Nursing potential safety hazard monitoring system and method for neurology department

Abstract

The invention discloses a neurology nursing potential safety hazard monitoring system and method, and relates to the technical field of medical information safety monitoring, and the system comprises a first collection processing unit which is used for collecting monitoring data of a target patient, inputting the monitoring data into a time sequence deep network for processing, and obtaining risk feature data; the second collecting and processing unit is used for collecting physiological state data of the target patient, constructing a risk assessment rule base based on a medical knowledge graph to analyze the physiological state data, and generating risk probability data; and the risk early warning unit is used for inputting the risk feature data and the risk probability data into a Bayesian fusion network, generating risk early warning information and sending the risk early warning information to the medical care terminal. According to the invention, collaborative analysis of multi-source information is realized through the Bayesian fusion network, the problem of low early warning accuracy of a traditional single monitoring method is solved, and the reliability of potential safety hazard monitoring is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information security monitoring, and in particular to a system and method for monitoring safety hazards in neurology nursing. Background Art

[0002] Monitoring of safety hazards in neurology nursing is a key link in ensuring the safety of patients with neurological diseases. Traditional nursing monitoring systems mainly rely on fixed threshold judgments and single parameter analysis, which are difficult to adapt to the complex changes in the condition of neurology patients and significant individual differences. In recent years, with the development of Internet of Things technology and artificial intelligence algorithms, intelligent and multi-dimensional nursing monitoring solutions have gradually attracted attention. Existing studies have shown that the efficiency of abnormal state recognition can be improved to a certain extent by collecting patient physiological parameters and environmental data through multi-source sensors and combining them with machine learning methods for analysis. However, these systems still have problems such as insufficient data analysis depth and limited knowledge-driven capabilities in practical applications.

[0003] The existing neurology nursing monitoring system has the following limitations in technical implementation: first, the data analysis method is mainly based on simple statistical analysis and threshold judgment, which is not effective when dealing with complex time series features; second, there is a lack of deep integration of professional medical knowledge in risk assessment, which leads to insufficient accuracy when dealing with multi-parameter interactions and difficulty in establishing an accurate mapping relationship between parameter abnormalities and clinical risks; third, the existing system mainly focuses on the patient's direct physiological indicators, and the collection and analysis of environmental parameters are relatively insufficient, which limits the system's ability to assess the impact of external factors on the patient's status; finally, the early warning mechanism mostly adopts deterministic judgment rules, lacks the support of probability and statistical theory, and lacks robustness when dealing with uncertainty and noise interference. These technical limitations affect the performance and clinical practicality of the nursing monitoring system. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the present invention provides a neurology nursing safety hazard monitoring system and method, which can solve the problems mentioned in the background technology.

[0006] The specific technical scheme is as follows: A method for monitoring safety hazards in neurology nursing care includes: a first acquisition and processing unit, for acquiring monitoring data of a target patient, inputting the monitoring data into a time series deep network for processing, and obtaining risk characteristic data; A second acquisition processing unit is used to collect physiological status data of the target patient, build a risk assessment rule base based on the medical knowledge graph to analyze the physiological status data, and generate risk probability data; The risk warning unit is used to input the risk characteristic data and the risk probability data into a Bayesian fusion network, generate risk warning information, and send the risk warning information to a medical terminal; the Bayesian fusion network is used to collaboratively analyze the risk characteristic data and the risk probability data; the risk warning information includes risk warning levels, abnormal indicators and disposal suggestions.

[0007] As a preferred solution of the neurology nursing safety hazard monitoring method of the present invention, wherein: the time series deep network includes a feature extraction module and an anomaly detection module; Inputting the monitoring data into the time series deep network for processing includes: The monitoring data is input into the feature extraction module, and the feature extraction module uses a long short-term memory network structure to extract the time series features of the monitoring data to obtain a feature time series matrix; The feature time series matrix is ​​input into the anomaly detection module, and the anomaly detection module calculates the weight distribution of each time series feature in the feature time series matrix based on the self-attention mechanism to generate the risk feature data.

[0008] As a preferred solution of the neurology nursing safety hazard monitoring method of the present invention, wherein: the anomaly detection module calculates the weight distribution of each time series feature in the feature time series matrix based on the self-attention mechanism to generate the risk feature data, including: Dividing the feature time series matrix into multiple time windows, and calculating an attention score for the time series features in each time window; Generate a feature weight matrix based on the attention score, wherein the feature weight matrix represents the importance between different temporal features; A short-term anomaly detection unit and a long-term anomaly detection unit are set, wherein: The short-term anomaly detection unit analyzes the feature weight matrix within the time window to identify a sudden anomaly pattern; The long-term anomaly detection unit analyzes the feature weight matrix across a plurality of the time windows to identify a gradual anomaly trend; The sudden abnormal pattern and the gradual abnormal trend are feature fused to generate the risk feature data.

[0009] As a preferred solution of the neurology nursing safety hazard monitoring method of the present invention, wherein: the physiological state data includes physical sign fluctuation data and environmental parameters; the medical knowledge graph includes a risk node set and an associated edge set; the risk node set includes physical sign nodes and environmental nodes, the physical sign nodes correspond to the physical sign fluctuation data, and the environmental nodes correspond to the environmental parameters; Wherein, the physical sign fluctuation data includes body temperature parameters and blood oxygen parameters; the environmental parameters include temperature parameters and light parameters; The risk probability data include the risk probability of physical sign fluctuation and the risk probability of environmental impact; The process of generating the risk probability data comprises the following steps: Mapping the physical sign fluctuation data to the physical sign node, and mapping the environmental parameter to the environmental node; Calculate the node risk value based on the status value of each node; if the status value of the physical sign node exceeds a first preset threshold, set the node risk value to a high risk level; otherwise, keep the calculated node risk value; if the status value of the environment node exceeds a second preset threshold, set the node risk value to a high risk level; otherwise, keep the calculated node risk value; The risk probability data is generated according to the node risk value.

[0010] As a preferred solution of the neurology nursing safety hazard monitoring method of the present invention, wherein: generating the risk probability data according to the node risk value comprises the following steps: Obtain the body temperature parameters and blood oxygen parameters of the target patient in a continuous time period, calculate the change trends of the body temperature parameters and the blood oxygen parameters, and when the change direction of the body temperature parameter is opposite to the change direction of the blood oxygen parameter, increase the risk probability of physical sign fluctuation according to the linear relationship according to the node risk value of the physical sign node, and generate physical sign fluctuation risk probability data; Obtaining the change rate of the body temperature parameter and the change rate of the blood oxygen parameter. When the change rate of the body temperature parameter is greater than its historical average change rate, and the change rate of the blood oxygen parameter is less than its historical average change rate, the physical sign fluctuation risk probability is increased according to the node risk value of the physical sign node in a linear relationship to generate physical sign fluctuation risk probability data; Obtain the temperature parameters and light parameters of the target patient's environment, calculate the change trends of the temperature parameters and light parameters, and when the temperature parameter continues to rise and the light parameter deviates from its preset reference interval, increase the environmental impact risk probability according to the node risk value of the environmental node in a linear relationship to generate environmental impact risk probability data; Obtaining the change trend of the ambient temperature parameter and the target patient's body temperature parameter, and when the change trend of the ambient temperature parameter is consistent with the change trend of the body temperature parameter, increasing the environmental impact risk probability according to a linear relationship based on the node risk value of the environmental node, and generating environmental impact risk probability data; The physical sign fluctuation risk probability data and the environmental impact risk probability data together constitute risk probability data, which are used for Bayesian fusion analysis with the risk feature data.

[0011] As a preferred solution of the neurology nursing safety hazard monitoring method of the present invention, the risk characteristic data and the risk probability data are input into a Bayesian fusion network to generate risk warning information, including the following steps: Constructing a Bayesian fusion network, specifically, constructing the sudden abnormal pattern and the gradual abnormal trend in the risk feature data as the feature layer nodes of the Bayesian fusion network; constructing the physical sign fluctuation risk probability and the environmental impact risk probability in the risk probability data as the probability layer nodes of the Bayesian fusion network; constructing a conditional probability table based on the association relationship between the feature layer nodes and the probability layer nodes, establishing the association between the sudden abnormal pattern and the physical sign fluctuation risk probability, and the association between the gradual abnormal trend and the environmental impact risk probability; Perform Bayesian probability fusion, specifically, establish a conditional probability matrix between feature layer nodes and probability layer nodes, calculate the contribution weight of sudden abnormal patterns to the risk probability of physical sign fluctuations, and the contribution weight of gradual abnormal trends to the risk probability of environmental impacts; use Bayesian reasoning to iteratively update the conditional probability matrix, use the association between sudden abnormal patterns and the risk probability of physical sign fluctuations as a comprehensive evaluation of physical sign abnormalities, and use the association between gradual abnormal trends and the risk probability of environmental impacts as a comprehensive evaluation of environmental impacts; use maximum a posteriori probability estimation to perform probability inference on the updated conditional probability matrix, combine the prior probabilities of feature layer nodes and the conditional probabilities of probability layer nodes, and calculate a comprehensive risk value reflecting the overall state of the patient; Determine the risk warning level based on the comprehensive risk value: If the posterior probability of the comprehensive risk value is greater than the high threshold, the risk warning level is determined to be high; If the posterior probability of the comprehensive risk value is between the low threshold and the high threshold, the risk warning level is determined to be medium; If the posterior probability of the comprehensive risk value is less than the low threshold, the risk warning level is determined to be low; Execute the abnormal index identification process, including calculating the sudden abnormal index and the gradual abnormal index; Calculate the product value of the conditional probability and the risk probability of each sudden abnormal indicator, and the product value of the conditional probability and the risk probability of each gradual abnormal indicator respectively; if the product value of the conditional probability and the risk probability of a sudden abnormal indicator or the product value of the conditional probability and the risk probability of a gradual abnormal indicator is not only the maximum value, and the difference between the maximum value and the second largest product value is greater than a preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum value is determined as an abnormal indicator; If the product value of the conditional probability and the risk probability of a sudden abnormal indicator or the product value of the conditional probability and the risk probability of a gradual abnormal indicator is a maximum value, but the difference between the maximum value and the second largest product value is less than or equal to the preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum product value and the sudden abnormal indicator or the gradual abnormal indicator corresponding to the second largest product value are simultaneously determined as abnormal indicators; If the product values ​​of the conditional probabilities and risk probabilities of the multiple sudden abnormal indicators or the product values ​​of the conditional probabilities and risk probabilities of the gradual abnormal indicators are equal and are the maximum value, then the sudden abnormal indicators or the gradual abnormal indicators with the same maximum product value are all determined as abnormal indicators; Selecting a corresponding disposal suggestion from a preset disposal suggestion library based on the risk warning level and the abnormality indicator; The risk warning level, the abnormal index and the treatment suggestion are packaged as the risk warning information and sent to the medical terminal.

[0012] As a preferred embodiment of the method for monitoring safety hazards in neurology nursing of the present invention, the monitoring data includes physiological parameters and behavioral parameters; wherein the physiological parameters include heart rate parameters, blood pressure parameters and respiratory parameters; and the behavioral parameters include body movement parameters and posture parameters; Calculate the node risk value, including: Acquire the state value of each node, wherein the state value of the vital sign node includes the normalized values ​​of the body temperature parameter and the blood oxygen parameter, and the state value of the environment node includes the normalized values ​​of the temperature parameter and the light parameter; Obtain the weight coefficient of each node in the medical knowledge graph; Multiply the state value of each node by the corresponding weight coefficient and sum them up to get the node risk value.

[0013] To further solve the above technical problems, the present invention provides the following technical solutions: A method for monitoring safety hazards in neurology nursing, comprising: Collecting monitoring data of target patients, inputting the monitoring data into a time series deep network for processing, and obtaining risk characteristic data; Collect physiological status data of target patients, and build a risk assessment rule base based on the medical knowledge graph to analyze the physiological status data and generate risk probability data; The risk characteristic data and the risk probability data are input into a Bayesian fusion network to generate risk warning information, and the risk warning information is sent to a medical terminal.

[0014] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the above-mentioned method for monitoring safety hazards in neurological nursing when executing the computer program.

[0015] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for monitoring safety hazards in neurology nursing.

[0016] Beneficial effects of the present invention: The present invention introduces a time series deep network through the first acquisition and processing unit, thereby realizing in-depth analysis of patient monitoring data, overcoming the limitation that traditional methods are difficult to distinguish between sudden abnormalities and gradual abnormalities; the second acquisition and processing unit constructs a risk assessment rule base based on the medical knowledge graph, converts professional medical knowledge into computable assessment rules, and improves the accuracy and interpretability of risk assessment; the risk warning unit uses a Bayesian fusion network to collaboratively analyze feature data and probability data, and generates risk warning information based on comprehensive consideration of multi-dimensional information, significantly reducing the false alarm rate. The overall solution realizes intelligent processing of the entire process from data acquisition to risk warning through the organic combination of three functional units, providing more reliable and efficient technical support for the monitoring of safety hazards in neurology nursing. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a schematic diagram of the overall structure of the neurology nursing safety hazard monitoring system proposed by the present invention; Figure 2 This is a schematic diagram of the overall process of the neurology nursing safety hazard monitoring method proposed in the present invention. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a method for monitoring safety hazards in neurology nursing.

[0022] Figure 1 The overall structural diagram of the neurology nursing safety hazard monitoring system is shown. The neurology nursing safety hazard monitoring system includes a first acquisition and processing unit, a second acquisition and processing unit and a risk warning unit.

[0023] Among them, the first acquisition and processing unit collects the physiological parameters and behavioral parameters of the target patient as monitoring data, and obtains risk characteristic data after processing through a time series deep network.

[0024] Specifically, the physiological parameters include heart rate parameters, blood pressure parameters and breathing parameters, while the behavioral parameters include body movement parameters and posture parameters.

[0025] It should be noted that the first acquisition processing unit of the present invention includes multiple sensor arrays and data processors, wherein the sensor arrays are composed of a physiological sensor group for acquiring physiological parameters and a behavioral sensor group for acquiring behavioral parameters.

[0026] Specifically, the physiological sensor group includes an electrocardiogram sensor for collecting the heart rate parameters of the target patient, a blood pressure sensor for collecting the systolic and diastolic blood pressure parameters of the target patient, and a respiratory sensor for collecting the respiratory rate parameters of the target patient. These sensors are connected to the data processor wirelessly to transmit the collected physiological parameters in real time. The behavioral sensor group includes a pressure sensor array arranged on the surface of the bed and a posture sensor fixed on the torso of the target patient. The pressure sensor array is used to collect the body motion parameters of the target patient, and the posture sensor is used to collect posture parameters such as motion acceleration and tilt angle. The behavioral sensor group uses low-power Bluetooth technology to transmit data with the data processor.

[0027] The data processor receives the raw data collected by each sensor and first performs signal preprocessing, including: bandpass filtering of the ECG signal to eliminate power frequency interference and baseline drift; median filtering of the blood pressure signal to remove abnormal pulses; wavelet transform noise reduction of the respiratory signal; and normalization of the body movement and posture signals.

[0028] Furthermore, the preprocessed monitoring data is input into the time series deep network for processing, and the time series deep network includes a feature extraction module and anomaly detection module. First, the feature extraction module uses the long short-term memory network structure to extract the time series features of the monitoring data and obtain the feature time series matrix; then, the feature time series matrix is ​​input into the anomaly detection module, and the anomaly detection module calculates the weight distribution of each time series feature in the feature time series matrix based on the self-attention mechanism to generate risk feature data, which specifically includes the following steps: divide the feature time series matrix into multiple time windows, and calculate the attention score for the time series features in each time window; generate a feature weight matrix based on the attention score, and the feature weight matrix represents the importance between different time series features; then, set a short-term anomaly detection unit and a long-term anomaly detection unit, where: The short-term anomaly detection unit analyzes the feature weight matrix within the time window to identify sudden anomaly patterns; The long-term anomaly detection unit analyzes the feature weight matrix across multiple time windows to identify gradually changing anomaly trends; The sudden abnormal pattern and the gradual abnormal trend are feature-fused to generate risk feature data.

[0029] It should be noted that the time series deep network in the present invention adopts a dual-stream structure, including a feature extraction module and an anomaly detection module. The feature extraction module adopts a long short-term memory network (LSTM) structure, which includes an input gate, a forget gate, and an output gate, and can effectively capture long-term dependencies in the monitoring data. The hidden layer dimension of the LSTM network is set to 128, and the time step is 60 to balance the computational efficiency and feature extraction capabilities. The LSTM network processes each type of monitoring data separately: for physiological parameters, its amplitude changes, trend characteristics, and waveform characteristics are extracted; for behavioral parameters, its frequency characteristics, duration, and change pattern are extracted. These features are organized in chronological order to form a feature time series matrix.

[0030] After the feature time series matrix is ​​input into the anomaly detection module, it is first divided into multiple time windows, and the window length can be adjusted according to the changing characteristics of different parameters. In each time window, the self-attention mechanism is used to calculate the correlation strength between different features. The self-attention mechanism generates attention scores through the interaction of the query matrix, key matrix and value matrix, and then forms a feature weight matrix.

[0031] The anomaly detection module includes a short-term anomaly detection unit and a long-term anomaly detection unit. The short-term anomaly detection unit focuses on the feature changes within a single time window and identifies sudden abnormal patterns by setting multi-level discrimination rules. For example, when a physiological parameter changes significantly in a short period of time, or an abnormal body movement pattern is detected, a sudden abnormal alarm will be triggered. The long-term anomaly detection unit identifies gradual abnormalities by tracking and analyzing the feature change trends in multiple continuous time windows. For example, when a physiological parameter does not exceed the alarm threshold but shows a trend of continuous deterioration, or when the behavior pattern gradually deviates from the normal range, the system will identify this gradual abnormal trend. Finally, the system fuses the features of the sudden abnormal pattern and the gradual abnormal trend. The fusion process takes into account the severity, development speed and mutual influence of different types of abnormalities. For example, when multiple parameters are detected to be slightly abnormal at the same time, the system will increase the overall risk level even if the abnormality of a single parameter is not high. The fusion result is the risk feature data, which serves as an important basis for subsequent risk assessment.

[0032] It should be noted that the feature fusion process of the sudden abnormal pattern and the gradual abnormal trend in the present invention can be implemented in the following ways but not limited to: a data fusion method based on Dempster-Shafer evidence theory can be adopted, and the short-term abnormal detection results and the long-term abnormal trend can be used as two sources of evidence. By constructing a basic probability distribution function, the conflict degree between the evidence is calculated, and then the risk feature data after fusion is obtained by using the evidence synthesis rule. A fuzzy comprehensive evaluation method can be used for feature fusion. By establishing a fuzzy relationship matrix, the detection results of short-term anomalies and long-term anomalies are mapped to the risk level space, and the weight coefficient is set according to the importance of different types of anomalies, and finally a comprehensive evaluation result is obtained. A multi-source data fusion method based on Kalman filtering can be adopted, and the time series data of short-term anomaly detection results and long-term abnormal trends can be used as observations. Through the state prediction and measurement update process, the dynamic fusion of multi-source abnormal information can be realized. A feature fusion method based on BP neural network can be adopted, and the short-term abnormal feature vector and the long-term abnormal feature vector are used as the input layer. By setting appropriate hidden layer nodes and activation functions, the network is trained to establish a mapping relationship from features to risk levels. A feature fusion method based on principal component analysis (PCA) can be used to reduce the dimension of the features of short-term abnormalities and long-term abnormalities, extract the main feature components, and evaluate the degree of abnormality based on the feature reconstruction error. It should be understood that the above feature fusion method is only an exemplary description, and those skilled in the art can choose other appropriate fusion methods according to actual needs, as long as they can effectively integrate the short-term abnormal detection results and the long-term abnormal trend to achieve an accurate assessment of the target patient's status.

[0033] Through this multi-dimensional monitoring and analysis method, the present invention can timely discover abnormal conditions of target patients and provide reliable early warning information for medical staff.

[0034] Furthermore, the second acquisition and processing unit collects physiological status data of the target patient, builds a risk assessment rule base based on the medical knowledge graph for analysis, and calculates risk probability data.

[0035] Specifically, physiological state data includes vital sign fluctuation data and environmental parameters. Among them, vital sign fluctuation data includes body temperature parameters and blood oxygen parameters, and environmental parameters include temperature parameters and light parameters. Risk probability data includes vital sign fluctuation risk probability and environmental impact risk probability. The medical knowledge graph includes a risk node set and an associated edge set; the risk node set includes vital sign nodes and environmental nodes, where vital sign nodes correspond to vital sign fluctuation data and environmental nodes correspond to environmental parameters.

[0036] It should be noted that the risk assessment rule base in the present invention is a set of rule mapping and risk calculation mechanisms based on the medical knowledge graph. Specifically, the risk assessment rule base includes two core parts: data mapping rules and risk assessment rules. The data mapping rules define how to map the collected physiological state data to the nodes of the medical knowledge graph. By mapping the physical sign fluctuation data to the physical sign node and mapping the environmental parameters to the environmental node, the system establishes an association relationship between the data and the knowledge graph. This mapping mechanism enables the system to analyze the correlation between different parameters based on the structural characteristics of the knowledge graph. The risk assessment rules mainly include node risk value calculation rules and risk probability generation rules. The node risk value calculation rules quantitatively evaluate the risk level of each node by combining the node state value and the weight coefficient. Among them, the state value is normalized to ensure the comparability of different parameters, and the weight coefficient is dynamically adjusted according to historical data and the degree of influence to ensure the accuracy of the evaluation results. The risk probability generation rules construct a complete risk probability calculation framework based on the preset threshold and node risk value. When an abnormality is detected, the system will enable corresponding evaluation rules according to different situations, such as the coordinated change rules of body temperature and blood oxygen parameters, the association analysis rules of environmental parameters and vital sign parameters, etc. The above rule base construction method based on medical knowledge graph organically combines clinical experience and data features, enabling the system to evaluate the patient's status from multiple dimensions, improving the accuracy and interpretability of risk assessment.

[0037] The process of generating risk probability data includes the following steps: Map the physical sign fluctuation data to the physical sign nodes, and map the environmental parameters to the environmental nodes; Calculate the node risk value, including: Obtain the status value of each node, which includes the following data: the status value of the vital sign node includes the normalized values ​​of body temperature parameters and blood oxygen parameters, and the status value of the environmental node includes the normalized values ​​of temperature parameters and light parameters; obtain the weight coefficient of each node in the medical knowledge graph, wherein the weight coefficient of the vital sign node is determined based on the historical physiological state data of the target patient, and the weight coefficient of the environmental node is determined based on the degree of influence of the environmental parameters on the target patient; multiply the status value of each node by the corresponding weight coefficient and sum them to obtain the node risk value.

[0038] Generate risk probability data based on node risk value: Obtain the target patient's body temperature parameters and blood oxygen parameters in a continuous period of time, calculate the change trends of the body temperature parameters and blood oxygen parameters, and when the change direction of the body temperature parameter is opposite to that of the blood oxygen parameter, increase the risk probability of physical sign fluctuation according to the linear relationship based on the node risk value of the physical sign node, and generate physical sign fluctuation risk probability data including the abnormal degree of body temperature parameters and the fluctuation characteristics of blood oxygen parameters; The change rate of body temperature parameters and the change rate of blood oxygen parameters are obtained. When the change rate of body temperature parameters is greater than its historical average change rate and the change rate of blood oxygen parameters is less than its historical average change rate, the risk probability of physical sign fluctuation is increased according to a linear relationship based on the node risk value of the physical sign node, and physical sign fluctuation risk probability data including the abnormal degree of body temperature parameters and the fluctuation characteristics of blood oxygen parameters are generated.

[0039] It should be noted that the degree of abnormality of body temperature parameters refers to a quantitative indicator of the deviation of the target patient's body temperature parameters from the normal range. In the present invention, this indicator not only takes into account the absolute value deviation of body temperature, but more importantly, combines the temperature change characteristics of neurological patients. For example, for patients with cerebral hemorrhage, even a slight increase in body temperature may indicate a change in the condition; and for patients with multiple sclerosis, diurnal fluctuations in body temperature may affect disease activity. Therefore, the calculation of the degree of abnormality of body temperature parameters needs to comprehensively consider multiple dimensions such as temperature deviation value, rate of change, and duration. The calculation of the abnormal degree of body temperature parameters in the present invention includes but is not limited to the following methods: a weighted scoring method based on multidimensional features, comprehensively considering the temperature baseline deviation index, the rate of change index and the duration index, wherein the temperature baseline deviation index uses the Z-score standardization method to calculate the degree of deviation of body temperature from the individual normal value, the rate of change index is obtained by calculating the derivative of the temperature change in unit time, and the duration index characterizes the proportion of the duration of abnormal body temperature in the observation period; a body temperature fluctuation feature extraction method based on time series analysis, using the autoregressive integral moving average model ARIMA or exponential smoothing method to analyze the change trend and periodic characteristics of body temperature; a body temperature complexity analysis method based on sample entropy, by calculating the entropy value of the body temperature time series to evaluate the stability of the body temperature regulation function. The weights of each calculation index can be adjusted according to the characteristics of different nervous system diseases. The fluctuation characteristics of blood oxygen parameters describe the dynamic change law of the target patient's blood oxygen saturation. In neurological nursing, the fluctuation of blood oxygen parameters may reflect a variety of pathological states, such as sleep apnea, swallowing dysfunction, etc. The present invention constructs a fluctuation characteristic index by analyzing the change amplitude, frequency and duration of blood oxygen parameters. Especially for patients at risk of impaired respiratory function, such as myasthenia gravis, motor neuron disease, etc., early identification of blood oxygen fluctuation characteristics is of great significance for preventing complications. The calculation of blood oxygen parameter fluctuation characteristics in the present invention includes but is not limited to the following methods: a blood oxygen fluctuation evaluation method based on multi-scale analysis, combined with fluctuation amplitude coefficient, frequency characteristic coefficient and duration coefficient, wherein the fluctuation amplitude coefficient is obtained by calculating the standard deviation of blood oxygen saturation, the frequency characteristic coefficient uses power spectrum density analysis to extract the main frequency components of blood oxygen changes, and the duration coefficient characterizes the cumulative time proportion of blood oxygen below the threshold; a blood oxygen abnormality pattern recognition method based on change point detection, using the CUSUM algorithm or the Bayesian change point detection method to identify the mutation point and gradual change interval of blood oxygen saturation; a blood oxygen curve analysis method based on morphological characteristics, by extracting the peak and valley values, slope and area of ​​the blood oxygen waveform and other characteristics to quantify the dynamic characteristics of blood oxygen changes. Each evaluation index can be adaptively adjusted according to the patient's respiratory function status and disease type.

[0040] The present invention adopts the above calculation method, combined with modern statistical analysis and signal processing technology, to achieve accurate quantification of abnormal characteristics of body temperature and blood oxygen parameters. It should be understood that the above calculation method is only an exemplary description, and those skilled in the art can select other appropriate calculation methods according to specific application requirements, as long as they can accurately evaluate the abnormal degree of body temperature and blood oxygen fluctuation characteristics of the target patient.

[0041] The temperature parameters and light parameters of the target patient's environment are obtained, and the changing trends of the temperature parameters and light parameters are calculated. When the temperature parameters continue to rise and the light parameters deviate from their reference intervals, the environmental impact risk probability is increased according to a linear relationship based on the node risk value of the environmental node, and environmental impact risk probability data including the environmental temperature impact intensity and the degree of interference of the light parameters are generated; the changing trends of the environmental temperature parameters and the target patient's body temperature parameters are obtained. When the changing trends of the environmental temperature parameters are consistent with the changing trends of the body temperature parameters, the environmental impact risk probability is increased according to a linear relationship based on the node risk value of the environmental node, and environmental impact risk probability data including the environmental temperature impact intensity and the degree of interference of the light parameters are generated; the vital sign fluctuation risk probability data and the environmental impact risk probability data together constitute the risk probability data, which are used for Bayesian fusion analysis with the risk feature data.

[0042] It should be noted that the intensity of the impact of ambient temperature reflects the actual effect of ambient temperature changes on the target patient. In neurological nursing, ambient temperature not only directly affects the patient's physiological comfort, but may also affect the manifestation of certain symptoms, such as temperature-sensitive patients with neuromyelitis optica may have their symptoms aggravated by increased temperature. The calculation of the intensity of the impact of ambient temperature in the present invention includes but is not limited to the following methods: a weighted combination calculation method based on temperature deviation coefficient, temperature persistence coefficient and body temperature correlation coefficient, wherein the temperature deviation coefficient characterizes the degree of deviation of the current temperature from the median of the appropriate temperature range, the temperature persistence coefficient characterizes the ratio of the duration exceeding the appropriate range to the reference time window, and the body temperature correlation coefficient uses the Pearson correlation coefficient to calculate the correlation between the ambient temperature sequence and the body temperature sequence; a temperature cumulative effect evaluation method based on a sliding time window, by setting time windows of different scales, analyzing the short-term and long-term effects of ambient temperature on the patient's body temperature; a temperature fluctuation feature extraction method based on frequency domain analysis, using fast Fourier transform or wavelet transform to analyze the frequency characteristics and energy distribution of temperature changes. The weight coefficients of each calculation method can be set by expert experience or optimized using machine learning methods.

[0043] The degree of interference of light parameters describes the level of influence of ambient light on the patient's condition. In the nursing practice of neurology, proper light management is of great significance for maintaining the patient's circadian rhythm and improving sleep quality. The calculation of the degree of interference of light parameters in the present invention includes but is not limited to the following methods: a comprehensive evaluation method based on light intensity deviation coefficient, circadian rhythm deviation coefficient and exposure duration coefficient, wherein the light intensity deviation coefficient represents the standardized deviation between the current light intensity and the recommended light intensity, the circadian rhythm deviation coefficient is obtained by comparing the spectrum analysis results of light changes with the standard 24-hour cycle, and the exposure duration coefficient represents the proportion of abnormal light exposure time in the observation time window; a light abnormality pattern recognition method based on local abnormal factors, which identifies abnormal light exposure patterns by calculating the local density characteristics of light intensity in the time series; a light interference classification model based on support vector machine or random forest, which maps the multidimensional features of light parameters to the interference degree scoring space. The parameters of each evaluation method can be dynamically adjusted according to the individual characteristics and clinical needs of the patient.

[0044] These calculation methods all use mature technologies in the fields of statistics and data analysis, combined with professional knowledge of neurology nursing, to achieve accurate quantitative evaluation of the degree of impact of environmental parameters. It should be understood that the above calculation methods are only exemplary, and those skilled in the art can choose other appropriate calculation methods according to actual application scenarios, as long as they can accurately evaluate the degree of impact of environmental parameters on target patients.

[0045] It should be noted that the first preset threshold is a risk level determination benchmark set for the state value of the vital sign node. In the monitoring of safety hazards in neurological nursing, when the state value of the vital sign node exceeds the first preset threshold, it will be directly determined as a high risk level. Specifically, the setting of the first preset threshold needs to comprehensively consider the normalized values ​​of body temperature parameters and blood oxygen parameters: for patients with cerebrovascular diseases who are at risk of fever, the weight of the body temperature parameter after normalization should be higher; for patients with neuromuscular diseases who are at risk of respiratory dysfunction, the weight of the blood oxygen parameter after normalization should be higher; the first preset threshold can be determined by analyzing previous clinical data and combining expert experience, and is preferably set to a weighted combination threshold of the normalized values ​​of body temperature parameters and blood oxygen parameters.

[0046] The second preset threshold is the environmental risk level determination standard set for the environmental node status value. When the environmental node status value exceeds the second preset threshold, it will be directly determined as a high risk level. The second preset threshold needs to consider the normalized values ​​of temperature parameters and lighting parameters at the same time: the normalized value of the ambient temperature parameter should be weighted based on the temperature range suitable for ward care; the normalized value of the lighting parameter should be weighted in accordance with the patient's treatment and rest needs; the characteristics of environmental changes in different seasons and time periods should also be included in the weight consideration range. Exemplarily, the second preset threshold is preferably set to between 0.75-0.85. The determination of this range is based on the following considerations: 1. Determination of the normalized value of the temperature parameter: The suitable temperature range for the ward is usually 20-26°C. When the temperature exceeds this range, the normalized value is about 0.8. Considering the impact of seasonal changes, it can be adjusted within the range of ±0.05. 2. Determination of normalized values ​​of light parameters: The recommended range of daytime illumination in wards is 200-300 lux, and the recommended range of nighttime illumination is less than 50 lux. When the illumination exceeds the recommended range, the normalized value is about 0.75. 3. Comprehensive weight consideration: The recommended weight of temperature parameters is 0.6; the recommended weight of light parameters is 0.4; after weighted combination, the optimal threshold range of 0.75-0.85 is obtained. This threshold range can ensure sensitivity to environmental anomalies while avoiding excessive warnings, and is suitable as a benchmark value for determining environmental risk levels.

[0047] Therefore, the core function of the two preset thresholds is to quickly identify high-risk conditions, trigger a direct increase in risk levels, and provide a basic judgment basis for subsequent risk probability calculations.

[0048] Preferably, the present invention aims to solve the technical problems in the existing neurology nursing monitoring system, such as the overly simple analysis of the patient's physiological status data, the lack of a knowledge-driven risk assessment mechanism, and the inability to effectively identify the impact of environmental factors on the patient's status. To this end, the present invention focuses on realizing the intelligent analysis of physiological status data by constructing a medical knowledge graph and a risk assessment rule base.

[0049] Specifically, the present invention uses a medical knowledge graph to model vital sign fluctuation data and environmental parameters, and characterizes the correlation between different parameters through a set of risk nodes and a set of associated edges. This graph-based modeling method effectively integrates clinical expert knowledge, enabling the system to evaluate the patient's status from multiple dimensions. For example, when evaluating the risk of vital sign fluctuations, the system first analyzes the risk assessment rule base based on the medical knowledge graph, directly identifies high-risk states by calculating node risk values, and generates risk probability data based on the changing trends and rates of changes in body temperature and blood oxygen parameters, thereby achieving rapid and accurate identification of abnormal states.

[0050] In terms of the assessment of environmental factors, the present invention achieves rapid determination of environmental risks by setting environmental nodes and their corresponding preset thresholds. The system calculates node risk values ​​based on the normalized numerical values ​​of environmental parameters, and generates environmental impact risk probability data by analyzing the coordinated change characteristics of temperature parameters and light parameters, as well as the correlation trend between ambient temperature and patient body temperature. This multi-level environmental risk assessment is particularly suitable for the nursing management of temperature-sensitive neuromyelitis optica patients or light-sensitive epilepsy patients.

[0051] The risk warning unit inputs the risk characteristic data and risk probability data into the Bayesian fusion network, generates risk warning information, and sends the risk warning information to the medical terminal.

[0052] Specifically, the Bayesian fusion network is used to collaboratively analyze risk feature data and risk probability data. Risk warning information includes risk warning level, abnormal indicators and disposal suggestions.

[0053] First, a Bayesian fusion network is constructed. The sudden abnormal pattern and gradual abnormal trend in the risk feature data are constructed as the feature layer nodes of the Bayesian fusion network; the physical sign fluctuation risk probability and environmental impact risk probability in the risk probability data are constructed as the probability layer nodes of the Bayesian fusion network; based on the association between the feature layer nodes and the probability layer nodes, a conditional probability table is constructed to establish the association between the sudden abnormal pattern and the physical sign fluctuation risk probability, and the association between the gradual abnormal trend and the environmental impact risk probability.

[0054] Secondly, perform Bayesian probability fusion. Establish a conditional probability matrix between feature layer nodes and probability layer nodes, calculate the contribution weight of sudden abnormal patterns to the risk probability of physical sign fluctuations, and the contribution weight of gradual abnormal trends to the risk probability of environmental impacts; use Bayesian reasoning to iteratively update the conditional probability matrix, and use the association between sudden abnormal patterns and the risk probability of physical sign fluctuations as a comprehensive evaluation of physical sign abnormalities, and use the association between gradual abnormal trends and the risk probability of environmental impacts as a comprehensive evaluation of environmental impacts. Use the maximum a posteriori probability estimation to perform probability inference on the updated conditional probability matrix, and combine the prior probability of the feature layer nodes and the conditional probability of the probability layer nodes to calculate the comprehensive risk value reflecting the overall status of the patient: ; in, Represents the comprehensive risk value, with a value range of [0,1]; Represents a set of risk features, including sudden abnormal patterns and gradual abnormal trend ; Represents a risk probability set, including the risk probability of physical sign fluctuations and environmental impact risk probability ; Indicates the risk probability data A risk probability value; Indicates that under the given comprehensive risk value R, the risk characteristics are observed The conditional probability of =1, Indicates the sudden abnormal mode ;when =2, Indicates a gradual abnormal trend .

[0055] Again, determine the risk warning level based on the comprehensive risk value: like Greater than high threshold θ h , then the risk warning level is judged to be high; like At low threshold θ l and high threshold θ h The risk warning level is medium. like Less than low threshold θ l , the risk warning level is judged to be low.

[0056] It should be noted that the high threshold and low threshold in the present invention are key criteria for distinguishing risk levels, which can be set to 0.85 and 0.65. This interval setting takes into account the sensitivity of patients with neurological diseases to fluctuations in physical signs, which can not only detect high-risk conditions in a timely manner, but also avoid excessive warnings.

[0057] Then, the abnormal indicator identification process is performed, which includes calculating the sudden abnormal indicator and the gradual abnormal indicator: ; ; in, Indicates sudden abnormal indicators. represents the gradient anomaly indicator, m Indicates the index of sudden anomaly indicators and gradual anomaly indicators.

[0058] Calculate the product value of the conditional probability and risk probability of each sudden abnormal indicator, and the product value of the conditional probability and risk probability of each gradual abnormal indicator respectively; if the product value of the conditional probability and risk probability of a sudden abnormal indicator or the product value of the conditional probability and risk probability of a gradual abnormal indicator is not only the maximum value, and the difference between the maximum value and the second largest product value is greater than the preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum value is determined as the abnormal indicator; if the product value of the conditional probability and risk probability of a sudden abnormal indicator or the conditional probability and risk probability of a gradual abnormal indicator is not only the maximum value, but also the difference between the maximum value and the second largest product value is greater than the preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum value is determined as the abnormal indicator; If the product value of the conditional probability of a plurality of sudden abnormal indicators and the risk probability is the maximum value, but the difference between the maximum value and the second largest product value is less than or equal to the preset difference threshold, the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum product value and the sudden abnormal indicator or the gradual abnormal indicator corresponding to the second largest product value are simultaneously determined as abnormal indicators; if the product values ​​of the conditional probability of a plurality of sudden abnormal indicators and the risk probability or the product values ​​of the conditional probability of a plurality of gradual abnormal indicators and the risk probability are equal and are the maximum values, then the sudden abnormal indicators or the gradual abnormal indicators with the same maximum product value are all determined as abnormal indicators.

[0059] It should be noted that the preset difference threshold is used to judge the significance level of abnormal indicators. Its calculation is based on the distribution of the product value of conditional probability and risk probability. It can be set to 0.2 in the present invention, that is, when the product value of a certain indicator exceeds other indicators by more than 0.2, it is judged to be a significant difference. This design based on multi-level thresholds fully considers the particularity of neurological patients. Through the dynamic combination of conditional probability and risk probability, it realizes the accurate identification of different types of abnormalities and provides a reliable early warning basis for medical staff.

[0060] Finally, based on the risk warning level and abnormal indicators, the corresponding treatment suggestion is selected from the preset treatment suggestion library; if it is determined to be a high risk level, the emergency intervention treatment suggestion is selected; if it is determined to be a medium risk level, the key monitoring treatment suggestion is selected; if it is determined to be a low risk level, the routine care treatment suggestion is selected; the risk level, abnormal indicators and treatment suggestions are packaged into risk warning information and sent to the medical terminal.

[0061] Exemplarily, the treatment suggestion library is a hierarchical medical care knowledge base, which sets corresponding treatment plans for different risk levels and abnormal indicator combinations: If the risk level is determined to be high, emergency intervention suggestions will be selected. Emergency intervention suggestions include but are not limited to: immediately notify the doctor on duty to assess the patient's condition; initiate the corresponding emergency plan based on abnormal indicators, such as taking physical cooling or heat preservation measures immediately when the body temperature is abnormal, and immediately giving oxygen or adjusting respiratory support when the blood oxygen is abnormal; increase the frequency of vital sign monitoring and continuously monitor important indicators; prepare emergency equipment and medicines to ensure that emergency items are complete and available; and record the patient's symptom changes and treatment process in detail.

[0062] If the risk level is determined to be medium, the key monitoring treatment recommendations will be selected. Key monitoring treatment recommendations include but are not limited to: appropriately increase the frequency of nursing inspections and closely observe changes in the patient's condition; take corresponding nursing intervention measures for abnormal indicators, such as adjusting the ward temperature, lighting and other environmental parameters; evaluate the patient's medication and treatment effects; strengthen basic nursing care to prevent complications; and keep relevant records to provide a basis for doctors to formulate treatment plans.

[0063] If the risk level is determined to be low, routine nursing treatment suggestions will be selected. Routine nursing treatment suggestions include but are not limited to: maintaining normal nursing inspection frequency; maintaining a suitable ward environment; keeping routine nursing records; observing the patient's general condition; and completing doctor's orders on time.

[0064] When selecting treatment suggestions, the system will make personalized adjustments to the treatment content based on specific abnormal indicators. For example, for patients with abnormal body temperature, the treatment suggestions will focus on temperature management; for patients with abnormal blood oxygen, respiratory management measures will be emphasized. In addition, the system will also consider factors such as the patient's disease type and age to provide more targeted suggestions.

[0065] Compared with the prior art, the present invention can more accurately identify different types of abnormal situations by dividing abnormal indicators into sudden abnormal indicators and gradual abnormal indicators, and calculating the product value of conditional probability and risk probability in a targeted manner. The prior art often uses a single threshold judgment method, which cannot effectively distinguish sudden abnormalities from gradual abnormalities, resulting in a high false alarm rate. The solution of the present invention significantly improves the accuracy of abnormality identification by considering the characteristics of different types of abnormalities.

[0066] Preferably, compared with the prior art, the present invention realizes the collaborative analysis of risk feature data and risk probability data through the Bayesian fusion network, overcoming the limitation of single threshold judgment in the traditional method. The system adopts the maximum a posteriori probability estimation method to calculate the comprehensive risk value, and improves the accuracy of risk warning through the dynamic adjustment of weight factors. For example, when it is detected that the ambient temperature continues to rise and the patient's body temperature shows an upward trend, the system will adjust the risk level in time, which is of great significance for preventing the aggravation of symptoms of patients with neurological diseases. At the same time, the Bayesian fusion network of the present invention has adaptive characteristics, and can continuously optimize the risk assessment effect through iterative updates of the conditional probability matrix, significantly improving the reliability and practicality of the system.

[0067] The intelligent analysis based on medical knowledge graph and Bayesian fusion network provided by the present invention has achieved significant results in improving the accuracy of neurological nursing safety monitoring, reducing false alarm rate, and realizing personalized monitoring through rapid judgment of risk assessment rule base and precise calculation of Bayesian probability fusion, and has important practical value for improving the quality of neurological nursing.

[0068] Example 2, reference Figure 2 , which is an embodiment of the present invention, provides a method for monitoring safety hazards in neurology nursing, comprising: Figure 2 The overall process diagram of the neurology nursing safety hazard monitoring method is shown, including the following steps: S1: Collect monitoring data of target patients and obtain risk characteristic data after processing through a time series deep network.

[0069] S2: Collect the physiological status data of the target patient, build a risk assessment rule base based on the medical knowledge graph for analysis, and calculate the risk probability data.

[0070] S3: Input the risk feature data and risk probability data into the Bayesian fusion network to generate risk warning information, and send the risk warning information to the medical terminal.

[0071] Embodiment 3 is an embodiment of the present invention, which is different from the previous embodiment in that: if the function is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0072] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0073] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0074] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Neurology nursing safety hazard monitoring system, characterized by: include: A first acquisition and processing unit is used to acquire monitoring data of a target patient, and input the monitoring data into a time series deep network for processing to obtain risk characteristic data; A second collection and processing unit is used to collect physiological status data of the target patient, build a risk assessment rule base based on the medical knowledge graph to analyze the physiological status data, and generate risk probability data; A risk warning unit, used for inputting the risk characteristic data and the risk probability data into a Bayesian fusion network, generating risk warning information, and sending the risk warning information to a medical terminal; The Bayesian fusion network is used to collaboratively analyze the risk feature data and the risk probability data; the risk warning information includes risk warning levels, abnormal indicators and disposal suggestions.

2. The neurology nursing safety hazard monitoring system according to claim 1, characterized in that: The temporal deep network includes a feature extraction module and an anomaly detection module; Inputting the monitoring data into the time series deep network for processing includes: The monitoring data is input into the feature extraction module, and the feature extraction module uses a long short-term memory network structure to extract the time series features of the monitoring data to obtain a feature time series matrix; The feature time series matrix is ​​input into the anomaly detection module, and the anomaly detection module calculates the weight distribution of each time series feature in the feature time series matrix based on the self-attention mechanism to generate the risk feature data.

3. The neurology nursing safety hazard monitoring system according to claim 2, characterized in that: The anomaly detection module calculates the weight distribution of each time series feature in the feature time series matrix based on the self-attention mechanism to generate the risk feature data, including: Dividing the feature time series matrix into multiple time windows, and calculating an attention score for the time series features in each time window; Generate a feature weight matrix based on the attention score, wherein the feature weight matrix represents the importance between different temporal features; A short-term anomaly detection unit and a long-term anomaly detection unit are set, wherein: The short-term anomaly detection unit analyzes the feature weight matrix within the time window to identify a sudden anomaly pattern; The long-term anomaly detection unit analyzes the feature weight matrix across a plurality of the time windows to identify a gradual anomaly trend; The sudden abnormal pattern and the gradual abnormal trend are feature fused to generate the risk feature data.

4. The neurology nursing safety hazard monitoring system according to claim 3, characterized in that: The physiological state data includes vital sign fluctuation data and environmental parameters; the medical knowledge graph includes a risk node set and an associated edge set; the risk node set includes vital sign nodes and environmental nodes, the vital sign nodes correspond to the vital sign fluctuation data, and the environmental nodes correspond to the environmental parameters; Wherein, the physical sign fluctuation data includes body temperature parameters and blood oxygen parameters; the environmental parameters include temperature parameters and light parameters; The risk probability data include the risk probability of physical sign fluctuation and the risk probability of environmental impact; The process of generating the risk probability data comprises the following steps: Mapping the physical sign fluctuation data to the physical sign node, and mapping the environmental parameter to the environmental node; Calculate the node risk value based on the status value of each node; if the status value of the physical sign node exceeds a first preset threshold, set the node risk value to a high risk level; otherwise, keep the calculated node risk value; if the status value of the environment node exceeds a second preset threshold, set the node risk value to a high risk level; otherwise, keep the calculated node risk value; The risk probability data is generated according to the node risk value.

5. The neurology nursing safety hazard monitoring system according to claim 4, characterized in that: Generating the risk probability data according to the node risk value comprises the following steps: Obtain the body temperature parameters and blood oxygen parameters of the target patient in a continuous time period, calculate the change trends of the body temperature parameters and the blood oxygen parameters, and when the change direction of the body temperature parameter is opposite to the change direction of the blood oxygen parameter, increase the risk probability of physical sign fluctuation according to the linear relationship according to the node risk value of the physical sign node, and generate physical sign fluctuation risk probability data; Obtaining the change rate of the body temperature parameter and the change rate of the blood oxygen parameter. When the change rate of the body temperature parameter is greater than its historical average change rate, and the change rate of the blood oxygen parameter is less than its historical average change rate, the physical sign fluctuation risk probability is increased according to the node risk value of the physical sign node in a linear relationship to generate physical sign fluctuation risk probability data; Obtain the temperature parameters and light parameters of the target patient's environment, calculate the change trends of the temperature parameters and light parameters, and when the temperature parameter continues to rise and the light parameter deviates from its preset reference interval, increase the environmental impact risk probability according to the node risk value of the environmental node in a linear relationship to generate environmental impact risk probability data; Obtaining the change trend of the ambient temperature parameter and the target patient's body temperature parameter, and when the change trend of the ambient temperature parameter is consistent with the change trend of the body temperature parameter, increasing the environmental impact risk probability according to a linear relationship based on the node risk value of the environmental node, and generating environmental impact risk probability data; The physical sign fluctuation risk probability data and the environmental impact risk probability data together constitute risk probability data, which are used for Bayesian fusion analysis with the risk feature data.

6. The neurology nursing safety hazard monitoring system according to claim 5, characterized in that: Inputting the risk characteristic data and the risk probability data into a Bayesian fusion network to generate risk warning information includes the following steps: Constructing a Bayesian fusion network, specifically constructing the sudden abnormal pattern and the gradual abnormal trend in the risk feature data as feature layer nodes of the Bayesian fusion network; Constructing the risk probability of physical sign fluctuation and the risk probability of environmental impact in the risk probability data as probability layer nodes of the Bayesian fusion network; Based on the association between feature layer nodes and probability layer nodes, a conditional probability table is constructed to establish the association between sudden abnormal patterns and the risk probability of physical sign fluctuations, as well as the association between gradual abnormal trends and the risk probability of environmental impacts; Perform Bayesian probability fusion, specifically, establish a conditional probability matrix between feature layer nodes and probability layer nodes, calculate the contribution weight of sudden abnormal patterns to the risk probability of physical sign fluctuations, and the contribution weight of gradual abnormal trends to the risk probability of environmental impacts; use Bayesian reasoning to iteratively update the conditional probability matrix, use the association between sudden abnormal patterns and the risk probability of physical sign fluctuations as a comprehensive evaluation of physical sign abnormalities, and use the association between gradual abnormal trends and the risk probability of environmental impacts as a comprehensive evaluation of environmental impacts; use maximum a posteriori probability estimation to perform probability inference on the updated conditional probability matrix, combine the prior probabilities of feature layer nodes and the conditional probabilities of probability layer nodes, and calculate a comprehensive risk value reflecting the overall state of the patient; Determine the risk warning level based on the comprehensive risk value: If the posterior probability of the comprehensive risk value is greater than the high threshold, the risk warning level is determined to be high; If the posterior probability of the comprehensive risk value is between the low threshold and the high threshold, the risk warning level is determined to be medium; If the posterior probability of the comprehensive risk value is less than the low threshold, the risk warning level is determined to be low; Execute the abnormal index identification process, including calculating the sudden abnormal index and the gradual abnormal index; Calculate the product value of the conditional probability and risk probability of each sudden abnormal indicator, and the product value of the conditional probability and risk probability of each gradual abnormal indicator respectively; If the product value of the conditional probability and the risk probability of a sudden abnormal indicator or the product value of the conditional probability and the risk probability of a gradual abnormal indicator is not only the maximum value, and the difference between the maximum value and the second largest product value is greater than the preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum value is determined as the abnormal indicator; If the product value of the conditional probability and the risk probability of a sudden abnormal indicator or the product value of the conditional probability and the risk probability of a gradual abnormal indicator is a maximum value, but the difference between the maximum value and the second largest product value is less than or equal to the preset difference threshold, then the sudden abnormal indicator or the gradual abnormal indicator corresponding to the maximum product value and the sudden abnormal indicator or the gradual abnormal indicator corresponding to the second largest product value are simultaneously determined as abnormal indicators; If the product values ​​of the conditional probability and the risk probability of the plurality of sudden abnormal indicators or the product values ​​of the conditional probability and the risk probability of the gradual abnormal indicators are equal and are the maximum value, then the sudden abnormal indicators or the gradual abnormal indicators with the same maximum product value are all determined as abnormal indicators; Selecting a corresponding disposal suggestion from a preset disposal suggestion library based on the risk warning level and the abnormality indicator; The risk warning level, the abnormal index and the treatment suggestion are packaged as the risk warning information and sent to the medical terminal.

7. The neurology nursing safety hazard monitoring system according to claim 6, characterized in that: The monitoring data includes physiological parameters and behavioral parameters; wherein the physiological parameters include heart rate parameters, blood pressure parameters and breathing parameters; the behavioral parameters include body movement parameters and posture parameters; Calculate the node risk value, including: Acquire the state value of each node, wherein the state value of the vital sign node includes the normalized values ​​of the body temperature parameter and the blood oxygen parameter, and the state value of the environment node includes the normalized values ​​of the temperature parameter and the light parameter; Obtain the weight coefficient of each node in the medical knowledge graph; Multiply the state value of each node by the corresponding weight coefficient and sum them up to get the node risk value.

8. A method for monitoring safety hazards in neurology nursing, based on the neurology nursing safety hazard monitoring system according to any one of claims 1 to 7, characterized in that: include: Collecting monitoring data of target patients, inputting the monitoring data into a time series deep network for processing, and obtaining risk characteristic data; Collect physiological status data of target patients, and build a risk assessment rule base based on the medical knowledge graph to analyze the physiological status data and generate risk probability data; The risk characteristic data and the risk probability data are input into a Bayesian fusion network to generate risk warning information, and the risk warning information is sent to a medical terminal.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for monitoring safety hazards in neurological nursing as described in claim 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the method for monitoring safety hazards in neurological nursing as described in claim 8 are implemented.

Images