Intelligent neurosurgery nursing monitoring method and system
Through intelligent neurosurgical nursing monitoring methods, multimodal data and generative adversarial networks are used to solve the problem of limited nursing monitoring efficiency and quality in the existing technology, and fast and accurate patient monitoring and risk prediction are achieved, improving nursing quality and efficiency.
Patent Information
- Application Number
- CN202411975163.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-04-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing neurosurgical nursing monitoring methods rely on manual observation and basic medical equipment, making it difficult to quickly collect and analyze patient condition data, resulting in limited nursing efficiency and disease analysis capabilities, affecting the quality of care.
Using intelligent neurosurgical nursing monitoring methods, we collect multimodal nursing data, identify physiological classification data, build a training generator, analyze the change trends of risk levels and physiological indicators, set a discrimination baseline, conduct cyclic adversarial training, build a generative adversarial network, monitor the patient's physical status and establish a feedback mechanism.
It realizes rapid and accurate monitoring of neurosurgery patients, identify risk factors in advance, improves nursing efficiency and quality, enhances data simulation accuracy and confrontation ability, promptly detects abnormal situations, and improves the processing efficiency of medical staff.
Smart Images
Figure CN119786065A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical care, and in particular to an intelligent neurosurgery care monitoring method and system. Background Art
[0002] Neurosurgical nursing monitoring refers to the process of using a variety of neuroelectrophysiological monitoring technologies to monitor the health status of neurosurgical patients who need to be monitored and managed in real time. By using the neurosurgical nursing monitoring system, medical staff can promptly detect changes in the patient's condition, such as intracranial hemorrhage, thereby reducing the occurrence of adverse events. With the development of my country's medical industry and the continuous improvement of computer technology, advanced imaging technology enables doctors to diagnose neurosurgical diseases more accurately, which has led to the rapid development of neurosurgical nursing needs. In order to improve the quality of care, hospitals have to upgrade the neurosurgical nursing monitoring system.
[0003] Existing neurosurgery nursing monitoring mainly relies on manual observation and basic medical equipment. This method evaluates and analyzes the condition of neurosurgery patients by using vital signs monitoring instruments, neurological assessments and nursing records to improve the quality of care. However, this method requires medical staff to regularly monitor patients' vital signs and frequently patrol wards, and it is difficult to quickly collect and analyze patient condition data, which limits the efficiency of neurosurgery nursing and the ability to analyze patients' conditions, thereby affecting the quality of care for neurosurgery patients. Summary of the invention
[0004] In order to solve the above problems, the present invention provides an intelligent neurosurgery nursing monitoring method and system, which can improve the nursing quality and nursing efficiency of neurosurgery patients.
[0005] In a first aspect, the present invention provides an intelligent neurosurgery nursing monitoring method, comprising: Collecting multimodal nursing data of neurosurgery patients, identifying physiological classification data of patients in the multimodal nursing data, and constructing a training generator of the multimodal nursing data according to the physiological classification data of patients; Analyze the risk level corresponding to the patient's physiological classification data, extract the physiological indicators corresponding to the risk level, analyze the time series change trend of the physiological indicators, set the normal change threshold of the patient's physiological classification data according to the time series change trend, set the discrimination baseline of the training generator based on the normal change threshold, and use the discrimination baseline to analyze the quality of the generated data of the training generator; Constructing a discriminant model of the quality of the generated data, setting the recognition degree of the discriminant model for the quality of the generated data, and based on the recognition degree, performing cyclic adversarial training on the training generator and the discriminant model to obtain a cyclic adversarial training result; According to the result of the cyclic adversarial training, the balance coefficient between the training generator and the discriminant model is calculated, and the recognition degree of the discriminant model to the output data of the training generator is calculated; combining the balance coefficient and the recognition degree, a generative adversarial network of the training generator and the discriminant model is constructed; based on the generative adversarial network, the physical state value of the neurosurgery patient during the nursing process is monitored, and a feedback mechanism for the physical state value is constructed; The generative adversarial network and the feedback mechanism are combined to perform nursing monitoring on the neurosurgery patient to obtain a nursing monitoring result.
[0006] In a possible implementation of the first aspect, identifying the patient physiological classification data in the multimodal nursing data includes: Formatting the multimodal nursing data to obtain formatted data; Analyzing the data source of the multimodal nursing data and identifying the creation time corresponding to the data in the multimodal nursing data; Setting a data tag of the formatted data according to the data source and the data creation time; Based on the data labels, the multimodal nursing data is classified to obtain patient physiological classification data.
[0007] In a possible implementation of the first aspect, constructing a training generator for the multimodal nursing data according to the patient physiological classification data includes: identifying abnormal missing values of the multimodal nursing data based on the patient physiological classification data; Analyzing statistical attributes of the multimodal nursing data, and analyzing data features of the multimodal nursing data based on the statistical attributes and the abnormal missing values; Identify the variation risk of the data features and set a risk suppression mechanism for the variation risk; analyzing a degree of simulation of generated data corresponding to the multimodal nursing data; A training generator for the multimodal nursing data is constructed based on the risk suppression mechanism and the simulation degree.
[0008] In a possible implementation manner of the first aspect, analyzing the risk level corresponding to the patient's physiological classification data includes: Analyzing physiological characteristics corresponding to the patient's physiological classification data; Calculating feature correlations between the patient's physiological classification data based on the physiological features; Based on the feature correlation, constructing a feature grading judgment matrix of the patient's physiological classification data; Calculate the indicator weight corresponding to the patient's physiological classification data: The risk level corresponding to the patient's physiological classification data is analyzed by combining the indicator weights and the feature classification judgment matrix.
[0009] In a possible implementation of the first aspect, analyzing the time series change trend of the physiological indicator includes: Collecting sample data corresponding to the physiological index, and performing cleaning and correction processing on the sample data to obtain target sample data; Analyze the significant differences of the target sample data at different times; According to the significant difference, evaluating the change amplitude of the target sample data over time; Under the variation range, analyzing the result attribute of the physiological indicator and constructing a time series evolution diagram of the result attribute; Based on the time series evolution diagram, the time series change trend of the physiological index is analyzed.
[0010] In a possible implementation manner of the first aspect, setting a discrimination baseline of the training generator based on the normal change threshold includes: Based on the normal change threshold, analyzing the recognition effect of the training generator on the abnormal physiological data of the patient; Evaluating the interpretability of the recognition effect; Defining a tolerance level for the simulation data of the training generator according to the explainability and the recognition effect; Based on the tolerance level, a discrimination baseline of the training generator is set.
[0011] In a possible implementation manner of the first aspect, constructing the discriminant model of the generated data quality includes: Collecting real data corresponding to the generated data, and extracting physiological index time series characteristics of the real data and the generated data; Analyzing and integrating the temporal characteristics of the physiological indicators to obtain integrated characteristics; Calculating the characteristic correlation coefficient between the physiological index time series characteristics according to the integrated characteristics; Based on the feature correlation coefficient, setting data partitions of the actual data and the generated data; Performing data discrimination processing on the data partition to obtain a discrimination output result; Combining the integrated features with the discriminant output results, a discriminant model for the quality of the generated data is constructed.
[0012] In a possible implementation manner of the first aspect, setting data partitions of the real data and the generated data based on the feature correlation coefficient includes: Based on the feature correlation coefficient, identifying the feature levels of the actual data and the generated data; According to the feature level, setting a partitioning method of the actual data and the generated data; Analyzing the division ratio between the actual data and the generated data; According to the partitioning method and the partitioning ratio, data partitions of the actual data and the generated data are set.
[0013] In a possible implementation manner of the first aspect, the constructing a feedback mechanism for the physical state value includes: Collecting physiological data in a generative adversarial network corresponding to the body state value, and analyzing the reliability of the physiological data; According to the reliability and the actual situation of the physical state value during the nursing process, setting a threshold value for abnormal physiological state of the neurosurgery patient; Setting a trigger alarm for an abnormal threshold of the physiological state; A feedback mechanism for the body state value is constructed based on the abnormal physiological state threshold and the trigger alarm.
[0014] In a second aspect, the present invention provides an intelligent neurosurgery nursing monitoring system, the system comprising: A model data generation module is used to collect multimodal nursing data of neurosurgery patients, identify patient physiological classification data in the multimodal nursing data, and construct a training generator for the multimodal nursing data based on the patient physiological classification data; A data discrimination module, used for analyzing the risk level corresponding to the patient's physiological classification data, extracting the physiological indicators corresponding to the risk level, analyzing the time series change trend of the physiological indicators, setting a normal change threshold of the patient's physiological classification data according to the time series change trend, setting a discrimination baseline of the training generator based on the normal change threshold, and using the discrimination baseline to analyze the quality of the generated data of the training generator; A model adversarial module, used for constructing a discriminant model of the quality of the generated data, setting the recognition degree of the discriminant model for the quality of the generated data, and performing cyclic adversarial training on the training generator and the discriminant model based on the recognition degree to obtain a cyclic adversarial training result; A data analysis module, for calculating the balance coefficient between the training generator and the discriminant model according to the result of the cyclic adversarial training, and calculating the degree of recognition of the discriminant model on the output data of the training generator, and constructing a generative adversarial network of the training generator and the discriminant model in combination with the balance coefficient and the degree of recognition, and monitoring the physical state value of the neurosurgery patient during the nursing process based on the generative adversarial network, and constructing a feedback mechanism for the physical state value; The nursing monitoring module is used to combine the generative adversarial network and the feedback mechanism to perform nursing monitoring on the neurosurgery patient and obtain nursing monitoring results.
[0015] Compared with the prior art, the technical principle and beneficial effects of this solution are: The embodiment of the present invention can provide data support for the subsequent construction of a patient data training generator by collecting multimodal nursing data of neurosurgery patients, help monitor the physiological data of neurosurgery patients, and by identifying the patient's physiological classification data in the multimodal nursing data, build a training model for the patient's physiological data according to different types of physiological data indicators of the patient. The abnormal physiological data of the patient in the nursing process is monitored by the training model, and the multimodal nursing data of the patient is simulated. The change trend of the patient's health data is analyzed, and the risk factors are identified in advance to improve the nursing efficiency; further, the embodiment of the present invention can quickly identify the patient's health risk according to the risk coefficient of the physiological indicator by extracting the physiological indicator corresponding to the risk level, help nursing staff monitor the severity and recovery of the patient's condition, and can promptly discover and prevent risks in the early stage, and then use the discriminant baseline to analyze the quality of generated data, improve the simulation accuracy of generated data, and enhance the training generator's ability to resist subsequent discriminant models; the embodiment of the present invention constructs the discriminant model of the generated data quality, which can accurately distinguish between real data and generated data to verify the simulation effect of the training generator and optimize the generated data, Improve the accuracy of generated data, and through cyclic adversarial training, help the training generator learn multiple change patterns of real data to optimize the performance of the training generator. At the same time, through adversarial training, a balance can be achieved between the training generator and the discriminant model, so as to avoid one party being too strong and causing unstable training, and improve the stability of the entire model training; the embodiment of the present invention monitors the physical state value of the neurosurgery patient in the nursing process based on the generative adversarial network, which can be used to supplement the actual collected patient nursing data, enhance the diversity and richness of the data set, and predict the abnormal risk of the patient in the nursing process, improve the nursing quality of the patient, and timely discover abnormal situations by constructing a feedback mechanism for the physical state value, and improve the processing efficiency of medical staff; further, the embodiment of the present invention monitors the nursing of the neurosurgery patient by combining the generative adversarial network and the feedback mechanism, and obtains the nursing monitoring results, which can timely discover the abnormal situation of the neurosurgery patient in the nursing process, and continuously track the changes in the patient's physical state, predict the patient's physical state, effectively reduce the risk of infection in the patient, and improve the quality of nursing monitoring. Therefore, the intelligent neurosurgery nursing monitoring method and system proposed in the embodiment of the present invention can improve the nursing quality and nursing efficiency of neurosurgery patients. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0018] Figure 1 A flowchart of an intelligent neurosurgery nursing monitoring method provided by one embodiment of the present invention; Figure 2 A module diagram of an intelligent neurosurgery nursing monitoring system provided by one embodiment of the present invention. DETAILED DESCRIPTION
[0019] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0020] The embodiment of the present invention provides an intelligent neurosurgical nursing monitoring method, and the execution subject of the intelligent neurosurgical nursing monitoring method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present invention. In other words, the intelligent neurosurgical nursing monitoring method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.
[0021] See also Figure 1 FIG. 1 is a flow chart of an intelligent neurosurgery nursing monitoring method provided by an embodiment of the present invention. Figure 1 The intelligent neurosurgical nursing monitoring methods described in the paper include: S1. Collect multimodal nursing data of neurosurgery patients, identify patient physiological classification data in the multimodal nursing data, and construct a training generator for the multimodal nursing data based on the patient physiological classification data.
[0022] The embodiments of the present invention can provide data support for the subsequent construction of a patient data training generator by collecting multimodal nursing data of neurosurgical patients, thereby helping to monitor the physiological data of neurosurgical patients. The neurosurgical patients refer to patients with brain, spinal cord and peripheral nervous system diseases or injuries who need health status monitoring and management. The multimodal nursing data refers to body monitoring data from different sensors and information sources during the patient care process.
[0023] Optionally, the multimodal nursing data collection of the neurosurgery patient can be obtained through the hospital's electronic health record system.
[0024] Furthermore, the embodiments of the present invention can identify the patient physiological classification data in the multimodal nursing data and construct a training model for the patient physiological data based on different types of physiological data indicators of the patient, monitor abnormal physiological data in the patient care process through the training model, and improve the quality of patient care. The patient physiological classification data refers to different data types of patient physiological indicators and health information generated by different monitoring instruments.
[0025] As an embodiment of the present invention, the identifying of the patient physiological classification data in the multimodal nursing data includes: formatting the multimodal nursing data to obtain formatted data; analyzing the data source of the multimodal nursing data and identifying the creation time corresponding to the data in the multimodal nursing data; setting a data tag of the formatted data according to the data source and the data creation time; and classifying the multimodal nursing data based on the data tag to obtain the patient physiological classification data.
[0026] Among them, the formatted data refers to the patient's multimodal nursing data in a unified format, the data source refers to the source of various data in the multimodal nursing data, the creation time refers to the time when the different monitoring data provided by each monitoring instrument is created, and the data label refers to the data detail label of each monitoring data corresponding to the multimodal nursing data, such as data format, source and creation time.
[0027] Optionally, the data source of the multimodal nursing data and the corresponding creation time analysis of the data can be obtained through basic equipment that monitors patient health data, such as an electroencephalogram monitor, and the data label setting of the formatted data can be implemented using the time when the data was created, the data format and the data source.
[0028] The embodiment of the present invention can simulate the patient's multimodal nursing data, analyze the changing trend of the patient's health data, identify risk factors in advance, and improve nursing efficiency by constructing a training generator for the patient's physiological classification data. The training generator refers to a synthetic data model that can generate a highly similar data to real patient data.
[0029] As an embodiment of the present invention, the construction of the training generator for the multimodal nursing data based on the patient physiological classification data includes: identifying abnormal missing values of the multimodal nursing data based on the patient physiological classification data; analyzing the statistical properties of the multimodal nursing data, and analyzing the data features of the multimodal nursing data based on the statistical properties and the abnormal missing values; identifying the variation risk of the data features, and setting a risk suppression mechanism for the variation risk; analyzing the simulation degree of the generated data corresponding to the multimodal nursing data; and constructing the training generator for the multimodal nursing data based on the risk suppression mechanism and the simulation degree.
[0030] Among them, the abnormal missing values refer to the errors and missing data values that appear in the process of integrating different physiological data of patients into multimodal nursing data, the statistical attributes refer to the quantitative indicator attributes used to describe and summarize the characteristics of multimodal nursing data, the data characteristics refer to the attributes or characteristics that can be observed in the data set, such as time series attributes, the variation risk refers to the risk caused by the instability or variability of the characteristics of the patient's physiological data, the risk suppression mechanism refers to the behavioral intervention mechanism taken to deal with data anomalies, the generated data refers to the multimodal nursing data of patients simulated by the training generator, and the simulation degree refers to the indicator degree used to evaluate the authenticity of the generated data.
[0031] Optionally, the identification of abnormal missing values of the multimodal nursing data based on the patient physiological classification data can be obtained through the Pandas library, the statistical attribute analysis of the multimodal nursing data can be implemented using scipy.stats, the identification of variation risk of the data features can be obtained through abnormal algorithm detection, such as the IQR algorithm, the risk suppression mechanism setting of the variation risk can be implemented using random forests, and the simulation degree analysis of the generated data corresponding to the multimodal nursing data can be obtained through the FID score.
[0032] S2. Analyze the risk level corresponding to the patient's physiological classification data, extract the physiological indicators corresponding to the risk level, analyze the time series change trend of the physiological indicators, set the normal change threshold of the patient's physiological classification data according to the time series change trend, set the discrimination baseline of the training generator based on the normal change threshold, and use the discrimination baseline to analyze the quality of the generated data of the training generator.
[0033] The embodiment of the present invention can form a graded judgment of the patient's physiological data by analyzing the risk level corresponding to the patient's physiological classification data, identify data with higher risk levels and give priority to processing, and reduce the safety hazards of patients in the nursing process. The risk level refers to the risk classification obtained by evaluating the health risks that the patient may face based on the patient's physiological indicators, clinical manifestations, nursing records and other data.
[0034] As an embodiment of the present invention, the analysis of the risk level corresponding to the patient's physiological classification data includes: analyzing the physiological characteristics corresponding to the patient's physiological classification data; calculating the feature correlation between the patient's physiological classification data based on the physiological characteristics; constructing a feature grading judgment matrix of the patient's physiological classification data based on the feature correlation; calculating the indicator weights corresponding to the patient's physiological classification data; and analyzing the risk level corresponding to the patient's physiological classification data in combination with the indicator weights and the feature grading judgment matrix.
[0035] Among them, the physiological characteristics refer to the patient's physical reactions and symptom characteristics after nervous system diseases or injuries, the characteristic correlation refers to the correlation between different physiological indicators and symptoms of the patient, the characteristic grading judgment matrix refers to a mathematical matrix for evaluating the patient's physiological state and the severity of the disease, and the indicator weight refers to the importance of different physiological data when evaluating the patient's physiological state and the severity of the disease.
[0036] Optionally, the feature grading judgment matrix construction of the patient physiological classification data based on the feature correlation can be obtained through Python.
[0037] Furthermore, in an optional embodiment of the present invention, the characteristic correlation between the patient physiological classification data is calculated using the following formula: ; Wherein, k represents the feature correlation between the patient physiological classification data, n represents the number of samples of the patient physiological classification data, and represents the observed values of the X and Y features of the physiological classification data of the i-th patient, and Represents the average value of the X and Y features of the physiological classification data of the i-th patient.
[0038] Furthermore, the embodiments of the present invention can quickly identify the patient's health risks and improve the efficiency of nursing intervention by extracting the physiological indicators corresponding to the risk levels according to the risk coefficients of the physiological indicators, wherein the physiological indicators refer to body parameters that measure the patient's physiological state and health status.
[0039] Optionally, the physiological index corresponding to the risk level can be extracted and obtained through the patient's nursing intervention instrument, such as a monitor.
[0040] Furthermore, the embodiments of the present invention can help nursing staff monitor the severity and recovery of patients' conditions by analyzing the time series change trends of the physiological indicators, and can promptly detect and prevent risks in the early stages. The time series change trends refer to the changing trends of patients' physiological indicators over time within a certain period of time, such as the changes in body temperature over time, which can evolve into a trend of infection or inflammation in patients.
[0041] As an embodiment of the present invention, the analysis of the time series change trend of the physiological indicator includes: collecting sample data corresponding to the physiological indicator, and cleaning and correcting the sample data to obtain target sample data; analyzing the significant differences of the target sample data at different times; based on the significant differences, evaluating the change amplitude of the target sample data over time; under the change amplitude, analyzing the result attributes of the physiological indicator, and constructing a time series evolution graph of the result attributes; based on the time series evolution graph, analyzing the time series change trend of the physiological indicator.
[0042] Among them, the sample data refers to part of the sample data of the physiological indicator monitoring data of the neurosurgery patient extracted, the target sample data refers to the sample data after data cleaning, abnormal data removal and format unification of the part of the sample data, the significant difference refers to the obvious change of the physiological indicator within a certain period of time, the amplitude of change refers to the rate of change of the physiological indicator within a certain period of time, the result attribute refers to the different result types represented by the change of the physiological indicator, such as the severity of the disease or the improvement of the disease, and the time series evolution graph refers to a trend graph compiled from the evolution data of the physiological indicator over time.
[0043] Optionally, the significant difference analysis of the target sample data at different times can be obtained through statistical tests, such as t-test tests. Based on the significant differences, the evaluation of the change amplitude of the target sample data over time can be achieved using the proportional formula of the rate of change, and the result attribute analysis of the physiological indicators can be obtained through a time series prediction model.
[0044] The embodiment of the present invention can monitor the patient's physical condition and predict the patient's possible physiological changes in the future by setting the normal change threshold of the patient's physiological classification data according to the change trend of the time series, thereby reducing the possibility of infection in the patient. The normal change threshold refers to the range value that represents the normal fluctuation of the patient's physiological indicators within a certain time range.
[0045] Optionally, according to the time series change trend, the normal change threshold of the physiological indicator can be obtained through the standard deviation and variance of the physiological indicator data.
[0046] Furthermore, the embodiment of the present invention sets the discrimination baseline of the training generator based on the normal change threshold, thereby ensuring that the generated data can better simulate the normal physiological changes of patients in real data, and at the same time can effectively identify and handle abnormal situations, thereby improving the simulation accuracy of the generated data. The discrimination baseline refers to the reference line for judging the authenticity of the data generated in the training generator.
[0047] As an embodiment of the present invention, the discrimination baseline of the training generator is set based on the normal change threshold, including: analyzing the recognition effect of the training generator on the abnormal physiological data of the patient based on the normal change threshold; evaluating the interpretability of the recognition effect; defining the tolerance level for the simulation data of the training generator according to the interpretability and the recognition effect; and setting the discrimination baseline of the training generator based on the tolerance level.
[0048] Among them, the recognition effect refers to the accuracy of the training generator in identifying abnormal data that deviates from the normal physiological data range, the explainability refers to the extent to which the recognition effect can be understood and explained by doctors or relevant personnel, such as why certain data are defined as abnormal, the simulation data refers to the data generated by the training generator that simulates the physiological indicators of real patients, and the tolerance level refers to the maximum degree of abnormality or deviation that the system can accept in the generated simulation data.
[0049] Optionally, based on the normal change threshold, the analysis of the recognition effect of the training generator on the patient's abnormal physiological data can be obtained through an anomaly detection algorithm, such as an Isolation Forest algorithm, and the interpretability evaluation of the recognition effect can be achieved using expert evaluation, and based on the interpretability and the recognition effect, the tolerance definition of the simulation data of the training generator can be obtained through a similarity measure between the simulation data and the real data, such as cosine similarity.
[0050] The embodiment of the present invention can improve the simulation accuracy of the generated data and enhance the training generator's ability to resist subsequent discriminant models by utilizing the discriminant baseline to analyze the quality of the generated data of the training generator. The quality of the generated data refers to the level of quality and reliability of the generated data compared with the real data.
[0051] Optionally, the quality analysis of the generated data of the training generator may be implemented using the discrimination baseline, such as using the tolerance level of the discrimination baseline to evaluate the compliance level of the generated data quality.
[0052] S3. Construct a discriminant model for the quality of the generated data, set the recognition degree of the discriminant model for the quality of the generated data, and based on the recognition degree, perform cyclic adversarial training on the training generator and the discriminant model to obtain a cyclic adversarial training result.
[0053] The embodiment of the present invention can accurately distinguish between real data and generated data by constructing a discriminant model of the generated data quality to verify the simulation effect of the training generator, optimize the generated data, and improve the accuracy of the generated data. The discriminant model refers to a model used to distinguish between real data and generated data.
[0054] As an embodiment of the present invention, the construction of the discriminant model of the quality of the generated data includes: collecting real data corresponding to the generated data, and extracting the physiological indicator time series characteristics of the real data and the generated data; analyzing and integrating the physiological indicator time series characteristics to obtain integrated characteristics; calculating the characteristic correlation coefficient between the physiological indicator time series characteristics based on the integrated characteristics; setting data partitions of the real data and the generated data based on the characteristic correlation coefficient; performing data discrimination processing on the data partitions to obtain discrimination output results; and constructing the discriminant model of the quality of the generated data by combining the integrated characteristics and the discrimination output results.
[0055] Among them, the real data refers to the original physiological data collected directly from neurosurgery patients, the generated data refers to the simulated physiological data created by training the generator, the physiological indicator time series feature refers to the characteristics of the physiological indicators extracted from the real data and the generated data changing over time, the integrated feature refers to the feature set obtained after analyzing and integrating the physiological indicator time series features of the real data and the generated data, the feature correlation coefficient refers to the correlation intensity coefficient between different physiological indicator time series features, the data partitioning refers to dividing the real data and the generated data into different data intervals, and the discrimination output result refers to the output result of evaluating the similarity between the generated data and the real data.
[0056] Optionally, the extraction of physiological indicator time series features of the real data and the generated data can be obtained through a convolutional neural network, the analysis and integration of the physiological indicator time series features can be achieved using a fully connected layer, and the data discrimination processing of the data partition can be achieved using a sigmoid function.
[0057] As another embodiment of the present invention, setting the data partitions of the real data and the generated data based on the feature correlation coefficient includes: identifying the feature levels of the real data and the generated data based on the feature correlation coefficient; setting the partition method of the real data and the generated data according to the feature level; analyzing the partition ratio of the real data and the generated data; and setting the data partitions of the real data and the generated data according to the partition method and the partition ratio.
[0058] Among them, the feature level refers to the level division of patient feature data of the real data and the generated data according to the feature correlation coefficient, the partitioning method refers to the method of how to divide the real data and the generated data, and the partition ratio refers to the ratio of the real data and the generated data to the system data set during the partitioning process.
[0059] Optionally, based on the feature correlation coefficient, the feature level identification of the real data and the generated data can be obtained by the absolute value of the feature correlation coefficient, the partitioning method setting of the real data and the generated data can be achieved by a random partitioning method, and the division ratio analysis of the real data and the generated data can be obtained by cross-validation.
[0060] Furthermore, the embodiment of the present invention can help understand the similarity between the generated data and the real data by setting the recognition degree of the discriminant model for the quality of the generated data, thereby ensuring the accuracy of the generated data. The recognition degree refers to a quantitative indicator of the recognition degree of the discriminant model for the quality of the generated data.
[0061] Optionally, the recognition degree setting of the discriminant model for the quality of the generated data can be obtained by training the output probability value of the discriminant model.
[0062] The embodiment of the present invention performs cyclic adversarial training on the training generator and the discriminant model based on the recognition degree to obtain a cyclic adversarial training result, which can help the training generator learn various change patterns of real data to optimize the performance of the training generator. At the same time, through adversarial training, a balance can be achieved between the training generator and the discriminant model to avoid one party being too powerful and causing unstable training, thereby improving the stability of the entire model training. The cyclic adversarial training result refers to the result obtained by alternately training the generator and the discriminator during the cyclic adversarial training process.
[0063] Optionally, the cyclic adversarial training of the training generator and the discriminant model based on the recognition degree can be implemented using PyTorch.
[0064] S4. According to the result of the cyclic adversarial training, the balance coefficient between the training generator and the discriminant model is calculated, and the recognition rate of the discriminant model to the output data of the training generator is calculated. In combination with the balance coefficient and the recognition degree, a generative adversarial network of the training generator and the discriminant model is constructed. Based on the generative adversarial network, the physical state values of the neurosurgical patients during the nursing process are monitored, and a feedback mechanism for the physical state values is constructed.
[0065] The embodiment of the present invention can dynamically adjust the training parameters by calculating the balance coefficient between the training generator and the discriminant model according to the result of the cyclic adversarial training, so as to avoid overfitting of the training generator to the existing data pattern, thereby improving the quality of the generated data. The balance coefficient refers to the average difference between the data generated by the training generator and the predicted value of the real data of the discriminant model.
[0066] Optionally, according to the cyclic adversarial training result, the balance coefficient between the training generator and the discriminant model is calculated using the following formula: ; Wherein, D represents the discriminator model, G represents the training generator, represents the balance coefficient between the training generator and the discriminant model, E represents the expected value of the balance between the training generator and the discriminant model, Represents the true data distribution in the discriminant model , z represents the noise vector of the spatial distribution Pz of the real data in the discriminant model, G(z) represents the sample generated by the training generator based on the real data noise vector z, D(x) represents the probability evaluation that the discriminant model sample x belongs to the real data distribution, Represents the natural logarithm of the equilibrium expected value of the training generator and the discriminant model.
[0067] Furthermore, the embodiments of the present invention can help identify the degree of similarity between the training generator and the discriminant model by calculating the recognition rate of the discriminant model for the output data of the training generator, thereby ensuring that the training generator develops in the right direction. The output data refers to the data samples generated by the training generator during the training process, and the recognition rate refers to the recognition rate of the discriminant model for the output data.
[0068] Optionally, the recognition rate calculation of the discriminant model on the output data of the training generator can be obtained through the Keras library.
[0069] The embodiment of the present invention combines the balance coefficient and the recognition degree to construct a generative adversarial network of the training generator and the discriminant model, thereby ensuring that the generated data can better reflect the physiological changes in the real world, helping medical staff to better understand the patient's condition, thereby effectively monitoring the patient's care process. The generative adversarial network refers to a network that includes the training generator and the discriminant model for generating synthetic data similar to real data.
[0070] Optionally, the construction of the generative adversarial network combining the balance coefficient and the recognition degree, the training generator and the discriminant model can be implemented using TensorFlow.
[0071] Furthermore, the embodiment of the present invention monitors the physical state values of the neurosurgery patients during the nursing process based on the generative adversarial network, which can be used to supplement the actually collected patient care data, enhance the diversity and richness of the data set, and at the same time predict the abnormal risks of the patients during the nursing process and improve the quality of care for the patients. The physical state value refers to the numerical values of various physiological indicators and health conditions monitored during the nursing process of neurosurgery patients.
[0072] Optionally, the monitoring of the physical state value of the neurosurgery patient during the nursing process based on the generative adversarial network can be obtained by transmitting the data of the monitoring device to the stream processing library in real time.
[0073] The embodiment of the present invention can detect abnormal situations in a timely manner by constructing a feedback mechanism for the physical status value, and immediately provide drug intervention or take other measures, thereby improving the processing efficiency of medical staff. The feedback mechanism refers to a mechanism for outputting feedback on the monitoring data of the patient's physical status during the care process of neurosurgery patients.
[0074] As an embodiment of the present invention, the feedback mechanism for constructing the body state value includes: collecting physiological data in a generative adversarial network corresponding to the body state value, and analyzing the reliability of the physiological data; setting an abnormal physiological state threshold for neurosurgery patients according to the reliability and the actual situation of the body state value during the nursing process; setting a trigger alarm for the abnormal physiological state threshold; and constructing a feedback mechanism for the body state value according to the abnormal physiological state threshold and the trigger alarm.
[0075] Among them, the physiological data refers to the simulated physiological data generated by the training generator in the generative adversarial network and discriminated by the discriminant model, the reliability refers to the consistency between the physiological data and the actual physiological data, the actual situation refers to the actual physiological condition of neurosurgical patients during the nursing process, the abnormal physiological state threshold refers to the limit value set according to the reliability of the physiological data and the actual situation of the patient, and the trigger alarm means that when the physiological indicator exceeds the set abnormal threshold, the system will automatically issue a prompt alarm.
[0076] Optionally, the reliability of the physiological data can be obtained by the recognition rate of the discriminant model to the output data of the training generator, and based on the reliability and the actual situation of the neurosurgical patient during the nursing process, the setting of the abnormal physiological state threshold of the neurosurgical patient can be implemented using the statistical function in SciPy, and the trigger alarm setting of the abnormal physiological state threshold can be obtained through a sound alarm.
[0077] S5. Combining the generative adversarial network and the feedback mechanism, perform nursing monitoring on the neurosurgery patient to obtain a nursing monitoring result.
[0078] The embodiment of the present invention combines the generative adversarial network and the feedback mechanism to perform nursing monitoring on the neurosurgical patients and obtain nursing monitoring results. It can timely discover abnormal conditions of neurosurgical patients during the nursing process, continuously track changes in the patient's physical state, predict the patient's physical state, effectively reduce the risk of infection in the patient, and improve the quality of nursing monitoring. The nursing monitoring results refer to the results obtained based on the monitored physiological data and generated data in the process of performing nursing monitoring on neurosurgical patients using the generative adversarial network and the feedback mechanism.
[0079] It can be seen that the embodiment of the present invention can provide data support for the subsequent construction of a patient data training generator by collecting multimodal nursing data of neurosurgery patients, help monitor the physiological data of neurosurgery patients, and by identifying the patient's physiological classification data in the multimodal nursing data, a training model for the patient's physiological data can be constructed according to different types of physiological data indicators of the patient. The abnormal physiological data of the patient in the nursing process can be monitored through the training model, and the multimodal nursing data of the patient can be simulated. The change trend of the patient's health data is analyzed, and the risk factors are identified in advance to improve the nursing efficiency; further, the embodiment of the present invention can quickly identify the patient's health risks according to the risk coefficient of the physiological indicators by extracting the physiological indicators corresponding to the risk level, help nursing staff monitor the severity and recovery of the patient's condition, and can promptly discover and prevent risks in the early stage, and then use the discriminant baseline to analyze the quality of generated data, improve the simulation accuracy of generated data, and enhance the training generator's ability to resist subsequent discriminant models; the discriminant model of the generated data quality constructed by the embodiment of the present invention can accurately distinguish between real data and generated data to verify the simulation effect of the training generator and optimize the generated data. According to the invention, the accuracy of the generated data is improved, and through the cyclic adversarial training, the training generator is helped to learn the various change patterns of the real data to optimize the performance of the training generator. At the same time, through adversarial training, the balance between the training generator and the discriminant model can be achieved, so as to avoid one party being too strong and causing unstable training, and improve the stability of the entire model training; the embodiment of the invention monitors the physical state value of the neurosurgery patient in the nursing process based on the generative adversarial network, which can be used to supplement the actual collected patient nursing data, enhance the diversity and richness of the data set, and predict the abnormal risk of the patient in the nursing process, improve the nursing quality of the patient, and timely discover the abnormal situation by constructing the feedback mechanism of the physical state value, and improve the processing efficiency of medical staff; further, the embodiment of the invention monitors the nursing of the neurosurgery patient by combining the generative adversarial network and the feedback mechanism, and obtains the nursing monitoring result, which can timely discover the abnormal situation of the neurosurgery patient in the nursing process, and continuously track the changes in the patient's physical state, predict the patient's physical state, effectively reduce the risk of infection in the patient, and improve the quality of nursing monitoring. Therefore, the intelligent neurosurgery nursing monitoring method and system proposed in the embodiment of the invention can improve the nursing quality and nursing efficiency of neurosurgery patients.
[0080] like Figure 2 Shown is a functional module diagram of the intelligent neurosurgery nursing monitoring system of the present invention.
[0081] The intelligent neurosurgery nursing monitoring system 200 of the present invention can be installed in an electronic device. According to the functions to be implemented, the intelligent neurosurgery nursing monitoring system can include a model data generation module 201, a data discrimination module 202, a model confrontation module 203, a data analysis module 204, and a nursing monitoring module 205. The module of the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.
[0082] In the embodiment of the present invention, the functions of each module / unit are as follows: The model data generation module 201 is used to collect multimodal nursing data of neurosurgery patients, identify the patient's physiological classification data in the multimodal nursing data, and construct a training generator for the multimodal nursing data according to the patient's physiological classification data; The data discrimination module 202 is used to analyze the risk level corresponding to the patient's physiological classification data, extract the physiological indicators corresponding to the risk level, analyze the time series change trend of the physiological indicators, set the normal change threshold of the patient's physiological classification data according to the time series change trend, set the discrimination baseline of the training generator based on the normal change threshold, and use the discrimination baseline to analyze the quality of the generated data of the training generator; The model confrontation module 203 is used to construct a discriminant model of the quality of the generated data, set the recognition degree of the discriminant model for the quality of the generated data, and based on the recognition degree, perform cyclic adversarial training on the training generator and the discriminant model to obtain a cyclic adversarial training result; The data analysis module 204 is used to calculate the balance coefficient between the training generator and the discriminant model according to the cyclic adversarial training result, and calculate the recognition degree of the discriminant model to the output data of the training generator, and construct a generative adversarial network of the training generator and the discriminant model based on the balance coefficient and the recognition degree, and monitor the physical state value of the neurosurgery patient during the nursing process based on the generative adversarial network, and construct a feedback mechanism for the physical state value; The nursing monitoring module 205 is used to combine the generative adversarial network and the feedback mechanism to perform nursing monitoring on the neurosurgery patient and obtain a nursing monitoring result.
[0083] In detail, each module in the intelligent neurosurgery nursing monitoring system 200 of the embodiment of the present invention is used in the same manner as described above. Figure 1 The intelligent neurosurgery nursing monitoring method described in the invention has the same technical means and can produce the same technical effects, so it will not be repeated here.
[0084] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.
[0085] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.
Claims
1. An intelligent neurosurgery nursing monitoring method, characterized in that: The method comprises: Collecting multimodal nursing data of neurosurgery patients, identifying physiological classification data of patients in the multimodal nursing data, and constructing a training generator of the multimodal nursing data according to the physiological classification data of patients; Analyze the risk level corresponding to the patient's physiological classification data, extract the physiological indicators corresponding to the risk level, analyze the time series change trend of the physiological indicators, set the normal change threshold of the patient's physiological classification data according to the time series change trend, set the discrimination baseline of the training generator based on the normal change threshold, and use the discrimination baseline to analyze the quality of the generated data of the training generator; Constructing a discriminant model of the quality of the generated data, setting the recognition degree of the discriminant model for the quality of the generated data, and based on the recognition degree, performing cyclic adversarial training on the training generator and the discriminant model to obtain a cyclic adversarial training result; According to the result of the cyclic adversarial training, the balance coefficient between the training generator and the discriminant model is calculated, and the recognition degree of the discriminant model to the output data of the training generator is calculated; combining the balance coefficient and the recognition degree, a generative adversarial network of the training generator and the discriminant model is constructed; based on the generative adversarial network, the physical state value of the neurosurgery patient during the nursing process is monitored, and a feedback mechanism for the physical state value is constructed; The generative adversarial network and the feedback mechanism are combined to perform nursing monitoring on the neurosurgery patient to obtain a nursing monitoring result.
2. The method according to claim 1, characterized in that The step of identifying the patient's physiological classification data in the multimodal nursing data comprises: Formatting the multimodal nursing data to obtain formatted data; Analyzing the data source of the multimodal nursing data and identifying the creation time corresponding to the data in the multimodal nursing data; Setting a data tag of the formatted data according to the data source and the data creation time; Based on the data labels, the multimodal nursing data is classified to obtain patient physiological classification data.
3. The method according to claim 1, characterized in that The step of constructing a training generator for the multimodal nursing data according to the patient physiological classification data includes: identifying abnormal missing values of the multimodal nursing data based on the patient physiological classification data; Analyzing statistical attributes of the multimodal nursing data, and analyzing data features of the multimodal nursing data based on the statistical attributes and the abnormal missing values; Identify the variation risk of the data features and set a risk suppression mechanism for the variation risk; analyzing a degree of simulation of generated data corresponding to the multimodal nursing data; A training generator for the multimodal nursing data is constructed based on the risk suppression mechanism and the simulation degree.
4. The method according to claim 1, characterized in that: The analyzing the risk level corresponding to the patient's physiological classification data includes: Analyzing physiological characteristics corresponding to the patient's physiological classification data; Calculating feature correlations between the patient's physiological classification data based on the physiological features; Based on the feature correlation, constructing a feature grading judgment matrix of the patient's physiological classification data; Calculate the indicator weight corresponding to the patient's physiological classification data: The risk level corresponding to the patient's physiological classification data is analyzed by combining the indicator weights and the feature classification judgment matrix.
5. The method according to claim 1, characterized in that The analyzing the time series variation trend of the physiological index comprises: Collecting sample data corresponding to the physiological index, and performing cleaning and correction processing on the sample data to obtain target sample data; Analyze the significant differences of the target sample data at different times; According to the significant difference, evaluating the change amplitude of the target sample data over time; Under the variation range, analyzing the result attribute of the physiological indicator and constructing a time series evolution diagram of the result attribute; Based on the time series evolution diagram, the time series change trend of the physiological index is analyzed.
6. The method according to claim 1, characterized in that The step of setting the discrimination baseline of the training generator based on the normal change threshold comprises: Based on the normal change threshold, analyzing the recognition effect of the training generator on the abnormal physiological data of the patient; Evaluating the interpretability of the recognition effect; Defining a tolerance level for the simulation data of the training generator according to the explainability and the recognition effect; Based on the tolerance level, a discrimination baseline of the training generator is set.
7. The method according to claim 1, characterized in that The step of constructing a discriminant model for generating data quality comprises: Collecting real data corresponding to the generated data, and extracting physiological index time series characteristics of the real data and the generated data; Analyzing and integrating the temporal characteristics of the physiological indicators to obtain integrated characteristics; Calculating the characteristic correlation coefficient between the physiological index time series characteristics according to the integrated characteristics; Based on the feature correlation coefficient, setting data partitions of the actual data and the generated data; Performing data discrimination processing on the data partition to obtain a discrimination output result; Combining the integrated features with the discriminant output results, a discriminant model for the quality of the generated data is constructed.
8. The method according to claim 7, characterized in that The step of setting data partitions of the actual data and the generated data based on the feature correlation coefficient comprises: Based on the feature correlation coefficient, identifying the feature levels of the actual data and the generated data; According to the feature level, setting a partitioning method of the actual data and the generated data; Analyzing the division ratio between the actual data and the generated data; According to the partitioning method and the partitioning ratio, data partitions of the actual data and the generated data are set.
9. The method according to claim 1, characterized in that: The feedback mechanism for constructing the body state value includes: Collecting physiological data in a generative adversarial network corresponding to the body state value, and analyzing the reliability of the physiological data; According to the reliability and the actual situation of the physical state value during the nursing process, setting a threshold value for abnormal physiological state of the neurosurgery patient; Setting a trigger alarm for an abnormal threshold of the physiological state; A feedback mechanism for the body state value is constructed based on the abnormal physiological state threshold and the trigger alarm.
10. An intelligent neurosurgery nursing monitoring system, the system implementing the method according to claim 1, characterized in that: The system comprises: A model data generation module is used to collect multimodal nursing data of neurosurgery patients, identify patient physiological classification data in the multimodal nursing data, and construct a training generator for the multimodal nursing data based on the patient physiological classification data; A data discrimination module, used for analyzing the risk level corresponding to the patient's physiological classification data, extracting the physiological indicators corresponding to the risk level, analyzing the time series change trend of the physiological indicators, setting a normal change threshold of the patient's physiological classification data according to the time series change trend, setting a discrimination baseline of the training generator based on the normal change threshold, and using the discrimination baseline to analyze the quality of the generated data of the training generator; A model adversarial module, used for constructing a discriminant model of the quality of the generated data, setting the recognition degree of the discriminant model for the quality of the generated data, and performing cyclic adversarial training on the training generator and the discriminant model based on the recognition degree to obtain a cyclic adversarial training result; A data analysis module, for calculating the balance coefficient between the training generator and the discriminant model according to the result of the cyclic adversarial training, and calculating the degree of recognition of the discriminant model on the output data of the training generator, and constructing a generative adversarial network of the training generator and the discriminant model in combination with the balance coefficient and the degree of recognition, and monitoring the physical state value of the neurosurgery patient during the nursing process based on the generative adversarial network, and constructing a feedback mechanism for the physical state value; The nursing monitoring module is used to combine the generative adversarial network and the feedback mechanism to perform nursing monitoring on the neurosurgery patient and obtain nursing monitoring results.