Operating room nursing task automatic scheduling system

Abstract

The invention relates to the technical field of medical care, and discloses an operating room nursing task automatic scheduling system, which is characterized in that a data acquisition module acquires initial data related to an operating room in real time; the task analysis module is used for receiving the initial data acquired by the data acquisition module and analyzing the nursing task; the task scheduling module is used for automatically distributing the nursing tasks according to a result of the task analysis module, generating a task scheduling scheme and sending the task scheduling scheme to corresponding nursing personnel and related departments; the real-time monitoring module performs data interaction with a mobile terminal carried by a nurse and monitoring equipment in an operating room to monitor the execution condition of a nursing task in real time, and when an abnormal condition occurs in the task execution process, an alarm is given out immediately; the feedback and optimization module evaluates and optimizes the task scheduling scheme and adjusts a task allocation strategy; the overall operation efficiency of the operating room is improved, and the professionality and high efficiency of operation cooperation work are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of medical care, and in particular to an automatic scheduling system for nursing tasks in an operating room. Background Art

[0002] Operating room nursing work involves complex tasks such as patient transfer, surgical coordination, and surgical instrument and equipment management, all of which require extremely high timeliness and accuracy. Traditional operating room nursing task scheduling relies heavily on manual arrangements, which presents numerous drawbacks. For example, manual scheduling is susceptible to subjective factors, leading to irrational task allocation, with some nurses overburdened and others idle. During the patient transfer process, manual coordination can easily lead to transfer delays, impacting the surgical process. For surgical instruments and equipment, manual management struggles to maintain real-time status and maintenance requirements, which can easily lead to untimely instrument preparation or equipment failures that impact the performance of surgery. With the advancement of medical technology and the increase in surgical volume, traditional manual scheduling methods are no longer able to meet the efficient and precise nursing needs of modern operating rooms. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and to design an automatic scheduling system for operating room nursing tasks.

[0004] The present invention provides an automatic scheduling system for operating room nursing tasks, the system comprising: The data acquisition module connects with the hospital's electronic medical record system, nursing staff management system, equipment management system, and operating room sensor equipment to obtain initial data related to the operating room in real time; The task analysis module is used to receive the initial data collected by the data acquisition module and analyze the nursing tasks; The task scheduling module is used to automatically allocate nursing tasks according to the results of the task analysis module, generate task scheduling plans, and send the task scheduling plans to the corresponding nursing staff and relevant departments; The real-time monitoring module monitors the execution of nursing tasks in real time by exchanging data with the mobile terminals carried by nursing staff and the monitoring equipment in the operating room. When any abnormal situation is found during the execution of the task, an alarm is immediately issued; The feedback and optimization module receives abnormal information from the real-time monitoring module and feedback from nursing staff on task execution, evaluates and optimizes the task scheduling plan, and adjusts the task allocation strategy.

[0005] Optionally, in a first implementation of the present invention, the data acquisition module includes: The deployment submodule is used to deploy edge computing nodes and build a lightweight stream processing framework composed of Apache Kafka and Apache Flink to obtain initial data related to the operating room in real time; The mapping submodule is used to build a knowledge graph in the medical field, identify various entities related to surgical care and the relationships between them, use natural language processing technology to extract information from unstructured text in electronic medical records, and map the extracted information and the time series data in the initial data into a unified semantic space through the entity and relationship mapping mechanism of the knowledge graph.

[0006] Optionally, in a second implementation of the present invention, the deployment submodule includes: The data stream collected from the operating room sensor equipment is transmitted to the edge computing node, and Apache Kafka is used for real-time data reception, buffering and distribution. Then, Apache Flink is used to perform real-time cleaning and preprocessing on the flowing data to remove noise and outliers in the data, and to interpolate missing data to obtain initial data.

[0007] Optionally, in a third implementation of the present invention, the task analysis module includes: The medical event chain parsing submodule receives the initial data transmitted by the data acquisition module, inputs the unstructured text data in the electronic medical record into the pre-trained medical large model BioBERT, identifies key events related to the surgery, sorts the extracted key events according to timestamps to form an ordered event sequence, and stores the constructed event sequence in a structured form to form a surgical process knowledge base; a causal reasoning submodule for identifying causal dependencies of nursing tasks through counterfactual analysis; The input submodule is used to construct a dynamic spatiotemporal graph and input it into the spatiotemporal graph neural network model. The spatial relationship between nodes is learned through graph convolution operations. The temporal changes of nodes are learned through time series analysis. The analysis is performed in different time periods and spatial regions to obtain early warning information for task execution. The integration submodule is used to integrate the surgical process knowledge base, causal dependencies and warning information of task execution, analyze the nursing tasks, and determine the task requirement information, wherein the task requirement information at least includes the priority of the task, the required resources and the execution order.

[0008] Optionally, in a fourth implementation of the present invention, the causal reasoning submodule includes: Based on the initial data and surgical process knowledge base, counterfactual analysis scenarios are set for nursing tasks. In each counterfactual analysis scenario, based on the statistical and logical relationships between the factors in the initial data, the causal effects between the nursing tasks and the factors in the initial data are calculated. Based on the calculated causal effects, the causal dependencies between different nursing tasks and between nursing tasks and other factors are determined.

[0009] Optionally, in a fifth implementation of the present invention, the input submodule includes: An operating room space-time graph is constructed with the personnel, equipment, and tasks in the operating room as nodes, and relationships including at least personnel movement paths, equipment usage associations, and task sequences as edges. The initial data collected by sensors is used to assign space-time attributes to the nodes and edges of the operating room space-time graph, forming a dynamic space-time graph.

[0010] Optionally, in a sixth implementation of the present invention, the task scheduling module includes: The collaborative decision-making submodule is used to abstract each nurse into an independent intelligent agent, construct a multi-dimensional vector containing the real-time status of the operating room, define the set of actions that the intelligent agent can perform, design a comprehensive reward mechanism, conduct a multi-agent collaborative decision-making process, and generate a dynamic allocation plan; The encoding submodule is used to construct a Hamiltonian that represents the task allocation goals and constraints, encode the initial state information of the current caregiver and task, explore the solution space using the tunneling effect of the quantum system, converge the quantum state to the lowest energy state by lowering the system temperature, introduce a penalty term in the quantum optimization process, and simulate the optimal solution search process for task allocation through the interaction between quantum bits; The weighted fusion submodule is used to weightedly fuse the dynamic allocation scheme and the optimal solution provided by the quantum algorithm to obtain the optimal task allocation scheme, convert the optimal task allocation scheme into scheduling instructions, and issue them.

[0011] Optionally, in a seventh implementation of the present invention, the collaborative decision-making submodule includes: Each agent obtains local state information and extracts features by observing the environment. The agent calculates the optimal action based on the current policy network and exchanges partial state information through the communication mechanism. When multiple agents compete for the same task, the conflict resolution mechanism is triggered to redistribute the task.

[0012] Optionally, in an eighth implementation of the present invention, the real-time monitoring module includes: The comparison submodule is used to obtain multimodal data from the mobile terminals carried by nurses and the monitoring equipment in the operating room, and perform multimodal comparative learning and analysis on the multimodal data; The calculation submodule is used to calculate the distance metric between the embedding vectors of each modality at the current moment. When the distance exceeds the set threshold, it is determined that there is a risk of cross-modal inconsistency; A determination submodule is used to identify cross-modal inconsistencies detected by multimodal contrastive learning and determine abnormal pattern characteristics; The generation submodule is used to obtain the abnormal type and severity according to the abnormal pattern characteristics and generate corresponding alarm information.

[0013] Optionally, in a ninth implementation of the present invention, the comparison submodule includes: A convolutional neural network is used to extract features from surveillance video frame images, and the images are mapped to the visual embedding space. The audio Mel-spectrogram is processed through a recurrent neural network combined with an attention mechanism to generate a speech embedding vector. The Transforme model is used to encode the text data to obtain a text embedding representation.

[0014] The technical solution provided by the present invention is to obtain the initial data related to the operating room in real time by connecting with the hospital's electronic medical record system, nursing staff management system, equipment management system and sensor equipment in the operating room; receive the collected initial data and analyze the nursing tasks; automatically allocate the nursing tasks according to the results of the task analysis, generate a task scheduling plan, and send the task scheduling plan to the corresponding nursing staff and relevant departments; through data interaction with the mobile terminals carried by the nursing staff and the monitoring equipment in the operating room, monitor the execution of the nursing tasks in real time, and immediately issue an alarm when an abnormal situation occurs during the execution of the task; receive the real-time monitoring model The system can evaluate and optimize the task scheduling plan and adjust the task allocation strategy based on the abnormal information fed back by the block and the feedback of the nursing staff on the task execution. The present invention realizes the reasonable allocation of nursing tasks, avoids the subjectivity and irrationality of manual scheduling, balances the workload of nursing staff, improves the work efficiency and satisfaction of nursing staff, reduces the waiting time of patients, ensures that the operation is carried out on time, improves the overall operation efficiency of the operating room, ensures the professionalism and efficiency of the operation cooperation work, improves the success rate of the operation and the safety of the patients, monitors the execution of tasks in real time, discovers and handles abnormal problems in time, continuously improves the task scheduling plan through feedback and optimization mechanism, and improves the adaptability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Various other advantages and benefits will become apparent to those skilled in the art by reading the following detailed description of the preferred embodiment.The accompanying drawings are only for the purpose of illustrating the preferred embodiment and are not to be considered as limiting the present invention.

[0016] Figure 1 A schematic diagram of the structure of an automatic scheduling system for operating room nursing tasks provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a task analysis module provided in an embodiment of the present invention; Figure 3 A structural diagram of a task scheduling module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The terms "first," "second," "third," "fourth," and so forth (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar items and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or device.

[0018] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A schematic diagram of the structure of an automatic scheduling system for operating room nursing tasks provided by an embodiment of the present invention includes: The data acquisition module connects with the hospital's electronic medical record system, nursing staff management system, equipment management system, and operating room sensor equipment to obtain initial data related to the operating room in real time; The task analysis module is used to receive the initial data collected by the data acquisition module and analyze the nursing tasks; The task scheduling module is used to automatically allocate nursing tasks according to the results of the task analysis module, generate task scheduling plans, and send the task scheduling plans to the corresponding nursing staff and relevant departments; The real-time monitoring module monitors the execution of nursing tasks in real time by exchanging data with the mobile terminals carried by nursing staff and the monitoring equipment in the operating room. When any abnormal situation is found during the execution of the task, an alarm is immediately issued; The feedback and optimization module receives abnormal information from the real-time monitoring module and feedback from nursing staff on task execution, evaluates and optimizes the task scheduling plan, and adjusts the task allocation strategy.

[0019] In this embodiment, the data acquisition module includes: The deployment submodule is used to deploy edge computing nodes and build a lightweight stream processing framework composed of Apache Kafka and Apache Flink to obtain initial data related to the operating room in real time; The mapping submodule is used to build a knowledge graph in the medical field, identify various entities related to surgical care and the relationships between them, use natural language processing technology to extract information from unstructured text in electronic medical records, and map the extracted information and the time series data in the initial data into a unified semantic space through the entity and relationship mapping mechanism of the knowledge graph.

[0020] In this embodiment, the deployment submodule includes: The data stream collected from the operating room sensor equipment is transmitted to the edge computing node, and Apache Kafka is used for real-time data reception, buffering and distribution. Then, Apache Flink is used to perform real-time cleaning and preprocessing on the flowing data to remove noise and outliers in the data, and to interpolate missing data to obtain initial data.

[0021] In this embodiment, please refer to Figure 2 , the task analysis module includes: The medical event chain parsing submodule receives the initial data transmitted by the data acquisition module, inputs the unstructured text data in the electronic medical record into the pre-trained medical large model BioBERT, identifies key events related to the surgery, sorts the extracted key events according to timestamps to form an ordered event sequence, and stores the constructed event sequence in a structured form to form a surgical process knowledge base; a causal reasoning submodule for identifying causal dependencies of nursing tasks through counterfactual analysis; The input submodule is used to construct a dynamic spatiotemporal graph and input it into the spatiotemporal graph neural network model. The spatial relationship between nodes is learned through graph convolution operations. The temporal changes of nodes are learned through time series analysis. The analysis is performed in different time periods and spatial regions to obtain early warning information for task execution. The integration submodule is used to integrate the surgical process knowledge base, causal dependencies and warning information of task execution, analyze the nursing tasks, and determine the task requirement information, wherein the task requirement information at least includes the priority of the task, the required resources and the execution order.

[0022] In this embodiment, the causal reasoning submodule includes: Based on the initial data and surgical process knowledge base, counterfactual analysis scenarios are set for nursing tasks. In each counterfactual analysis scenario, based on the statistical and logical relationships between the factors in the initial data, the causal effects between the nursing tasks and the factors in the initial data are calculated. Based on the calculated causal effects, the causal dependencies between different nursing tasks and between nursing tasks and other factors are determined.

[0023] In this embodiment, the input submodule includes: An operating room space-time graph is constructed with the personnel, equipment, and tasks in the operating room as nodes, and relationships including at least personnel movement paths, equipment usage associations, and task sequences as edges. The initial data collected by sensors is used to assign space-time attributes to the nodes and edges of the operating room space-time graph, forming a dynamic space-time graph.

[0024] In this embodiment, please refer to Figure 3 , the task scheduling module includes: The collaborative decision-making submodule is used to abstract each nurse into an independent intelligent agent, construct a multi-dimensional vector containing the real-time status of the operating room, define the set of actions that the intelligent agent can perform, design a comprehensive reward mechanism, conduct a multi-agent collaborative decision-making process, and generate a dynamic allocation plan; The encoding submodule is used to construct a Hamiltonian that represents the task allocation goals and constraints, encode the initial state information of the current caregiver and task, explore the solution space using the tunneling effect of the quantum system, converge the quantum state to the lowest energy state by lowering the system temperature, introduce a penalty term in the quantum optimization process, and simulate the optimal solution search process for task allocation through the interaction between quantum bits; The weighted fusion submodule is used to weightedly fuse the dynamic allocation scheme and the optimal solution provided by the quantum algorithm to obtain the optimal task allocation scheme, convert the optimal task allocation scheme into scheduling instructions, and issue them.

[0025] In this embodiment, the collaborative decision-making submodule includes: Each agent obtains local state information and extracts features by observing the environment. The agent calculates the optimal action based on the current policy network and exchanges partial state information through the communication mechanism. When multiple agents compete for the same task, the conflict resolution mechanism is triggered to redistribute the task.

[0026] In this embodiment, the real-time monitoring module includes: The comparison submodule is used to obtain multimodal data from the mobile terminals carried by nurses and the monitoring equipment in the operating room, and perform multimodal comparative learning and analysis on the multimodal data; The calculation submodule is used to calculate the distance metric between the embedding vectors of each modality at the current moment. When the distance exceeds the set threshold, it is determined that there is a risk of cross-modal inconsistency; A determination submodule is used to identify cross-modal inconsistencies detected by multimodal contrastive learning and determine abnormal pattern characteristics; The generation submodule is used to obtain the abnormal type and severity according to the abnormal pattern characteristics and generate corresponding alarm information.

[0027] In this embodiment, the comparison submodule includes: A convolutional neural network is used to extract features from surveillance video frame images, and the images are mapped to the visual embedding space. The audio Mel-spectrogram is processed through a recurrent neural network combined with an attention mechanism to generate a speech embedding vector. The Transforme model is used to encode the text data to obtain a text embedding representation.

[0028] In this embodiment, various operating room-related data is collected, including patient information (such as patient name, age, surgery type, surgery time, and preoperative status), nursing staff information (such as nurse name, qualifications, skills, and current work status), surgical instrument and equipment information (such as instrument name, quantity, usage status, and maintenance cycle), and operating room usage information (such as operating room number, current occupancy, and expected idle time). The data collection module connects with the hospital's electronic medical record system, nursing staff management system, equipment management system, and operating room sensor equipment to obtain accurate data in real time. It also analyzes nursing tasks such as patient transfer, surgical coordination, and surgical instrument and equipment management. For patient transfer tasks, a reasonable transfer time and route are calculated based on the patient's surgery time, location, and operating room layout. For surgical coordination tasks, the required number of nursing staff and their responsibilities are determined based on the surgery type and nursing staff skills. For surgical instrument and equipment management tasks, an instrument preparation and equipment allocation plan is developed based on the surgical requirements and instrument and equipment usage status. An intelligent scheduling algorithm is then used to automatically allocate nursing tasks. When assigning tasks, the system comprehensively considers factors such as nurses' workload, skill matching, and task urgency to ensure rational and balanced task allocation. For example, complex surgical procedures are assigned to experienced and well-skilled nurses. For urgent surgical patient transfers, currently available nurses are prioritized. The task scheduling module also generates a detailed task schedule, including the assigned task recipient, execution time, and execution requirements, and sends this schedule to the appropriate nurses and relevant departments. The system monitors the execution of nursing tasks in real time by interacting with nurses' mobile devices (such as smart bracelets and tablets) and operating room monitoring equipment to obtain real-time status information on task execution, including the location and progress of surgical patient transfers, nurses' work status, and the use of surgical instruments and equipment. If anomalies are detected during task execution, such as transfer delays, nurses' emergencies, or instrument failures, the real-time monitoring module immediately issues an alert. Based on this feedback, the system adjusts the task allocation strategy and reschedules tasks to ensure the smooth progress of surgical nursing work. At the same time, the feedback and optimization module will analyze historical task scheduling data, summarize experiences and lessons, and continuously optimize the intelligent scheduling algorithm to improve the accuracy and efficiency of task scheduling.

[0029] In this embodiment, the system is connected to the hospital's electronic medical record system, nursing staff management system, equipment management system, and sensor equipment in the operating room. By embedding an adaptive data acquisition interface in each system or setting a data synchronization module, the initial data related to the operating room can be obtained in real time. Specifically, it includes surgical patient information (name, age, type of surgery, surgery time, preoperative status, etc.), nursing staff information (name, qualifications, skills, current working status, etc.), surgical instrument equipment information (instrument name, quantity, usage status, maintenance cycle, etc.), operating room usage status information (operating room number, current occupancy, expected idle time, etc.); edge computing nodes are deployed locally in the operating room, and a lightweight stream processing framework composed of Apache Kafka and Apache Flink is built; the data stream collected from the operating room sensor equipment is transmitted to the edge computing node, and Apache Kafka receives, buffers, and distributes data in real time, and then uses Apache Flink to clean and preprocess the data in real time; for example, it removes noise and outliers from the data, interpolates missing data, and converts the data into a format that the system can recognize, reducing the delay in data transmission to the cloud and providing high-quality real-time data for subsequent data processing; under the premise of protecting hospital data privacy, distributed model training is carried out based on the federated learning framework; each participating hospital (data provider) retains the original data locally and does not transmit it centrally; the system deploys model training clients locally in each hospital and sends the initial task scheduling model parameters to each client; each client performs model training locally based on local electronic medical records, sensors and other multi-source heterogeneous data, and calculates the update gradient of the model parameters; the updated gradient is then uploaded to the central server, which collects the gradient information of all clients, performs aggregation calculations, updates the global model parameters, and updates the updated model parameters again It is sent to each client; the above process is repeated until the model converges, thereby realizing distributed training that integrates multi-source heterogeneous data and avoiding the privacy risks brought by the centralized storage of the original data set; constructing a knowledge graph in the medical field, sorting out various entities related to surgical care (such as patients, nurses, surgical instruments, surgery types, etc.) and their relationships (such as patients-surgery types, nurses-skill expertise, surgical instruments-surgery types, etc.); using natural language processing technology to extract information from unstructured texts in electronic medical records (such as surgical records), and mapping the extracted key information and time series data collected by sensors (such as surgical instrument usage time series, patient vital signs data series) to a unified semantic space through the entity and relationship mapping mechanism of the knowledge graph; for example, the name of the surgical instrument mentioned in the surgical record and the usage time data of the instrument collected by the sensor are associated and integrated through the "surgical instrument" entity in the knowledge graph to solve the alignment problem of multimodal data and form structured and semantically fused data.

[0030] In this example, the initial data obtained from the electronic medical record system is cleaned to remove duplicate, erroneous, or incomplete data records, unify the data format, for example, standardize time records in different formats into a unified timestamp format, and perform word segmentation on the text information to prepare for subsequent analysis. The pre-trained medical large-scale model BioBERT is used to scan the pre-processed data sentence by sentence. The model learns the semantic associations between words in a large amount of medical text data and identifies key event expressions related to surgery, such as "patient enters the operating room," "anesthesia begins," and "surgical instrument preparation is completed." etc., and extract these key events; sort the extracted key events according to the order of appearance of key events in the electronic medical record and combine with professional knowledge in the medical field to construct an ordered sequence of key events; at the same time, mark the time information and related entities such as patients, medical staff, etc. for each key event; store the constructed key event sequence in a structured form to form a surgical process knowledge base; in the knowledge base, clarify the logical relationship between each key event, such as sequence, parallel relationship, etc., to facilitate subsequent query and use; collect all the collected initial data, including nursing task data, equipment status data, patient condition data, etc., and organize these data into a format suitable for causal reasoning algorithm processing, such as constructing a data set containing multiple variables, each variable represents a factor, such as nursing task type, equipment fault code, patient vital sign indicator, etc.; for each nursing task-related factor, propose a counterfactual hypothesis; for example, for "specific equipment failure" This factor, assuming that the equipment is operating normally, how will the execution of nursing tasks change? Create multiple virtual scenarios by setting different counterfactual conditions. In each virtual scenario, calculate the causal effect between nursing tasks and other factors based on the statistical and logical relationships between the factors in the data set. Use relevant methods of causal reasoning, such as propensity score matching and instrumental variable method, to eliminate the influence of confounding factors and accurately assess the degree of causal dependence between factors. Determine the causal logical relationship between different nursing tasks and between nursing tasks and other factors based on the calculated causal effect. Record these causal logical relationships in the form of charts or rules, such as drawing a causal relationship diagram to clarify each The direction and degree of influence of factors on nursing tasks; the nursing staff, equipment, and nursing tasks in the operating room are defined as nodes respectively; initial attributes are assigned to each node, such as the nursing staff's skill level, current location, and working status; the type, location, and usage status of the equipment; the type, expected duration, and priority of the nursing task; based on the association between the nursing staff and the equipment, the nursing task, and the relationship between the equipment and the nursing task, edges are constructed between the nodes; for example, if a nursing staff is responsible for operating a certain equipment to complete a specific nursing task, a connecting edge is established between the nursing staff node, the equipment node, and the nursing task node, and weights are assigned to the edges based on the actual situation, such as the operating proficiency and the urgency of the task;The movement path data of the nursing staff, the frequency and location data of the equipment usage, and the time information are integrated into the graph structure; the node status and edge relationship at different time points are updated in chronological order to form a dynamic spatiotemporal relationship graph; for example, at regular time intervals, the new position of the nursing staff is recorded and the connection relationship with other nodes is updated; the dynamic spatiotemporal relationship graph is analyzed using the spatiotemporal graph neural network model; the spatiotemporal correlation between each node is calculated through the graph convolution operation and time series analysis operation in the model; for example, the relationship between the movement path of the nursing staff and the equipment usage hotspot in different time periods is analyzed to predict the bottleneck position and time point that may appear during the task execution; the surgical process knowledge base generated by the medical event chain parsing algorithm, the causal logical relationship obtained by the causal reasoning algorithm, and the task execution spatiotemporal relationship constructed by the spatiotemporal graph neural network algorithm are integrated into the graph structure; the movement path data of the nursing staff and the equipment usage hotspot in different time periods are analyzed to predict the bottleneck position and time point that may appear during the task execution; the surgical process knowledge base generated by the medical event chain parsing algorithm, the causal logical relationship obtained by the causal reasoning algorithm, and the task execution spatiotemporal relationship constructed by the spatiotemporal graph neural network algorithm are integrated into the graph structure; the movement path data of the nursing staff and the equipment usage hotspot are updated and the connection relationship between the nodes is updated; the spatiotemporal graph neural network model is used to analyze the dynamic spatiotemporal relationship between the nursing staff and the equipment usage hotspot in different time periods; the spatiotemporal graph neural network model is used to analyze the dynamic spatiotemporal relationship between the nursing staff and the equipment usage hotspot in Models are uniformly converted into the same data format to facilitate data fusion; related information is associated based on the same entities (such as nursing tasks, equipment, and nursing staff) in the results of each algorithm; for example, the time sequence information of nursing tasks in the surgical process knowledge base is matched and integrated with the information on the influence of other factors on the nursing tasks in the causal logic relationship, as well as the spatial location information of the execution of the tasks in the spatiotemporal model; from the integrated data, key elements such as the nature of the nursing task (such as basic nursing, specialized nursing, etc.), priority (determined by causal relationships and surgical processes), required resources (including manpower, equipment, etc.), and time requirements (determined based on surgical processes and spatiotemporal models) are extracted; the extracted key elements are output in a structured form to form a comprehensive nursing task analysis report, providing a detailed and accurate basis for subsequent task allocation.

[0031] In this embodiment, each nurse is abstracted as an independent intelligent agent with the ability to observe the environment, perform actions and learn strategies; a multidimensional vector containing the real-time status of the operating room is constructed, such as the patient's condition priority, equipment availability, nurse location / skills / load, task urgency, etc.; a set of actions that the agent can perform is defined, such as discrete actions such as accepting tasks, handing over tasks, requesting support, or continuous actions such as adjusting task execution efficiency; a comprehensive reward mechanism is designed, including multi-dimensional weighted rewards such as task completion time, resource utilization, fairness indicators (such as workload balance), and patient satisfaction estimates; the nursing task allocation problem is converted into a quantum optimization problem, such as mapping the task-personnel allocation relationship into quantum bit states ; Construct a Hamiltonian that represents the task allocation goals and constraints, including constraints such as the shortest task completion time, minimum resource conflict, and priority for emergency tasks; initialize the quantum system state and encode the initial state information of the current nursing staff and tasks; the digital twin-driven simulation environment constructs a three-dimensional space model based on parameters such as the operating room layout, equipment distribution, and nursing staff movement speed; converts nursing task processes, equipment operation specifications, medical collaboration models, etc. into behavioral rules of the digital twin system; uses historical surgical data and real-time collected data to calibrate the digital twin model to ensure that the simulation environment is highly consistent with the physical operating room; each intelligent agent obtains local state information by observing the environment, and extracts task priorities, resource requirements, and time windows The intelligent agent calculates the optimal action based on the current strategy network and exchanges some state information (such as load status and skill advantages) through the communication mechanism. When multiple intelligent agents compete for the same resource or task, the conflict resolution mechanism is triggered and the task is redistributed through negotiation or a central coordinator. The tunneling effect of the quantum system is used to explore the solution space, and the quantum state converges to the lowest energy state by reducing the system temperature, corresponding to the optimal task allocation scheme. Penalty terms are introduced in the quantum optimization process to ensure that the task allocation meets the conditions of time constraints, skill matching, resource constraints, etc. The optimal state of the quantum system is decoded into the actual task-personnel allocation scheme and converted into executable scheduling instructions. Multiple candidates are run in parallel in the digital twin environment. Select scheduling plans and simulate different nursing staff collaboration modes and task execution paths; evaluate the robustness of each plan by injecting random perturbations (such as equipment failures and new task insertions) and calculate the performance fluctuations of the plan under different scenarios; build a Bayesian optimization model based on simulation results and update scheduling parameters such as task allocation weights and agent collaboration strategies; weightedly fuse the dynamic allocation plan generated by MADRL, the global optimal solution fragment provided by the quantum algorithm, and the robustness strategy verified by the digital twin; dynamically adjust the allocation plan to prioritize key tasks in response to real-time constraints such as sudden emergency insertions and equipment failures; convert the optimal plan generated by the algorithm into scheduling instructions that medical staff can understand, such as task lists, execution order, resource requirements, etc.Task instructions are pushed to nursing staff's mobile terminals in batches, along with necessary patient information and operating instructions, based on the urgency and execution time of the task. Collaboration requests are sent to relevant systems, such as the equipment management system and drug supply department, to ensure that the necessary resources are in place in a timely manner. For high-risk tasks, emergency plans are simultaneously issued, including information such as the location of backup equipment and emergency contact information.

[0032] In this embodiment, task execution status, location coordinates, operation records and other data are obtained from the mobile terminal carried by the nursing staff; video streams and audio data are collected through the operating room monitoring equipment; real-time update data of the electronic medical record system, as well as the instrument usage status and sensor monitoring data of the equipment management system are synchronized; the video stream is decomposed into frame images, and the audio data is converted into Mel spectrum graphs; the electronic medical record text is preprocessed by word segmentation, part-of-speech tagging, etc.; the numerical data collected by the equipment sensor is normalized and the data format is unified for subsequent analysis; accurate timestamps are added to all collected data, and multi-source data are aligned based on the timestamps to ensure consistency of different types of data in the time dimension; according to the characteristics of the operating room sensor data, a pulse neural network architecture is designed, including an input layer, a hidden layer and an output layer; the input layer neurons correspond to different types of sensor data channels, such as temperature sensors, position sensors, etc.; the continuous data collected by the sensor is converted into a pulse sequence; when the sensor data change exceeds a preset threshold, the corresponding neuron generates a pulse, and the data features are encoded with the pulse emission time and frequency to achieve efficient processing of asynchronous data; using SNN The neural pulse transmission mechanism is used to extract features of the input pulse sequence in the hidden layer to capture the dynamic change pattern and abnormal fluctuation characteristics of the sensor data; the output layer outputs the status evaluation result of the current sensor data based on the extracted features to determine whether there is an abnormal trend; a convolutional neural network (CNN) is used to extract features from the monitoring video frame image and map the image to the visual embedding space; a recurrent neural network (RNN) is used in combination with the attention mechanism to process the audio Mel spectrogram and generate a speech embedding vector; a Transformer is used to extract features of the monitoring video frame image and map the image to the visual embedding space; a recurrent neural network (RNN) is used to process the audio Mel spectrogram and generate a speech embedding vector; a Transformer is used to extract features of the monitoring video frame image and map the image to the visual embedding space ... The model encodes the electronic medical record text to obtain a text embedding representation; a contrastive learning loss function is designed to jointly train the embedding vectors of the three modalities; during the training process, the embedding representation distances of the same event in different modalities are shortened, and the embedding representation distances of different events are extended, so that the embedding spaces of the three modalities are aligned with each other, and cross-modal consistency features are learned; the distance metric (such as cosine distance) between the embedding vectors of each modality at the current moment is calculated in real time; when the distance exceeds the set threshold, it is determined that there is a risk of cross-modal inconsistency, that is, an abnormal situation may occur, such as the operation displayed on the surveillance video is inconsistent with the electronic medical record, and the content of the medical conversation does not match the actual operation; medical operation specifications, surgical process standards, instrument management rules and other knowledge are modeled in the form of symbolic logic to construct a knowledge base; for example, it is defined as "the operation can only begin after the instrument inventory is completed" and "the transfer of instruments during surgery must follow the principle of sterility" and other rules; use deep learning models (such as convolutional neural networks and recurrent neural networks) to further perform pattern recognition on sensor abnormality results output by spiking neural networks and cross-modal inconsistencies detected by multimodal contrastive learning, and extract abnormal pattern features; convert the identified abnormal pattern features into symbolic logic representations, and match and reason with the rules in the knowledge base;Through forward and reverse reasoning, the system determines whether the abnormal situation complies with medical operation standards and verifies the accuracy of abnormality detection results. For example, if abnormal instrument use is detected, logical reasoning is used to determine whether it violates instrument management rules and whether it will affect surgical safety. The system makes a final abnormality decision by integrating the sensor abnormality results of the spiking neural network, the cross-modal inconsistency detection results of multimodal contrastive learning, and the logical reasoning verification results of the neural symbolic system. When any abnormality condition is met, it is determined that an abnormality occurred during task execution. Based on the type and severity of the abnormality, corresponding alarm information is generated, including the time, location, type, and possible impact of the abnormality. The alarm information is sent through the system to relevant nursing staff, surgeons, and management personnel so that timely measures can be taken to deal with the abnormality.

[0033] In this embodiment, workflow features, such as task type distribution, equipment usage patterns, and peak hour characteristics, are extracted from historical scheduling data of different departments to construct a personalized feature vector for each department. Based on a meta-learning framework (such as MAML or Reptile), the parameters of a meta-model that can quickly adapt to the new department environment are initialized. The model includes components such as task priority evaluation and resource allocation strategy. For newly added departments, a small amount of historical data is used to fine-tune the model parameters, and the meta-model is quickly adapted to the department's specific workflow pattern through gradient updating. The task execution data of the current department is analyzed in real time to identify its workflow pattern (such as the emergency response type of cardiac surgery vs. the planned process of orthopedics), and the scheduling policy parameters are dynamically adjusted. Based on the historical operation data of nursing staff, a generative adversarial network (GAN) is used to generate adversarial samples that simulate different operating habits, such as fluctuations in task execution speed and changes in resource usage preferences. The adversarial samples are mixed into the training data, and the robustness of the scheduling model to changes in operating habits is enhanced through adversarial training to ensure the stable performance of the model when facing new operating patterns. The distribution changes of nursing staff's operation data are monitored in real time, and statistical process control (SPC) or KL is used to predict the distribution of the data. The system uses methods such as divergence to detect concept drift and trigger the model update mechanism; when significant changes in operating habits are detected, the scheduling model is updated using an incremental learning algorithm to retain the learned knowledge while adapting to the new operating mode; multiple optimization objectives and their mathematical expressions are clarified, including task completion time, nurse workload balance, equipment energy consumption, patient satisfaction prediction value, etc.; constraints such as nursing resource limitations, task priority relationships, and equipment maintenance cycles are converted into mathematical expressions and incorporated into the optimization framework; the multi-objective optimization problem is decomposed into multiple single-objective sub-problems using the MOEA / D algorithm, and a set of Pareto optimal solutions is generated through iterative search using an evolutionary algorithm; the weight vectors of each optimization objective are dynamically adjusted based on real-time feedback data (such as an increase in the proportion of sudden emergency tasks) to prioritize key objectives; multi-source real-time data such as feedback from nursing staff's mobile terminals, monitoring equipment data, and patient status changes are integrated to build a unified data view; algorithms such as isolation forests and autoencoders are used to detect abnormal events such as task timeouts, excessive resource consumption, and operational process deviations; and the system uses the nature of the abnormal event (such as equipment failure, human error) and the degree of impact (minor, moderate, severe) to determine the optimal solution. ) for classification and grading; analyze the impact propagation path of abnormal events on other tasks and overall scheduling based on the causal graph model, and assess potential risks; calculate the performance indicators of the current scheduling plan for each optimization goal, such as average task response time, nurse workload variance, total equipment energy consumption, etc.; compare the current plan with the historical optimal plan and Pareto optimal solution set to evaluate the relative advantages and disadvantages of the plan; identify bottlenecks in the scheduling process through process mining technology, such as excessive utilization of specific equipment and overload of certain nursing staff tasks; use meta-models to predict the scheduling effect in the absence of abnormal events and quantify the impact of abnormal events on overall performance;Based on the Pareto optimal solution set generated by the multi-objective evolutionary algorithm, multiple candidate optimization solutions are generated in combination with real-time abnormal situations and department workflow characteristics; risk assessment is performed on each candidate solution to predict its performance fluctuations and potential risks in different scenarios; for high-risk scenarios, flexible scheduling strategies are designed, such as pre-allocating backup resources and formulating cross-team collaboration plans; explainable decision support reports are generated to show the advantages and disadvantages of each candidate solution and the applicable scenarios for reference by managers; incremental learning or meta-learning algorithms are used to update scheduling model parameters and integrate new optimization strategies into the model; based on the urgency of the task and the execution stage, the scheduling of tasks that have not yet started is selectively updated to avoid interfering with tasks in progress; strategy verification is performed in a digital twin environment, and A / B Test and compare the performance of new and old strategies, and gradually promote the optimized strategy; establish a strategy version control system to record each updated parameter and performance indicator, allowing for rapid rollback when problems arise; store each optimization process and results in an experience library to build a playback buffer for reinforcement learning; update the meta-knowledge of the meta-learning model to improve the model's adaptability to new departments and workflow changes; convert medical experts' feedback on the optimization plan into domain knowledge and incorporate it into the next round of optimization; regularly perform parameter compression and knowledge distillation on the continuously learning model to maintain efficient model operation.

[0034] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An automatic scheduling system for operating room nursing tasks, characterized by: The system includes: The data acquisition module connects with the hospital's electronic medical record system, nursing staff management system, equipment management system, and operating room sensor equipment to obtain initial data related to the operating room in real time; The task analysis module is used to receive the initial data collected by the data acquisition module and analyze the nursing tasks; The task scheduling module is used to automatically allocate nursing tasks according to the results of the task analysis module, generate task scheduling plans, and send the task scheduling plans to the corresponding nursing staff and relevant departments; The real-time monitoring module monitors the execution of nursing tasks in real time by exchanging data with the mobile terminals carried by nursing staff and the monitoring equipment in the operating room. When any abnormal situation is found during the execution of the task, an alarm is immediately issued; The feedback and optimization module receives abnormal information from the real-time monitoring module and feedback from nursing staff on task execution, evaluates and optimizes the task scheduling plan, and adjusts the task allocation strategy.

2. The automatic scheduling system for operating room nursing tasks according to claim 1, characterized in that: The data acquisition module includes: The deployment submodule is used to deploy edge computing nodes and build a lightweight stream processing framework composed of Apache Kafka and Apache Flink to obtain initial data related to the operating room in real time; The mapping submodule is used to build a knowledge graph in the medical field, identify various entities related to surgical care and the relationships between them, use natural language processing technology to extract information from unstructured text in electronic medical records, and map the extracted information and the time series data in the initial data into a unified semantic space through the entity and relationship mapping mechanism of the knowledge graph.

3. The automatic scheduling system for operating room nursing tasks according to claim 2, characterized in that: The deployment submodule includes: The data stream collected from the operating room sensor equipment is transmitted to the edge computing node, and Apache Kafka is used for real-time data reception, buffering and distribution. Then, Apache Flink is used to perform real-time cleaning and preprocessing on the flowing data to remove noise and outliers in the data, and to interpolate missing data to obtain initial data.

4. The automatic scheduling system for operating room nursing tasks according to claim 1, characterized in that: The task analysis module includes: The medical event chain parsing submodule receives the initial data transmitted by the data acquisition module, inputs the unstructured text data in the electronic medical record into the pre-trained medical large model BioBERT, identifies key events related to the surgery, sorts the extracted key events according to timestamps to form an ordered event sequence, and stores the constructed event sequence in a structured form to form a surgical process knowledge base; a causal reasoning submodule for identifying causal dependencies of nursing tasks through counterfactual analysis; The input submodule is used to construct a dynamic spatiotemporal graph and input it into the spatiotemporal graph neural network model. The spatial relationship between nodes is learned through graph convolution operations. The temporal changes of nodes are learned through time series analysis. The analysis is performed in different time periods and spatial regions to obtain early warning information for task execution. The integration submodule is used to integrate the surgical process knowledge base, causal dependencies and warning information of task execution, analyze the nursing tasks, and determine the task requirement information, wherein the task requirement information at least includes the priority of the task, the required resources and the execution order.

5. The automatic scheduling system for operating room nursing tasks according to claim 4, characterized in that: The causal reasoning submodule includes: Based on the initial data and surgical process knowledge base, counterfactual analysis scenarios are set for nursing tasks. In each counterfactual analysis scenario, based on the statistical and logical relationships between the factors in the initial data, the causal effects between the nursing tasks and the factors in the initial data are calculated. Based on the calculated causal effects, the causal dependencies between different nursing tasks and between nursing tasks and other factors are determined.

6. The automatic scheduling system for operating room nursing tasks according to claim 4, characterized in that: The input submodule includes: An operating room space-time graph is constructed with the personnel, equipment, and tasks in the operating room as nodes, and relationships including at least personnel movement paths, equipment usage associations, and task sequences as edges. The initial data collected by sensors is used to assign space-time attributes to the nodes and edges of the operating room space-time graph, forming a dynamic space-time graph.

7. The automatic scheduling system for operating room nursing tasks according to claim 1, characterized in that: The task scheduling module includes: The collaborative decision-making submodule is used to abstract each nurse into an independent intelligent agent, construct a multi-dimensional vector containing the real-time status of the operating room, define the set of actions that the intelligent agent can perform, design a comprehensive reward mechanism, conduct a multi-agent collaborative decision-making process, and generate a dynamic allocation plan; The encoding submodule is used to construct a Hamiltonian that represents the task allocation goals and constraints, encode the initial state information of the current caregiver and task, explore the solution space using the tunneling effect of the quantum system, converge the quantum state to the lowest energy state by lowering the system temperature, introduce a penalty term in the quantum optimization process, and simulate the optimal solution search process for task allocation through the interaction between quantum bits; The weighted fusion submodule is used to weightedly fuse the dynamic allocation scheme and the optimal solution provided by the quantum algorithm to obtain the optimal task allocation scheme, convert the optimal task allocation scheme into scheduling instructions, and issue them.

8. The automatic scheduling system for operating room nursing tasks according to claim 7, characterized in that: The collaborative decision-making submodule includes: Each agent obtains local state information and extracts features by observing the environment. The agent calculates the optimal action based on the current policy network and exchanges partial state information through the communication mechanism. When multiple agents compete for the same task, the conflict resolution mechanism is triggered to redistribute the task.

9. The automatic scheduling system for operating room nursing tasks according to claim 1, characterized in that: The real-time monitoring module includes: The comparison submodule is used to obtain multimodal data from the mobile terminals carried by nurses and the monitoring equipment in the operating room, and perform multimodal comparative learning and analysis on the multimodal data; The calculation submodule is used to calculate the distance metric between the embedding vectors of each modality at the current moment. When the distance exceeds the set threshold, it is determined that there is a risk of cross-modal inconsistency; A determination submodule is used to identify cross-modal inconsistencies detected by multimodal contrastive learning and determine abnormal pattern characteristics; The generation submodule is used to obtain the abnormal type and severity according to the abnormal pattern characteristics and generate corresponding alarm information.

10. The automatic scheduling system for operating room nursing tasks according to claim 9, characterized in that: The comparison submodule includes: A convolutional neural network is used to extract features from surveillance video frame images, and the images are mapped to the visual embedding space. The audio Mel-spectrogram is processed through a recurrent neural network combined with an attention mechanism to generate a speech embedding vector. The Transforme model is used to encode the text data to obtain a text embedding representation.

Images