and supplementing" when required fields are missing, call the AI ​​recognition unified adaptation module to re-acquire third-party data; (2) Outlier identification and handling: For outlier data, mark and log it, and provide real-time feedback to technical personnel; (3) Format error identification and handling: Validate data type and format and automatically convert to a repairable format; (4) Relationship Validation: Validate the relationship between fields across views, and mark "data orphaned" if the relationship fails; In addition, the AI ​​assistant records "new error types" during the cleaning process, automatically updates the "cleaning rule base," and synchronizes it to the smart ward data docking document maintenance records.

[0007] Preferably, the core steps of AI view retrieval and retrieval verification and adjustment in the AI ​​recognition unified adaptation module are as follows: S201: Demand Analysis Section. The AI ​​assistant receives the user's natural language demand and extracts key features through Natural Language Processing (NLP). S202: The AI ​​assistant calls the smart ward system to connect to the document view library and filters candidate templates through feature matching; S203: The AI ​​assistant searches the historical case database, calculates the similarity between the requirements and the cases, and thus confirms the target template; S204: The AI ​​assistant automatically generates field mapping relationships based on the view docking rules and the field definitions in the smart ward data docking document; S205: Check whether the mapping relationship in S204 meets the docking requirements of the smart ward system docking document; S206: If the requirements do not match the documentation for the smart ward system, the AI ​​assistant will provide feedback that "the bed code format must be standardized according to the documentation for the smart ward system". S207: Technicians make fine-tuning adjustments based on the AI ​​results, and the adjusted records are automatically stored in the case library to optimize the accuracy of subsequent AI retrieval.

[0008] Preferably, the AI ​​medical intelligent analysis and model training module is used to enhance the business training of the AI ​​model, and its specific logical steps are as follows: S3011: Data Collection: Collect metadata of 32 types of views from historical docking cases and smart ward docking documents over the past 2 years, exclude invalid samples, remove cases with "vague requirements and no docking results", and finally retain 1,000 valid samples. S3012: Data preprocessing: Clean the data using AI and divide it into training set, validation set and test set in a 7:2:1 ratio to ensure that the test set samples cover all core views of the smart ward data docking document; S3013: Algorithm Integration: The decision tree splitting criterion is adopted. For each feature, the Gini coefficient after splitting is calculated, the feature with the smallest coefficient is selected as the splitting node, and the supervised learning algorithm Random Forest is selected to handle complex relationships and a large number of features. S3014: Constructing a CNN architecture: Designing convolutional and pooling layers to effectively capture the spatial and temporal features of the input data, and introducing fully connected layers to map the features output by the convolutional layers to the final prediction results; S3015: Model Training: Use historical data to train the model, divide it into training and validation sets, and select a loss function to optimize according to the problem type; S3016: Parameter tuning: Iteratively tune the hyperparameters of the CNN using regularization methods, including kernel size, pooling size, and learning rate; S3017: Model Validation and Evaluation: Evaluate model performance on the validation set, focusing on accuracy and recall metrics, analyzing the model's performance across different categories, and understanding its ability to handle and predict different needs.

[0009] Preferably, the AI ​​medical intelligent analysis and model training module is used to enhance the dialogue of the AI ​​model, and its specific logical steps are as follows: S3021: User Request Initiation: Technical / non-technical personnel submit and initiate a request; S3022: Request reception and intent recognition: AI automatically determines that the user's request belongs to the "query / fault / consultation" scenario, matches the corresponding template, and stores key information from recent conversations; S3023: Layered prompt word matching and generation: Perform basic layer calling, personalized layer generation and scene layer filling; S3024: Dialogue content generation and output: Generate and output one of the following: fault repair plan, data operation steps, status feedback information, and output language style that is automatically adapted according to the user role; S3025: User Feedback Collection: Set up a simple feedback entry point, including: Technical personnel: feedback on "insufficient details" or "whether the solution is compliant", Non-technical personnel: feedback on "don't understand" or "whether it has been resolved"; S3026: Feedback-driven optimization: Simplify evaluation metrics: Focus on "scene adaptation rate", "output compliance rate" and "user satisfaction", avoid too many numbers, regularly convert effective feedback into training data, update prompt word templates and models, and ensure continuous optimization of dialogue effects.

[0010] Preferably, the operational logic steps of the remote intelligent monitoring and control central module are as follows: S401: Model Integration and Deployment: Deploy the trained AI model to the server and connect it to the data interaction center to achieve unified flow of model data. The AI ​​assistant monitors the model deployment status in real time and provides anomaly warnings. S402: Model Prediction Value Acquisition: The web interface allows users to configure data acquisition strategies through the AI ​​assistant. Once triggered, the web client sends a request to the data interaction center, which then uniformly calls the model API to obtain the prediction values. The AI ​​assistant performs preliminary cleaning on the obtained prediction values, which are then synchronized to the web client by the data interaction center for monitoring the real-time status of the data. At the same time, historical prediction value data is stored in the data center for archiving. S403: Remote monitoring: The core data source is the model prediction value synchronized by the data interaction center. The web interface uses charts to display the device status. A real-time connection is established with the data interaction center through AJAX+WebSockets technology to realize dynamic updates of front-end data. The AI ​​assistant assists in analyzing the trend of prediction values, marks abnormal data points on the web interface, and provides trend prediction to improve the intelligence of monitoring. S404: Remote Control: The web interface is designed as the device control operation area. After the user initiates a control command, the AI ​​assistant first verifies the validity of the command. If the verification is successful, the command is transmitted to the data interaction center via the API interface. The data interaction center standardizes the format of the command and then forwards it to the server. Finally, the server sends the command to the target device. At the same time, the AI ​​assistant feeds back the command execution result to the web interface and synchronizes it to the data center for archiving. S405: Alarms and Notifications: Based on the model prediction values ​​synchronized by the data interaction center, multi-level alarm rules are set. When the prediction value triggers the threshold, the data interaction center pushes the alarm signal to the AI ​​assistant. The AI ​​assistant intelligently determines the alarm level and generates personalized notification content based on the user's preset notification preferences. At the same time, it links the data interaction center to synchronize alarm information to the web terminal to ensure that users respond to device abnormalities in a timely manner. S406: Log Recording and Auditing: All user operations are synchronized to the data interaction center in real time. The data center completes the structured storage of operation logs. The AI ​​assistant regularly performs intelligent analysis of the logs, identifies abnormal operations, generates audit reports, and works with the log management system to perform regular cleanup and encrypted backup. The data interaction center ensures the integrity and traceability of the log archive. S407: Remote Diagnosis and Maintenance: Integrates remote diagnostic tools and interfaces with the data interaction center. After technicians initiate a diagnostic request via the web, the device's real-time operating data is transmitted to the diagnostic tool through the data interaction center. The AI ​​assistant analyzes the data simultaneously to help technicians quickly locate faults. When the device is remotely upgraded, the AI ​​assistant verifies the integrity and compatibility of the firmware / software package. The upgrade package is transmitted to the device in segments through the data interaction center to ensure a stable upgrade process. At the same time, the AI ​​assistant records the upgrade progress and provides feedback to the web interface.

[0011] Preferably, the AI-enhanced interactive visualization platform and intelligent reporting module includes a data visualization interface, an alarm and suggestion system, and a report generation interface; The data visualization interface is designed using development frameworks such as React, Vue.js, and Angular to build an intuitive user interface. CSS frameworks are used to ensure responsive design, adapting to different devices and screen sizes. Chart libraries are used to present various chart formats, ensuring clear and attractive visualizations. User interactivity utilizes interactive components and state management tools in front-end technologies to ensure users can flexibly customize time ranges, select specific devices, and observe data changes in real time. The alarm and suggestion system is divided into real-time alarms and detailed maintenance suggestions. Real-time alarms are integrated into the real-time alarm system. When the AI ​​model detects potential maintenance needs or equipment abnormalities, it delivers alarms to users in real time through the interface and notifications. Combined with backend services and frontend frameworks, it uses WebSocket real-time communication technology to ensure the push of real-time alarms. Detailed maintenance suggestions provide detailed maintenance advice on the interface, explaining possible causes of failures and providing recommended maintenance measures. The report generation interface automatically summarizes the core quantitative indicators of AI assistant dialogue interaction within the period, including total number of dialogue requests, text noise reduction accuracy, prompt word template matching rate, and dialogue response accuracy. The indicator data are all from the dialogue interaction logs recorded by the system in real time, and are presented intuitively through data visualization to help users quickly evaluate the overall usability and operational stability of the AI ​​assistant dialogue function.

[0012] Compared with existing technologies, the beneficial effects of this invention are: 1. The AI ​​recognition unified adaptation module automatically identifies user needs, matches business specifications, and generates docking logic. The AI ​​intelligent data processing module automatically completes data cleaning and standardization, eliminating the need for repetitive manual operations, significantly reducing the workload of technical personnel, and shortening the docking cycle. 2. By leveraging the AI ​​medical intelligent analysis and model training module, a fault prediction model is constructed using algorithms such as random forest, realizing the transformation of data processing from "manual" to "AI automatic analysis and prediction", reducing errors caused by manual intervention, improving the intelligence and accuracy of data processing, and solving the limitations of existing solutions that rely on fixed interface protocols and require manual code debugging. 3. The AI-enhanced interactive visualization platform provides an intuitive visual interface and simplified operation. Combined with a hierarchical prompt system, it simplifies terminology and operation steps for non-technical personnel, provides detailed parameters and specifications for technical personnel, and supports fault warning viewing and quick location, thereby improving the operation efficiency and experience of users with different roles. 4. The remote intelligent monitoring and control center module monitors server resource load and equipment operating status in real time, provides early warning of potential problems based on the fault prediction model, and pushes repair solutions in real time. This eliminates the need to wait for business personnel to discover data loss and provide feedback, shortens fault response time, and reduces the impact on medical services. 5. By embedding business specifications into the docking logic through AI, the system automatically aligns with requirements such as view field mapping and interface parameters. At the same time, the AI-identified unified adaptation module can automatically switch protocols and templates to adapt to the docking needs of multiple scenarios such as hospital HIS systems and medical equipment. This eliminates the need for repeated development or configuration modifications, thereby improving system flexibility and the accuracy of specification execution. 6. Automatically record all operation logs and mark the source of the standards to facilitate quick identification of the root cause of the problem; at the same time, the AI ​​model will regularly optimize itself through new cases and updated standards, and the accuracy of matching needs and identifying problems will continue to improve in long-term use, reducing maintenance costs and ensuring the long-term efficient operation of the system.

[0013] This invention reduces manual intervention by using an AI-based unified adaptation module and an AI-based intelligent data processing module. These modules automatically complete requirement identification, template matching, data cleaning, and connection logic generation, improving efficiency and shortening the cycle time. An AI-based medical intelligent analysis and model training module constructs a fault prediction model. Combined with the real-time monitoring and multi-level early warning functions of a remote intelligent monitoring and control center module, it identifies potential problems in advance and pushes repair solutions, enhancing system stability and reducing business delays. An AI-enhanced interactive visualization platform and a hierarchical prompting system simplify operations and terminology for non-technical personnel and provide detailed parameters for technical personnel, optimizing the user experience for all roles and lowering the operational threshold. Simultaneously, the system automatically aligns with business specifications to reduce errors, adapts to multiple scenarios without redundant development, and provides traceable operation logs for easy problem troubleshooting. Furthermore, the AI ​​model can be continuously optimized through new cases and specifications, improving efficiency and reducing maintenance costs in the long term. Ultimately, this achieves the core goals of AI-driven cost reduction and efficiency improvement, standardized alignment to reduce risks, and ease of use to meet the needs of all roles. Attached Figure Description

[0014] Figure 1 This is a block diagram of a data interaction system for intelligent ward AI assistant collaborative view docking proposed in this invention; Figure 2 This is a flowchart illustrating the operation of the AI ​​intelligent data processing module in a data interaction system for AI assistant collaborative view docking in a smart ward, as proposed in this invention. Figure 3 This is a flowchart illustrating the operation of the AI ​​recognition unified adaptation module in a data interaction system for AI assistant collaborative view docking in a smart ward, as proposed in this invention. Figure 4This invention presents a flowchart of the AI ​​model enhancement and business training process in a data interaction system for AI assistant collaborative view docking in a smart ward. Figure 5 This is a flowchart of the AI ​​model enhanced dialogue in a data interaction system for AI assistant collaborative view docking in a smart ward, as proposed in this invention. Figure 6 This is a flowchart illustrating the operation of the remote intelligent monitoring and control central module in a data interaction system for AI assistant collaboration view docking proposed in this invention. Detailed Implementation

[0015] The present invention will be further explained below with reference to specific embodiments.

[0016] Example Reference Figure 1-6 This embodiment proposes a data interaction system for AI assistant collaborative view docking in smart wards, including an AI intelligent data processing module, an AI recognition unified adaptation module, an AI medical intelligent analysis and model training module, a remote intelligent monitoring and control center module, and an AI-enhanced interactive visualization platform and intelligent reporting module; The AI-powered intelligent data processing module is used for data cleaning and standardization. The specific operational logic steps of the AI ​​intelligent data processing module are as follows: S101: Obtain third-party interface data through AI recognition unified adaptation module, and load smart ward interface documents as cleaning rule base; The cleaning rule base includes cleaning types and AI automatic cleaning logic. The cleaning types include: missing value identification and handling, outlier identification and handling, format error identification and handling, and correlation verification. The corresponding AI automatic cleaning logic is as follows: (1) Missing value identification and processing: Mark missing data in required fields, mark and log them for feedback, use non-required fields to fill and record in the log for feedback, and trigger "backtracking and supplementing" when required fields are missing, call the AI ​​recognition unified adaptation module to re-acquire third-party data; (2) Outlier identification and handling: For outlier data, mark and log it, and provide real-time feedback to technical personnel; (3) Format error identification and handling: Validate data type and format and automatically convert to a repairable format; (4) Relationship Validation: Validate the relationship between fields across views, and mark "data orphaned" if the relationship fails; In addition, the AI ​​assistant records "new error types" during the cleaning process, automatically updates the "cleaning rule base", and synchronizes it to the smart ward data docking document maintenance record; S102: AI automatically executes cleaning data and determines the cleaning pass rate; S103: If the cleaning pass rate is >95%, a report will be generated and the data will enter the integration process. The generated report includes the amount of cleaned data, the distribution of error types, the repair rate, and a list of unrepaired data. S104: If the cleaning pass rate is less than 95%, then unrepaired data will be filtered out, and the AI ​​assistant will push the solution in the docking document. After the user repairs the data, the AI ​​cleaning will be triggered again. The AI ​​recognition unified adaptation module is used to automatically retrieve the smart ward data docking document view library and view docking based on the smart ward data docking document docking specification, automatically identify the protocol type of external devices and systems, and generate adaptation logic. In the unified adaptation module for AI recognition, the core steps of AI view retrieval and retrieval verification and adjustment are as follows: S201: Demand Analysis Section. The AI ​​assistant receives the user's natural language demand and extracts key features through Natural Language Processing (NLP). S202: The AI ​​assistant calls the smart ward system to connect to the document view library and filters candidate templates through feature matching; S203: The AI ​​assistant searches the historical case database, calculates the similarity between the requirements and the cases, and thus confirms the target template; S204: The AI ​​assistant automatically generates field mapping relationships based on the view docking rules and the field definitions in the smart ward data docking document; S205: Check whether the mapping relationship in S204 meets the docking requirements of the smart ward system docking document; S206: If the requirements do not match the documentation for the smart ward system, the AI ​​assistant will provide feedback that "the bed code format must be standardized according to the documentation for the smart ward system". S207: Technicians make fine-tuning adjustments based on AI results, and the adjusted records are automatically stored in the case library to optimize the accuracy of subsequent AI retrieval. The AI ​​medical intelligent analysis and model training module is used to receive standardized data and generate equipment failure prediction models through data preprocessing, feature extraction, AI algorithm integration and model training verification. The AI ​​medical intelligent analysis and model training module is used to enhance the AI ​​model and strengthen business training. Its specific logical steps are as follows: S3011: Data Collection: Collect metadata of 32 types of views from historical docking cases and smart ward docking documents over the past 2 years, exclude invalid samples, remove cases with "vague requirements and no docking results", and finally retain 1,000 valid samples. S3012: Data preprocessing: Clean the data using AI and divide it into training set, validation set and test set in a 7:2:1 ratio to ensure that the test set samples cover all core views of the smart ward data docking document; S3013: Algorithm Integration: The decision tree splitting criterion is adopted. For each feature, the Gini coefficient after splitting is calculated, the feature with the smallest coefficient is selected as the splitting node, and the supervised learning algorithm Random Forest is selected to handle complex relationships and a large number of features. S3014: Constructing a CNN architecture: Designing convolutional and pooling layers to effectively capture the spatial and temporal features of the input data, and introducing fully connected layers to map the features output by the convolutional layers to the final prediction results; S3015: Model Training: Use historical data to train the model, divide it into training and validation sets, and select a loss function to optimize according to the problem type; S3016: Parameter tuning: Iteratively tune the hyperparameters of the CNN using regularization methods, including kernel size, pooling size, and learning rate; S3017: Model Validation and Evaluation: Evaluate model performance on the validation set, focusing on accuracy and recall metrics, analyzing the model's performance across different categories, and understanding its ability to handle and predict different needs. The AI ​​medical intelligent analysis and model training module is used to enhance the dialogue of AI models. Its specific logical steps are as follows: S3021: User Request Initiation: Technical / non-technical personnel submit and initiate a request; S3022: Request reception and intent recognition: AI automatically determines that the user's request belongs to the "query / fault / consultation" scenario, matches the corresponding template, and stores key information from recent conversations; S3023: Layered prompt word matching and generation: Perform basic layer calling, personalized layer generation and scene layer filling; S3024: Dialogue content generation and output: Generate and output one of the following: fault repair plan, data operation steps, status feedback information, and output language style that is automatically adapted according to the user role; S3025: User Feedback Collection: Set up a simple feedback entry point, including: Technical personnel: feedback on "insufficient details" or "whether the solution is compliant", Non-technical personnel: feedback on "don't understand" or "whether it has been resolved"; S3026: Feedback-driven optimization: Simplify evaluation metrics: Focus on "scene adaptability", "output compliance" and "user satisfaction", avoid too many numbers, regularly convert effective feedback into training data, update prompt word templates and models, and ensure continuous optimization of dialogue effects; The remote intelligent monitoring and control central module deploys a fault prediction model to obtain prediction results in real time, enabling remote monitoring, control, alarm, and log recording of equipment; The operational logic steps of the remote intelligent monitoring and control hub module are as follows: S401: Model Integration and Deployment: Deploy the trained AI model to the server and connect it to the data interaction center to achieve unified flow of model data. The AI ​​assistant monitors the model deployment status in real time and provides anomaly warnings. S402: Model Prediction Value Acquisition: The web interface allows users to configure data acquisition strategies through the AI ​​assistant. Once triggered, the web client sends a request to the data interaction center, which then uniformly calls the model API to obtain the prediction values. The AI ​​assistant performs preliminary cleaning on the obtained prediction values, which are then synchronized to the web client by the data interaction center for monitoring the real-time status of the data. At the same time, historical prediction value data is stored in the data center for archiving. S403: Remote monitoring: The core data source is the model prediction value synchronized by the data interaction center. The web interface uses charts to display the device status. A real-time connection is established with the data interaction center through AJAX+WebSockets technology to realize dynamic updates of front-end data. The AI ​​assistant assists in analyzing the trend of prediction values, marks abnormal data points on the web interface, and provides trend prediction to improve the intelligence of monitoring. S404: Remote Control: The web interface is designed as the device control operation area. After the user initiates a control command, the AI ​​assistant first verifies the validity of the command. If the verification is successful, the command is transmitted to the data interaction center via the API interface. The data interaction center standardizes the format of the command and then forwards it to the server. Finally, the server sends the command to the target device. At the same time, the AI ​​assistant feeds back the command execution result to the web interface and synchronizes it to the data center for archiving. S405: Alarms and Notifications: Based on the model prediction values ​​synchronized by the data interaction center, multi-level alarm rules are set. When the prediction value triggers the threshold, the data interaction center pushes the alarm signal to the AI ​​assistant. The AI ​​assistant intelligently determines the alarm level and generates personalized notification content based on the user's preset notification preferences. At the same time, it links the data interaction center to synchronize alarm information to the web terminal to ensure that users respond to device abnormalities in a timely manner. S406: Log Recording and Auditing: All user operations are synchronized to the data interaction center in real time. The data center completes the structured storage of operation logs. The AI ​​assistant regularly performs intelligent analysis of the logs, identifies abnormal operations, generates audit reports, and works with the log management system to perform regular cleanup and encrypted backup. The data interaction center ensures the integrity and traceability of the log archive. S407: Remote Diagnosis and Maintenance: Integrates remote diagnostic tools with a data interaction center. After technicians initiate a diagnostic request via the web, the device's real-time operating data is transmitted to the diagnostic tool through the data interaction center. The AI ​​assistant simultaneously analyzes the data to help technicians quickly locate faults. During remote device upgrades, the AI ​​assistant verifies the integrity and compatibility of the firmware / package. The upgrade package is transmitted to the device in segments through the data interaction center to ensure a stable upgrade process. At the same time, the AI ​​assistant records the upgrade progress and provides feedback to the web interface. The AI-enhanced interactive visualization platform and intelligent reporting module provide managers and medical staff with a data visualization interface and maintenance suggestions, supporting fault warning viewing and maintenance plan formulation; The AI-enhanced interactive visualization platform and intelligent reporting module include a data visualization interface, an alarm and suggestion system, and a report generation interface. The data visualization interface is designed using development frameworks such as React, Vue.js, and Angular to build an intuitive user interface. CSS frameworks are used to ensure responsive design, adapting to different devices and screen sizes. Chart libraries are used to present various chart formats, ensuring clear and attractive visualizations. User interactivity utilizes interactive components and state management tools in front-end technologies to ensure users can flexibly customize time ranges, select specific devices, and observe data changes in real time. The alarm and suggestion system is divided into real-time alarms and detailed maintenance suggestions. Real-time alarms are integrated with the real-time alarm system. When the AI ​​model detects potential maintenance needs or equipment abnormalities, it delivers alerts to users in real time through the interface and notifications. Combined with backend services and frontend frameworks, it uses WebSocket real-time communication technology to ensure the push of real-time alarms. Detailed maintenance suggestions provide detailed maintenance advice on the interface, explaining possible causes of failures and providing recommended maintenance measures. The report generation interface automatically summarizes the core quantitative indicators of AI assistant dialogue interactions within the period, including total dialogue requests, text noise reduction accuracy, prompt word template matching rate, and dialogue response accuracy. All indicator data comes from the system's real-time recorded dialogue interaction logs and is presented intuitively through data visualization, helping users quickly assess the overall usability and operational stability of the AI ​​assistant's dialogue function. This embodiment reduces manual intervention through the AI ​​recognition unified adaptation module and the AI ​​intelligent data processing module, automatically completing demand identification, template matching, data cleaning, and docking logic generation, improving docking efficiency and shortening the cycle. The AI ​​medical intelligent analysis and model training module constructs a fault prediction model, combined with remote intelligent monitoring... The real-time monitoring and multi-level early warning functions of the control and management central module can identify potential problems in advance and push repair solutions, enhancing system stability and reducing business delays. The AI-enhanced interactive visualization platform and hierarchical prompt word system simplify operations and terminology for non-technical personnel and provide detailed parameters for technical personnel, optimizing the user experience for all roles and lowering the operating threshold. At the same time, the system can automatically align with business specifications to reduce errors, adapt to multiple scenarios without repeated development, and the operation logs are traceable to facilitate problem troubleshooting. Moreover, the AI ​​model can be continuously optimized through new cases and specifications, improving efficiency and reducing maintenance costs in the long term. Ultimately, it achieves the core goals of AI-driven cost reduction and efficiency improvement, standard alignment to reduce risks, and ease of use to meet the needs of all roles.

[0017] In this embodiment, the system first relies on the AI-based unified adaptation module to receive user requests, extracts key features using NLP technology, retrieves the smart ward data docking document view library and historical docking case library, automatically identifies the protocol type of external devices / systems and generates adaptation logic, and simultaneously completes automatic generation of field mapping and rule verification to achieve rapid access to third-party docking data. Subsequently, the AI ​​intelligent data processing module obtains the third-party docking data, uses the smart ward docking document as a cleaning rule library, performs missing value, outlier, and format error identification and processing, and performs correlation verification, generates a cleaning report and pushes repair suggestions, and transmits the data to the next module after completing data standardization processing. Then, the AI ​​medical intelligent analysis and model training module receives the standardized data, uses 1,000 valid historical docking cases from the past two years as training data, and after data preprocessing (such as TF-IDF encoded text features, extraction and merging of structured view features), it uses the random forest algorithm (containing 50 decision trees) to build an equipment fault prediction model, and uses the Adam optimizer and early shutdown policy. After model training and validation are completed, the generated model provides an API interface for subsequent modules to call. Then, the remote intelligent monitoring and control center module deploys this fault prediction model to obtain prediction results in real time. Based on the "HTTP+WebSocket" dual protocol, it links with the data interaction center to realize remote monitoring of equipment (status display in charts), control (command verification and issuance), alarm (multi-level rule triggering and personalized notification), and log recording (structured storage and auditing). At the same time, relying on the interface template of the unified AI recognition module, the device data is synchronized to the hospital's existing system. Finally, the AI-enhanced interactive visualization platform and intelligent reporting module provide managers and medical staff with a data visualization interface to display fault warning information and maintenance suggestions, support maintenance plan formulation, and also collect statistics on the core indicators of AI assistant dialogue interaction, reducing the operation threshold for non-technical personnel. Moreover, the various modules are connected through standardized data interaction processes to adapt to the docking requirements of the smart ward system. At the same time, the system will iteratively optimize the model and prompt word template based on user feedback to improve the long-term use effect.

[0018] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A data interaction system for intelligent ward AI assistant collaborative view docking, characterized in that, It includes an AI intelligent data processing module, an AI recognition unified adaptation module, an AI medical intelligent analysis and model training module, a remote intelligent monitoring and control center module, and an AI-enhanced interactive visualization platform and intelligent reporting module; The AI ​​intelligent data processing module is used for data cleaning and standardization. The AI ​​recognition and unified adaptation module is used to automatically retrieve the smart ward data docking document view library and view docking based on the smart ward data docking document docking specification, automatically identify the protocol type of external devices and systems, and generate adaptation logic. The AI ​​medical intelligent analysis and model training module is used to receive standardized data and generate equipment fault prediction models through data preprocessing, feature extraction, AI algorithm integration and model training verification. The remote intelligent monitoring and control central module is equipped with a fault prediction model to obtain prediction results in real time, enabling remote monitoring, control, alarm, and log recording of equipment. The AI-enhanced interactive visualization platform and intelligent reporting module provide managers and medical staff with a data visualization interface and maintenance suggestions, supporting fault warning viewing and maintenance plan formulation.

2. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, The specific operational logic steps of the AI ​​intelligent data processing module are as follows: S101: Obtain third-party interface data through AI recognition unified adaptation module, and load smart ward interface documents as cleaning rule base; S102: AI automatically executes cleaning data and determines the cleaning pass rate; S103: If the cleaning pass rate is >95%, a report will be generated and the data will enter the docking process; S104: If the cleaning pass rate is less than 95%, then unrepaired data will be filtered out, and the AI ​​assistant will push the solution in the docking document. After the user repairs the data, the AI ​​cleaning will be triggered again.

3. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 2, characterized in that, In step S103, the generated report includes the amount of cleaned data, the distribution of error types, the repair rate, and a list of unrepaired data. The cleaning rule base in S101 includes cleaning types and AI automatic cleaning logic. The cleaning types include: missing value identification and processing, outlier identification and processing, format error identification and processing, and correlation verification. The corresponding AI automatic cleaning logic is as follows: (1) Missing value identification and processing: Mark missing data in required fields, mark and log them for feedback, use non-required fields to fill and record in the log for feedback, and trigger "backtracking and supplementing" when required fields are missing, call the AI ​​recognition unified adaptation module to re-acquire third-party data; (2) Outlier identification and handling: For outlier data, mark and log it, and provide real-time feedback to technical personnel; (3) Format error identification and handling: Validate data type and format and automatically convert to a repairable format; (4) Relationship Validation: Validate the relationship between fields across views, and mark "data orphaned" if the relationship fails; In addition, the AI ​​assistant records "new error types" during the cleaning process, automatically updates the "cleaning rule base," and synchronizes it to the smart ward data docking document maintenance records.

4. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, In the AI ​​recognition unified adaptation module, the core steps of AI view retrieval and retrieval verification and adjustment are as follows: S201: Demand Analysis Section. The AI ​​assistant receives the user's natural language demand and extracts key features through Natural Language Processing (NLP). S202: The AI ​​assistant calls the smart ward system to connect to the document view library and filters candidate templates through feature matching; S203: The AI ​​assistant searches the historical case database, calculates the similarity between the requirements and the cases, and thus confirms the target template; S204: The AI ​​assistant automatically generates field mapping relationships based on the view docking rules and the field definitions in the smart ward data docking document; S205: Check whether the mapping relationship in S204 meets the docking requirements of the smart ward system docking document; S206: If the requirements do not match the documentation for the smart ward system, the AI ​​assistant will provide feedback that "the bed code format must be standardized according to the documentation for the smart ward system". S207: Technicians make fine-tuning adjustments based on the AI ​​results, and the adjusted records are automatically stored in the case library to optimize the accuracy of subsequent AI retrieval.

5. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, The AI ​​medical intelligent analysis and model training module is used to enhance the AI ​​model and strengthen business training. Its specific logical steps are as follows: S3011: Data Collection: Collect metadata of 32 types of views from historical docking cases and smart ward docking documents in the past 2 years, exclude invalid samples, remove cases with "vague requirements and no docking results", and finally retain 1,000 valid samples. S3012: Data preprocessing: Clean the data using AI and divide it into training set, validation set and test set in a 7:2:1 ratio to ensure that the test set samples cover all core views of the smart ward data docking document; S3013: Algorithm Integration: The decision tree splitting criterion is adopted. For each feature, the Gini coefficient after splitting is calculated, the feature with the smallest coefficient is selected as the splitting node, and the supervised learning algorithm Random Forest is selected to handle complex relationships and a large number of features. S3014: Constructing a CNN architecture: Designing convolutional and pooling layers to effectively capture the spatial and temporal features of the input data, and introducing fully connected layers to map the features output by the convolutional layers to the final prediction results; S3015: Model Training: Use historical data to train the model, divide it into training and validation sets, and select a loss function to optimize according to the problem type; S3016: Parameter tuning: Iteratively tune the hyperparameters of the CNN using regularization methods, including kernel size, pooling size, and learning rate; S3017: Model Validation and Evaluation: Evaluate model performance on the validation set, focusing on accuracy and recall metrics, analyzing the model's performance across different categories, and understanding its ability to handle and predict different needs.

6. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, The AI ​​medical intelligent analysis and model training module is used to enhance the AI ​​model's dialogue capabilities. Its specific logical steps are as follows: S3021: User Request Initiation: Technical / non-technical personnel submit and initiate a request; S3022: Request reception and intent recognition: AI automatically determines that the user's request belongs to the "query / fault / consultation" scenario, matches the corresponding template, and stores key information from recent conversations; S3023: Layered prompt word matching and generation: Perform basic layer calling, personalized layer generation and scene layer filling; S3024: Dialogue content generation and output: Generate and output one of the following: fault repair plan, data operation steps, status feedback information, and output language style that is automatically adapted according to the user role; S3025: User Feedback Collection: Set up a simple feedback entry point, including: Technical personnel: feedback on "insufficient details" or "whether the solution is compliant", Non-technical personnel: feedback on "don't understand" or "whether it has been resolved"; S3026: Feedback-driven optimization: Simplify evaluation metrics: Focus on "scene adaptation rate", "output compliance rate" and "user satisfaction", avoid too many numbers, regularly convert effective feedback into training data, update prompt word templates and models, and ensure continuous optimization of dialogue effects.

7. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, The operational logic steps of the remote intelligent monitoring and control central module are as follows: S401: Model Integration and Deployment: Deploy the trained AI model to the server and connect it to the data interaction center to achieve unified flow of model data. The AI ​​assistant monitors the model deployment status in real time and provides anomaly warnings. S402: Model Prediction Value Acquisition: The web interface allows users to configure data acquisition strategies through the AI ​​assistant. Once triggered, the web client sends a request to the data interaction center, which then uniformly calls the model API to obtain the prediction values. The AI ​​assistant performs preliminary cleaning on the obtained prediction values, which are then synchronized to the web client by the data interaction center for monitoring the real-time status of the data. At the same time, historical prediction value data is stored in the data center for archiving. S403: Remote monitoring: The core data source is the model prediction value synchronized by the data interaction center. The web interface uses charts to display the device status. A real-time connection is established with the data interaction center through AJAX+WebSockets technology to realize dynamic updates of front-end data. The AI ​​assistant assists in analyzing the trend of prediction values, marks abnormal data points on the web interface, and provides trend prediction to improve the intelligence of monitoring. S404: Remote Control: The web interface is designed as the device control operation area. After the user initiates a control command, the AI ​​assistant first verifies the validity of the command. If the verification is successful, the command is transmitted to the data interaction center via the API interface. The data interaction center standardizes the format of the command and then forwards it to the server. Finally, the server sends the command to the target device. At the same time, the AI ​​assistant feeds back the command execution result to the web interface and synchronizes it to the data center for archiving. S405: Alarms and Notifications: Based on the model prediction values ​​synchronized by the data interaction center, multi-level alarm rules are set. When the prediction value triggers the threshold, the data interaction center pushes the alarm signal to the AI ​​assistant. The AI ​​assistant intelligently determines the alarm level and generates personalized notification content based on the user's preset notification preferences. At the same time, it links the data interaction center to synchronize alarm information to the web terminal to ensure that users respond to device abnormalities in a timely manner. S406: Log Recording and Auditing: All user operations are synchronized to the data interaction center in real time. The data center completes the structured storage of operation logs. The AI ​​assistant regularly performs intelligent analysis of the logs, identifies abnormal operations, generates audit reports, and works with the log management system to perform regular cleanup and encrypted backup. The data interaction center ensures the integrity and traceability of the log archive. S407: Remote Diagnosis and Maintenance: Integrates remote diagnostic tools and interfaces with the data interaction center. After technicians initiate a diagnostic request via the web, the device's real-time operating data is transmitted to the diagnostic tool through the data interaction center. The AI ​​assistant analyzes the data simultaneously to help technicians quickly locate faults. When the device is remotely upgraded, the AI ​​assistant verifies the integrity and compatibility of the firmware / software package. The upgrade package is transmitted to the device in segments through the data interaction center to ensure a stable upgrade process. At the same time, the AI ​​assistant records the upgrade progress and provides feedback to the web interface.

8. The data interaction system for intelligent ward AI assistant collaborative view docking according to claim 1, characterized in that, The AI-enhanced interactive visualization platform and intelligent reporting module include a data visualization interface, an alarm and suggestion system, and a report generation interface; The data visualization interface is designed using a development framework to build an intuitive user interface and uses a CSS framework to ensure responsive design that adapts to different devices and screen sizes. Charts are presented using a chart library to ensure clear and attractive visualizations. User interactivity utilizes interactive components and state management tools in front-end technologies to ensure that users can flexibly customize time ranges, select specific devices, and observe data changes in real time. The alarm and suggestion system is divided into real-time alarms and detailed maintenance suggestions. Real-time alarms are integrated into the real-time alarm system. When the AI ​​model detects potential maintenance needs or equipment abnormalities, it delivers alarms to users in real time through the interface and notifications. Combined with backend services and frontend frameworks, it uses WebSocket real-time communication technology to ensure the push of real-time alarms. Detailed maintenance suggestions provide detailed maintenance advice on the interface, explaining possible causes of failures and providing recommended maintenance measures. The report generation interface automatically summarizes the core quantitative indicators of AI assistant dialogue interaction within the period, including total number of dialogue requests, text noise reduction accuracy, prompt word template matching rate, and dialogue response accuracy. The indicator data are all from the dialogue interaction logs recorded by the system in real time, and are presented intuitively through data visualization to help users quickly evaluate the overall usability and operational stability of the AI ​​assistant dialogue function.

Images