An operating room nursing monitoring method and system

Abstract

The application provides an operating room nursing monitoring method and system, generates a multi-source time sequence synchronization data stream of a unified time reference between patient physiological parameters and surgical operations in an operating room through multi-source monitoring data of patient physiological parameters and personnel operation behaviors, and extracts a multi-modal feature tensor set representing patient physiological characteristics and surgical operation characteristics; determines a time-varying correlation weight matrix of a correlation between the patient physiological parameters and the surgical operations based on historical multi-source monitoring data, constructs a perioperative dynamic knowledge graph of patient risks in the operating room through the time-varying correlation weight matrix; determines a collaborative reasoning result representing potential risks of the patient according to the perioperative dynamic knowledge graph and the multi-modal feature tensor set; and generates grading warning information of physiological abnormal risks of the patient and matching clinical intervention guidelines based on the collaborative reasoning result. According to the application, the patient in the operating room can be nursed and monitored based on the correlation between the surgical operations and the patient physiological parameters.

Description

Technical Field

[0001] This application relates to the field of nursing monitoring technology, and more specifically, to a method and system for operating room nursing monitoring. Background Technology

[0002] Nursing monitoring is an important part of the medical and nursing process. It refers to the continuous or periodic collection and analysis of patients' health data, such as vital signs, changes in condition, and psychological state, by nursing staff using professional equipment, clinical observation, and assessment tools, in order to promptly detect abnormalities, evaluate the effectiveness of nursing care, and provide a basis for medical decision-making.

[0003] Currently, operating room nursing monitoring mainly relies on manual monitoring and experience-based judgment by medical staff. Patient physiological parameters and surgical personnel's operational behaviors are collected independently and in a scattered manner, without unified temporal calibration and multimodal fusion, making it impossible to form coherent and correlated time-series data. Historical monitoring data has not been deeply analyzed, making it difficult to quantify and summarize the inherent patterns between surgical procedures and patient physiological changes. Furthermore, a systematic and dynamically updated risk assessment knowledge system is lacking. Traditional methods can only issue simple alarms for single physiological indicators exceeding limits, failing to combine surgical procedures for coupled analysis and forward reasoning. This results in delayed risk identification, poor targeted early warnings, and low levels of intelligence, failing to meet the requirements of refined, safe, and intelligent nursing monitoring in operating rooms. Therefore, how to conduct nursing monitoring of patients in the operating room based on the correlation between surgical procedures and patient physiological parameters has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for monitoring nursing care in the operating room, which can monitor the nursing care of patients in the operating room based on the correlation between surgical procedures and patients' physiological parameters.

[0005] In a first aspect, this application provides a method for monitoring nursing care in the operating room, comprising the following steps: Multiple types of medical sensor devices are deployed in the operating room, and multi-source monitoring data of patients' physiological parameters and personnel's operational behavior are collected based on the deployed medical sensor devices. Multimodal fusion is performed on the multi-source monitoring data to generate a multi-source time-series synchronous data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room. A multimodal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics is extracted from the multi-source time-series synchronous data stream. Historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room are acquired. Temporal causal correlation mining is performed on the historical multi-source monitoring data to obtain a time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operation. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the boundary of the surgical operation standard matching the target surgical procedure. Based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, collaborative reasoning of the abnormal risk of the patient's physiological state in the operating room is performed to obtain collaborative reasoning results that characterize the patient's potential risk. Based on the results of the collaborative reasoning, a graded early warning information on the risk of physiological abnormalities in patients and corresponding clinical intervention guidelines are generated.

[0006] In some embodiments, multimodal fusion of the multi-source monitoring data to generate a multi-source time-series synchronized data stream with a unified time reference between patient physiological parameters and surgical procedures in the operating room specifically includes: The multi-source monitoring data is preprocessed to obtain preprocessed multi-source monitoring data; Time matching is performed on the preprocessed multi-source monitoring data to obtain matched multi-source monitoring data; The matched multi-source monitoring data are fused into a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical procedure in the operating room.

[0007] In some embodiments, extracting a multimodal feature tensor set characterizing patient physiological characteristics and surgical operation characteristics from the multi-source time-series synchronized data stream specifically includes: The multi-source time-synchronous data stream is segmented to obtain multiple time-related data segments; Patient physiological characteristics and surgical operation characteristics are extracted from each time-series related data segment to obtain a multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics.

[0008] In some embodiments, performing time-series causal association mining on the historical multi-source monitoring data to obtain a time-varying association weight matrix of the relationship between patient physiological parameters and surgical procedures specifically includes: The historical multi-source monitoring data is preprocessed to obtain preprocessed historical multi-source monitoring data; Time-series correlation mining was performed on the preprocessed historical multi-source monitoring data to obtain the correlation between different operations and physiological fluctuations. A time-varying correlation weight matrix is ​​constructed based on all the correlations to determine the relationship between patient physiological parameters and surgical procedures.

[0009] In some embodiments, constructing a perioperative dynamic knowledge graph of patient risk in the operating room through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the boundary of surgical operation procedures matched with the target surgical procedure specifically includes: Obtain the preoperatively generated individualized dynamic physiological safety thresholds for patients and the boundaries of surgical operation procedures that match the target surgical procedure; Patient physiological parameters, surgical procedure type, risk event type, risk level, and intervention strategy are set as nodes in the knowledge graph; The directed association edges and relation weights between each node are determined based on the time-varying association weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the surgical operation specification boundary that matches the target surgical procedure. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed by using directed edges and relational weights between nodes and each node.

[0010] In some embodiments, the collaborative reasoning of the physiological state of patients in the operating room based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, to obtain a collaborative reasoning result characterizing the patient's potential risk, specifically includes: Obtain the physiological abnormality judgment rules for surgical procedure adaptation in the operating room; Using the physiological abnormality judgment rules adapted to the surgical procedure as constraints, the abnormal risk of the multimodal feature tensor set is judged to obtain the basic judgment results of whether the patient's physiological parameters deviate from the threshold and whether the surgical operation exceeds the standard boundary. Establish a collaborative reasoning mechanism that combines knowledge graph logical reasoning with temporal feature probabilistic reasoning; Based on the perioperative dynamic knowledge graph and the collaborative reasoning mechanism, the basic judgment results are collaboratively analyzed to obtain collaborative reasoning results that characterize the patient's potential risks.

[0011] In some embodiments, generating graded early warning information on the risk of patient physiological abnormalities and corresponding clinical intervention guidelines based on the collaborative reasoning results specifically includes: Obtain the preset risk classification and early warning strategy; Based on the preset risk grading and early warning strategy, the collaborative reasoning results are used to make early warning judgments, thereby obtaining graded early warning information on the patient's physiological abnormality risk and corresponding clinical intervention guidelines.

[0012] Secondly, this application provides an operating room nursing monitoring system, comprising: The acquisition module is used to deploy various types of medical sensor devices in the operating room and to collect multi-source monitoring data of patients' physiological parameters and personnel's operational behavior based on the deployed medical sensor devices. The processing module is used to perform multimodal fusion on the multi-source monitoring data, generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room, and extract a multimodal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics from the multi-source time-series synchronized data stream; The processing module is also used to acquire historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room, perform time-series causal correlation mining on the historical multi-source monitoring data, obtain a time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operation, and construct a perioperative dynamic knowledge graph of patient risk in the operating room through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the boundary of the surgical operation standard matching the target surgical procedure. The processing module is also used to perform collaborative reasoning on the abnormal risk of the patient's physiological state in the operating room based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, so as to obtain collaborative reasoning results that characterize the patient's potential risk. The execution module is used to generate graded early warning information on the risk of physiological abnormalities in patients and corresponding clinical intervention guidelines based on the collaborative reasoning results.

[0013] Thirdly, this application provides a computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described operating room nursing monitoring method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described operating room nursing monitoring method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The operating room nursing monitoring method and system provided in this application first deploys multiple types of medical sensor devices in the operating room, and collects multi-source monitoring data of patient physiological parameters and personnel operation behaviors based on the deployed medical sensor devices; performs multimodal fusion on the multi-source monitoring data to generate a multi-source time-series synchronous data stream with a unified time reference between patient physiological parameters and surgical operations in the operating room, and extracts a multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics from the multi-source time-series synchronous data stream; acquires historical multi-source monitoring data corresponding to each type of medical sensor device in the operating room, and performs time-series causality analysis on the historical multi-source monitoring data. Association mining is used to obtain a time-varying association weight matrix of the relationship between patient physiological parameters and surgical procedures. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed using this time-varying association weight matrix, preoperatively generated individualized dynamic physiological safety thresholds for patients, and surgical procedure standard boundaries matching the target surgical procedure. Based on this perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, collaborative reasoning of abnormal risks in the patient's physiological state is performed to obtain collaborative reasoning results characterizing potential patient risks. Based on these collaborative reasoning results, graded early warning information and corresponding clinical intervention guidelines for patient physiological abnormality risks are generated.

[0016] Therefore, this application, in the process of operating room nursing monitoring, collects multi-source monitoring data of patient physiological parameters and personnel operational behaviors through multiple types of sensors, which can provide comprehensive and complete raw data support for the correlation analysis between surgical operations and patient physiological parameters; by performing multimodal fusion on multi-source monitoring data and extracting multimodal feature tensor sets, unified temporal alignment and feature condensation of physiological parameters and surgical operations can be achieved; by constructing a time-varying correlation weight matrix and a perioperative dynamic knowledge graph through correlation mining of historical data, the inherent correlation patterns between surgical operations and patient physiological parameters can be discovered and solidified, forming a dynamically updated correlation risk knowledge base, thus solving the problem of two This addresses the problem of unquantifiable and reusable correlations between surgical procedures and patients. By combining perioperative dynamic knowledge graphs, multimodal feature tensor sets, and pre-defined rules for collaborative reasoning of abnormal risks, a comprehensive assessment of potential patient risks can be achieved based on the correlation between surgical procedures and patient physiological parameters. This transforms single-indicator monitoring into correlated and coupled reasoning, significantly improving the accuracy and foresight of risk identification. Finally, based on the collaborative reasoning results, tiered early warning information and corresponding clinical intervention guidelines are generated. More targeted early warning content can be output based on the correlation between these two, truly realizing operating room nursing monitoring centered on the intrinsic correlation between surgical procedures and physiology, effectively improving the safety and intelligence level of intraoperative monitoring. Using this proposed solution, nursing monitoring of patients in the operating room can be conducted based on the correlation between surgical procedures and patient physiological parameters. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of an operating room nursing monitoring method according to some embodiments of this application; Figure 2 This is an exemplary flowchart illustrating the determination of a multi-source timing synchronization data stream according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the determination of a time-varying correlation weight matrix according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of an operating room nursing monitoring system according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing an operating room nursing monitoring method according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The figure is an exemplary flowchart of an operating room nursing monitoring method according to some embodiments of this application. The operating room nursing monitoring method mainly includes the following steps: In step 101, multiple types of medical sensor devices are deployed in the operating room, and multi-source monitoring data of the patient's physiological parameters and personnel's operational behavior are collected based on the deployed medical sensor devices.

[0020] In practice, based on the spatial layout of the operating room, the division of surgical operation areas, and the patient monitoring needs, various types of medical sensor devices are deployed in different zones within the operating room. Sensors for physiological parameters such as heart rate, blood pressure, blood oxygen, body temperature, and electrocardiogram are attached to the corresponding monitoring sites on the patient's body surface. Cameras, motion capture, instrument status, and behavioral perception sensors are placed on the ceiling of the operating room, around the operating table, and on surgical instruments. All medical sensor devices are uniformly calibrated and configured with the same communication protocol and acquisition clock. Through these various medical sensor devices, real-time data on the patient's various physiological parameters, as well as data on the surgical operations, instrument use, and adherence to behavioral norms by medical staff, are collected synchronously, forming multi-source monitoring data covering both patient physiology and operating room operational behavior.

[0021] It should be noted that the multi-source monitoring data in this application reflects the patient's real-time physiological state, vital sign fluctuations, and tolerance level throughout the entire surgical process. It also objectively reflects the surgical operation behavior of medical staff, the standardization of the operation procedure, the use of instruments, and the correspondence between the surgical operation and the patient's physiological changes, providing a comprehensive, true, and objective basis for operating room nursing safety monitoring, abnormal risk identification, and early warning.

[0022] In step 102, the multi-source monitoring data is fused in a multi-modal manner to generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room. A multi-modal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics is extracted from the multi-source time-series synchronized data stream.

[0023] In some embodiments, reference Figure 2 The figure is an exemplary flowchart of determining a multi-source time-series synchronized data stream in some embodiments of this application. In this embodiment, multi-modal fusion of the multi-source monitoring data to generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and surgical operations in the operating room can be achieved by the following steps: In step 1021, the multi-source monitoring data is preprocessed to obtain preprocessed multi-source monitoring data; In step 1022, time matching is performed on the preprocessed multi-source monitoring data to obtain matched multi-source monitoring data; In step 1023, the matched multi-source monitoring data is fused in a multi-modal manner to form a multi-source time-series synchronous data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room.

[0024] In specific implementation, the multi-source monitoring data is preprocessed to obtain preprocessed multi-source monitoring data. This can be achieved by the following method: processing the multi-source monitoring data using data preprocessing methods in this field, removing redundant and erroneous data generated during sensor acquisition through data cleaning, eliminating abnormal data exceeding the normal range using threshold judgment, filling in missing data that occurred during acquisition using linear interpolation algorithms, and performing unified format conversion and dimension normalization processing on monitoring data of different formats and dimensions to obtain preprocessed multi-source monitoring data. Other methods can also be used in other embodiments, which are not limited here.

[0025] In addition, in specific implementation, time matching is performed on the preprocessed multi-source monitoring data to obtain the matched multi-source monitoring data. This can be achieved in the following way: using the unified system clock of the operating room monitoring system as a reference, timestamp calibration and time matching are performed on the preprocessed physiological parameter data and surgical operation behavior data. Multi-source data with different acquisition frequencies and different transmission delays are aligned to the same time reference, and the patient's physiological data at the same moment is associated with the corresponding surgical operation data to eliminate the time offset between multi-source data and obtain the matched multi-source monitoring data. Other methods can also be used in other embodiments, which are not limited here.

[0026] In addition, in specific implementation, the multi-modal data fusion of the matched multi-source monitoring data to form a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and surgical operations in the operating room can be achieved in the following way: using existing spatiotemporal data fusion methods, the multi-type monitoring data that have completed time matching are associated and integrated in the time and spatial dimensions, and the physiological parameters and surgical operation data corresponding to the same time node are bound and merged to form a data unit containing three core information elements: timestamp, physiological parameters, and surgical operation. Then, all data units are arranged continuously in chronological order to finally form a multi-source time-series synchronized data stream with a unified time reference that can synchronously reflect the correspondence between the patient's physiological parameters and surgical operations. Other methods can also be used in other embodiments, which are not limited here.

[0027] It should be noted that the multi-source time-series synchronous data stream in this application represents a time-series data set of patient physiological parameters and surgical operation behavior, reflecting the patient's vital signs and physiological indicators at every moment during the entire surgical process, as well as the surgical operation content and behavior of medical staff at the corresponding moment, intuitively demonstrating the synchronous correspondence, time-series change pattern and intrinsic linkage characteristics between surgical operation and patient physiological fluctuations.

[0028] In some embodiments, extracting a multimodal feature tensor set characterizing patient physiological characteristics and surgical operation characteristics from the multi-source time-series synchronized data stream can be achieved using the following steps: The multi-source time-synchronous data stream is segmented to obtain multiple time-related data segments; Patient physiological characteristics and surgical operation characteristics are extracted from each time-series related data segment to obtain a multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics.

[0029] In specific implementation, the multi-source time-series synchronized data stream is segmented to obtain multiple time-series associated data segments. This can be achieved in the following way: the multi-source time-series synchronized data stream is segmented based on the phased characteristics of the operating room surgical procedure and the temporal continuity of routine monitoring data. It is implemented by combining two segmentation methods. First, based on the core nodes of the surgical operation, the continuous multi-source time-series synchronized data stream is divided into independent data segments corresponding to the surgical stage, such as division according to surgical steps such as skin incision, lesion exposure, hemostasis, suturing, and incision closure. Second, a fixed-duration time window is preset, such as 10 seconds per time window, and the data stream of the incomplete stage is segmented. To avoid the breakage of feature information between adjacent data segments, the time window is overlapped for segmentation, with an overlap duration of 2 seconds. Through the above methods, the complete multi-source time-series synchronized data stream is divided into multiple time-series associated data segments of regular length that cover complete surgical information. Each data segment contains all physiological parameters and surgical operation behavior data within the corresponding time period. Other methods can also be used in other embodiments, which are not limited here.

[0030] In addition, in specific implementation, the extraction of patient physiological characteristics and surgical operation characteristics from each time-series associated data segment to obtain a multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics can be achieved in the following way: extract patient physiological characteristics and surgical operation characteristics from each time-series associated data segment separately; for the numerical physiological parameter data in each time-series associated data segment, use statistical analysis methods in this field to extract features; specifically calculate the average, maximum, minimum, rate of change, and fluctuation amplitude of physiological parameters such as heart rate, blood pressure, and blood oxygen saturation in that time-series associated data segment; at the same time, identify the peak and trough times of physiological parameters in that data segment and their corresponding values, as well as the duration of physiological parameters exceeding the normal clinical range, thereby forming a set of physiological features representing the patient's physiological state. For surgical operation data such as images and actions within each time-series associated data segment, firstly, invalid action segments caused by sensor jitter and environmental interference are removed using conventional image frame extraction and action sequence sorting methods. Then, the type features, frequency features, amplitude features, and continuity features of the operation sequence are extracted to form a set of surgical operation features representing the surgical operation behavior. Finally, the physiological feature set and surgical operation feature set extracted from each data segment are dimensionally unified and structurally integrated. Tensor encoding is performed according to a fixed structure of time window identifier, physiological feature dimension, and surgical operation feature dimension to finally generate a multimodal feature tensor set that can completely represent the patient's physiological change pattern and surgical operation behavior pattern. Other methods can be used in other embodiments, which are not limited here.

[0031] It should be noted that the time-series associated data segment in this application refers to a continuous sub-data set with local time ranges, containing patient physiological parameters and surgical operation behaviors within the corresponding time period, obtained by dividing multi-source time-series synchronous data streams according to fixed duration windows or surgical procedure nodes. It reflects the short-term changes and fluctuations of the patient's vital signs within a certain local time period, as well as the specific surgical operation behaviors and characteristics of the corresponding medical staff, demonstrating the real-time correspondence between surgical operations and the patient's physiological state within a short period. The multimodal feature tensor set reflects the core information of the fluctuation amplitude, trend of change, abnormal state of the patient's physiological parameters, and the type, frequency, amplitude, and standardization of surgical operations. It eliminates redundant and interfering content in the original data and highly condenses the key information related to patient safety and surgical standardization, which can be used to provide standardized and high-quality feature data support for abnormal risks of physiology and operation.

[0032] In step 103, historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room are acquired. Time-series causal correlation mining is performed on the historical multi-source monitoring data to obtain a time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operations. The perioperative dynamic knowledge graph of patient risk in the operating room is constructed through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the boundary of the surgical operation standard matching the target surgical procedure.

[0033] It should be noted that the historical multi-source monitoring data in this application refers to multimodal, time-stamped historical monitoring data collected and archived by various medical sensor devices during multiple surgeries of the same type of surgery and patients with similar conditions. This data includes patient physiological parameters, medical staff operation behaviors, and surgical procedure nodes. It reflects the routine changes in patients' physiological indicators under similar surgeries, the characteristics of typical abnormalities, the objective correlation between different surgical operations and physiological fluctuations, safe operating ranges, and the triggering conditions of historical risk events. This provides a real and reliable prior data foundation for exploring the intrinsic relationship between surgical operations and patient physiology and establishing risk assessment criteria.

[0034] In some embodiments, reference Figure 3 The figure is an exemplary flowchart for determining the time-varying correlation weight matrix in some embodiments of this application. In this embodiment, the time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operations can be obtained by performing time-series causal correlation mining on the historical multi-source monitoring data using the following steps: In step 1031, the historical multi-source monitoring data is preprocessed to obtain preprocessed historical multi-source monitoring data; In step 1032, time-series correlation mining is performed on the preprocessed historical multi-source monitoring data to obtain the correlation between different operations and physiological fluctuations; In step 1033, a time-varying correlation weight matrix is ​​constructed based on all the correlations to determine the relationship between the patient's physiological parameters and the surgical procedure.

[0035] In specific implementation, the historical multi-source monitoring data is preprocessed to obtain the preprocessed historical multi-source monitoring data. This can be achieved in the following way: First, the historical multi-source monitoring data is screened to remove invalid data samples caused by surgical interruptions, data missing rates exceeding 10%, and sensor malfunctions, retaining only valid samples with complete surgical procedures and continuous data acquisition. Then, the screened valid samples are cleaned, manually removing redundant and duplicate data and garbled data. Thresholds are set based on clinically recognized normal ranges of physiological parameters, and a threshold judgment method is used to remove abnormal data exceeding reasonable ranges. For the small amount of missing data in the samples, [further processing is required]. Linear interpolation is used for data imputation, which involves selecting the average of two adjacent valid data points before and after the missing data and using this average as the imputation value. Simultaneously, physiological parameter data of different dimensions are normalized using a scaling method, converting all data into dimensionless data between 0 and 1. Surgical operation behavior data such as images and actions are format-standardized and converted into structured data that can be associated with numerical physiological parameter data. Finally, preprocessed historical multi-source monitoring data with complete data, no abnormal noise, uniform format, and consistent dimensions is obtained. Other methods can be used in other embodiments, which are not limited here.

[0036] In addition, in specific implementation, the temporal correlation mining of the preprocessed historical multi-source monitoring data to obtain the correlation between different operations and physiological fluctuations can be achieved in the following way: First, the preprocessed historical data is grouped according to surgical type and surgical stage to ensure the comparability of data in the same group. Then, for each group of data, with surgical operation behavior as the independent variable, changes in patient physiological parameters as the dependent variable, and anesthesia depth, intraoperative blood loss, and arterial blood gas indicators as control variables, partial correlation analysis is used to calculate the partial correlation coefficient between each surgical operation and changes in each physiological parameter. False correlations caused by control variables are eliminated to clarify the true magnitude and duration of changes in physiological parameters before and after the operation. The larger the absolute value of the partial correlation coefficient, the stronger the correlation between the surgical operation and the corresponding physiological parameter change. At the same time, the frequency of occurrence of this correlation in the historical sample is counted, and weak correlations with a frequency of less than 5% are eliminated. Finally, the correlation between different surgical operations and various physiological parameter fluctuations is obtained, clarifying the direction, magnitude, and correlation strength of the influence of each operation on physiological parameters. Other methods can also be used in other embodiments, which are not limited here.

[0037] In addition, in specific implementation, the time-varying correlation weight matrix that constructs the correlation between patient physiological parameters and surgical operations based on all the correlations can be implemented in the following way: First, define the rows and columns of the matrix, with the surgical operation type as the matrix row and the patient physiological parameter index as the matrix column. Use the partial correlation coefficient between each surgical operation obtained from time-series correlation mining and the corresponding physiological parameter as the element value of the corresponding position in the matrix. The positive or negative value of the element value indicates the direction of the correlation. A positive value indicates that the physiological parameter increases after the operation is performed, and a negative value indicates that the physiological parameter decreases after the operation is performed. The absolute value of the element value indicates the correlation strength. At the same time, set a matrix update mechanism. Every time a certain amount of historical monitoring data is added, repeat the above preprocessing and time-series correlation mining steps, recalculate the partial correlation coefficient, and replace the element value of the corresponding position in the matrix to realize the dynamic update of the matrix. Finally, a time-varying correlation weight matrix that can reflect the changes in the correlation between patient physiological parameters and surgical operations in real time is constructed. Other methods can also be used in other embodiments, which are not limited here.

[0038] It should be noted that the correlation in this application represents the quantitative correspondence between different surgical procedures and fluctuations in patients' physiological parameters in terms of temporal influence, correlation strength, and change patterns. It reflects the direction, magnitude, and linkage characteristics of the influence of various surgical procedures on patients' physiological indicators such as heart rate, blood pressure, and blood oxygen. The time-varying correlation weight matrix reflects the dynamic changes in the correlation strength and influence weight between different surgical procedures and patients' physiological parameters. It can adapt to the regular updates of different surgical scenarios and new data in real time. It can be used to provide the correlation weights and logical constraints between nodes for the perioperative dynamic knowledge graph of the operating room. It is the core data carrier for realizing the risk assessment of the linkage between operation and physiology.

[0039] In some embodiments, constructing a perioperative dynamic knowledge graph of patient risk in the operating room through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the boundary of surgical operation procedures matched with the target surgical procedure can be achieved through the following steps: Obtain the preoperatively generated individualized dynamic physiological safety thresholds for patients and the boundaries of surgical operation procedures that match the target surgical procedure; Patient physiological parameters, surgical procedure type, risk event type, risk level, and intervention strategy are set as nodes in the knowledge graph; The directed association edges and relation weights between each node are determined based on the time-varying association weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the surgical operation specification boundary matching the target surgical procedure. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed by using directed edges and relational weights between nodes and each node.

[0040] In practice, the first step is to obtain the patient's individualized dynamic physiological safety threshold generated preoperatively and the surgical operation standard boundary matching the target surgical procedure. These preset parameters are determined based on clinical nursing standards, anesthesia monitoring standards, and surgical operation industry guidelines. The patient's individualized dynamic physiological safety threshold generated preoperatively is an initial threshold generated based on the patient's age, underlying diseases, preoperative baseline examination results, and anesthesia classification. During the operation, it is dynamically adjusted in real time based on the depth of anesthesia, surgical progress, and blood gas analysis results. Specifically, it includes the normal range, abnormal threshold, and critical threshold of heart rate, blood pressure, blood oxygen saturation, body temperature, and electrocardiogram. The surgical operation standard boundary matching the target surgical procedure is generated based on the target surgical type, standard surgical procedure, and the patient's preoperative imaging anatomical data. It is adjusted in stages according to the surgical stage, including the range of motion, execution sequence, and operation area boundary of skin incision, hemostasis, and suturing, as well as the usage specifications and transfer procedures of surgical instruments. For example, skin incision operation requires controlling the incision depth within a preset range, and hemostasis operation requires being performed within a designated area and the range of motion does not exceed the safety limit. Other methods can be used in other embodiments, which are not limited here.

[0041] Furthermore, in practical implementation, setting patient physiological parameters, surgical operation types, risk event types, risk levels, and intervention strategies as nodes in the knowledge graph can be achieved in the following way: Using the knowledge graph entity node construction method of this field, the core nodes of the knowledge graph are clearly defined. Specifically, the patient physiological parameter node includes all monitored physiological indicators such as heart rate, blood pressure, blood oxygen saturation, and body temperature, with each physiological parameter node associated with a corresponding individualized dynamic physiological safety threshold as an attribute. The surgical operation type node specifically includes various operations in the entire surgical process, such as skin incision, lesion exposure, hemostasis, suturing, and instrument transfer, with each operation node associated with a corresponding target surgical procedure. The matching surgical procedure standard boundary is used as an attribute; the risk event type node specifically includes categories such as abnormal physiological parameters, operational violations, and coupling risks, and each risk event node is associated with specific abnormal manifestations as an attribute; the risk level node is divided into three levels: high risk, medium risk, and low risk, and each level node is associated with the corresponding risk judgment standard as an attribute; the intervention strategy node includes clinical treatment measures and emergency operation procedures for different risks, and each strategy node is associated with the corresponding applicable scenario as an attribute. Through the above settings, a complete entity node framework of the knowledge graph is formed, ensuring that the attributes of each node are clear and traceable. Other methods can be used in other embodiments, which are not limited here.

[0042] In addition, in specific implementation, the directed association edges and relational weights between each node can be determined based on the time-varying association weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the surgical operation specification boundary matching the target surgical procedure. This can be achieved in the following way: For surgical operation type nodes and patient physiological parameter nodes, the association coefficient between the corresponding surgical operation and physiological parameter in the time-varying association weight matrix is ​​used as the relational weight of the directed association edge between them. The direction of the directed edge is from the surgical operation type node to the patient physiological parameter node, which intuitively reflects the influence of the surgical operation on the physiological parameter. For example, the weight of the edge from the skin incision operation node to the heart rate node is the association coefficient between the skin incision operation and heart rate in the time-varying association weight matrix. The larger the absolute value of the association coefficient, the higher the weight and the stronger the association. For patient physiological parameter nodes and risk event type and risk level nodes, the actual value of the physiological parameter is compared with the preoperatively generated individualized dynamic physiological safety threshold for patients. If the physiological parameter exceeds the normal range but does not reach the critical threshold, a physiological parameter node is established pointing to the "abnormal physiological parameter" event. The directed edges connecting nodes and "medium-risk" level nodes are weighted according to the extent of the exceedance; the greater the exceedance, the higher the weight. If physiological parameters reach the critical threshold, a directed edge is established pointing to the "abnormal physiological parameters" event node and the "high-risk" level node, with the weight set to the highest value. For surgical operation type nodes and risk event type and risk level nodes, the actual execution of the surgical operation is compared with the boundary of the surgical operation standard matching the target surgical procedure. If the operation exceeds the standard boundary, a directed edge is established from the surgical operation node to the "operation violation" event node and the corresponding risk level node, with the weight set according to the degree of exceedance; the more severe the exceedance, the higher the weight. For risk level nodes and intervention strategy nodes, a directed edge is established from the risk level node to the corresponding intervention strategy node based on the corresponding clinical intervention measures matched to different risk levels, with the weights all set to fixed values ​​to ensure accurate correspondence between intervention strategies and risk levels. Through the above methods, the directed edges and corresponding weights between all nodes are clearly defined. Other methods can be used in other embodiments, which are not limited here.

[0043] Furthermore, in specific implementation, the perioperative dynamic knowledge graph of patient risk in the operating room, constructed by the directed edges and relation weights between nodes, can be achieved in the following way: Using traditional knowledge graph topology construction methods in this field, the nodes, directed edges, and relation weights are structurally integrated and logically bound. They are then categorized and organized according to node type and relational relationships to construct a graph topology structure containing a node layer, a relational layer, and a weight layer, ensuring clear relational logic between nodes and reasonable weight allocation. Simultaneously, a dynamic update mechanism is set up, where the time-varying relational weight matrix is ​​iteratively updated with newly added historical monitoring data, and physiological thresholds are updated according to clinical data. When adjusting the surgical procedure specifications or optimizing the boundaries of surgical operation specifications for matching the target surgical procedure, the attribute information, the direction of the directed edges, and the relation weights of the corresponding nodes are updated synchronously. For example, after adding surgical data, the correlation coefficient between a certain type of operation and physiological parameters in the time-varying correlation weight matrix changes. That is, the weights of the edges between the operation node and the corresponding physiological parameter node are updated synchronously to ensure that the knowledge graph can adapt to the changes in the surgical scenario in real time. Finally, a perioperative dynamic knowledge graph that can evolve dynamically and can represent the inherent logical relationship between the physiological parameters of patients in the operating room, surgical operations, and safety risks is constructed. Other methods can also be used in other embodiments, which are not limited here.

[0044] It should be noted that the preset physiological safety thresholds in this application represent the pre-set normal, abnormal, and critical ranges of physiological indicators such as heart rate, blood pressure, blood oxygen, and body temperature, reflecting whether the patient's vital signs are in a safe and stable state. They can be used to directly determine whether physiological indicators are abnormal and provide quantitative judgment standards for risk identification. The surgical operation standard boundary for target procedure matching represents the pre-set compliance boundaries of various surgical operations in terms of spatial location, range of motion, execution sequence, and instrument use, reflecting whether the surgical operation conforms to the standard procedures and safety requirements. It can be used to determine whether the operation is in violation of regulations and to provide a basis for identifying operation-related risks. The directed association edge represents a logical connection in the knowledge graph that connects different entities with a clear direction, reflecting the unidirectional causal or correspondence between the surgical operation and physiological parameters, the correspondence between physiological abnormalities and risks, and the matching of risks and intervention measures. The relation weight represents the quantitative value assigned to each directed association edge, reflecting the strength of the association, the magnitude of the influence, and the degree of risk between different nodes. The perioperative dynamic knowledge graph reflects the complete, real-time, and dynamic internal logic between surgical operations, patient physiology, risk events, and intervention measures, and can be used to provide knowledge support for operating room safety early warning and nursing decision-making.

[0045] In step 104, the physiological state of the patient in the operating room is subjected to collaborative reasoning of abnormal risk based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, so as to obtain the collaborative reasoning result characterizing the patient's potential risk.

[0046] In some embodiments, the collaborative reasoning of the physiological state of patients in the operating room based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, to obtain a collaborative reasoning result characterizing the patient's potential risk, can be achieved through the following steps: Obtain the physiological abnormality judgment rules for surgical procedure adaptation in the operating room; Using the physiological abnormality judgment rules adapted to the surgical procedure as constraints, the abnormal risk of the multimodal feature tensor set is judged to obtain the basic judgment results of whether the patient's physiological parameters deviate from the threshold and whether the surgical operation exceeds the standard boundary. Establish a collaborative reasoning mechanism that combines knowledge graph logical reasoning with temporal feature probabilistic reasoning; Based on the perioperative dynamic knowledge graph and the collaborative reasoning mechanism, the basic judgment results are collaboratively analyzed to obtain collaborative reasoning results that characterize the patient's potential risks.

[0047] It should be noted that the physiological abnormality judgment rules for surgical procedure adaptation in this application represent the pre-established judgment criteria and constraints for physiological parameters, reflecting the safety boundaries of the patient's vital signs, the compliance bottom line of surgical operations, and the distinction standards for different degrees of abnormality. They serve as the objective basis for judging whether the state is normal, abnormal, or dangerous. The content includes the normal range, abnormal threshold, critical threshold, and duration of continuous abnormality judgment conditions for various physiological parameters such as heart rate, blood pressure, blood oxygen, and body temperature. It also covers the normative boundaries and violation judgment standards for surgical operations such as the range of motion, execution sequence, and operation area. Furthermore, it includes the graded judgment conditions for individual abnormalities of physiological parameters, operational violations, and abnormalities of the coupling of the two. It can be directly used as the basis for judging abnormalities in the patient's physiological state and surgical behavior, performing compliance verification on real-time feature data, and outputting basic abnormality judgment results.

[0048] In specific implementation, using the physiological abnormality judgment rules adapted to the surgical procedure as constraints, the multimodal feature tensor set is subjected to feature judgment of abnormal risk to obtain basic judgment results on whether the patient's physiological parameters deviate from the threshold and whether the surgical operation exceeds the standard boundary. This can be achieved in the following way: For the sub-tensors representing the patient's physiological characteristics in the multimodal feature tensor set, the core feature values ​​of the average value, maximum fluctuation value, and duration of abnormality of physiological parameters are extracted, and each value is compared with the normal range, abnormal threshold, critical threshold, and duration standard in the judgment table of the physiological abnormality judgment rules adapted to the surgical procedure. If the physiological parameter value exceeds the abnormal threshold, the judgment is made accordingly. If the value meets the duration requirement, it is judged as "abnormal physiological parameters", and the abnormality type and the magnitude of the deviation are recorded. For the sub-tensor that centrally represents the surgical operation characteristics of the multimodal feature tensor, the core feature information such as the amplitude of the operation action, the execution sequence, and the regional displacement are extracted. These are compared dimension by dimension with the surgical operation standard boundary that matches the target surgical procedure in the judgment table. If the amplitude of the operation action deviates from the standard boundary or the timing is lagging, it is judged as "operation violation", and the violation type and the degree of violation are recorded. Finally, all comparison results are integrated to obtain a clear basic judgment result. Other methods can be used in other embodiments, which are not limited here.

[0049] Furthermore, in the specific implementation, a collaborative reasoning mechanism combining knowledge graph logical reasoning and temporal feature probabilistic reasoning is determined. This collaborative reasoning mechanism is a well-known and conventional technique in the field of operating room risk reasoning, specifically divided into two major reasoning modules with clearly defined integration logic: First, the knowledge graph logical reasoning module, relying on the node associations, relation weights, and causal transmission paths of surgical operation-physiological parameters-risk events-intervention strategies stored in the perioperative dynamic knowledge graph, conducts causal deduction according to the unidirectional logical chain of "operational behavior-physiological state change-risk event occurrence-intervention strategy adaptation". For example, starting from the hemostasis operation node, its influence weight on blood oxygen saturation is found through the association edge, and then the abnormal blood oxygen saturation node is associated with the circulatory system risk event node, clarifying the source and path of risk transmission; Second, the temporal feature probabilistic reasoning module... The reasoning module, based on the temporal continuity of the multimodal feature tensor set, uses temporal statistical methods in this field to calculate the probability of anomalies. Specifically, it calculates the probability of the persistence of abnormal physiological parameters, the probability of the coupling between operational violations and physiological abnormalities, and the probability of the development trend of risk events. For example, through historical data statistics, it is found that "the coupling probability of a slight overstep in hemostasis operation accompanied by abnormal blood oxygenation is 65%". At the same time, a probability threshold is set, which is judged as high probability above 70%, medium probability between 50% and 70%, and low probability below 50%. Finally, the fusion rules are clarified, and the causal logic derived from the knowledge graph is integrated with the probability of temporal feature probability reasoning. Based on causal logic and with probability values ​​as an aid, the core logical framework of the collaborative reasoning mechanism is jointly constructed. Other methods can be used in other embodiments, which are not limited here.

[0050] Furthermore, in specific implementation, the basic judgment results, based on the perioperative dynamic knowledge graph and the collaborative reasoning mechanism, are collaboratively analyzed to obtain collaborative reasoning results characterizing the patient's potential risks. This can be achieved in the following way: First, the basic judgment results are input into the knowledge graph logic reasoning module of the collaborative reasoning mechanism. Through the associated edges and weights of the graph nodes in the perioperative dynamic knowledge graph, the specific types of risk events, inducing factors, and potential transmission paths are traced, while matching the corresponding preliminary range of risk levels. Then, the basic judgment results are input into the temporal feature probability reasoning module of the collaborative reasoning mechanism. Combined with the temporal data of the multimodal feature tensor set, the specific probability and development trend of various risks occurring under the current abnormal state are calculated. For example, it is concluded that "the probability of current abnormal blood oxygenation accompanied by hemostasis exceeding the limit leading to circulatory system failure is 68%, showing a continuous upward trend." Subsequently, the analysis results of the two modules are fused and verified to exclude isolated physiological parameter waves. Dynamic interference, such as a brief drop in blood oxygen saturation without accompanying operational abnormalities, is judged as accidental interference. If multi-dimensional abnormal coupling exists, the authenticity of the risk is confirmed. Then, based on the correlation between risk level and intervention strategy in the perioperative dynamic knowledge graph, corresponding emergency treatment measures are matched. Finally, the risk type, risk level, inducing factors, development trend, and intervention recommendations are integrated to form a complete collaborative reasoning result that characterizes the patient's potential risk. For example, "The patient has a high risk of circulatory system caused by a moderate abnormality in blood oxygen saturation coupled with a slight overstep of the hemostasis operation. The inducing factor is that the amplitude of the hemostasis operation deviates from the standard boundary by 2mm during the operation. The risk development trend is that if the operation amplitude is not adjusted in time, it will continue to deteriorate. The corresponding intervention strategy is to immediately reduce the amplitude of the hemostasis operation, continuously monitor blood pressure changes, and prepare vasopressors." This provides accurate and comprehensive decision-making basis for operating room safety monitoring and emergency treatment. Other methods can also be used in other embodiments, which are not limited here.

[0051] It should be noted that the basic judgment results in this application reflect whether the patient's current physiological parameters deviate from the corresponding thresholds, the magnitude of the deviation, and the duration of the deviation, as well as whether the surgical operation exceeds the standard boundaries, the degree of transgression, and the timing violations. It only reflects whether there is an abnormality in a single dimension and does not involve risk correlation or trend analysis. The collaborative reasoning mechanism represents a comprehensive risk analysis model composed of knowledge graph logical reasoning and temporal feature probabilistic reasoning. It reflects the complete reasoning logic that integrates and judges the causal relationships, safety rule constraints, and real-time temporal features in the perioperative dynamic knowledge graph. It can be used to conduct multi-dimensional coupling analysis on the basic judgment results to achieve accurate judgment from single abnormality to comprehensive risk. The collaborative reasoning results reflect the potential risk types, risk levels, inducing factors, risk transmission paths, development trends, and supporting intervention suggestions generated by the coupling of the patient's physiology and surgical operation. It fully reflects the patient's current overall safety status and can be used to directly generate the patient's risk classification early warning information and supporting clinical intervention guidelines, providing accurate and reliable decision-making basis for operating room safety monitoring and clinical emergency response.

[0052] In step 105, based on the collaborative reasoning results, a graded early warning information on the patient's physiological abnormality risk and corresponding clinical intervention guidelines are generated.

[0053] In some embodiments, generating graded early warning information on the risk of patient physiological abnormalities and corresponding clinical intervention guidelines based on the collaborative reasoning results can be achieved through the following steps: Obtain the preset risk classification and early warning strategy; Based on the preset risk grading and early warning strategy, the collaborative reasoning results are used to make early warning judgments, thereby obtaining graded early warning information on the patient's physiological abnormality risk and corresponding clinical intervention guidelines.

[0054] It should be noted that the risk grading and early warning strategy preset in this application reflects the severity, urgency, and warning priority of different physiological abnormality risks, and clarifies the unified execution standard from risk assessment to early warning output. Its content includes complete configuration rules such as the basis for risk level classification, triggering conditions for different risk levels, intensity of audio-visual prompts, interface warning colors, text broadcast content, information push targets, and early warning display format. It can be directly used to standardize and normalize the early warning level judgment of collaborative reasoning results, match the corresponding warning form, guide the structured generation of graded early warning information and supporting clinical intervention guidelines, ensure that the early warning content is clear and hierarchical, and provide intuitive, unified, and timely risk prompts for operating room medical staff.

[0055] In specific implementation, the risk grading and early warning strategy is used to make early warning judgments on the collaborative reasoning results to obtain graded early warning information and corresponding clinical intervention guidelines for the patient's physiological abnormality risk. This can be achieved in the following way: Based on the preset risk grading and early warning strategy, the collaborative reasoning results are analyzed and graded one by one. First, key contents such as risk type, risk level, inducing factors, abnormal physiological parameters, violations, and risk development trends are extracted from the collaborative reasoning results. Then, the extracted information is matched one by one with the early warning trigger conditions of each level in the strategy to accurately determine the early warning level corresponding to the current risk. Graded early warning information of the corresponding level is generated according to the format specified by the strategy. At the same time, according to the risk level and risk type determined by the matching, the corresponding clinical intervention guidelines pre-bound in the strategy are automatically retrieved. The graded early warning information and the corresponding clinical intervention guidelines are integrated and standardized and packaged. Finally, complete early warning and guidance information containing risk level, abnormal cause, risk trend, and treatment suggestions is generated. This information can be directly displayed on the operating room monitoring equipment and pushed to medical staff. The entire process is implemented using existing conventional technologies such as rule matching, condition comparison, information retrieval and packaging. The judgment logic is clear and the execution process is stable and reliable.

[0056] It should be noted that the graded early warning information in this application refers to standardized warning information divided into different warning levels according to the degree of risk and urgency. It reflects the severity, urgency and development trend of the risk of coupling between patient physiological abnormalities and surgical procedures. It can be used to provide tiered and differentiated risk reminders to medical staff, intuitively distinguish the severity of risks and achieve accurate early warning. The supporting clinical intervention guidelines reflect the standardized treatment measures, operation adjustment suggestions and emergency treatment procedures for various intraoperative abnormal risks. They can be used to directly provide medical staff with timely, clear and executable clinical treatment guidance, quickly carry out targeted interventions and improve the standardization and timeliness of intraoperative emergency treatment.

[0057] In another aspect, in some embodiments, this application provides an operating room nursing monitoring system, with reference to... Figure 4 The figure is a schematic diagram of the structure of an operating room nursing monitoring system according to some embodiments of this application. The operating room nursing monitoring system 400 includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below: The acquisition module 401 in this application is mainly used to deploy multiple types of medical sensor devices in the operating room, and to collect multi-source monitoring data of the patient's physiological parameters and personnel's operating behavior based on the deployed medical sensor devices. Processing module 402, in this application, is used to perform multimodal fusion on the multi-source monitoring data, generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room, and extract a multimodal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics from the multi-source time-series synchronized data stream; It should be noted that the processing module 402 in this application is also used to acquire historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room, perform time-series causal correlation mining on the historical multi-source monitoring data, obtain a time-varying correlation weight matrix of the correlation between patient physiological parameters and surgical operation, and construct a perioperative dynamic knowledge graph of patient risk in the operating room through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the surgical operation standard boundary matching the target surgical procedure. In addition, it should be noted that the processing module 402 in this application is also used to perform collaborative reasoning on the abnormal risk of the patient's physiological state in the operating room based on the perioperative dynamic knowledge graph, the multimodal feature tensor set and the physiological abnormality judgment rules of the surgical procedure adaptation in the operating room, so as to obtain the collaborative reasoning result characterizing the patient's potential risk. The execution module 403 in this application is mainly used to generate graded early warning information on the risk of physiological abnormalities in patients and supporting clinical intervention guidelines based on the collaborative reasoning results.

[0058] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described operating room nursing monitoring method.

[0059] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing an operating room nursing monitoring method according to some embodiments of this application. The operating room nursing monitoring method in the above embodiments can... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0060] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0061] The communication bus 502 can be used to transmit information between the aforementioned components.

[0062] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0063] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The method used in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0064] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0065] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0066] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0067] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described operating room nursing monitoring method.

[0068] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0069] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for monitoring nursing care in the operating room, characterized in that, Includes the following steps: Multiple types of medical sensor devices are deployed in the operating room, and multi-source monitoring data of patients' physiological parameters and personnel's operational behavior are collected based on the deployed medical sensor devices. Multimodal fusion is performed on the multi-source monitoring data to generate a multi-source time-series synchronous data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room. A multimodal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics is extracted from the multi-source time-series synchronous data stream. Historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room are acquired. Temporal causal correlation mining is performed on the historical multi-source monitoring data to obtain a time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operation. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the boundary of the surgical operation standard matching the target surgical procedure. Based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, collaborative reasoning of the abnormal risk of the patient's physiological state in the operating room is performed to obtain collaborative reasoning results that characterize the patient's potential risk. Based on the results of the collaborative reasoning, a graded early warning information on the risk of physiological abnormalities in patients and corresponding clinical intervention guidelines are generated.

2. The method as described in claim 1, characterized in that, The multi-source monitoring data is fused in multiple modes to generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical procedure in the operating room. This specifically includes: The multi-source monitoring data is preprocessed to obtain preprocessed multi-source monitoring data; Time matching is performed on the preprocessed multi-source monitoring data to obtain matched multi-source monitoring data; The matched multi-source monitoring data are fused into a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical procedure in the operating room.

3. The method as described in claim 1, characterized in that, The multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics extracted from the multi-source time-series synchronized data stream specifically includes: The multi-source time-synchronized data stream is segmented to obtain multiple time-related data segments; Patient physiological characteristics and surgical operation characteristics are extracted from each time-series related data segment to obtain a multimodal feature tensor set representing patient physiological characteristics and surgical operation characteristics.

4. The method as described in claim 1, characterized in that, The time-varying correlation weight matrix obtained by performing time-series causal association mining on the historical multi-source monitoring data specifically includes: The historical multi-source monitoring data is preprocessed to obtain preprocessed historical multi-source monitoring data; Time-series correlation mining was performed on the preprocessed historical multi-source monitoring data to obtain the correlation between different operations and physiological fluctuations. A time-varying correlation weight matrix is ​​constructed based on all the correlations to determine the relationship between patient physiological parameters and surgical procedures.

5. The method as described in claim 1, characterized in that, The perioperative dynamic knowledge graph of patient risk in the operating room is constructed by using the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety thresholds for patients, and the surgical operation standard boundaries that match the target surgical procedure. Specifically, this includes: Obtain the preoperatively generated individualized dynamic physiological safety thresholds for patients and the boundaries of surgical operation procedures that match the target surgical procedure; Patient physiological parameters, surgical procedure type, risk event type, risk level, and intervention strategy are set as nodes in the knowledge graph; The directed association edges and relation weights between each node are determined based on the time-varying association weight matrix, the preoperatively generated individualized dynamic physiological safety threshold for patients, and the surgical operation specification boundary that matches the target surgical procedure. A perioperative dynamic knowledge graph of patient risk in the operating room is constructed by using directed edges and relational weights between nodes and each node.

6. The method as described in claim 1, characterized in that, Based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, collaborative reasoning is performed on the abnormal risk of the patient's physiological state in the operating room. The collaborative reasoning results characterizing the patient's potential risk specifically include: Obtain the rules for determining physiological abnormalities in surgical procedure adaptation within the operating room; Using the physiological abnormality judgment rules adapted to the surgical procedure as constraints, the abnormal risk of the multimodal feature tensor set is judged to obtain the basic judgment results of whether the patient's physiological parameters deviate from the threshold and whether the surgical operation exceeds the standard boundary. Establish a collaborative reasoning mechanism that combines knowledge graph logical reasoning with temporal feature probabilistic reasoning; Based on the perioperative dynamic knowledge graph and the collaborative reasoning mechanism, the basic judgment results are collaboratively analyzed to obtain collaborative reasoning results that characterize the patient's potential risks.

7. The method as described in claim 1, characterized in that, The generation of graded early warning information and corresponding clinical intervention guidelines for patients' physiological abnormality risks based on the aforementioned collaborative reasoning results specifically includes: Obtain the preset risk classification and early warning strategy; Based on the preset risk grading and early warning strategy, the collaborative reasoning results are used to make early warning judgments, thereby obtaining graded early warning information on the patient's physiological abnormality risk and corresponding clinical intervention guidelines.

8. An operating room nursing monitoring system, characterized in that, include: The acquisition module is used to deploy various types of medical sensor devices in the operating room and to collect multi-source monitoring data of patients' physiological parameters and personnel's operational behavior based on the deployed medical sensor devices. The processing module is used to perform multimodal fusion on the multi-source monitoring data, generate a multi-source time-series synchronized data stream with a unified time reference between the patient's physiological parameters and the surgical operation in the operating room, and extract a multimodal feature tensor set representing the patient's physiological characteristics and the surgical operation characteristics from the multi-source time-series synchronized data stream; The processing module is also used to acquire historical multi-source monitoring data corresponding to various types of medical sensor devices in the operating room, perform time-series causal correlation mining on the historical multi-source monitoring data, obtain a time-varying correlation weight matrix of the relationship between patient physiological parameters and surgical operation, and construct a perioperative dynamic knowledge graph of patient risk in the operating room through the time-varying correlation weight matrix, the preoperatively generated individualized dynamic physiological safety threshold of the patient, and the boundary of the surgical operation standard matching the target surgical procedure. The processing module is also used to perform collaborative reasoning on the abnormal risk of the patient's physiological state in the operating room based on the perioperative dynamic knowledge graph, the multimodal feature tensor set, and the physiological abnormality judgment rules for surgical procedure adaptation in the operating room, so as to obtain collaborative reasoning results that characterize the patient's potential risk. The execution module is used to generate graded early warning information on the risk of physiological abnormalities in patients and corresponding clinical intervention guidelines based on the collaborative reasoning results.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the operating room nursing monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the operating room nursing monitoring method as described in any one of claims 1 to 7.

Images