Anesthesia ventilator monitoring system

Abstract

The present invention relates to the field of ventilator data monitoring technology, specifically to an anesthesia ventilator monitoring system, the system comprising: a real-time data anomaly monitoring module that collects patient monitoring data of anesthesia ventilators in real time, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate, compares the patient monitoring data with a preset safety threshold, and generates an abnormal data monitoring result. In the present invention, by real-time monitoring of the patient's airway pressure, tidal volume, and respiratory rate, combined with safety threshold comparison, it is possible to timely identify abnormal states and generate time series signals of abnormal data, calculate the fractal dimension of the time series signal and analyze the fluctuation characteristics, quantify the complexity and stability of the abnormal state, so as to increase the accuracy of abnormal evaluation, and use a Bayesian network model to associate current patient data with abnormal data, so as to predict future trend changes in patient data.

Description

Technical Field

[0001] The invention relates to the technical field of ventilator data monitoring, and in particular to an anesthesia ventilator monitoring system. Background Art

[0002] The anesthesia ventilator monitoring system refers to a real-time monitoring and data analysis system specifically used for anesthesia ventilators. By continuously monitoring the operating parameters of the anesthesia ventilator and the patient's respiratory status, it can provide real-time feedback on airway pressure, tidal volume, respiratory rate, airflow and other data. It can also detect abnormal situations in a timely manner through the alarm function, such as ventilator failure or patient breathing discomfort, which helps doctors understand the patient's respiratory status in real time.

[0003] Existing technologies make it difficult to conduct in-depth analysis of monitoring data and can only provide feedback on real-time monitoring data. For example, when airway pressure is abnormal, the system can only give real-time abnormal values ​​and cannot further analyze the persistence or changing trend of the abnormality. At the same time, it is difficult for the system to analyze whether the gradual abnormal changes in the data will continue to intensify or stabilize based on the monitoring data, and it is easy to lag behind in the changes of key data nodes. In addition, for abnormal responses, short-term abnormalities and prediction deviations in the same time period are difficult to quickly and effectively distinguish, which increases the burden on medical personnel to interpret and make decisions on data, which may affect the quality of patient monitoring and the efficiency of equipment monitoring. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an anesthesia breathing machine monitoring system.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: an anesthesia ventilator monitoring system comprises:

[0006] The real-time data abnormality monitoring module collects the patient monitoring data of the anesthesia ventilator in real time, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate, compares the patient monitoring data with the preset safety threshold, and generates abnormal data monitoring results;

[0007] The abnormal fluctuation assessment module converts all patient monitoring data into time series signals, extracts abnormal data marked in the abnormal data monitoring results from the time series signals, analyzes the complexity and abnormal fluctuation characteristics of the abnormal data time series signals corresponding to the abnormal data, obtains abnormal fluctuation complexity analysis results, evaluates the stability of the abnormal data time series signals corresponding to the abnormal data according to the abnormal fluctuation complexity analysis results, and generates abnormal fluctuation assessment results;

[0008] The respiratory state prediction module extracts all current patient monitoring data to construct a Bayesian network model, defines the current patient monitoring data as an input node of the Bayesian network model, extracts abnormal data of the same type as the input node in the abnormal fluctuation assessment result, and defines it as an observation node of the Bayesian network model, calculates the state change probability of the input node, analyzes the trend change of the input node in the future time period, and generates a respiratory state prediction result;

[0009] The respiratory status warning module generates a first warning signal based on the abnormal data monitoring results when abnormal data occurs. At the same time, the trend change of similar input nodes in the predicted future time period in the respiratory status prediction results is compared with the numerical values ​​of the actual monitoring observation nodes in the time period, and a second warning signal is generated based on the comparison results. Different types of warning signals are classified to generate an anesthesia ventilator data monitoring warning report.

[0010] As a further solution of the present invention, the steps of obtaining the abnormal data monitoring results are specifically as follows:

[0011] Based on real-time collection of patient monitoring data of anesthesia ventilators, by comparing the patient monitoring data at each time point with a preset safety threshold, the patient monitoring data corresponding to the time point that exceeds the safety threshold is marked as abnormal data, indicating that the patient's vital signs are abnormal, and the screened abnormal data is generated;

[0012] Based on real-time collection of patient monitoring data of anesthesia ventilators, the patient monitoring data corresponding to time points that do not exceed the safety threshold are marked as normal data, so that the normal data can be distinguished from the screened abnormal data, and abnormal data monitoring results are generated.

[0013] As a further solution of the present invention, the steps for obtaining the abnormal fluctuation complexity analysis result are specifically as follows:

[0014] Convert all patient monitoring data into time series signals and perform normalization processing, extract abnormal data marked in the abnormal data monitoring results from the time series signals, and generate abnormal data time series signals;

[0015] Based on the abnormal data time series signal, the formula is adopted:

[0016] ;

[0017] Calculate the fractal dimension of anomalous data time series signals , get the fractal dimension information;

[0018] in, is in scale In order to cover the minimum number of intervals required for the abnormal data time series signal, Indicates the scale of different abnormal data time series signal division areas, It is a weight parameter set based on different abnormal data. It is an adjustment coefficient set based on the fluctuation of abnormal data time series signal. is the natural logarithm, Used to linearize abnormal data time series signals, Representation scale The natural logarithm of the reciprocal;

[0019] Based on the fractal dimension information, the complexity and abnormal fluctuation characteristics of the target abnormal data time series signal are analyzed. The abnormal fluctuation characteristics include the fluctuation amplitude of the abnormal data time series signal and the distribution density of the abnormal values, and the abnormal fluctuation complexity analysis results are generated.

[0020] As a further solution of the present invention, the step of obtaining the abnormal fluctuation assessment result is specifically as follows:

[0021] Based on the abnormal fluctuation complexity analysis results, the stability of the abnormal data time series signal is evaluated, and the stability evaluation result is generated by judging the impact of the fluctuation characteristics of the abnormal data time series signal on the overall trend;

[0022] Based on the stability assessment results, all information of the abnormal data time series signal is integrated and displayed in a visual manner to generate abnormal fluctuation assessment results.

[0023] As a further solution of the present invention, the step of acquiring the state change probability of the input node is specifically as follows:

[0024] Extract all current patient monitoring data, combine with the abnormal data in the abnormal fluctuation assessment results, analyze the correlation and causal relationship between input nodes and observation nodes, and construct a Bayesian network model;

[0025] Based on the data in the Bayesian network model, the formula is adopted:

[0026] ;

[0027] Compute observation nodes After the corresponding abnormal data appears, the same type of input node The probability of state change , generate the state change probability of the input node;

[0028] in, Indicates that in the same type of input node Under the known state, according to the observation node The probability of abnormal data being recorded, Indicates that the observation node has not started to be observed In the case of the same type of input node The probability of a state change, Represents the observation node The corresponding probability of abnormal data occurrence is Based on the observation node and similar input nodes Set the dynamic correlation coefficient.

[0029] As a further solution of the present invention, the steps of obtaining the respiratory state prediction result are specifically:

[0030] The state change probability of each input node is sorted according to the time series, and divided into abnormal state and normal state, a change trend curve is constructed, the potential period of abnormal state in the future is clarified, and the input node change trend analysis result is generated;

[0031] Based on the input node change trend analysis result, by combining the input node with the time change trend, recording the start time period, duration and impact range of the immediate abnormal state and normal state, a respiratory state prediction result is generated.

[0032] As a further solution of the present invention, the steps of obtaining the first warning signal and the second warning signal are specifically:

[0033] According to the abnormal data monitoring result, abnormal data exceeding the safety threshold range is obtained, and a first warning signal is generated;

[0034] The respiratory state prediction result is compared with the state of the actual monitoring input node during the predicted period to determine whether the comparison result exceeds a preset warning range threshold. If so, a second warning signal is generated.

[0035] As a further solution of the present invention, the steps for obtaining the anesthesia ventilator data monitoring early warning report are specifically as follows:

[0036] Based on the information of the first warning signal and the second warning signal, matching and integrating the input nodes corresponding to the first warning signal and the second warning signal to generate a signal summary result;

[0037] Based on the signal aggregation result, different types of warning signals are classified, the first warning signal is classified as warning data for real-time monitoring, and the second warning signal is classified as warning data for prediction deviation, and an anesthesia ventilator data monitoring warning report is generated.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are:

[0039] In the present invention, by real-time monitoring of the patient's airway pressure, tidal volume and respiratory rate, combined with safety threshold comparison, it is possible to timely identify abnormal conditions and generate time series signals of abnormal data. The fractal dimension calculation and fluctuation characteristic analysis of the time series signal quantify the complexity and stability of the abnormal state to increase the accuracy of abnormal evaluation. The Bayesian network model is used to associate the current patient data with the abnormal data, and the trend changes of future patient data can be predicted, so that monitoring is extended from static analysis to dynamic prediction, and potential abnormalities are captured in advance. By integrating and classifying real-time monitoring signals and prediction signals, an early warning report is generated, which not only provides real-time abnormal data, but also includes trend change analysis, thereby enhancing the guiding role and decision-making support capabilities of data monitoring, making anesthesia ventilator data monitoring more intelligent and able to provide accurate respiratory status data support for medical personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a system flow chart of the present invention;

[0041] Figure 2 A flow chart of obtaining abnormal data monitoring results for the present invention;

[0042] Figure 3 A flow chart of obtaining abnormal fluctuation complexity analysis results for the present invention;

[0043] Figure 4 A flow chart for obtaining abnormal fluctuation evaluation results of the present invention;

[0044] Figure 5 A flow chart of obtaining the state change probability of an input node according to the present invention;

[0045] Figure 6 A flowchart of obtaining a respiratory state prediction result according to the present invention;

[0046] Figure 7 A flow chart of obtaining a first warning signal and a second warning signal for the present invention;

[0047] Figure 8 The present invention is a flow chart for obtaining anesthesia ventilator data monitoring and early warning report. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0050] See also Figure 1 , an anesthesia ventilator monitoring system comprising:

[0051] The real-time data abnormality monitoring module collects the patient monitoring data of the anesthesia ventilator in real time, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate, compares the patient monitoring data with the preset safety threshold, and generates abnormal data monitoring results;

[0052] The abnormal fluctuation assessment module converts all patient monitoring data into time series signals, extracts abnormal data marked in the abnormal data monitoring results from the time series signals, analyzes the complexity and abnormal fluctuation characteristics of the abnormal data time series signals corresponding to the abnormal data, obtains abnormal fluctuation complexity analysis results, evaluates the stability of the abnormal data time series signals corresponding to the abnormal data according to the abnormal fluctuation complexity analysis results, and generates abnormal fluctuation assessment results;

[0053] The respiratory state prediction module extracts all current patient monitoring data to build a Bayesian network model. By defining the current patient monitoring data as the input node of the Bayesian network model, the abnormal data of the same type as the input node in the abnormal fluctuation evaluation results are extracted and defined as the observation node of the Bayesian network model. By calculating the state change probability of the input node and analyzing the trend change of the input node in the future period, the respiratory state prediction result is generated.

[0054] The respiratory status warning module generates a first warning signal based on the abnormal data monitoring results when abnormal data appears. At the same time, it compares the trend changes of similar input nodes in the predicted future time period in the respiratory status prediction results with the values ​​of the actual monitoring observation nodes in the time period, and generates a second warning signal based on the comparison results. It classifies different types of warning signals and generates an anesthesia ventilator data monitoring warning report.

[0055] See also Figure 2 , the specific steps for obtaining abnormal data monitoring results are:

[0056] Based on real-time collection of patient monitoring data of anesthesia ventilators, by comparing the patient monitoring data at each time point with a preset safety threshold, the patient monitoring data corresponding to the time point that exceeds the safety threshold is marked as abnormal data, indicating that the patient's vital signs are abnormal, and the screened abnormal data is generated;

[0057] The patient monitoring data of the anesthesia ventilator is collected in real time, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate. For each monitoring data point, it is first necessary to compare it with the preset safety threshold, where the safety threshold is usually determined by clinical guidelines, the patient's physiological condition and historical data. For example, the safety threshold of airway pressure may be 20-30cmH2O, the safety threshold of tidal volume is 300-600mL, and the safety threshold of respiratory rate is 12-20 times / minute. The data collection process is usually monitored in real time using the built-in sensors or additional equipment of the anesthesia ventilator. The system collects patient data once every certain period of time (such as every second or every minute). The collected real-time data will be compared with these preset safety thresholds item by item. If a data point exceeds the safety range, the system will mark the monitoring data at that time point as abnormal data, indicating that the value of the monitoring point does not conform to the normal physiological range, which may mean that the patient's vital signs are abnormal.

[0058] Based on the real-time collection of patient monitoring data of anesthesia ventilators, the patient monitoring data corresponding to the time point that does not exceed the safety threshold is marked as normal data, so that the normal data is distinguished from the screened abnormal data, and the abnormal data monitoring results are generated;

[0059] Patient monitoring data that does not exceed the safety threshold is considered normal data. The system will classify and organize normal data and abnormal data, and archive the monitoring data of all time points in a unified manner through data integration for subsequent viewing and analysis. The integrated data includes two categories: one is normal data, and the other is data marked as abnormal. At this stage, the system generates abnormal data monitoring results, including the category of data at each time point (normal or abnormal), and indicates the specific time period and related parameters of the abnormal data. The data will be generated in the form of a report for medical personnel to conduct subsequent processing and judgment.

[0060] See also Figure 3 , the specific steps for obtaining the abnormal fluctuation complexity analysis results are:

[0061] Convert all patient monitoring data into time series signals and perform normalization processing, extract abnormal data marked in abnormal data monitoring results from the time series signals, and generate abnormal data time series signals;

[0062] All patient monitoring data are standardized and converted into time series signals. First, all patient monitoring data, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate monitoring data, will be standardized using MATLAB so that the data range is scaled to between 0 and 1. The standardized data is represented as a set of dimensionless time series signals. Then, the normalize function in MATLAB is used to convert the standardized data into time series signals. The conversion process will arrange the monitoring data in chronological order into a time series signal. Through the time series signal, further data complexity analysis and abnormal fluctuation feature identification are performed.

[0063] Based on the abnormal data time series signal, the formula is used:

[0064] ;

[0065] Calculate the fractal dimension of anomalous data time series signals , get the fractal dimension information;

[0066] in, is in scale In order to cover the minimum number of intervals required for the abnormal data time series signal, the point distribution area of ​​the abnormal data is divided and the minimum number of spheres required for coverage is calculated. Represents the scale of different abnormal data time series signal division areas. Assuming that the length of the time series is When choosing the scale When , the entire sequence is divided into intervals, in each interval, use a minimum sphere (or covering set) to cover all points, and count the number of spheres required for coverage The purpose of dividing the region is to quantify the complexity of the signal, that is, to observe how the structural characteristics of the signal change at different resolutions. It is a weight parameter set based on different abnormal data, which is used to adjust the relative importance of abnormal time series signals. It is calculated as the inverse of the standard deviation of the monitoring signal. ,in Indicates abnormal data The standard deviation of It is the adjustment coefficient set based on the fluctuation of abnormal data time series signal. The calculation formula is: , is the average value of the abnormal data, is the maximum value of the abnormal data, is the natural logarithm, Used to linearize abnormal data time series signals, express The natural logarithm of the reciprocal.

[0067] Taking the abnormal tidal volume time series signal as an example, assuming that abnormal tidal volume is detected in a certain monitoring, the data of the abnormal period is standardized and processed as follows: , calculate the parameters:

[0068] average value , the maximum value , adjustment coefficient , standard deviation , calculated as:

[0069] ; Weight parameter .

[0070] Select two scales : , the number of coverage intervals , adjusted quantity ; , the number of coverage intervals , adjusted quantity .

[0071] calculate and :

[0072] , , ;

[0073] , , .

[0074] Compute the slope of the fractal dimension using two points:

[0075] Calculating fractal dimension .

[0076] The results show that the fractal dimension of abnormal tidal volume data time series signal .

[0077] Based on the fractal dimension information, the complexity and abnormal fluctuation characteristics of the target abnormal data time series signal are analyzed. The abnormal fluctuation characteristics include the fluctuation amplitude of the abnormal data time series signal and the distribution density of the abnormal values, and the abnormal fluctuation complexity analysis results are generated;

[0078] Analyze the complexity and abnormal fluctuation characteristics of abnormal tidal volume data. Abnormal tidal volume data may be manifested as dense distribution of abnormal values ​​or fluctuation range beyond the normal range, as follows:

[0079] Fractal dimension The level of directly reflects the complexity and abnormal fluctuation characteristics of the time series signal. Generally speaking, : Indicates that the signal has higher complexity and irregular fluctuations, usually corresponding to more significant abnormalities or dynamic changes; : Indicates that the signal is relatively regular, with low fluctuation or relatively stable; :Usually corresponds to a signal close to a single regular pattern, such as periodic fluctuations or very small anomalies. It belongs to the lower complexity range, indicating that the fluctuation range of abnormal tidal volume data is limited and tends to be regular. For the characteristic analysis of abnormal tidal volume data: Since the standardized time series data is: ;

[0080] Mean and maximum value , by calculating The relative fluctuation range is 26.3%, indicating that the fluctuation range of the high value of 1.2 is relatively isolated; the data distribution is concentrated in the range of 0.8 to 1, with only one obvious outlier (1.2), calculated by the outlier ratio , assuming that the proportion of outliers in the normal abnormal signal distribution is lower than 30%, the complexity is low, and the standard deviation The ratio of the mean , assuming that the normal tidal volume time series usually has significant abnormal characteristics only when the fluctuation amplitude ratio exceeds 20%, it shows that the overall fluctuation amplitude is small, which belongs to the high range of the normal tidal volume fluctuation range; the high value 1.2 is isolated and distributed near other points close to the mean, indicating that the abnormal data may be a short-term occasional fluctuation phenomenon, such as the instantaneous increase in tidal volume during the patient's inhalation process, rather than a continuous pathological abnormality. In addition, assuming that the subsequent tidal volume abnormal data is: ;

[0081] The tidal volume rises rapidly in a short period of time and remains in a high range, reaching 1.8, which is much higher than the normal range. This fluctuation is not just an isolated abnormality of a single data point, but a continuous and drastic fluctuation over a period of time, indicating that the patient's tidal volume is at risk of overventilation, which may be caused by excessive anesthesia dose, improper ventilator settings, or the patient's physiological response. In addition, the airway pressure data at a certain moment is far away from the normal range, which may be due to instantaneous equipment failure, sensor error, or acute pathological changes.

[0082] Such isolated values ​​may not directly affect the overall patient status, but if they occur frequently or last for a long time, they may indicate a device failure or a potential problem with the respiratory system. If the airway pressure data continues to fluctuate violently over a period of time, it means that the patient's respiratory system may be facing a large pressure load, which may be caused by factors such as airway obstruction, respiratory infection, and the effects of anesthetic drugs. This situation usually requires immediate intervention to prevent the patient from respiratory failure or other serious complications. Abnormal isolated values ​​of respiratory rate usually indicate a transient physiological response or a transient error in the sensor.

[0083] If an isolated value occurs under normal physiological conditions, it usually does not have a significant impact on the patient. However, if such isolated values ​​occur repeatedly, it may indicate that the patient has experienced some kind of stress response, such as short-term shortness of breath or temporary apnea. Continuous and severe abnormalities in respiratory rate mean that the patient may have experienced an acute physiological event, such as acute respiratory distress, excessive anesthetic concentration, or respiratory abnormalities caused by other pathological conditions. This continuous fluctuation indicates that the patient's respiratory system is experiencing greater pressure, and emergency adjustment of anesthesia or respiratory support may be required to prevent further deterioration.

[0084] Combined with the above analysis, through the results of fractal dimension and the specific abnormal fluctuation characteristics of abnormal data, it is judged that the abnormal tidal volume is a non-sustained, isolated high-value fluctuation abnormality. The patient's respiratory status can be comprehensively evaluated by further combining the fractal dimension of other indicators such as airway pressure and respiratory rate.

[0085] See also Figure 4 , the specific steps for obtaining abnormal fluctuation assessment results are:

[0086] Based on the abnormal fluctuation complexity analysis results, the stability of the abnormal data time series signal is evaluated, and the stability evaluation results are generated by judging the impact of the fluctuation characteristics of the abnormal data time series signal on the overall trend;

[0087] The stability of the abnormal data time series signal is evaluated based on the abnormal fluctuation complexity analysis results. The stability of the signal mainly depends on the fluctuation amplitude characteristics and complexity. If the signal distribution is regular, the fluctuation amplitude is small, and the outliers are isolated, the signal can be considered to have high stability. On the contrary, if the signal fluctuation amplitude is large, the outliers are widely distributed or last for a long time, the signal stability is poor. For abnormal tidal volume data, short-term high-value fluctuations that do not significantly affect the overall trend can be judged as relatively stable anomalies. If the complexity increases and the outliers are densely distributed, further attention may be needed to the patient's respiratory status to assess potential risks.

[0088] Based on the stability assessment results, all information of the abnormal data time series signal is integrated and displayed in a visual way to generate abnormal fluctuation assessment results;

[0089] The information is integrated to generate abnormal fluctuation evaluation results. By analyzing the fractal dimension, abnormal fluctuation characteristics and distribution law of the abnormal data time series signal, the relevant information of abnormal tidal volume is summarized. The specific operations include: The low complexity of the representation, combined with the fluctuation amplitude of 26.3% and the isolated distribution characteristics of outliers, is used to create an abnormal fluctuation assessment report. The report content is divided into multiple parts, including the analysis conclusion of the fractal dimension, the amplitude, duration and distribution law of the abnormal fluctuation. The single-point high-value anomaly is recorded as a short-term sporadic phenomenon and classified as a non-persistent anomaly. At the same time, it is compared with the fractal dimension range of the normal signal to indicate the deviation range and fluctuation characteristics of the current signal. Finally, all analysis results are organized into tables or visualizations for quick viewing and use by medical personnel.

[0090] See also Figure 5 , the specific steps for obtaining the state change probability of the input node are:

[0091] Extract all current patient monitoring data, combine the abnormal data in the abnormal fluctuation assessment results, analyze the correlation and causal relationship between input nodes and observation nodes, and build a Bayesian network model;

[0092] Extract the current patient monitoring data, analyze the relationship between the current patient monitoring data and the abnormal fluctuation characteristics of the abnormal data time series signal in the abnormal fluctuation assessment results, and realize the quantitative analysis of the data relationship by constructing a Bayesian network model. The specific operations include the following steps: First, use data processing and statistical analysis software such as R or Python (combined with the pgmpy library or bnlearn package) to build a Bayesian network model, standardize the collected patient monitoring data (such as airway pressure, tidal volume, respiratory rate) into a time series signal and input it into the model. When defining the model node, use each type of monitoring data as an independent input node. For example, the "tidal volume fluctuation node" is used to describe the changing characteristics of the tidal volume, and the "respiratory frequency node" represents the change in the monitoring value of the respiratory frequency. Combined with the abnormal fluctuation assessment results, the time points marked as abnormal are introduced into the model as observation nodes. By analyzing the conditional independence between these observation nodes and other nodes, the probability distribution of the model is trained using historical data, and the causal relationship edges between nodes are determined through a network structure learning algorithm (such as K2 or HillClimbing algorithm) to complete the structure construction of the Bayesian network.

[0093] Based on the data in the Bayesian network model, the formula is used:

[0094] ;

[0095] Compute observation nodes After the corresponding abnormal data appears, the same type of input node The probability of state change , generate the state change probability of the input node;

[0096] in, Indicates that in the same type of input node Under the known state, according to the observation node The probability of abnormal data being recorded is calculated by counting similar input nodes. (abnormal status) and (normal state) data records, and filter the observation nodes from them Record the time point of the anomaly and calculate the frequency of occurrence. The calculation method is:

[0097] ;

[0098] ;

[0099] Indicates that the observation node has not started to be observed In the case of the same type of input node The probability of state change is calculated by counting the similar input nodes in the historical data. The ratio of normal to abnormal states is obtained by:

[0100] ;

[0101] ;

[0102] Represents the observation node The corresponding probability of abnormal data occurrence is obtained by calculating the proportion of abnormal data in the abnormal fluctuation assessment results. The calculation method is:

[0103] ;

[0104] Based on the observation node and similar input nodes The set dynamic correlation coefficient is used to quantify the ratio of abnormal data strength to overall signal strength. The calculation method is:

[0105] ;

[0106] is an abnormal data time series signal, such as the normalized value of tidal volume in an abnormal time period, It is a time series signal of all patient monitoring data, including normal and abnormal data, obtained through historical records.

[0107] For example, if the observation node The abnormal data recorded indicates abnormal fluctuations in the patient's tidal volume. Indicates whether the tidal volume is abnormal in the next 10 seconds (status 1 is abnormal, status 0 is normal). Based on the monitoring data and abnormal fluctuation evaluation results, the parameters are assumed to be as follows: , , , , , , conditional probability calculation:

[0108] calculate :

[0109] ;

[0110] ;

[0111] calculate :

[0112] ;

[0113] ;

[0114] Verify the probability and: ;

[0115] Among them, the remaining 0.2 represents the probability not covered in the input data or model structure.

[0116] The results show that P(T=1|E)=0.512: when the abnormal data recorded by the observation node E appears, the probability that the tidal volume state will be abnormal in the next 10 seconds is 51.2%. Because it does not mean that it will definitely be abnormal in the future, but the probability of abnormality is higher, so P(T=0|E)=0.288 is calculated: it is used to indicate that the probability of the tidal volume state converting to normal is 28.8%.

[0117] See also Figure 6 , the specific steps for obtaining the respiratory state prediction results are:

[0118] Sort the state change probability of each input node according to the time series and divide it into abnormal state and normal state, build a change trend curve chart, clarify the potential period of abnormal state in the future, and generate the input node change trend analysis results;

[0119] By extracting the change probability of each input node calculated by the Bayesian network model, the change probability of all nodes is sorted according to the time series, and the change probability of different input nodes is classified, and the probability data is divided into abnormal state and normal state. Assume that the abnormal state range is: when When , the input node is considered to be in an abnormal state; assuming the normal state range: when When the probability of both is lower than 0.5, the input node is considered to be in a normal state; when the probability of both is lower than 0.5, the time trend can be further analyzed. For example, according to the calculation results, when the abnormal data recorded by the observation node E appears, the probability of abnormal tidal volume in the next 10 seconds is P(T=1|E)=0.512, which is in the abnormal state range; and the probability of normal tidal volume is P(T=0|E)=0.288, which is lower than the normal state range. It shows that the node state may show a potential abnormal risk trend. Combined with the time interval analysis, if the state probability of the node gradually increases over time, it may further point to the abnormal risk trend in the future. In this process, it is necessary to adjust the influence of each node in combination with the weight parameter, form a trend curve chart through MATLAB, clarify the potential time period and change range of the abnormal state in the future, and provide time dimension support for respiratory state prediction.

[0120] Based on the analysis results of the input node change trend, by combining the input node with the trend of time change, recording the start period, duration and impact range of the immediate abnormal state and normal state, the respiratory state prediction result is generated;

[0121] The time series state data of the input nodes are sorted, the state data of each time period is matched with the trend change curve in the trend curve chart, the starting time points of the abnormal state and the normal state in the time series are extracted, the duration of each state and its impact on the trend change are counted, and the frequency, duration and position of the abnormal state are further marked on the trend curve chart by comparing the distribution characteristics of different states on the time axis. The potential rules of the occurrence of abnormal states are analyzed with reference to the key points of the trend change curve, such as peaks and troughs, and cross-validated with the state data of historical observation nodes, providing a reliable reference for medical personnel to quickly evaluate and judge the patient's respiratory status.

[0122] See also Figure 7 The steps for obtaining the first warning signal and the second warning signal are specifically as follows:

[0123] According to the abnormal data monitoring results, abnormal data exceeding the safety threshold range is obtained to generate a first warning signal;

[0124] First, all abnormal data that exceed the safety threshold range are extracted, and the time point and monitoring parameters corresponding to each abnormal data are confirmed, such as the time point and specific value exceeding the airway pressure safety threshold, and the record of exceeding the tidal volume or respiratory rate safety threshold. After data classification, the system generates the first warning signal based on the abnormal data at each time point, and stores the detailed information of these abnormal data in the content of the warning signal, including the specific time when the abnormality occurred, the relevant monitoring parameters and their actual values. At the same time, the degree of exceeding the safety range is recorded, and the level of the warning signal is marked by comparing the deviation between the abnormal data and the safety threshold. After the warning signal is generated, the system automatically saves these data in a standardized format to the alarm log for subsequent analysis and processing by medical personnel.

[0125] Compare the respiratory state prediction result with the state of the actual monitoring input node during the arrival prediction period, determine whether the comparison result exceeds the preset warning range threshold, and if so, generate a second warning signal;

[0126] By extracting the abnormal state probability value and normal state probability value within the prediction period, and comparing these probability values ​​with the preset warning range threshold, for the matching results of the predicted probability and the actual observation at each time point, the system calculates whether the predicted probability and the actual state are consistent, and marks the time point when the deviation exceeds the threshold as an abnormal deviation item, further classifies and records the time point of the deviation item, the corresponding input node and the abnormal state probability value, and generates the second warning signal through these records. Specifically, assuming that the warning range threshold is an interval, the warning range threshold of the abnormal state is defined as , the warning range threshold of the normal state is defined as In a certain prediction period, the predicted probability of abnormal state is P(T=1|E)=0.75, the predicted probability of normal state is P(T=0|E)=0.25, and the actual observed state is "normal". First, the predicted probability of abnormal state P(T=1|E)=0.75 is compared with the warning range threshold of abnormal state. The results show , so the time period is marked as a high-risk abnormal state prediction; then, the predicted normal state probability P(T=0|E)=0.25 is compared with the warning range threshold of the normal state, and the results show , further indicating that the probability of the predicted result being in the normal state is low. The actual monitored state is "normal", which is different from the predicted abnormal state probability P(T=1|E)=0.75. The system calculates the predicted deviation value as:

[0127] ;

[0128] The deviation value exceeds the normal state warning range threshold [0,0.4], so it is recorded as a deviation exceeding the threshold. The system marks the time point as an abnormal deviation item and records the detailed data of the time period, including the time point, the predicted abnormal state probability P(T=1|E)=0.75, the actual state is "normal", the deviation value 0.75, and the corresponding input node information. Based on the above deviation exceeding the warning range threshold, the system generates a second warning signal, which details the time point, predicted probability, actual state and deviation value of the deviation, and marks it as a time period that needs to be focused on for further analysis by medical personnel.

[0129] See also Figure 8 , the specific steps for obtaining the anesthesia ventilator data monitoring early warning report are:

[0130] Based on the information of the first warning signal and the second warning signal, the input nodes corresponding to the first warning signal and the second warning signal are matched and integrated to generate a signal summary result;

[0131] Integrating the information of the first warning signal and the second warning signal, the system aggregates the two types of signal data into dedicated anesthesia ventilator monitoring software, such as Carestation Insights or similar medical equipment management platforms. The signals are matched and integrated through timestamps and input nodes. Repeated status information is automatically merged and marked as multi-frequency anomalies. The integration results are displayed in a visual form on the software interface. Abnormal situations are marked with different colors or icons to highlight their corresponding time periods and abnormal levels. At the same time, a unified signal prompt is generated to remind medical personnel in the form of sound or interface pop-up windows.

[0132] Based on the signal summary results, different types of warning signals are classified, the first warning signal is classified as the warning data of real-time monitoring, and the second warning signal is classified as the warning data of prediction deviation, and an anesthesia ventilator data monitoring warning report is generated;

[0133] Different types of signals are classified to generate anesthesia ventilator data monitoring and early warning report. The system classifies the first warning signal as a real-time monitoring signal into one category, and the second warning signal as a prediction deviation signal into another category. Independent classification labels are created in Carestration Insights, and the number, source and key indicators of each type of signal are marked. The classification results are presented in tabular and graphical forms. The signal report lists in detail the occurrence time, corresponding node and status information of each type of signal, and finally generates an anesthesia ventilator data monitoring and early warning report containing classification statistics and signal detailed information.

[0134] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An anesthesia ventilator monitoring system, characterized in that: The system comprises: The real-time data abnormality monitoring module collects the patient monitoring data of the anesthesia ventilator in real time, including the patient's airway pressure, the patient's tidal volume, and the patient's respiratory rate, compares the patient monitoring data with the preset safety threshold, and generates abnormal data monitoring results; The abnormal fluctuation assessment module converts all patient monitoring data into time series signals, extracts abnormal data marked in the abnormal data monitoring results from the time series signals, analyzes the complexity and abnormal fluctuation characteristics of the abnormal data time series signals corresponding to the abnormal data, obtains abnormal fluctuation complexity analysis results, evaluates the stability of the abnormal data time series signals corresponding to the abnormal data according to the abnormal fluctuation complexity analysis results, and generates abnormal fluctuation assessment results; The respiratory state prediction module extracts all current patient monitoring data to construct a Bayesian network model, defines the current patient monitoring data as an input node of the Bayesian network model, extracts abnormal data of the same type as the input node in the abnormal fluctuation assessment result, and defines it as an observation node of the Bayesian network model, calculates the state change probability of the input node, analyzes the trend change of the input node in the future time period, and generates a respiratory state prediction result; The steps for obtaining the state change probability of the input node are specifically as follows: Extract all current patient monitoring data, combine with the abnormal data in the abnormal fluctuation assessment results, analyze the correlation and causal relationship between input nodes and observation nodes, and construct a Bayesian network model; Based on the data in the Bayesian network model, the formula is adopted: Calculate the state change probability P(T|E) of the same type of input node T after the abnormal data corresponding to the observation node E appears, and generate the state change probability of the input node; Among them, P(E|T) represents the probability of abnormal data appearing according to the record of observation node E under the known state of the same type input node T, P(T) represents the probability of state change of the same type input node T without starting to observe the observation node E, P(E) represents the probability of abnormal data appearing corresponding to the observation node E, and α is the dynamic correlation coefficient set according to the observation node E and its same type input node T; The steps for obtaining the respiratory state prediction result are specifically as follows: The state change probability of each input node is sorted according to the time series, and divided into abnormal state and normal state, a change trend curve is constructed, the potential period of abnormal state in the future is clarified, and the input node change trend analysis result is generated; Based on the input node change trend analysis result, by combining the input node with the time change trend, recording the start time period, duration and impact range of the immediate abnormal state and normal state, a respiratory state prediction result is generated; The respiratory status warning module generates a first warning signal based on the abnormal data monitoring results when abnormal data occurs. At the same time, the trend change of similar input nodes in the predicted future time period in the respiratory status prediction results is compared with the numerical values ​​of the actual monitoring observation nodes in the time period, and a second warning signal is generated based on the comparison results. Different types of warning signals are classified to generate an anesthesia ventilator data monitoring warning report.

2. The anesthesia breathing machine monitoring system according to claim 1, characterized in that: The steps for obtaining the abnormal data monitoring results are specifically as follows: Based on real-time collection of patient monitoring data of anesthesia ventilators, by comparing the patient monitoring data at each time point with a preset safety threshold, the patient monitoring data corresponding to the time point that exceeds the safety threshold is marked as abnormal data, indicating that the patient's vital signs are abnormal, and the screened abnormal data is generated; Based on real-time collection of patient monitoring data of anesthesia ventilators, the patient monitoring data corresponding to time points that do not exceed the safety threshold are marked as normal data, so that the normal data can be distinguished from the screened abnormal data, and abnormal data monitoring results are generated.

3. The anesthesia breathing machine monitoring system according to claim 1, characterized in that: The steps for obtaining the abnormal fluctuation complexity analysis results are specifically as follows: Convert all patient monitoring data into time series signals and perform normalization processing, extract abnormal data marked in the abnormal data monitoring results from the time series signals, and generate abnormal data time series signals; Based on the abnormal data time series signal, the formula is adopted: Calculate the fractal dimension D of the abnormal data time series signal to obtain fractal dimension information; Where N(r) is the minimum number of intervals required to cover the abnormal data time series signal at scale r, r represents the scale of different abnormal data time series signals to divide the area, W is the weight parameter set based on different abnormal data, A is the adjustment coefficient set based on the fluctuation of the abnormal data time series signal, ln is the natural logarithm, lnN(r) is used to linearize the abnormal data time series signal, and ln(1 / r) represents the natural logarithm of the reciprocal of scale r; Based on the fractal dimension information, the complexity and abnormal fluctuation characteristics of the target abnormal data time series signal are analyzed. The abnormal fluctuation characteristics include the fluctuation amplitude of the abnormal data time series signal and the distribution density of the abnormal values, and the abnormal fluctuation complexity analysis results are generated.

4. The anesthesia breathing machine monitoring system according to claim 1, characterized in that: The steps for obtaining the abnormal fluctuation assessment result are specifically as follows: Based on the abnormal fluctuation complexity analysis result, the stability of the abnormal data time series signal is evaluated, and the stability evaluation result is generated by judging the influence of the fluctuation characteristics of the abnormal data time series signal on the overall trend; Based on the stability assessment results, all information of the abnormal data time series signal is integrated and displayed in a visual manner to generate abnormal fluctuation assessment results.

5. The anesthesia breathing machine monitoring system according to claim 1, characterized in that: The steps of obtaining the first warning signal and the second warning signal are specifically as follows: According to the abnormal data monitoring result, abnormal data exceeding the safety threshold range is obtained, and a first warning signal is generated; The respiratory state prediction result is compared with the state of the actual monitoring input node during the predicted period to determine whether the comparison result exceeds a preset warning range threshold. If so, a second warning signal is generated.

6. The anesthesia breathing machine monitoring system according to claim 1, characterized in that: The steps for obtaining the anesthesia ventilator data monitoring early warning report are specifically as follows: Based on the information of the first warning signal and the second warning signal, matching and integrating the input nodes corresponding to the first warning signal and the second warning signal to generate a signal summary result; Based on the signal aggregation result, different types of warning signals are classified, the first warning signal is classified as warning data for real-time monitoring, and the second warning signal is classified as warning data for prediction deviation, and an anesthesia ventilator data monitoring warning report is generated.

Images