A nursing sterile operation management system

Abstract

The application discloses a kind of nursing sterile operation management and control system, it is related to nursing control technical field, including multimodal data acquisition module, AI intelligent identification module, early warning intervention module, collaborative control module and quality control training module;Multimodal data acquisition module is used to obtain the image data and environmental parameter of sterile operation scene, all-element data of sterile article information, is transmitted to AI intelligent identification module after pre-processing;AI intelligent identification module is used to identify and analyze image data, judge illegal behavior and generate identification result and push to early warning intervention module;Early warning intervention module is used to execute hierarchical early warning intervention according to identification result, meanwhile, the data related to disposal backflow to AI intelligent identification module forms micro closed loop;Collaborative control module is used to integrate closed loop data to realize multi-department intercommunication and sterile article whole life cycle traceability;Quality control training module is used to generate quality control report based on whole process data and push customized training resources.

Description

Technical Field

[0001] This invention relates to the field of nursing management technology, and in particular to a nursing aseptic operation management system. Background Technology

[0002] In current hospital aseptic operation management scenarios, different departments such as operating rooms, ICUs, and general wards have significantly different aseptic operation standards due to differences in diagnosis and treatment needs. For example, the requirements for surgical instrument transfer and disinfection time thresholds vary. Moreover, medical data privacy protection regulations are becoming increasingly strict, and the risk of leakage of sensitive data such as patient personal information and department-specific operating procedures is enormous. Traditional management models mainly rely on manual on-site verification and post-event traceability, which is not only inefficient and unable to cover the entire operation process, but also suffers from problems such as missed judgments and high false judgment rates. In existing intelligent management systems, if data is shared across departments to optimize models, it is easy to lead to the leakage of sensitive information. If data is strictly isolated, the AI ​​recognition models of each department will only be adapted to local scenarios, with weak generalization ability and poor adaptability, which cannot meet the needs of accurate and safe aseptic operation management under a unified standard across the hospital. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by proposing a nursing aseptic operation control system.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A nursing aseptic operation control system, characterized in that it includes a multimodal data acquisition module, an AI intelligent recognition module, an early warning and intervention module, a collaborative control module, and a quality control training module; The multimodal data acquisition module is used to acquire image data and environmental parameters of the aseptic operation scenario, as well as all elements of aseptic item information, and transmits them to the AI ​​intelligent recognition module after preprocessing. The AI ​​intelligent recognition module is used to identify and analyze image data, determine violations, and generate recognition results that are pushed to the early warning and intervention module. The early warning and intervention module is used to perform graded early warning and intervention based on the identification results, and at the same time, the relevant data of the handling is fed back to the AI ​​intelligent identification module to form a micro closed loop; The collaborative management module is used to integrate closed-loop data to achieve inter-departmental communication and full life-cycle traceability of sterile items. The quality control training module is used to generate quality control reports based on full-process data and push customized training resources.

[0005] The above technical solution further includes: Specifically, the multimodal data acquisition module includes: a high-definition AI camera, an RFID reader / writer, a temperature and humidity sensor, and a smart bracelet; The high-definition AI camera is used to capture image data of the aseptic operation scenario, as well as the operation behavior of medical staff and the environmental images of the operation area; The RFID reader / writer is used to bind and collect information about sterile items; The temperature and humidity sensor is used to monitor parameters related to the cleanliness of the operating environment; The smart bracelet is used to receive early warning information and coordinate with medical staff; The aforementioned devices achieve real-time data aggregation and timestamp alignment through an edge computing gateway.

[0006] Specifically, the multimodal data acquisition module is also equipped with a data preprocessing unit; The data preprocessing unit uses a combination of AI pre-annotation and manual precise verification to process the collected image data and environmental parameter-related data, and introduces a federated learning framework to share annotated data across departments while protecting the data privacy of each department.

[0007] Specifically, the AI ​​intelligent recognition module also includes a personalized compliance standard library; The personalized compliance standard library was compiled by nursing quality control experts in conjunction with the aseptic operation standards of various departments. It contains quantitative parameters that can be recognized by AI. These quantitative parameters include at least the disinfection time threshold, the boundary coordinate range of the aseptic area, the glove wearing action node, and the expiration date threshold of sterile items.

[0008] Specifically, the AI ​​intelligent recognition module also incorporates a deep learning model that integrates CNN convolutional neural networks and pose estimation algorithms; The deep learning model is deployed on edge computing nodes to perform AI image recognition and analysis on the image data, subdividing violations into static violations and dynamic violations and achieving real-time analysis. The static violations include not wearing masks or hats and using damaged sterile packs, while the dynamic violations include crossing the sterile area with one's hands and improper handling of instruments.

[0009] Specifically, the early warning intervention module constructs a three-level early warning response mechanism, specifically used for: When a minor violation is detected, the smart bracelet will send a vibration reminder and standardized operating instructions to the medical staff. When a moderate violation is detected, an early warning message is simultaneously pushed to the head nurse's quality control terminal and the violation details are automatically recorded. The head nurse's quality control terminal is a supporting terminal for the early warning and intervention module. When a serious violation is detected, the on-site audible and visual alarm device is triggered and the violation video segment is locked. At the same time, the relevant information is pushed to the hospital quality control platform. The on-site audible and visual alarm device and the hospital quality control platform are respectively the early warning execution device and the information aggregation platform of the early warning intervention module.

[0010] Specifically, the early warning intervention module also has an early warning feedback data feedback function, which is specifically used for: The system transmits the corrective actions and early warning results of medical staff’s violations to the AI ​​intelligent recognition module in real time, which serves as the basis for optimizing the parameters of the deep learning model, thus improving the micro-closed loop of recognition, early warning, correction and optimization.

[0011] Specifically, the quality control training module is used for: Automatically generate three-dimensional compliance reports for departments, individuals, and operation types; use data mining to identify high-frequency violation scenarios and high-risk groups. Customized training resources are pushed based on individual violation records, and VR technology is used to transform high-frequency violation scenarios into simulated training scenarios.

[0012] Specifically, the collaborative management module is used for: By connecting with the hospital's sterile supplies supply center management system, the sterile item tag information read by the RFID reader is combined with AI image recognition to achieve full-chain traceability of sterile items from sterilization, storage, transportation, use to recycling; Establish data links between operating rooms, ICUs, and general wards to achieve seamless aseptic control before, during, and after surgery, and integrate with the hospital's infection control system to incorporate compliance rate into infection control assessment indicators.

[0013] Specifically, it also includes a data visualization module; The data visualization module displays data on the aseptic operation compliance rate, common violation types, and early warning and handling efficiency of each department in real time through the quality control center's large visualization screen.

[0014] The present invention has the following beneficial effects: In this invention, by integrating the federated learning framework with the AI ​​pre-annotation-human verification data processing mode, the value sharing of cross-departmental annotated data and the collaborative optimization of AI recognition models are achieved while ensuring the privacy and security of the original data of aseptic operations in each department. This solution not only avoids the privacy leakage risk in traditional cross-departmental data sharing, but also improves the accuracy of AI in identifying violations through iteratively optimized annotation models, solving the recognition adaptability problem caused by differences in aseptic operation standards in different departmental scenarios. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the process of a nursing aseptic operation control system proposed in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example

[0017] like Figure 1 As shown, the present invention proposes a nursing aseptic operation control system, which includes a multimodal data acquisition module, an AI intelligent recognition module, an early warning and intervention module, a collaborative control module, and a quality control training module. The multimodal data acquisition module is used to acquire image data and environmental parameters of aseptic operation scenarios, as well as full-element data of aseptic items. After preprocessing, the data is transmitted to the AI ​​intelligent recognition module. The AI ​​intelligent recognition module is used to identify and analyze image data, determine violations, and generate recognition results that are pushed to the early warning and intervention module. The early warning and intervention module is used to execute tiered early warning and intervention based on the identification results, while simultaneously feeding back relevant data to the AI ​​intelligent identification module to form a micro-closed loop; The collaborative management module is used to integrate closed-loop data to achieve inter-departmental communication and full lifecycle traceability of sterile items; The quality control training module is used to generate quality control reports based on full-process data and push customized training resources.

[0018] Furthermore, in the aseptic operation scenario of the operating room, the multimodal data acquisition module captures image data such as the transfer of surgical instruments by medical staff and the operation in the aseptic area through a high-definition AI camera. It obtains environmental parameters such as temperature (22-25℃) and humidity (40%-60%) of the operating environment with the help of temperature and humidity sensors. It reads information such as sterilization batch and expiration date of surgical instruments, sterile dressings and other items through an RFID reader. After preprocessing these full-element data by image denoising, data format standardization and timestamp alignment, the data is transmitted to the AI ​​intelligent recognition module in real time. The AI ​​intelligent recognition module calls the preset operating room aseptic operation compliance standards, analyzes the received image data frame by frame, identifies whether medical staff have violated regulations such as not wearing sterile gloves or crossing the sterile area with their hands, generates recognition results including the type of violation, the time of the violation, and a screenshot of the violation, and pushes them to the early warning and intervention module. The early warning and intervention module executes tiered early warning and intervention based on the identification results: For minor violations (such as misplaced sterile items), a vibration alert will be sent to the medical staff operating the device via smart bracelet. If the violation is moderate (such as insufficient disinfection time), an alert will be sent to the head nurse's terminal simultaneously. If it is a serious violation (such as using expired sterile items), an on-site audible and visual alarm will be triggered, and relevant data such as the violation correction status and feedback from the handling personnel will be fed back to the AI ​​intelligent recognition module in real time. The collaborative management module integrates this micro closed-loop data, enabling data sharing and interoperability among multiple departments such as operating rooms, ICUs, and general wards. On the other hand, it achieves full lifecycle traceability of sterile items from sterilization, warehousing, requisition, use to recycling through RFID tracking and data association. Based on the aforementioned full-process data, the quality control training module automatically generates quality control reports covering the violation rate of each department, high-frequency violation types, and individual operational compliance. At the same time, it pushes customized training resources such as operation specification videos and online quizzes to relevant medical staff for high-frequency violation scenarios (such as non-standard instrument transfer), ensuring continuous optimization of aseptic operation management.

[0019] The multimodal data acquisition module specifically includes: a high-definition AI camera, an RFID reader / writer, a temperature and humidity sensor, and a smart bracelet; High-definition AI cameras are used to capture image data of aseptic operation scenarios, as well as the operation behavior of medical staff and the environmental images of the operation area; RFID readers are used to bind and collect information about sterile items; Temperature and humidity sensors are used to monitor parameters related to the cleanliness of the operating environment; Smart bracelets are used to receive early warning information and coordinate with medical staff; The aforementioned devices achieve real-time data aggregation and timestamp alignment through an edge computing gateway.

[0020] Furthermore, all components of the multimodal data acquisition module start working synchronously: A high-definition AI camera is aimed at the medical staff's hand operation area and infusion operation table, capturing image data such as disinfection actions before puncture and the process of putting on sterile gloves in real time. At the same time, it records environmental images of the operation area to preserve the complete operation scene. The RFID reader is pre-bound to the RFID tags of sterile items such as infusion sets and sterile dressings to be used. When medical staff take the items, the reader automatically reads the core information stored in the tag, such as sterilization batch, sterilization time, and expiration date. Temperature and humidity sensors continuously monitor cleanliness-related parameters of the operating environment, collecting data on the current temperature of 23°C and relative humidity of 45% in the ward. The smart bracelet is in standby receiving mode, ready to receive warning information from the subsequent warning and intervention module to coordinate with medical staff. All image data, sterile item information, environmental parameters, and other data collected by the devices are aggregated in real time through the edge computing gateway. The gateway performs timestamp alignment processing on the data transmitted by the above devices (uniformly marked with the precise time of the operation, such as "2024-05-20 09:30:05") to ensure the spatiotemporal consistency of data from different sources, providing a standardized data foundation for subsequent data preprocessing and identification analysis.

[0021] The multimodal data acquisition module is also equipped with a data preprocessing unit; The data preprocessing unit uses a combination of AI pre-annotation and manual precise verification to process the collected image data and environmental parameter-related data. It also introduces a federated learning framework to share labeled data across departments while protecting the data privacy of each department.

[0022] Furthermore, in scenarios where multiple departments in a hospital collaborate to manage aseptic procedures, after the preprocessing unit of the multimodal data acquisition module starts working, it first receives multimodal equipment acquisition data from departments such as the operating room, ICU, and general wards. This includes image data of aseptic procedures performed by medical staff (such as images of surgical instrument transfer and intravenous infusion disinfection) and environmental parameters related to the corresponding operating environment, such as temperature, humidity, and cleanliness. Subsequently, the data is processed using a combination of AI pre-annotation and precise manual verification. The AI ​​pre-labeled formula is: ; Formula parameter description parameter Meaning (Adaptation to aseptic operation scenarios) I Input images of aseptic procedures (such as operating room footage, infusion sterilization images). Lauto(I) AI-generated image tag collection N The number of target categories to be labeled in the image (e.g., medical staff, sterile packages, and sterile areas are 3 categories). Ci Category labels for the i-th type of target (e.g., C1="Medical staff", C2="Sterile packages", C3="Sterile area") Bi The bounding box coordinates of the i-th type of target are in the format (xmin, ymin, xmax, ymax) (e.g., the bounding box of the sterile area, the box for the placement of sterile packages). M The number of key points for a single type of target (e.g., two key points for a healthcare worker's wrist and elbow). Ki,j The coordinates of the j-th key point of the i-th target are in the format (x, y) (e.g., wrist coordinates K1,1=(120,250)). In the scenario of aseptic operation control in multiple departments of a hospital, the multimodal data acquisition module of each department's edge node first uploads the pre-processed operation image I (such as the operating room instrument transfer screen, the ward infusion disinfection screen) and the corresponding environmental parameters (temperature, humidity, cleanliness values) to the local data preprocessing unit. The preprocessing unit calls a pre-trained target detection and key point localization model to generate an AI automatic label set Lauto(I) through a formula. Ci labels target categories such as "medical staff", "sterile package", and "sterile area", Bi outputs the bounding box coordinates of each target, and Ki,j locates key action points such as the wrist and elbow of medical staff. At the same time, environmental parameters are timestamped with image labels (e.g., "2024-06-15 10:20:05 Temperature 23℃"). Subsequently, the hospital's quality control department formed a manual verification team to check and correct each Lauto(I), focusing on correcting the misjudgment issues in AI annotation (such as mislabeling "not wearing gloves" as "wearing gloves" and offsetting the coordinates of the sterile area boundary box), and supplementing the environmental parameter association annotation (such as associating "humidity 65%" with "risk of sterile package getting damp"), and finally generating an accurate set of manually corrected labels Lmanual(I); Next, the preprocessing unit introduces the federated learning framework. Each department retains its local original image data and corrected label data, does not transmit sensitive data to the outside world, and only encrypts and uploads the model training gradient parameters corresponding to the label set to the federated learning center node. Federated learning is a distributed machine learning architecture that protects data privacy. Its core operating logic is as follows: raw sensitive data such as aseptic operation images, AI-automated labels, and manually verified accurate labels from each department (operating room, ICU, general ward) are stored on the local server of the department and are not transmitted across departments. Each department only uses local data to train the AI ​​labeling model and extracts the gradient parameters required for model optimization. The central node aggregates gradient parameters from various departments, optimizes and upgrades the AI ​​annotation model, and then distributes the optimized model parameters to the local nodes of each department. Each department uses the upgraded model to perform AI pre-annotation, forming a closed loop of "automatic annotation - manual verification - federated optimization - iterative annotation". This not only realizes the value sharing of cross-departmental annotated data and protects the privacy data of each department, but also outputs standardized and high-precision annotated data for subsequent use by the AI ​​intelligent recognition module.

[0023] The AI ​​intelligent recognition module also includes a personalized compliance standard library; The personalized compliance standard library was compiled by nursing quality control experts in conjunction with the aseptic operation standards of various departments. It contains quantitative parameters that can be recognized by AI. The quantitative parameters include at least the disinfection time threshold, the boundary coordinate range of the aseptic area, the glove wearing action node, and the expiration date threshold of sterile items.

[0024] During the construction phase of the hospital nursing aseptic operation control system, the development of the personalized compliance standard library for the AI ​​intelligent recognition module proceeded according to the following process: First, led by hospital nursing quality control experts, senior head nurses from various departments such as the operating room, ICU, and general wards comprehensively reviewed the specific aseptic operation standards for different departments and clarified personalized control requirements based on the differences in clinical operation scenarios in each department. Subsequently, the expert team transformed the refined standards into quantitative parameters that could be recognized by AI. For the operating room, they set a disinfection time threshold of ≥30 seconds and a sterile area boundary coordinate range of 0-80cm x-axis and 0-60cm y-axis on the operating table. For the ICU, they specified that the glove-wearing action node was "the duration of non-sterile hand contact with the glove-wearing area is ≥5 seconds" and the sterile item expiration date threshold was "≤7 days from expiration". For general wards, the disinfection time threshold for aseptic intravenous infusion procedures is set at ≥15 seconds, and the distance between sterile and non-sterile instruments is set at ≥15cm. Finally, the quantitative parameters of these departments are classified and entered into the standard library to form a structured and callable personalized compliance judgment basis, which is then integrated into the AI ​​intelligent recognition module to provide a unified and scenario-appropriate judgment standard for subsequent AI to accurately identify aseptic operation violations in different departments.

[0025] The AI ​​intelligent recognition module also has a built-in deep learning model that integrates CNN convolutional neural networks and pose estimation algorithms; Deep learning models are deployed on edge computing nodes to perform AI image recognition and analysis on image data, subdividing violations into static and dynamic violations and achieving real-time analysis. Static violations include not wearing masks or hats and using damaged sterile packs, while dynamic violations include crossing the sterile area with your hands and improper handling of instruments.

[0026] Furthermore, in the AI ​​intelligent recognition module of the nursing aseptic operation control system, the pose estimation algorithm parameters are tailored to the hand / limb movement recognition scenarios of medical staff, as detailed below: 1. Key point confidence formula (locating key points of movement such as wrist and fingertips); ; : The probability that the image coordinate (x,y) is a key point of the kth class (k=1 represents the wrist, k=2 represents the fingertip); ( , ): The standard location of this type of critical point (e.g., the coordinates of the compliant area where the wrist should be during aseptic operation); σ=5 (Adapts to small-range motion recognition on the workbench) 2. Limb association formula (connecting key points to form arm trajectory): ; : The direction vector of the arm at coordinates (x, y); ( , =Wrist coordinates, ( , = Elbow coordinates, used to determine whether the arm crosses the boundary of the sterile area.

[0027] The deep learning model built into the AI ​​intelligent recognition module, which integrates CNN convolutional neural network and pose estimation algorithm, starts running on the edge computing node and receives surgical operation image data preprocessed by the multimodal data acquisition module in real time. The model first extracts key information such as the limb features of medical staff, the outline of sterile items, and the markings of sterile areas in the image through a CNN convolutional neural network. Then, it calls the pose estimation algorithm to perform analysis and uses the key point confidence formula to locate the coordinates of key points such as the wrist (x1, y1) and elbow (x2, y2) of medical staff. It generates the arm trajectory vector through the limb association formula and makes a judgment by combining the quantitative parameters in the personalized compliance standard library. If the image shows medical staff not wearing masks or the sterile packs on the operating table are damaged, it will be judged as a static violation. If the coordinates of the arm trajectory vector are found to be outside the boundary coordinate range of the sterile area (e.g., 0-80cm x-axis and 0-60cm y-axis of the operating table) by the attitude estimation algorithm, it is determined to be a dynamic violation of the hand's trajectory crossing the sterile area. If the relative position and movement trajectory of the wrist and the instrument during the instrument transfer process do not conform to the standard action nodes, it is judged as a dynamic violation of the non-standard instrument transfer action. The entire recognition and analysis process is completed locally on the edge computing node, taking ≤500ms, achieving real-time and accurate judgment of violations.

[0028] The early warning and intervention module constructs a three-level early warning response mechanism, specifically used for: When a minor violation is detected, a vibration reminder and standardized operating instructions will be pushed to the operating medical staff via smart bracelet; When a moderate violation is detected, an early warning message is simultaneously pushed to the head nurse's quality control terminal and the violation details are automatically recorded. The head nurse's quality control terminal is a supporting terminal for the early warning and intervention module. When a serious violation is detected, the on-site audible and visual alarm device is triggered and the violation video segment is locked. At the same time, the relevant information is pushed to the hospital quality control platform. The on-site audible and visual alarm device and the hospital quality control platform are respectively the early warning execution device and the information aggregation platform of the early warning intervention module.

[0029] The rule matching algorithm of the three-level early warning response mechanism uses condition judgment logic to correspond the violation level with the early warning action; Furthermore, in the aseptic dressing change operation scenario in the hospital operating room, the AI ​​intelligent recognition module first extracts features such as the mask wearing status of medical staff, the integrity of sterile packages, and the contours of hands and limbs in the image through a CNN convolutional neural network. Then, it calls the pose estimation algorithm and uses the key point confidence formula to locate key points of movement such as wrists and elbows. It generates arm movement trajectory vectors through limb association formulas and combines quantitative parameters in the personalized compliance standard library (such as the boundary coordinates of the sterile area and the disinfection time threshold) to identify three types of violations: "slight deviation of sterile package placement", "operation without sterile gloves", and "use of expired sterile gauze". According to the preset standards, they are judged to be minor, moderate and serious violations respectively. The rule matching algorithm of the early warning and intervention module receives the violation behavior feature value V and level label output by the AI ​​intelligent recognition module, calls the built-in hierarchical early warning rule library, and executes the condition judgment logic: When V∈V1 (minor violation set) is determined, the R1 response action is triggered, and a vibration reminder and standardized placement instructions are pushed to the operating medical staff through the smart bracelet; When V∈V2 (the set of moderate violations), the R2 response action is triggered, and an early warning information is pushed to the head nurse's quality control terminal simultaneously, and details such as the time and personnel involved in the violation are automatically recorded. When V∈V3 (the set of serious violations), the R3 response action is triggered, which activates the on-site audible and visual alarm device, locks the violation video segment, and pushes the information to the hospital's quality control platform, thereby achieving seamless integration of violation identification and tiered early warning. R1, R2, and R3 respectively represent the complete set of early warning response actions corresponding to minor, medium, and severe violations. Their specific meanings in the context of the system are as follows: R1: The combination of warning response actions for minor violations is to push vibration reminders and standardized operation guidelines to the operating medical staff through a smart bracelet. For example, when a slight deviation in the position of the sterile towel is detected, this action is triggered, and only the operating medical staff are reminded to correct the deviation.

[0030] R2: The warning response action combination for moderate violations is as follows: push reminders to the smart wristbands of medical staff performing the operation + simultaneously push warning information to the head nurse's quality control terminal + the system automatically records details such as the time, personnel, and behavior of the violation. For example, if the system detects that a medical staff member is preparing to perform an operation without wearing sterile gloves, this action is triggered, both reminding the medical staff member and simultaneously notifying the head nurse to intervene and manage the situation.

[0031] R3: The combination of early warning response actions for serious violations, specifically triggering the on-site audible and visual alarm device, automatically locking video clips before and after the violation, and pushing violation details and video links to the hospital's quality control platform. For example, when expired sterile instruments are detected, this action is triggered to issue an on-site warning, preserve evidence, and report to the hospital's quality control department for supervision.

[0032] The early warning and intervention module also has an early warning feedback data feedback function, specifically used for: The system transmits the corrective actions and early warning results of medical staff’s violations to the AI ​​intelligent recognition module in real time, which serves as the basis for optimizing the parameters of the deep learning model, thus improving the micro-closed loop of recognition, early warning, correction and optimization.

[0033] Furthermore, corrective actions by healthcare workers refer to rectified behaviors that comply with aseptic operation standards, taken by healthcare workers after receiving a warning alert in response to violations identified by the system. These actions must be matched with the corresponding violation level and type, as shown in the following example: Corrective actions for minor violations (triggering an R1 warning) When the smart bracelet sends reminders such as "sterile pack misplaced" or "suction tube not covered back in sterile box in time", medical staff must immediately adjust the placement of the sterile pack to the compliant area, put the suction tube back in place and cover the sterile box tightly to ensure that the operating environment meets the sterile requirements.

[0034] Corrective actions for moderate violations (triggering an R2 warning) When a reminder is received from the wristband and a simultaneous warning is issued by the head nurse's terminal (such as "Preparing to operate without wearing sterile gloves" or "Preparing to catheterize without disinfecting the urethral opening"), medical staff must immediately stop the current operation, put on sterile gloves according to the standard, and re-disinfect the urethral opening. After the rectification is completed, the staff must report the rectification to the head nurse and wait for the head nurse to confirm compliance before continuing the operation.

[0035] Corrective actions for serious violations (triggering an R3 warning) When an on-site audible and visual alarm is triggered (such as "using expired sterile instruments" or "damaged catheterization kit not replaced"), medical staff must immediately stop all operations, replace the instruments or catheterization kits with compliant sterile ones, and cooperate with hospital quality control personnel to retrieve videos of violations and explain the reasons for the violations. They must also participate in targeted aseptic operation training and pass the assessment to prevent similar problems from happening again.

[0036] The early warning response result refers to the final state achieved after intervention and correction of the violation following the triggering of the three-level early warning response mechanism, along with the total number of response records retained by the system. This includes both the rectification effectiveness of on-site operations and the closed-loop records between the system and the management end, specifically divided into three levels as follows: The outcome of minor violations (triggered by R1): After receiving a reminder from the wristband, medical staff immediately correct the violation (such as adjusting a misaligned sterile towel or covering an unsealed sterile container), and the operation returns to a compliant state; the system automatically records a simple log of "warning trigger time - correction completion time - violation type", without needing to report to the management end.

[0037] The outcome of handling a moderate violation (triggered R2): Medical staff stop the violation and complete the rectification (such as wearing sterile gloves again and re-performing the disinfection procedure). The head nurse checks the rectification status through the quality control terminal and signs to confirm it. The system retains a complete handling record, including "violation details - early warning push record - head nurse's verification opinion - screenshot of rectification completion", which is included in the department's monthly quality control statistics.

[0038] The outcome of handling serious violations (triggered R3): Upon the on-site audible and visual alarm, medical staff immediately stop using the non-compliant items (such as expired instruments or damaged sterile packs) and replace them with compliant consumables. The hospital's quality control platform simultaneously receives the violation information and the locked video clips.

[0039] The quality control training module is specifically used for: Automatically generate three-dimensional compliance reports for departments, individuals, and operation types; use data mining to identify high-frequency violation scenarios and high-risk groups. Customized training resources are pushed based on individual violation records, and VR technology is used to transform high-frequency violation scenarios into simulated training scenarios.

[0040] Furthermore, this module first connects to the system's backend storage of data such as departmental violation records, early warning handling results, and medical staff operation logs, automatically generating a three-dimensional compliance report of department-individual-operation type: The statistics are compiled from the perspective of departments, including operating rooms, ICUs, and general wards, to show the compliance rate and distribution of violation types of aseptic operations; from the perspective of individuals, the number of violations, the level of violations, and repeated violations of each medical staff member are recorded. We analyzed high-frequency violations in procedures such as aseptic dressing changes, catheterization, and sputum suction by operation type. At the same time, we used data mining technology to identify high-frequency violation scenarios (such as nighttime emergency aseptic catheterization and instrument transfer on the operating table) and high-risk groups (such as new nurses with less than one year of service and medical staff who frequently participate in emergency operations), thus forming a precise problem profile.

[0041] Next, based on each medical staff member's personal violation record, the module pushes customized training resources: for medical staff who repeatedly perform "operation without sterile gloves", it pushes videos on the standard wearing of sterile protective equipment and a question bank for operation assessment; for those who frequently "cross the boundary of the sterile area", it pushes diagrams of standardized limb movements and typical case analysis documents.

[0042] Finally, the module uses VR technology to transform the selected high-frequency violation scenarios into immersive simulation training scenarios. For example, it recreates real violation scenarios such as "damage to a sterile catheterization kit in an emergency" and "hands crossing sterile areas during surgical instrument transfer." Medical staff can wear VR devices to enter the simulation scenario and practice how to properly handle sudden violations and standardize their own operating actions in a virtual environment. Through repeated practice, they can strengthen their memory of compliant operations. At the same time, the module records the operation data of medical staff in the VR scenario to evaluate the training effect and form a quality control training closed loop of "data-based problem identification - customized training - simulation training - effect evaluation".

[0043] The collaborative management module is specifically used for: By connecting with the hospital's sterile supplies supply center management system, the information on sterile item tags read by RFID readers is combined with AI image recognition to achieve full-chain traceability of sterile items from sterilization, storage, transportation, use to recycling. Establish data links between operating rooms, ICUs, and general wards to achieve seamless aseptic control before, during, and after surgery, and integrate with the hospital's infection control system to incorporate compliance rate into infection control assessment indicators.

[0044] Furthermore, this module first establishes a real-time connection with the hospital's sterile supplies supply center management system. With the help of RFID readers deployed in the supply center's sterilization room, storage warehouse, departmental transport vehicles, ward usage points, and recycling temporary storage area, it reads the unique identification information in the tag of each sterile item (such as sterile surgical packs, urinary catheter packs, and suction catheters), including basic data such as sterilization time, sterilization batch, expiration date, storage conditions, and transport route. At the same time, these RFID data are combined with image data collected by the AI ​​intelligent recognition module for verification. In the supply center, the integrity of the sterile package seal is confirmed by image recognition and compared with the "intact" status of the RFID tag. During departmental transfers, image recognition is used to monitor whether the transfer boxes are properly sealed, and cross-verification is performed with the transfer time and route recorded by RFID. In clinical use, image recognition is used to confirm that the label information of sterile items used by medical staff matches the patient's surgical / treatment needs and is within the validity period; In the recycling process, image recognition confirms that used sterile items are classified according to medical waste standards and is linked to RFID recycling records. This enables full-chain traceability of sterile items from sterilization production → warehouse storage → departmental transfer → clinical use → waste recycling. If any abnormality occurs in any link (such as damaged sterile packs, expired use, or overdue transfer), the module will immediately trigger an alert to prevent unauthorized transfer.

[0045] Building upon this foundation, the module further integrates the aseptic control data links across three core scenarios: the operating room, ICU, and general wards, achieving seamless integration from pre-operative to intra-operative to post-operative care. Before the operation, the operating room retrieved the RFID data of sterile supplies from the supply center through a module, and combined it with AI image recognition to confirm that the surgical pack was in place and in compliance with regulations. At the same time, the patient's basic information was synchronized to the ICU and ward. During the operation, the module collects aseptic operation compliance data in the operating room in real time (such as whether the boundary of the sterile area is crossed and whether the instrument transfer is standardized) and pushes it to the hospital infection control system. After surgery, when the patient is transferred to the ICU or a general ward, the module will synchronize the intraoperative aseptic control record to the corresponding department. The ICU and ward can retrieve the information on the sterile items used by the patient during the operation through the module, continue the postoperative aseptic nursing control, and avoid control gaps when connecting across scenarios.

[0046] It also includes a data visualization module; The data visualization module displays real-time data on the aseptic operation compliance rate, common violation types, and early warning and handling efficiency of each department through the quality control center's large visualization screen.

[0047] Furthermore, by integrating and connecting the backend data from modules such as AI intelligent recognition, early warning intervention, collaborative management and control, and quality control training, the hospital cleans and summarizes core indicators such as the compliance rate of aseptic operation, common violation types, number of early warning triggers and handling efficiency, and abnormal traceability of aseptic items in all departments. This data is then transformed into intuitive visualization charts such as bar charts, line charts, pie charts, and heat maps, which are displayed in real time on the quality control center's visualization screen. It also supports data interaction and drill-down. Quality control personnel can click on any department or violation type data on the screen to quickly view the corresponding violation details, high-incidence groups, handling records, and rectification suggestions, thereby providing intuitive and dynamic data support for the hospital's aseptic control decisions.

[0048] In this embodiment, the specific implementation is as follows: after the aseptic operation scenario of each department (operating room, ICU, general ward) is activated, the high-definition AI camera of the multimodal data acquisition module captures images of medical staff's instrument transfer, glove wearing, and other operations in real time. The RFID reader reads the sterilization batch and expiration information of items such as sterile surgical packs and infusion sets in real time. The temperature and humidity sensor continuously monitors the temperature and humidity of the operating environment (such as 22-25℃ and 40%-60% humidity in the operating room). All equipment data are aggregated in real time through the edge computing gateway and the timestamps are aligned (the precise time of the operation is uniformly marked). The data is then transmitted to the data preprocessing unit. The unit first generates an automatic label set containing target categories, bounding boxes, and key points using AI pre-labeling formulas. After the manual verification team corrects mislabeling issues and supplements environmental parameter-related labels, the local data of each department is uploaded to the central node with encryption, ensuring that the data does not leave the domain and only the model gradient parameters are uploaded. The central node aggregates the parameters, optimizes the AI ​​labeling model, and then distributes it to each department, forming an iterative optimization closed loop. Finally, standardized, high-precision data is output to the AI ​​intelligent recognition module. The AI ​​intelligent recognition module calls upon the department's personalized compliance standard library (including quantitative parameters such as disinfection time threshold and sterile area boundary coordinates). Through a fusion model of CNN convolutional neural network and pose estimation algorithm deployed on edge computing nodes, it first extracts information such as the limb features of medical staff and the outline of sterile items in the image. Then, it locates action nodes such as wrists and fingertips through the key point confidence formula. It generates arm trajectory vectors with the help of limb association formula. It judges static violations such as not wearing masks and damaged sterile packages in real time, as well as dynamic violations such as hands crossing sterile areas and non-standard instrument transfer. It classifies violations into minor, medium and severe levels and then pushes the recognition results to the early warning and intervention module. After receiving the results, the early warning and intervention module compares the results with the hierarchical early warning rule library through a rule matching algorithm and triggers the corresponding early warning response: if it is a minor violation such as the misplacement of sterile packs, it immediately pushes a vibration reminder and standardized placement instructions to the smart wristband of the operating medical staff. For moderate violations such as not wearing sterile gloves during preparation for procedures, a reminder will be sent to the medical staff's wristbands, and a warning message containing the time, person, and screenshot of the violation will be sent to the head nurse's quality control terminal, and the details will be automatically recorded. In the event of serious violations such as the use of expired sterile instruments, the on-site audible and visual alarm device (flashing red warning light + voice warning) is immediately triggered, locking the video clips one minute before and after the violation. At the same time, the violation details and video link are pushed to the hospital's quality control platform. Subsequently, medical staff will take corresponding corrective actions according to different levels of violation (immediate adjustment for minor violations, cessation of operation and rectification after confirmation by the head nurse for moderate violations, and discontinuation of use of prohibited items and cooperation with quality control investigation for serious violations). The system will simultaneously record the warning and handling results and feed them back to the AI ​​intelligent recognition module, forming a micro closed loop of "recognition-warning-correction-optimization". The collaborative management module integrates this micro closed-loop data. On the one hand, it enables data exchange between the operating room, ICU, and general ward for aseptic operations. Before surgery, the operating room can access the RFID data of the sterile supplies supply center through the module and confirm the compliance and arrival of the surgical pack through AI image recognition. During surgery, compliance data is collected in real time and pushed to the hospital infection control system. After surgery, the intraoperative management records are synchronized to the ICU or general ward for continued nursing management. On the other hand, by combining RFID readers and AI image recognition, the entire process of sterile items from sterilization, warehousing, application, use to recycling can be traced, and abnormal situations can be alerted in real time. Based on data accumulated throughout the entire process, the quality control training module automatically generates a three-dimensional compliance report, including compliance rate at the departmental level, violation records at the individual level, and high-frequency violation points at the operation type level. Through data mining, it identifies high-frequency violation scenarios such as emergency catheterization at night and high-risk groups such as new nurses. It pushes customized training resources for recurring violation types (such as pushing videos on proper glove-wearing for ungloved operations and action diagrams for crossing sterile areas). It also restores high-frequency violation scenarios to VR simulation training scenarios, allowing medical staff to practice rectification procedures through VR devices. The module records training data and evaluates the effect to form a training closed loop. The data visualization module integrates core data from all modules, including AI recognition, early warning intervention, collaborative management and control, and quality control training. After cleaning and summarizing, the data is transformed into intuitive charts such as bar charts (comparison of departmental compliance rates), pie charts (percentage of common violation types), and heat maps (areas with high violation rates). These charts are displayed in real time on the quality control center's visualization screen, allowing quality control personnel to click on any data to drill down and view violation details, high-risk groups, treatment records, and rectification suggestions. This provides accurate and dynamic data support for the hospital's aseptic operation management decisions throughout the entire process.

[0049] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

Claims

1. A nursing aseptic operation control system, characterized in that, It includes a multimodal data acquisition module, an AI intelligent recognition module, an early warning and intervention module, a collaborative management and control module, and a quality control training module; The multimodal data acquisition module is used to acquire image data and environmental parameters of the aseptic operation scenario, as well as all elements of aseptic item information, and transmits them to the AI ​​intelligent recognition module after preprocessing. The AI ​​intelligent recognition module is used to identify and analyze image data, determine violations, and generate recognition results that are pushed to the early warning and intervention module. The early warning and intervention module is used to perform graded early warning and intervention based on the identification results, and at the same time, the relevant data of the handling is fed back to the AI ​​intelligent identification module to form a micro closed loop; The collaborative management module is used to integrate closed-loop data to achieve inter-departmental communication and full life-cycle traceability of sterile items. The quality control training module is used to generate quality control reports based on full-process data and push customized training resources.

2. The nursing aseptic operation control system according to claim 1, characterized in that, The multimodal data acquisition module specifically includes: a high-definition AI camera, an RFID reader, a temperature and humidity sensor, and a smart bracelet; The high-definition AI camera is used to capture image data of the aseptic operation scenario, as well as the operation behavior of medical staff and the environmental images of the operation area; The RFID reader / writer is used to bind and collect information about sterile items; The temperature and humidity sensor is used to monitor parameters related to the cleanliness of the operating environment; The smart bracelet is used to receive early warning information and coordinate with medical staff; The aforementioned devices achieve real-time data aggregation and timestamp alignment through an edge computing gateway.

3. The nursing aseptic operation control system according to claim 2, characterized in that, The multimodal data acquisition module is also equipped with a data preprocessing unit; The data preprocessing unit uses a combination of AI pre-annotation and manual precise verification to process the collected image data and environmental parameter-related data, and introduces a federated learning framework to share annotated data across departments while protecting the data privacy of each department.

4. The nursing aseptic operation control system according to claim 1, characterized in that, The AI ​​intelligent recognition module also includes a personalized compliance standard library; The personalized compliance standard library was compiled by nursing quality control experts in conjunction with the aseptic operation standards of various departments. It contains quantitative parameters that can be recognized by AI. These quantitative parameters include at least the disinfection time threshold, the boundary coordinate range of the aseptic area, the glove wearing action node, and the expiration date threshold of sterile items.

5. The nursing aseptic operation control system according to claim 4, characterized in that, The AI ​​intelligent recognition module also incorporates a deep learning model that integrates CNN convolutional neural networks and pose estimation algorithms; The deep learning model is deployed on edge computing nodes to perform AI image recognition and analysis on the image data, subdividing violations into static violations and dynamic violations and achieving real-time analysis. The static violations include not wearing masks or hats and using damaged sterile packs, while the dynamic violations include crossing the sterile area with one's hands and improper handling of instruments.

6. The nursing aseptic operation control system according to claim 1, characterized in that, The early warning intervention module constructs a three-level early warning response mechanism, specifically used for: When a minor violation is detected, the smart bracelet will send a vibration reminder and standardized operating instructions to the medical staff. When a moderate violation is detected, an early warning message is simultaneously pushed to the head nurse's quality control terminal and the violation details are automatically recorded. The head nurse's quality control terminal is a supporting terminal for the early warning and intervention module. When a serious violation is detected, the on-site audible and visual alarm device is triggered and the violation video segment is locked. At the same time, the relevant information is pushed to the hospital quality control platform. The on-site audible and visual alarm device and the hospital quality control platform are respectively the early warning execution device and the information aggregation platform of the early warning intervention module.

7. The nursing aseptic operation control system according to claim 6, characterized in that, The early warning intervention module also has an early warning feedback data return function, specifically used for: The system transmits the corrective actions and early warning results of medical staff’s violations to the AI ​​intelligent recognition module in real time, which serves as the basis for optimizing the parameters of the deep learning model, thus improving the micro-closed loop of recognition, early warning, correction and optimization.

8. The nursing aseptic operation control system according to claim 1, characterized in that, The quality control training module is specifically used for: Automatically generate three-dimensional compliance reports for departments, individuals, and operation types; use data mining to identify high-frequency violation scenarios and high-risk groups. Customized training resources are pushed based on individual violation records, and VR technology is used to transform high-frequency violation scenarios into simulated training scenarios.

9. The nursing aseptic operation control system according to claim 1, characterized in that, The collaborative management module is specifically used for: By connecting with the hospital's sterile supplies supply center management system, the sterile item tag information read by the RFID reader is combined with AI image recognition to achieve full-chain traceability of sterile items from sterilization, storage, transportation, use to recycling; Establish data links between operating rooms, ICUs, and general wards to achieve seamless aseptic control before, during, and after surgery, and integrate with the hospital's infection control system to incorporate compliance rate into infection control assessment indicators.

10. The nursing aseptic operation control system according to claim 1, characterized in that, It also includes a data visualization module; The data visualization module displays data on the aseptic operation compliance rate, common violation types, and early warning and handling efficiency of each department in real time through the quality control center's large visualization screen.

Images