Intensive care data acquisition and early warning system

Abstract

The invention discloses an intensive care data acquisition and early warning system, and relates to the technical field of medical early warning, abnormal points in data can be effectively identified through a data processing module, an abnormal data set is generated, and the method is beneficial to screening out possible instantaneous abnormal conditions. And on the basis of the anomaly sets, corresponding monitoring time periods are identified, the relevance between the behavior state and the physiological parameters of the patient is further analyzed, influence factors are calculated, and the relevance-based analysis method effectively solves the problem of false positive anomaly which is difficult to distinguish in a traditional intensive care system. On the basis, the hierarchical response module constructs a screening model by using a convolutional neural network technology, fits and outputs a screening index, distinguishes abnormal types according to the numerical value of the index and sends out a corresponding early warning instruction, and meanwhile, ensures that dynamic adjustment and switching of the early warning instruction are more intelligent and accurate through a preset time window mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of medical early warning, and in particular to an intensive care data collection and early warning system. Background Art

[0002] In the field of medical monitoring and data acquisition systems, with the development of information technology, medical monitoring systems have shifted from traditional manual monitoring to automated intelligent monitoring, especially in the field of intensive care (ICU), where automated data acquisition and real-time monitoring systems are essential. In the ICU environment, patients' vital signs and behavioral data are crucial monitoring indicators. In this scenario, one of the key challenges is how to effectively distinguish between true abnormalities and false positive abnormalities in patients' physiological data. False positive abnormalities are usually caused by equipment noise, natural physiological fluctuations of patients, or short-term movements, and are not true pathological changes.

[0003] At present, sensor technology and intelligent algorithms have been widely used in data collection and analysis in intensive care units. However, existing monitoring systems often generate a large number of false positive warnings due to patients' mood swings, slight movements or sensor failures. For false positive anomaly identification of patient data in intensive care units, existing systems usually rely on simple threshold trigger mechanisms, which are prone to triggering unnecessary alarms due to short-term abnormal fluctuations or equipment failures, resulting in a high false alarm rate. This not only increases the workload of medical staff, but may also delay the response to real critical situations. Therefore, the interference of false positive warnings needs to be considered. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides an intensive care data collection and early warning system, which solves the problems in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a critical care data acquisition and early warning system, including an indoor data acquisition module, a data processing module, a false positive monitoring analysis module and a graded response module;

[0006] The indoor data acquisition module is used to deploy several groups of monitoring equipment in the intensive care unit in advance, and based on the several groups of monitoring equipment, monitor the patient's vital signs data information and the patient's related behavior data information in real time;

[0007] The data processing module is used to pre-process the vital sign data information and the patient's related behavior data information, and identify abnormal data points in combination with the box plot method, and generate an abnormal set through statistics;

[0008] The false positive monitoring analysis module is used to identify the corresponding monitoring period according to the abnormal set, and analyze whether there is a correlation between the patient's behavioral state and physiological parameters based on the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz;

[0009] The hierarchical response module is used to fit the output screening index Sczs using the screening model constructed using convolutional neural network technology if there is a correlation, and distinguish false positive abnormalities based on the numerical value of the screening index Sczs to issue corresponding warning instructions, while presetting a time window to determine the switching of the warning instructions.

[0010] Preferably, the indoor data acquisition module includes a first acquisition unit and a second acquisition unit;

[0011] The first acquisition unit is used to monitor the vital signs data information of patients in the intensive care unit in real time, wherein the vital signs data information includes skin conductivity value GSR, heart rate variability value HRV, oxygen intake VO2 and heart rate value Xz;

[0012] The second acquisition unit is used to monitor relevant behavior data information of patients in the intensive care unit in real time, wherein the relevant behavior data information includes tidal volume Cqz and action frequency Dzpz.

[0013] Preferably, the data processing module includes a preprocessing unit and an anomaly extraction unit;

[0014] The preprocessing unit is used to remove noise in the data collected by several groups of monitoring devices through a filtering algorithm, identify missing values ​​in the data collected by the monitoring devices, and standardize parameters of different units to generate a monitoring data set.

[0015] Preferably, the abnormality extraction unit is used to sort the corresponding parameters of each monitoring time point in the monitoring data set by numerical value to form a time series of the corresponding parameters, and in combination with the box plot method, the time series of the corresponding parameters is divided into four parts according to percentage to obtain the first quartile Q1 and the third quartile Q3 respectively, and based on the first quartile Q1 and the third quartile Q3, the upper limit a and the lower limit b in the time series of the corresponding parameters are obtained, and the specific acquisition method is as follows:

[0016]

[0017] In the formula, IQR represents the interquartile range, Cz max Expressed as the maximum value in the time series of the corresponding parameter, Cz min Expressed as the minimum value in the time series of the corresponding parameter, Cz avg Expressed as the average value in the time series of the corresponding parameter;

[0018] The physiological parameters at each monitoring time point in the time series of the corresponding parameters are compared with their upper limit a and lower limit b respectively to identify abnormal data points. The specific identification method is as follows:

[0019] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is greater than its upper limit value a, the physiological parameter at the detection time point is regarded as an abnormal data point;

[0020] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is less than its lower limit value b, the physiological parameter at the detection time point is regarded as an abnormal data point;

[0021] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter falls within its lower limit value b and upper limit value a, the physiological parameter at the detection time point is regarded as a normal data point;

[0022] Using statistical algorithms, several groups of abnormal data points are aggregated to generate anomaly sets.

[0023] Preferably, the false positive monitoring analysis module includes a behavior analysis unit, a physiological fluctuation analysis unit and a correlation unit;

[0024] The behavior analysis unit is used to identify the monitoring period of the physiological parameters in the time series of each parameter according to the abnormal set, and calculate and obtain the situation estimation coefficient Qgxs in the corresponding monitoring period after linear normalization processing in combination with the relevant behavior data information of the patient. The situation estimation coefficient Qgxs is specifically obtained by the following method:

[0025]

[0026] Wherein, Dzpz represents the action frequency, Cqz represents the tidal volume, α and β are the weight values ​​of the action frequency Dzpz and the tidal volume Cqz respectively, and P represents the first correction constant, wherein the specific values ​​of α and β are set by the user according to the situation.

[0027] Preferably, the physiological fluctuation analysis unit is used to calculate and obtain the external interference coefficient Wgxs in the corresponding monitoring period according to the patient's vital sign data information and the monitoring period of the physiological parameters in the time series of each identified parameter after linear normalization processing. The external interference coefficient Wgxs is specifically obtained in the following way:

[0028]

[0029] In the formula, F1, F2 and F3 are weight values, HRV represents the heart rate variability value, GSR represents the skin conductivity value, VO2 represents the oxygen uptake, C represents the second correction constant, and the specific values ​​of F1, F2 and F3 are set by the user according to the situation.

[0030] Preferably, the association unit analyzes whether there is a correlation between the patient's behavioral state and physiological parameters based on the situation estimation coefficient Qgxs obtained in the behavior analysis unit and the physiological fluctuation analysis unit and the external interference coefficient Wgxs in the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz, which is specifically obtained in the following manner:

[0031]

[0032] In the formula, Qgxs i It is expressed as the situation estimation coefficient at the i-th monitoring time point, Wgxs i Expressed as the external interference coefficient at the i-th monitoring time point, Qgxs σ It is expressed as the standard deviation of the situation estimation coefficients involved in the detection cycle in the anomaly set, Wgxs σ It is expressed as the standard deviation of the internal and external interference coefficients of the detection cycle involved in the abnormal set, where cov(Qgxs i , Wgxs i ) is expressed as Qgxs i and Wgxs i The covariance of .

[0033] Preferably, the hierarchical response module includes a comparison unit, a screening and analysis unit, and an early warning trigger unit;

[0034] The comparison unit is used to compare and analyze the influencing factor Yyz with a preset influencing threshold value to determine whether there is a correlation between the patient's behavioral state and the physiological parameter. The specific determination content is as follows:

[0035] If the influencing factor Yyz falls within the influencing threshold, it will be determined that there is no correlation between the patient's behavioral state and the physiological parameters; if the influencing factor Yyz does not fall within the influencing threshold, it will be determined that there is a correlation between the patient's behavioral state and the physiological parameters.

[0036] Preferably, the screening analysis unit is used to input the monitoring data set into the screening model when there is a correlation between the patient's behavioral state and the physiological parameters, so as to fit and output the screening index Sczs, wherein the screening index Sczs is obtained by the following formula:

[0037]

[0038] In the formula, f XzRepresents the heart rate monitoring result value, Wgxs avg It is expressed as the average external interference coefficient;

[0039] Among them, the heart rate monitoring result value f Xz There are constraints, which are as follows:

[0040]

[0041] Among them, Xz represents the heart rate value, Xz min and Xz max Represented as the two edge values ​​of the normal heart rate range.

[0042] Preferably, the warning trigger unit is used to pre-set two groups of level thresholds, namely, a first threshold q1 and a second threshold q2, and the first threshold q1 is greater than the second threshold q2, and the screening index Sczs is compared with the three groups of level thresholds to issue a preliminary warning instruction:

[0043] If the screening index Sczs ≤ the second threshold q2, it means that the current patient is in a false positive abnormal state, and no warning instruction will be issued at this time;

[0044] If the second threshold q2 < screening index Sczs ≤ the first threshold q1, it means that the current patient is in a false positive abnormal state, but an orange warning instruction will be issued at this time;

[0045] If the screening index Sczs> the first threshold q1, it means that the current patient is not in a false positive abnormal state, and a red warning instruction will be continuously issued to the outside;

[0046] When an orange warning instruction is received, the switching mode of the warning instruction is determined in combination with the set time window T. The specific contents are as follows:

[0047] Rising switching condition: If Sczs>q currentlevel And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the superior command;

[0048] Downward switching condition: If Sczs<q currentlevel And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the lower level command;

[0049] in, It is expressed as the rate of change of the screening index, Sczs is expressed as the screening index, q currentlevel Represents the threshold of the current warning level, T durationIt is expressed as the duration that the value of the screening index Sczs continues to decrease, and T is expressed as the set time window.

[0050] The present invention provides a critical care data collection and early warning system, which has the following beneficial effects:

[0051] (1) The system ensures the multi-faceted and accurate monitoring data by collecting indoor data. This multi-source data collection provides a rich foundation for subsequent data analysis, making the real-time tracking of patient status more detailed and comprehensive. The data processing module can effectively identify abnormal points in the data and generate abnormal data sets. This method helps to screen out possible transient abnormalities. Based on these abnormal sets, the corresponding monitoring period is identified, and the correlation between the patient's behavioral state and physiological parameters is further analyzed to calculate the impact factor Yyz. This correlation-based analysis method effectively solves the problem of false positive abnormalities that are difficult to distinguish in traditional intensive care systems, reduces false alarms caused by slight movements or emotional fluctuations of patients, and improves the monitoring accuracy of the system. On this basis, the hierarchical response module uses convolutional neural network technology to construct a screening model, fits the output screening index Sczs, distinguishes the abnormal type according to the numerical value of the index, and issues corresponding warning instructions. At the same time, through the preset time window mechanism, it ensures that the dynamic adjustment and switching of the warning instructions are more intelligent and accurate, further avoiding unnecessary frequent switching of warning levels and improving the response efficiency of the warning system. The overall system can automatically determine whether an abnormal situation is a false positive and implement an automatic switching warning instruction function to further reduce unnecessary medical interventions, thereby improving the work efficiency of medical staff and optimizing the patient's treatment process.

[0052] (2) The behavior analysis unit analyzes and calculates the situational estimation coefficient Qgxs based on the abnormal set. This coefficient further reflects the patient's behavioral state during a specific monitoring period. Based on the patient's vital signs data and the changes in physiological parameters during the monitoring period, the external interference coefficient Wgxs is calculated. This coefficient identifies the patient's physiological changes during the corresponding period, which helps the system distinguish whether the physiological fluctuations are caused by non-abnormal conditions caused by external interference or by pathological changes in the patient's own body, thereby reducing the false alarm rate.

[0053] (3) When the screening analysis unit identifies the correlation between the patient's behavioral state and physiological parameters, it inputs the monitoring data set into the screening model and fits the output screening index Sczs. This process can accurately evaluate the current patient's physiological state. By combining the patient's heart rate monitoring results with the external interference coefficient, the system can dynamically adjust the screening index to accurately reflect the patient's physiological abnormality level, especially under the condition that the heart rate monitoring results are limited to the normal heart rate range, to ensure that the heart rate abnormality only triggers the warning in the real situation, effectively reducing the false positives caused by short-term fluctuations or slight movements, thereby improving the accuracy of screening. Secondly, the warning trigger unit realizes a multi-level warning instruction issuance mechanism through the preset two groups of level thresholds. This hierarchical warning system can flexibly determine whether the warning instruction needs to be triggered according to the numerical value of the screening index Sczs. In addition, the design controls the dynamic switching of the warning instruction by setting the time window T to ensure the rationality and continuity of the warning instruction. The system determines whether the warning level needs to be upgraded or downgraded according to the change rate and duration of the screening index, thereby avoiding the interference caused by frequent switching of the warning level. In short, the system further realizes the accurate distinction between false positive abnormalities and true abnormalities of patients through the synergistic effect of the screening and analysis unit and the early warning trigger unit, and at the same time provides a flexible early warning level switching mechanism to reduce the occurrence of false alarms and missed alarms. This not only improves the work efficiency of medical staff and reduces the burden on medical staff, but also improves the response speed to abnormal conditions of patients, providing patients in intensive care with a safer and more intelligent monitoring system.

[0054] (4) The warning switching in this system will not be triggered immediately based on the instantaneous value of the screening index Sczs. Instead, in order to further avoid false alarms caused by short-term fluctuations (for example, short-term mood swings or slight movements of the patient), a time window is set to observe the changing trend of the screening index Sczs during this period, thereby achieving accuracy in the warning switching. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 The present invention is a block diagram of a critical care data acquisition and early warning system. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] Example 1

[0058] See also Figure 1 , the present invention provides a critical care data acquisition and early warning system, including an indoor data acquisition module, a data processing module, a false positive monitoring analysis module and a graded response module;

[0059] The indoor data acquisition module is used to deploy several groups of monitoring equipment in the intensive care unit in advance, and based on the several groups of monitoring equipment, monitor the patient's vital signs data information and the patient's related behavior data information in real time;

[0060] The data processing module is used to pre-process the vital sign data information and the patient's related behavior data information, and identify abnormal data points in combination with the box plot method, and generate an abnormal set through statistics;

[0061] The false positive monitoring analysis module is used to identify the corresponding monitoring period according to the abnormal set, and analyze whether there is a correlation between the patient's behavioral state and physiological parameters based on the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz;

[0062] The hierarchical response module is used to fit the output screening index Sczs using the screening model constructed using convolutional neural network technology if there is a correlation, and distinguish false positive abnormalities based on the numerical value of the screening index Sczs to issue corresponding warning instructions, while presetting a time window to determine the switching of the warning instructions.

[0063] During the operation of this system, the system further realizes real-time monitoring of patients' vital signs (such as heart rate, respiration, blood pressure, etc.) and behavioral information (such as body movement, posture, etc.) through the indoor data acquisition module. Through the pre-deployment of several groups of monitoring equipment in the intensive care unit, the wide coverage and timely acquisition of data are ensured, further avoiding the limitations of single monitoring data, thereby improving the control of the patient's all-round status. The data processing module in the system generates anomaly sets by identifying abnormal data points, and eliminates data noise and redundancy by preprocessing vital signs and behavioral data, further reducing the false alarm rate and improving the accuracy of anomaly detection. The false positive monitoring analysis module plays a key role in the system. By analyzing the anomaly set, the correlation between the patient's behavior and physiological data during the monitoring period is identified, and the influence factor Yyz is further calculated to distinguish between true anomalies and false positive anomalies caused by non-pathological factors such as emotional fluctuations and slight movements. This module further reduces false alarms caused by sensor errors or non-pathological factors, so that medical staff can focus on pathological abnormalities that really need intervention. The hierarchical response module introduces convolutional neural network (CNN) technology to build a screening model, and further screens abnormal data through the fitting value of the screening index Sczs. This mechanism can intelligently distinguish false positives from real abnormalities based on the numerical value of Sczs, and judge the switching of warning instructions based on the preset time window to ensure the timeliness and accuracy of the warning. At the same time, the hierarchical warning mechanism can flexibly adjust the response level according to the severity of the abnormality and optimize the workflow of medical staff. In short, the system effectively improves the accuracy of abnormal data identification through multi-level data collection and processing methods, further reduces the false alarm rate, and improves the accuracy of warning and response efficiency in the intensive care environment, thereby better ensuring the safety of patients.

[0064] Example 2

[0065] Please refer to Figure 1 ,Specifically: the indoor data acquisition module includes a first acquisition unit and a second acquisition unit;

[0066] The first acquisition unit is used to monitor the vital signs data information of patients in the intensive care unit in real time, wherein the vital signs data information includes skin conductivity value GSR, heart rate variability value HRV, oxygen intake VO2 and heart rate value Xz;

[0067] The second acquisition unit is used to monitor relevant behavior data information of patients in the intensive care unit in real time, wherein the relevant behavior data information includes tidal volume Cqz and action frequency Dzpz.

[0068] The data processing module includes a preprocessing unit and an abnormality extraction unit;

[0069] The preprocessing unit is used to remove the noise in the data collected by several groups of monitoring devices through a filtering algorithm, ensure the smoothness and accuracy of the data, reduce the random errors introduced by electromagnetic interference, environmental noise or equipment problems during the collection process of the monitoring equipment, and maintain the stability of the signal; identify the missing values ​​in the data collected by the monitoring equipment, and standardize the parameters of different units for subsequent analysis, and generate a monitoring data set to fill or remove the missing values ​​in the collected data to ensure data continuity; ensure that all parameters have the same dimension through standardization, which is convenient for subsequent multi-parameter fusion analysis. In order to ensure that different physiological parameters (such as heart rate, respiratory rate, skin conductivity, etc.) are comparable in subsequent analysis, it is necessary to standardize the data and convert it into a dimensionless value. The commonly used standardization method is Z-score standardization. Z-score standardization adjusts the data to a dimensionless value with a mean of 0 and a standard deviation of 1, which allows parameters of different units (such as BPM of heart rate and micro-Siemens of skin conductance) to be compared and fused under the same dimension.

[0070] In this embodiment, the system realizes multi-dimensional collection of the patient's physiological and behavioral data through the first acquisition unit and the second acquisition unit; the first acquisition unit monitors the patient's skin conductivity value, heart rate variability value, oxygen intake and other vital signs in real time, which can accurately reflect the patient's physiological state and the activity of the autonomic nervous system. At the same time, the second acquisition unit monitors the patient's behavioral data, such as tidal volume and movement frequency, and provides the system with real-time feedback on the patient's activity status. Through multi-dimensional physiological and behavioral data collection, the system can perceive the patient's physical condition more comprehensively, laying a solid foundation for subsequent abnormal analysis. Secondly, the preprocessing unit and the abnormality extraction unit in the data processing module effectively remove noise through filtering algorithms to ensure the smoothness and accuracy of the data, which reduces the impact of external factors such as equipment noise and environmental interference on the monitoring data, and ensures the stability and authenticity of the signal. At the same time, the preprocessing unit maintains the continuity of the data by identifying and filling in the missing values ​​in the data, further avoiding data faults in the analysis process. In addition, the system uses the Z-score standardization method to convert physiological parameters of different units into dimensionless values, further ensuring the comparability and consistency of each parameter in the subsequent multi-parameter fusion analysis. This standardization process enables different physiological indicators to be compared and analyzed on a unified scale, improving the scientificity and accuracy of data processing. In short, the system effectively improves the quality and availability of critical care data through multi-dimensional data collection, precise preprocessing and standardization methods, reduces monitoring errors, and provides a high-precision data foundation for subsequent abnormal identification and early warning response, ultimately improving the efficiency and accuracy of critical care.

[0071] Example 3

[0072] Please refer to Figure 1 Specifically: the abnormal extraction unit is used to sort the corresponding parameters of each monitoring time point in the monitoring data set by numerical value to form a time series of the corresponding parameters, and combined with the box plot method, the time series of the corresponding parameters is divided into four parts according to percentage to obtain the first quartile Q1 and the third quartile Q3 respectively, and based on the first quartile Q1 and the third quartile Q3, the upper limit a and the lower limit b in the time series of the corresponding parameters are obtained. The specific acquisition method is as follows:

[0073]

[0074] In the formula, IQR represents the interquartile range, Cz max Expressed as the maximum value in the time series of the corresponding parameter, Cz min Expressed as the minimum value in the time series of the corresponding parameter, Cz avg It is expressed as the average value in the time series of the corresponding parameter, where the interquartile range IQR = Q1-Q3;

[0075] The physiological parameters at each monitoring time point in the time series of the corresponding parameters are compared with their upper limit a and lower limit b respectively to identify abnormal data points. The specific identification method is as follows:

[0076] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is greater than its upper limit a, the physiological parameter at the detection time point is regarded as an abnormal data point;

[0077] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is less than its lower limit value b, the physiological parameter at the detection time point is regarded as an abnormal data point;

[0078] If the physiological parameter at each monitoring time point in the time series of the corresponding parameter falls within its lower limit value b and upper limit value a, the physiological parameter at the detection time point is regarded as a normal data point;

[0079] Using statistical algorithms, several groups of abnormal data points are aggregated to generate anomaly sets.

[0080] In this embodiment, the abnormal extraction unit in the system adopts an analysis method based on time series, combined with the box plot method, which can accurately identify and filter out abnormal data points in the patient's vital signs, thereby improving the analysis accuracy of intensive care data and the efficiency of abnormal detection. First, the system forms a time series by numerically sorting the physiological parameters at each monitoring time point in the monitoring data set. This time series analysis helps to capture the changing trend of physiological parameters on the time axis and ensure the integrity and timing of data processing. Combined with the box plot method, the system divides the data into four parts according to percentages through quartile calculations, and obtains the first quartile and the third quartile, and then calculates the upper limit a and the lower limit b. This boundary setting based on statistics can effectively identify outliers in the data, that is, those abnormal values ​​that exceed the upper or lower limit range. Through clear rules, when the monitoring data is greater than the upper limit a or less than the lower limit b, the system identifies these data points as abnormal data points; when the data falls between the upper limit and the lower limit, it is judged as a normal data point. This method can accurately distinguish between normal and abnormal changes in vital signs, further avoiding the problems of false positives and omissions in traditional analysis methods. The system further uses statistical algorithms to summarize these abnormal data points and generate an abnormal set, thereby providing high-quality data support for subsequent false positive monitoring and early warning systems. In general, the system's ability to identify abnormal data is improved through the abnormal extraction process, which can reduce misjudgments caused by data fluctuations, equipment noise, etc.

[0081] Example 4

[0082] Please refer to Figure 1 ,Specifically: the false positive monitoring analysis module includes a behavior analysis unit, a physiological fluctuation analysis unit and a correlation unit;

[0083] The behavior analysis unit is used to identify the monitoring period of the physiological parameters in the time series of each parameter according to the abnormal set, and calculate and obtain the situation estimation coefficient Qgxs in the corresponding monitoring period after linear normalization processing in combination with the relevant behavior data information of the patient. The situation estimation coefficient Qgxs is specifically obtained by the following method:

[0084]

[0085] Wherein, Dzpz represents the action frequency, Cqz represents the tidal volume, α and β are the weight values ​​of the action frequency Dzpz and the tidal volume Cqz respectively, and P represents the first correction constant, wherein the specific values ​​of α and β are set by the user according to the situation.

[0086] The above-mentioned movement frequency Dzpz can be monitored and obtained by an accelerometer or a gyroscope. The accelerometer can detect the patient's movement frequency and amplitude, and is usually used to monitor body movement, turning over, etc. The gyroscope can be used to detect more detailed angle changes and movement directions. Wearable devices (such as smart bracelets and smart watches) are often integrated with such sensors to monitor the patient's movement behavior.

[0087] The tidal volume Cqz can be monitored and obtained through a spirometer or a respiratory sensor. The spirometer calculates the tidal volume by measuring the gas flow of each breath. The respiratory sensor can monitor the patient's chest expansion through a chest belt and then infer the respiratory volume. The tidal volume is usually measured by a special lung function tester, or by monitoring daily breathing conditions through a respiratory belt and airflow sensor.

[0088] The physiological fluctuation analysis unit is used to calculate and obtain the external interference coefficient Wgxs in the corresponding monitoring period according to the patient's vital sign data information and the monitoring period of the physiological parameters in the time series of each identified parameter after linear normalization processing. The external interference coefficient Wgxs is specifically obtained in the following way:

[0089]

[0090] In the formula, F1, F2 and F3 are weight values, HRV represents the heart rate variability value, GSR represents the skin conductivity value, VO2 represents the oxygen uptake, C represents the second correction constant, and the specific values ​​of F1, F2 and F3 are set by the user according to the situation.

[0091] The above-mentioned heart rate variability value HRV can be monitored and obtained through an electrocardiogram and a photoplethysmogram sensor. The electrocardiogram is a device that measures the electrical activity of the heart and calculates the heart rate variability by the change in the RR interval. The PPG sensor calculates the heart rate variability by detecting the change in blood flow. The electrocardiograph is a standard device commonly used in hospitals.

[0092] The skin conductivity value GSR can be monitored and obtained through a skin conductivity meter (GSR sensor). The skin conductivity meter reflects the patient's sweat gland activity and emotional state by detecting changes in the conductivity of the skin. The conductivity of the skin increases when the patient is nervous or anxious. This sensor is often used in emotion monitoring and stress testing equipment, and can also be integrated into wearable devices.

[0093] Oxygen uptake VO2 can be monitored and obtained through an indirect calorimeter or a respiratory gas analyzer; among them, by measuring the oxygen and carbon dioxide concentrations in the respiratory gas, the oxygen uptake reflects the patient's metabolic level and cardiopulmonary function status.

[0094] In this embodiment, the behavior analysis unit identifies the time series of each parameter according to the abnormal set, conducts an in-depth analysis of each monitoring period, and calculates the situation estimation coefficient Qgxs in combination with the patient's behavior data. This coefficient is a linear normalization calculation based on the action frequency and tidal volume data. In order to accurately reflect the impact of the patient's behavior on physiological fluctuations, the system uses weight values ​​to adjust the relative importance of the action frequency and tidal volume, and further accurately adjusts the calculation results by introducing correction constants. This quantitative evaluation method of the situation can effectively distinguish normal physiological changes caused by behavioral activities, and further avoid misjudging them as pathological abnormalities. Secondly, the physiological fluctuation analysis unit calculates the external interference coefficient Wgxs based on the patient's vital signs data and the time series of the monitoring period. This coefficient is processed by linear normalization, and comprehensively analyzes parameters such as heart rate variability, skin conductivity and oxygen intake, and uses weight values ​​and correction constants to adjust the relative contribution of each physiological parameter in the external interference. By calculating the external interference coefficient Wgxs, the system can accurately capture non-pathological fluctuations in the patient's physiological data, such as physiological changes caused by emotional fluctuations or environmental changes, and further filter out unnecessary false positive abnormalities. In short, through the synergistic effect of the situational estimation coefficient Qgxs and the external interference coefficient Wgxs, the system can more accurately analyze the correlation between the patient's behavior and physiological fluctuations, and avoid false alarms caused by normal behavioral changes or external interference. This sophisticated false positive screening mechanism significantly improves the system's reliability and warning accuracy, reduces the interference of medical staff, and improves the work efficiency of the intensive care unit.

[0095] Example 5

[0096] Please refer to Figure 1 Specifically: the association unit analyzes whether there is a correlation between the patient's behavior state and physiological parameters based on the situation estimation coefficient Qgxs obtained in the behavior analysis unit and the physiological fluctuation analysis unit and the external interference coefficient Wgxs in the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz, which is obtained specifically in the following manner:

[0097]

[0098] In the formula, Qgxs i It is expressed as the situation estimation coefficient at the i-th monitoring time point, Wgxs i Expressed as the external interference coefficient at the i-th monitoring time point, Qgxs σ It is expressed as the standard deviation of the situation estimation coefficients involved in the detection cycle in the anomaly set, Wgxs σ It is expressed as the standard deviation of the internal and external interference coefficients of the detection cycle involved in the abnormal set, where cov(Qgxs i , Wgxs i) is expressed as Qgxs i and Wgxs i The covariance of .

[0099] The hierarchical response module includes a comparison unit, a screening and analysis unit, and an early warning trigger unit;

[0100] The comparison unit is used to compare and analyze the influencing factor Yyz with a preset influencing threshold value to determine whether there is a correlation between the patient's behavioral state and the physiological parameter. The specific determination content is as follows:

[0101] If the influencing factor Yyz falls within the influencing threshold, it will be determined that there is no correlation between the patient's behavioral state and the physiological parameters; if the influencing factor Yyz does not fall within the influencing threshold, it will be determined that there is a correlation between the patient's behavioral state and the physiological parameters.

[0102] In this embodiment, first, the association unit can comprehensively analyze the patient's behavior and physiological fluctuations by combining the situation estimation coefficient Qgxs and the external interference coefficient Wgxs obtained from the behavior analysis unit and the physiological fluctuation analysis unit. By calculating the influence factor Yyz, the system can further accurately evaluate whether there is a correlation between the patient's behavior and physiological parameters in different monitoring periods. The calculation of the influence factor Yyz is based on a statistical method, which integrates the situation estimation coefficients and external interference coefficients of multiple monitoring time points, and combines the standard deviation and covariance to comprehensively evaluate the data of each monitoring period. Through such precise calculations, the system can distinguish between false positive abnormalities caused by patient behavior (such as slight movement) and true pathological abnormalities. Secondly, the hierarchical response module compares the calculated influence factor Yyz with the pre-set influence threshold through its comparison unit to further determine the correlation between the patient's behavioral state and physiological parameters. If the influencing factor Yyz falls within the threshold range, the system will determine that the patient's behavior has no obvious correlation with the physiological parameters, thereby reducing false alarms; if the influencing factor Yyz exceeds the influencing threshold, the system will determine that there is a correlation, indicating that the patient's behavior may have caused physiological abnormalities. This comparison process enables the system to accurately identify abnormal conditions caused by behavior, further reducing the occurrence of false positive warnings. Through the above design, the system can more accurately identify false positive abnormalities, reduce false alarms caused by minor behaviors, environmental changes, or non-pathological physiological fluctuations, and the screening and analysis unit further screens abnormal data that really need attention, while the warning trigger unit ensures that the system only issues an alarm when there is indeed a pathological risk. This hierarchical response mechanism maximizes the accuracy and reliability of the early warning system, provides more accurate decision-making support for medical staff, and reduces unnecessary intervention.

[0103] Example 6

[0104] Please refer to Figure 1 Specifically: the screening analysis unit is used to input the monitoring data set into the screening model when there is a correlation between the patient's behavioral state and the physiological parameters, so as to fit and output the screening index Sczs, and the screening index Sczs is obtained by the following formula:

[0105]

[0106] In the formula, f Xz Represents the heart rate monitoring result value, Wgxs avg It is expressed as the average external interference coefficient;

[0107] Among them, the heart rate monitoring result value f Xz There are constraints, which are as follows:

[0108]

[0109] Among them, Xz represents the heart rate value, Xz min and Xz max Represented as the two edge values ​​of the normal heart rate range.

[0110] The above-mentioned heart rate value Xz can be monitored and obtained through an electrocardiogram or a photoplethysmogram sensor.

[0111] The warning trigger unit is used to pre-set two groups of level thresholds, namely, the first threshold q1 and the second threshold q2, and the first threshold q1 is greater than the second threshold q2, and the screening index Sczs is compared with the three groups of level thresholds to issue a preliminary warning instruction:

[0112] If the screening index Sczs ≤ the second threshold q2, it means that the current patient is in a false positive abnormal state, and no warning instruction will be issued at this time;

[0113] If the second threshold q2 < screening index Sczs ≤ the first threshold q1, it means that the current patient is in a false positive abnormal state, but an orange warning instruction will be issued at this time;

[0114] If the screening index Sczs> the first threshold q1, it means that the current patient is not in a false positive abnormal state, and a red warning instruction will be continuously issued to the outside;

[0115] The setting of the two groups of level thresholds mentioned above can be obtained by the following method: according to the method of obtaining the screening index Sczs, the screening index Sczs at each monitoring time point is obtained, and the average screening index Sczs is obtained respectively by combining the statistical algorithm. and the standard deviation of the screening index σSczs, based on the average screening index and the standard deviation of the screening index σSczs to set two groups of level thresholds, threshold No. Threshold No. 2 k is a constant, usually taking a value of 1-3, corresponding to different confidence levels, and the specific value is set by the user (according to actual conditions); according to the setting methods of these two groups of thresholds, similarly, the above-mentioned setting of the impact threshold is also applicable.

[0116] When an orange warning instruction is received, the switching mode of the warning instruction is determined in combination with the set time window T. The specific contents are as follows:

[0117] Rising switching condition: If Sczs>q currentlevel And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the superior command;

[0118] Downward switching condition: If Sczs<q currentlevel And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the lower level command;

[0119] in, It is expressed as the rate of change of the screening index, Sczs is expressed as the screening index, q currentlevel Represents the threshold of the current warning level, T duration It is expressed as the duration that the value of the screening index Sczs continues to decrease, and T is expressed as the set time window.

[0120] In this embodiment, the screening analysis unit calculates the screening index Sczs using the screening model according to the correlation between the patient's behavioral state and the physiological parameters. By comprehensively considering the heart rate monitoring results and the average external interference coefficient, the system can quantitatively evaluate the current patient's physiological state, and ensure that the calculated screening index Sczs has high comparability according to the constraint of the heart rate. This screening process provides accurate input data for the early warning system to ensure the rationality of the early warning instruction. Secondly, the early warning trigger unit further realizes the graded early warning of the patient's state by setting two groups of grade thresholds. According to the size of the screening index Sczs, the system can dynamically determine whether the patient is in a false positive state, thereby deciding whether to issue an early warning instruction. Specifically: when the screening index Sczs is lower than the second threshold q2, it indicates a false positive abnormality, and the system does not issue an early warning for the time being, further avoiding unnecessary alarms caused by slight fluctuations or sensor noise, and reducing the workload of medical staff. When the screening index Sczs is between the two thresholds, the system issues an orange early warning to prompt medical staff to pay attention, but does not trigger an emergency response, reflecting the system's sensitivity to slight abnormalities. When the screening index Sczs exceeds the first threshold q1, the system determines it as a non-false positive state and issues a red warning, indicating that there is a serious abnormality, which may be a true physiological abnormality, further ensuring timely response in emergency situations. By monitoring the rate of change, the system can automatically switch the warning level according to the persistence and change trend of the warning signal. The rising and falling switching conditions ensure that the system can quickly adjust the warning level when the patient's condition deteriorates or improves, and only avoid responding too early or too late, so as to ensure that medical staff can make timely adjustments according to the actual condition of the patient. This flexible warning mechanism reduces false alarms and missed alarms, and improves the response efficiency of the intensive care system. In short, the system not only improves the ability to identify false positive abnormalities through the dynamic calculation of the screening index Sczs and the linkage with the intelligent warning system, but also provides medical staff with more accurate and graded warning instructions, avoiding unnecessary intervention, improving the efficiency and accuracy of the warning system, and ensuring the safety of patients in intensive care.

[0121] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A critical care data collection and early warning system, characterized in that: It includes indoor data acquisition module, data processing module, false positive monitoring analysis module and graded response module; The indoor data acquisition module is used to deploy several groups of monitoring equipment in the intensive care unit in advance, and based on the several groups of monitoring equipment, monitor the patient's vital signs data information and the patient's related behavior data information in real time; The data processing module is used to pre-process the vital sign data information and the patient's related behavior data information, and identify abnormal data points in combination with the box plot method, and generate an abnormal set through statistics; The false positive monitoring analysis module is used to identify the corresponding monitoring period according to the abnormal set, and analyze whether there is a correlation between the patient's behavioral state and physiological parameters based on the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz; The hierarchical response module is used to fit the output screening index Sczs using the screening model constructed using convolutional neural network technology if there is a correlation, and distinguish false positive abnormalities based on the numerical value of the screening index Sczs to issue corresponding warning instructions, while presetting a time window to determine the switching of the warning instructions.

2. The intensive care data acquisition and early warning system according to claim 1, characterized in that: The indoor data acquisition module includes a first acquisition unit and a second acquisition unit; The first acquisition unit is used to monitor the vital signs data information of patients in the intensive care unit in real time, wherein the vital signs data information includes skin conductivity value GSR, heart rate variability value HRV, oxygen intake VO2 and heart rate value Xz; The second acquisition unit is used to monitor relevant behavior data information of patients in the intensive care unit in real time, wherein the relevant behavior data information includes tidal volume Cqz and action frequency Dzpz.

3. The intensive care data acquisition and early warning system according to claim 2 is characterized in that: The data processing module includes a preprocessing unit and an abnormality extraction unit; The preprocessing unit is used to remove noise in the data collected by several groups of monitoring devices through a filtering algorithm, identify missing values ​​in the data collected by the monitoring devices, and standardize parameters of different units to generate a monitoring data set.

4. The intensive care data acquisition and early warning system according to claim 3, characterized in that: The abnormal extraction unit is used to sort the corresponding parameters of each monitoring time point in the monitoring data set by numerical value to form a time series of the corresponding parameters, and in combination with the box plot method, divide the time series of the corresponding parameters into four parts according to percentage to obtain the first quartile Q1 and the third quartile Q3 respectively, and based on the first quartile Q1 and the third quartile Q3, obtain the upper limit a and the lower limit b in the time series of the corresponding parameters, and the specific acquisition method is as follows: In the formula, IQR represents the interquartile range, Cz max Expressed as the maximum value in the time series of the corresponding parameter, Cz min Expressed as the minimum value in the time series of the corresponding parameter, Cz avg Expressed as the average value in the time series of the corresponding parameter; The physiological parameters at each monitoring time point in the time series of the corresponding parameters are compared with their upper limit a and lower limit b respectively to identify abnormal data points. The specific identification method is as follows: If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is greater than its upper limit a, the physiological parameter at the detection time point is regarded as an abnormal data point; If the physiological parameter at each monitoring time point in the time series of the corresponding parameter is less than its lower limit value b, the physiological parameter at the detection time point is regarded as an abnormal data point; If the physiological parameter at each monitoring time point in the time series of the corresponding parameter falls within its lower limit value b and upper limit value a, the physiological parameter at the detection time point is regarded as a normal data point; Using statistical algorithms, several groups of abnormal data points are aggregated to generate anomaly sets.

5. The intensive care data acquisition and early warning system according to claim 4, characterized in that: The false positive monitoring analysis module includes a behavior analysis unit, a physiological fluctuation analysis unit and a correlation unit; The behavior analysis unit is used to identify the monitoring period of the physiological parameters in the time series of each parameter according to the abnormal set, and calculate and obtain the situation estimation coefficient Qgxs in the corresponding monitoring period after linear normalization processing in combination with the relevant behavior data information of the patient. The situation estimation coefficient Qgxs is specifically obtained by the following method: Wherein, Dzpz represents the action frequency, Cqz represents the tidal volume, α and β are the weight values ​​of the action frequency Dzpz and the tidal volume Cqz respectively, and P represents the first correction constant, wherein the specific values ​​of α and β are set by the user according to the situation.

6. The intensive care data acquisition and early warning system according to claim 5, characterized in that: The physiological fluctuation analysis unit is used to calculate and obtain the external interference coefficient Wgxs in the corresponding monitoring period according to the patient's vital sign data information and the monitoring period of the physiological parameters in the time series of each identified parameter after linear normalization processing. The external interference coefficient Wgxs is specifically obtained in the following way: In the formula, F1, F2 and F3 are weight values, HRV represents the heart rate variability value, GSR represents the skin conductivity value, VO2 represents the oxygen uptake, C represents the second correction constant, and the specific values ​​of F1, F2 and F3 are set by the user according to the situation.

7. The intensive care data acquisition and early warning system according to claim 6, characterized in that: The association unit analyzes whether there is a correlation between the patient's behavior state and physiological parameters based on the situation estimation coefficient Qgxs obtained in the behavior analysis unit and the physiological fluctuation analysis unit and the external interference coefficient Wgxs in the corresponding monitoring period, so as to analyze and calculate the influencing factor Yyz, which is specifically obtained in the following manner: In the formula, Qgxs i It is expressed as the situation estimation coefficient at the i-th monitoring time point, Wgxs i Expressed as the external interference coefficient at the i-th monitoring time point, Qgxs σ It is expressed as the standard deviation of the situation estimation coefficients involved in the detection cycle in the anomaly set, Wgxs σ It is expressed as the standard deviation of the internal and external interference coefficients of the detection cycle involved in the abnormal set, where cov(Qgxs i , Wgxs i ) is expressed as Qgxs i and Wgxs i The covariance of .

8. The intensive care data acquisition and early warning system according to claim 2, characterized in that: The hierarchical response module includes a comparison unit, a screening and analysis unit, and an early warning trigger unit; The comparison unit is used to compare the impact factor Y yz Compare and analyze with the pre-set impact threshold to determine whether there is a correlation between the patient's behavioral state and physiological parameters. The specific judgment content is as follows: If the impact factor Y yz If the influencing factor Y yz If the patient's behavior status does not fall within the influence threshold, it will be determined that there is a correlation between the patient's behavior status and the physiological parameters.

9. The intensive care data acquisition and early warning system according to claim 8, characterized in that: The screening analysis unit is used to input the monitoring data set into the screening model when there is a correlation between the patient's behavioral state and the physiological parameters, so as to fit and output the screening index Sczs, wherein the screening index Sczs is obtained by the following formula: In the formula, f Xz Represents the heart rate monitoring result value, Wgxs avg It is expressed as the average external interference coefficient; Among them, the heart rate monitoring result value f Xz There are constraints, which are as follows: Among them, Xz represents the heart rate value, Xz min and Xz max Represented as the two edge values ​​of the normal heart rate range.

10. The intensive care data acquisition and early warning system according to claim 9, characterized in that: The warning trigger unit is used to pre-set two groups of level thresholds, namely, the first threshold q1 and the second threshold q2, and the first threshold q1 is greater than the second threshold q2, and the screening index Sczs is compared with the three groups of level thresholds to issue a preliminary warning instruction: If the screening index Sczs ≤ the second threshold q2, it means that the current patient is in a false positive abnormal state, and no warning instruction will be issued at this time; If the second threshold q2 < screening index Sczs ≤ the first threshold q1, it means that the current patient is in a false positive abnormal state, but an orange warning instruction will be issued at this time; If the screening index Sczs> the first threshold q1, it means that the current patient is not in a false positive abnormal state, and a red warning instruction will be continuously issued to the outside; When an orange warning instruction is received, the switching mode of the warning instruction is determined in combination with the set time window T. The specific contents are as follows: Rising switching condition: If Sczs>q current level And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the superior command; Downward switching condition: If Sczs<q current level And T duration >T, at this time, the early warning command will be switched and the switching operation will be carried out to the lower level command; in, It is expressed as the rate of change of the screening index, Sczs is expressed as the screening index, q current level Represents the threshold of the current warning level, T duration It is expressed as the duration that the value of the screening index Sczs continues to decrease, and T is expressed as the set time window.

Images