A mechanical ventilation analysis and early warning method and system

Abstract

A mechanical ventilation analysis early warning method and system, comprising the following steps: real-time acquisition, synchronous analysis of mechanical ventilation monitoring data and data quality analysis; according to the quality qualified monitoring data, the respiratory cycle, the trigger mode, the ventilation type and the control mode are divided; according to the monitoring data after the multi-condition interval division, a multi-level decision model is established to automatically monitor and warn the mechanical ventilation abnormal event; based on the specific sub-condition after the multi-condition interval division, the linear fitting characteristic sequence is monitored and warned in parallel to monitor and warn the respiratory system mechanics change abnormal event. The application identifies and warns all common mechanical ventilation abnormal events in clinic, judges the ventilator setting item parameters through the multi-condition interval processing of the monitoring data, distinguishes the trigger mode, establishes a multi-level decision model and dynamically analyzes the respiratory system mechanics change, so as to improve the mechanical ventilation treatment effect and reduce the risk of ventilator use.

Description

Technical Field

[0001] This application relates to a mechanical ventilation analysis and early warning method and system. Background Technology

[0002] A ventilator is a complex computer-controlled life support device that precisely regulates the gas delivered to patients with respiratory failure through pressure and flow sensors, thereby buying precious time for treating the underlying disease. While effectively replacing the respiratory organs to provide ventilation and maintain stable vital signs, ventilator-assisted systems must also adhere to the fundamental principles of mechanical ventilation: "life safety," "treatment effectiveness," and "patient comfort," thereby minimizing the harm caused by ventilator use to patients.

[0003] To achieve the above goals, intelligent data processing of mechanical ventilation equipment is required. This involves providing early warnings of potential respiratory distress events that could cause respiratory injury, analyzing them in real time, and promptly capturing changes in the respiratory system's mechanical structure during treatment. This will improve ventilator performance and reduce the risks associated with ventilator use. While performing invasive ventilation, clinicians can adjust the ventilator mode and parameters, as well as regulate pressure and flow, based on early warning information. This allows for the development of individualized ventilation strategies for different patients, thereby improving the effectiveness of mechanical ventilation and minimizing the risk of adverse reactions.

[0004] Currently, ventilators both domestically and internationally only integrate threshold alarms and technical alarms; their derivative products only alarm for one or a few abnormal events, lacking a comprehensive capability for identifying and warning of clinical abnormal events. Furthermore, ventilators only display monitoring values ​​for airway resistance and lung compliance, failing to assess changes in respiratory mechanics and the patient's pathological state during treatment. Traditional identification methods require clinicians to record numerical changes and analyze waveform data characteristics in conjunction with ventilation modes and multiple parameters to assess the patient's overall condition. This presents significant limitations due to the limited time clinicians have for bedside observation. Therefore, establishing a system that provides timely warnings for all common mechanical ventilation abnormalities, dynamically analyzes respiratory mechanics changes during mechanical ventilation, improves ventilator performance, reduces ventilator use risks, prompts physician and nurse interventions, and assists physicians in comprehensively assessing the patient's pathological state, thereby improving the efficacy of mechanical ventilation therapy, is a pressing issue in this field. Summary of the Invention

[0005] To address the aforementioned issues, this application proposes a mechanical ventilation analysis and early warning method, characterized by the following steps: Step 1: Real-time acquisition and synchronous analysis of mechanical ventilation monitoring data, followed by data quality analysis; Step 2: If the monitoring data quality is satisfactory, multi-condition interval division is performed based on respiratory cycle, triggering method, ventilation type, and control mode; Step 3: Based on the monitoring data after multi-condition interval division, a multi-level decision model is established for automatic monitoring of life-threatening abnormal events, mechanical ventilation quality abnormal events, and patient-ventilator asynchrony abnormal events; based on specific sub-conditions after multi-condition interval division, respiratory system mechanical changes are monitored in parallel according to the linearly fitted feature sequence; Step 4: Early warning information is pushed according to the alarm levels of life-threatening abnormal events, mechanical ventilation quality abnormal events, and patient-ventilator asynchrony abnormal events, while simultaneously pushing early warning information for respiratory system mechanical change abnormal events. This application aims to identify and warn of all common mechanical ventilation abnormal events in clinical practice. Through multi-condition interval processing of monitoring data, it judges the ventilator's own setting parameters, distinguishes between patient triggering and machine triggering, and maximizes the differentiation between the general rules of normal respiratory waveforms and the specific characteristics of abnormal event waveforms. Based on the waveform morphology and change characteristics of abnormal events, a multi-level decision-making model is established. Based on specific sub-conditions in multi-condition intervalization, respiratory system mechanical changes are analyzed. This model can detect abnormal mechanical ventilation events at an early stage, provide timely warnings, and prompt doctors and nurses to intervene, thereby improving the efficacy of mechanical ventilation treatment and reducing the risks of ventilator use.

[0006] Preferably, the monitoring data includes pressure waveform data, capacity waveform data, and flow velocity waveform data. While acquiring the monitoring data, data quality analysis is performed, including whether the waveform data has drifted, moved, or interfered with; whether the waveform data values ​​are abnormal; if there are data quality problems, an early warning information on data quality problems is directly pushed and monitoring data continues to be collected and analyzed.

[0007] Preferably, the respiratory cycle division is based on a maximum and minimum value search algorithm to identify the extreme values ​​of the monitoring data waveform curve and divide the respiratory interval; the duration of the monitoring data is not less than 60 seconds.

[0008] Preferably, the multi-condition interval division refers to establishing a multi-model integrated architecture and dividing the system according to triggering method, ventilation type, and control mode:

[0009] S1: Based on the triggering method classification rules, the machine triggering identification module further classifies the monitoring data into machine triggering and patient triggering;

[0010] S2: Based on the ventilation type classification rules, the control ventilation identification module further classifies the monitoring data into control ventilation, assisted ventilation and spontaneous ventilation;

[0011] S3: Based on the control mode classification rules, the control identification module further classifies the monitoring data into capacity control type and pressure control type;

[0012] This application converts waveform data into units based on respiratory cycles to compare the waveform morphology and variation characteristics of different respiratory cycles, eliminates periodic patterns, and locates abnormal events. A multi-model integrated architecture is used to classify triggering methods, ventilation types, and control modes. The performance characteristics of waveform data are mathematically derived to determine whether the ventilator's own settings are correct and to identify patient-triggered and machine-triggered types, laying the foundation for subsequent abnormal event analysis, monitoring, and early warning.

[0013] Preferably, this application uses a maximum and minimum value search algorithm based on the volume waveform curve, and uses the positions of the maximum and minimum values ​​as the basis for interval division. It then divides the pressure, volume, and flow rate time series within a unit monitoring time into respiratory cycle segments. The resulting pressure time series are denoted as P1, P2, ..., P... t-2 P t-1 P t The pressure waveform data of the t-th respiratory cycle is denoted as in Let V1 be the pressure value corresponding to the nth data sampling point in the tth respiratory cycle; the divided volume time series are denoted as V1, V2, ..., V3. t-2 V t-1 V t The volume waveform data of the t-th respiratory cycle is denoted as in The volume value corresponding to the nth data sampling point in the t-th respiratory cycle; flow rate time series F1, F2, ..., F t-2 F t-1 F t The flow rate waveform data for the t-th respiratory cycle is denoted as in Insp represents the flow rate value corresponding to the nth data sampling point in the tth respiratory cycle; t Let t be the inspiratory time of the t-th respiratory cycle, Exp t Let t be the exhalation time of the t-th respiratory cycle.

[0014] Preferably, the life safety abnormal events include air leakage events, airway circuit obstruction events, and gas trapping events; mechanical ventilation quality abnormal events include condensation in the tubing, suctioning, and CPR; periodic asynchrony events of human-machine asynchrony abnormal events include flow rate starvation, premature switching, delayed switching, and over-firing; frequency asynchrony events of human-machine asynchrony abnormal events include automatic triggering, invalid triggering, reverse triggering, and repeated triggering; the monitoring of internal events of the three major categories of abnormal events adopts a parallel monitoring method; the three-category classification in this application is for graded early warning and elimination of interference between events.

[0015] Preferably, the multi-level decision model consists of a life safety abnormal event decision tree as the first layer, a mechanical ventilation quality abnormal event decision tree as the second layer, and a human-machine asynchrony abnormal event decision tree as the third layer. The decision tree selects waveform data for preferred monitoring, and establishes complete logical judgment rules based on mathematical calculation formulas, fitting waveform curves to judge function characteristics, and rule-based thresholds. It establishes a correspondence between waveform morphology characteristics and change patterns and the performance characteristics of abnormal events, thereby identifying and classifying specific abnormal events.

[0016] Preferably, the monitoring of gas leakage events and gas trapping events in the aforementioned life safety anomalies specifically includes the following steps:

[0017] Select capacity time series and flow rate time series for abnormal event identification:

[0018] S1: First, determine the flow rate based on the flow rate time series F1, F2, ..., Ft-2, Ft-1, Ft, using the criterion of whether Ft reaches zero at the end of expiration.

[0019] S2: If the flow rate value reaches zero at the end of expiration, then proceed to step S4;

[0020] S3: Conversely, if the flow rate value does not return to zero at the end of exhalation, continue to judge according to Expt=0: if Expt=0 is true, then jump to step S4; otherwise, jump to step S5.

[0021] S4: Then, based on the volume time series V1, V2, ..., Vt-2, Vt-1, Vt, determine whether Vt shows a vertical decreasing trend at the end of the respiration period and satisfies the following conditions. If both conditions are met, proceed to step S6; otherwise, terminate monitoring.

[0022] S5: The end-expiratory flow rate values ​​of Ft-2, Ft-1, and Ft did not return to zero for at least three respiratory cycles. At this time, a gas trapping event may occur during mechanical ventilation, which will be judged as an abnormal life safety event.

[0023] S6: If at least three respiratory cycles Vt-2, Vt-1, and Vt satisfy two of the conditions in S4, then an air leakage event may occur during mechanical ventilation, which will be judged as an abnormal life safety event.

[0024] Preferably, the monitoring of pipeline condensate events in the aforementioned mechanical ventilation quality abnormality events specifically includes the following steps:

[0025] Select flow velocity time series for abnormal event identification:

[0026] First, based on the flow velocity time series F1, F2, ..., F... t-2 F t-1 F t Make a judgment: Calculate F during the expiratory phase t First-order difference F t ′ and second-order difference F t ";in Second-order difference The location and number of local maxima and local minima in the expiratory phase are determined by the first-order and second-order differences.

[0027] The flow velocity time series F1, F2, ..., F t-2 F t-1 F t Based on the fluctuation pattern of rising and then falling during the exhalation phase, the duration of each fluctuation is obtained by subtracting the adjacent minimum and maximum values. If the frequency of oscillation is greater than a preset rule threshold and the duration of each oscillation is less than the preset rule threshold, then there may be a condensation event in the mechanical ventilation equipment pipeline, which is determined to be an abnormal mechanical ventilation quality event.

[0028] Preferably, the determination of human-machine asynchrony abnormal events specifically includes the following steps:

[0029] S1: Based on the definition rules of mechanical ventilation cycle and frequency, the mechanical ventilation cycle and frequency module further divides the monitoring data into cycle asynchrony or frequency asynchrony. If it is cycle asynchrony, the cycle asynchrony sub-decision tree is used to identify and classify cycle asynchrony abnormal events; if it is frequency asynchrony, the frequency asynchrony sub-decision tree is used to identify and classify frequency asynchrony abnormal events.

[0030] S2: A human-machine asynchrony index (AI) exceeding a set threshold (10%) is considered a serious human-machine asynchrony event.

[0031]

[0032] The total number of respiratory cycles during the monitoring period is statistically analyzed, including the total number of mechanical and spontaneous respiratory cycles, the number of respiratory cycles with human-machine asynchrony events, and the total number of respiratory cycles with invalid trigger events. Based on the above formula, if the human-machine asynchrony index exceeds a set threshold, a relevant alarm will be triggered.

[0033] The monitoring of flow rate starvation anomaly and repetitive triggering anomaly events in the aforementioned human-machine asynchrony anomaly events specifically includes the following steps:

[0034] Selecting pressure time series for abnormal event identification:

[0035] S1: Sub-decision tree with asynchronous execution cycles for flow rate starvation abnormal events; S2: Sub-decision tree with asynchronous execution frequencies for repeated triggering abnormal events; S5: Sub-decision tree with asynchronous execution frequencies for repeated triggering abnormal events.

[0036] S2: First, based on the pressure time series P1, P2, ..., P... t-2 P t-1 P t The judgment is based on the "concave" state of the pressure-time series during the intake phase. The waveform characteristics are converted into a mathematical expression: the P in the intake phase is calculated. t First-order difference P t ′ and second-order difference P t ";in Second-order difference

[0037] S3: If the second difference of the inspiratory phase pressure time series is non-negative and the pressure waveform fitting curve approximately satisfies the characteristics of a concave function, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6.

[0038] S4: If there are outlier points in the second difference of the inspiratory phase pressure time series, it indicates that the rate of increase of the inspiratory phase pressure value has changed. Using the outlier points as the dividing points, if the fitting curves of the two waveform segments approximately satisfy the convex function characteristics, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchronous abnormal event, and the process will proceed to step S6.

[0039] S5: If Exp t =0 is true or Exp t ≤0.5*mean(Insp1,...,Insp t If the expiratory time is less than half of the average inspiratory time, then a repeated triggering event may occur during mechanical ventilation. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6.

[0040] S6: The human-machine asynchrony index is calculated based on the AI ​​formula. When the value exceeds the set threshold, a relevant alarm is pushed.

[0041] Preferably, the linearly fitted feature sequences are used to monitor respiratory system mechanical changes in parallel. This monitoring of respiratory system mechanical changes is independent and parallel to monitoring life-threatening events, mechanical ventilation quality issues, and patient-ventilator asynchrony events. The main operational steps include:

[0042] S1: Select the triggering method based on the multi-condition interval division of the multi-model integration architecture;

[0043] S2: Select the ventilation type for control ventilation based on the multi-condition interval division of the multi-model integrated architecture;

[0044] S3: Based on the multi-condition interval division of the multi-model integrated architecture, determine the capacity control type and pressure control type;

[0045] S4: Perform adaptive change trend identification:

[0046] S4_1: In capacity-controlled ventilation mode, the characteristic sequence representing compliance trend is denoted as:

[0047]

[0048] Plateau pressure during the m-th respiratory cycle Based on the position at the beginning of the expiratory phase, the average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop is taken as the plateau pressure value for this respiratory cycle.

[0049] Positive end-expiratory pressure (PEP) in the m-th respiratory cycle m The peep value for this respiratory cycle is the average of all pressure sampling points within the stable trend range of the pressure waveform before the end of the expiratory phase, based on the position at which the expiratory phase ends.

[0050] The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.1.

[0051] P Δ2 It is inversely proportional to the compliance value, and P Δ2 After linear fitting of the feature sequences, the fitting curve can reflect the changing trend of compliance;

[0052] S4_2: In pressure-controlled ventilation mode, the characteristic sequence representing compliance trend, the effective inspiratory time sequence is denoted as:

[0053] Δt=[Δt1, Δt2,..., Δt m , ..., Δt z];

[0054] The sequence of peak velocity angles is denoted as:

[0055] θ = [θ1, θ2, ..., θ m , ..., θ z ];

[0056] Δt and θ are positively correlated with compliance values. After linearly fitting the Δt feature sequence and the θ feature sequence respectively, the fitted curve can reflect the changing trend of compliance.

[0057] S5: Perform airway resistance change trend identification:

[0058] S5_1: In volume-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance is denoted as:

[0059]

[0060] Peak pressure during the m-th respiratory cycle The peak pressure of this respiratory cycle is defined as the maximum value of all pressure sampling data points from the beginning of the inspiratory phase to the end of the expiratory phase.

[0061] Plateau pressure during the m-th respiratory cycle The average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop at the beginning of the expiratory phase is taken as the plateau pressure value for this respiratory cycle.

[0062] The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.1.

[0063] P Δ1 It is directly proportional to the airway resistance value, and P Δ1 After linear fitting of the characteristic sequence, the fitting curve can reflect the changing trend of airway resistance.

[0064] S5_2: In pressure-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance, with the peak flow rate sequence denoted as:

[0065]

[0066] Peak flow rate in the m-th respiratory cycle The peak flow rate for this respiratory cycle is defined as the maximum value of all flow rate sampling points from the beginning of the inspiratory phase to the end of the expiratory phase.

[0067] f peak It is inversely proportional to the airway resistance value, f peak After linear fitting of the characteristic sequence, the fitted curve can be used to reflect the changing trend of airway resistance;

[0068] S6: Perform dynamic changes analysis of the respiratory system. If the trend curve shows the following, it is determined to be an abnormal event of respiratory system mechanical changes:

[0069] In capacity-controlled ventilation mode, when the curve reflecting the change in compliance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment:

[0070] In pressure-controlled ventilation mode, when the two trend curves reflecting compliance change per unit time show a continuous downward trend at a certain moment, and the decrease is 4% of the corresponding value at that moment:

[0071] In volume-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment;

[0072] In pressure-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous downward trend at a certain moment, and the decrease is 6% of the corresponding value at that moment.

[0073] Preferably, the linear fitting feature sequences reflecting the trends of airway resistance and compliance can be directly calculated from the performance characteristics of pressure-time waveforms and flow-time waveforms. Increased airway resistance or decreased compliance indicates that the patient's pathological condition has deteriorated during mechanical ventilation. Increased airway resistance indicates that there may be lesions in the patient's upper respiratory tract, trachea, and bronchi. Decreased compliance indicates that there may be lesions in the patient's chest, lungs, and other organs, which may induce related diseases. Therefore, clinicians can comprehensively assess the patient's pathological condition based on the alarm type information, conduct relevant organ function examinations, and provide targeted treatment to effectively guide the implementation of mechanical ventilation.

[0074] Preferably, while monitoring all common abnormal events of mechanical ventilation in real time, it provides multi-level alarms for abnormal events of life safety, abnormal events of mechanical ventilation quality, and abnormal events of patient-ventilator asynchrony. Based on the basic treatment principles of clinical guidelines, it first ensures the patient's life safety during treatment, secondly ensures the effectiveness of mechanical ventilation treatment, and thirdly meets the patient's comfort during treatment, so as to minimize the occurrence of patient-ventilator asynchrony.

[0075] Following a priority order of warning and handling for life-threatening emergencies, timely control of mechanical ventilation quality abnormalities, and precise alerts for potential patient-ventilator asynchrony abnormalities, this application simultaneously pushes alerts for respiratory system biomechanical changes throughout the entire mechanical ventilation process. This application combines multi-level early warning with parallel alerts for respiratory system biomechanical changes during clinical implementation. During monitoring, all abnormal events are screened, and alarms are triggered in a hierarchical order, prioritizing alarms for critical events to ensure efficient handling of life-threatening emergencies, while also considering other types of abnormal events.

[0076] On the other hand, this application also discloses a mechanical ventilation analysis and early warning system, including: a data acquisition module for real-time acquisition, synchronous analysis of mechanical ventilation monitoring data and data quality analysis; a data preprocessing module for dividing respiratory cycle, triggering method, ventilation type and control mode into multi-condition intervals, provided the monitoring data quality is qualified; an event monitoring module for automatically monitoring life safety abnormal events, mechanical ventilation quality abnormal events, and human-ventilator asynchrony abnormal events based on the monitoring data after multi-condition interval division and establishing a multi-level decision model; based on specific sub-conditions after multi-condition interval division, parallel monitoring of respiratory system mechanical changes according to linearly fitted feature sequences; and an event early warning module for pushing early warning information according to the alarm levels of life safety abnormal events, mechanical ventilation quality abnormal events, and human-ventilator asynchrony abnormal events, while simultaneously pushing early warning information for respiratory system mechanical change abnormal events.

[0077] This application can bring the following beneficial effects:

[0078] 1. This application uses multi-condition interval processing of monitoring data to judge the ventilator's own setting parameters and distinguish between patient triggering and machine triggering. It maximizes the distinction between the general rules of normal respiratory waveforms and the specific characteristics of abnormal event waveforms. Based on the manifestation characteristics of abnormal event waveform morphology and changes, a multi-level decision model is established. Based on specific sub-conditions in the multi-condition interval processing, respiratory system mechanical changes are analyzed. This can detect abnormal mechanical ventilation events at an early stage, provide timely warnings, and prompt doctors and nurses to intervene, thereby improving the efficacy of mechanical ventilation treatment and reducing the risks of ventilator use.

[0079] 2. This application converts waveform data into units based on respiratory cycles to compare the waveform morphology and changes of different respiratory cycles, eliminates periodicity, locates abnormal events, classifies triggering methods, ventilation types, and control modes using a multi-model integrated architecture, mathematically derives the characteristics of waveform data, determines whether the ventilator's own settings are correct, and identifies patient-triggered and machine-triggered types, laying the foundation for subsequent abnormal event analysis, monitoring, and early warning.

[0080] 3. This application, in conjunction with specific clinical circumstances, establishes a multi-level early warning system and a parallel early warning system for abnormal events related to respiratory system mechanics changes. During the monitoring process, all abnormal events are screened, and an alarm sequence is set up to prioritize alarms for critical events, ensuring the efficiency of handling life-threatening events, while also taking into account other types of abnormal events. Attached Figure Description

[0081] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0082] Figure 1 This is a schematic diagram of Example 1;

[0083] Figure 2 This is a schematic diagram of Example 2. Detailed Implementation

[0084] To clearly illustrate the technical features of this solution, the following detailed description of specific implementation methods will be provided.

[0085] The solution has already been described in the invention description. To better illustrate the technical solution of this application, this application will elaborate on a more specific aspect of each event:

[0086] Example 1: A mechanical ventilation analysis and early warning method, such as Figure 1 As shown, it includes the following steps:

[0087] S101: Real-time acquisition and synchronous analysis of mechanical ventilation monitoring data and data quality analysis;

[0088] The monitoring data includes pressure waveform data, capacity waveform data, and flow velocity waveform data. While acquiring the monitoring data, data quality analysis is performed: including whether there is drift, movement, or interference in the waveform data; whether there are abnormalities in the waveform data values; if there are data quality problems, an early warning information on data quality problems is directly pushed and monitoring data continues to be collected and analyzed.

[0089] S102: If the monitoring data quality is qualified, then perform multi-condition interval division of respiratory cycle, triggering mode, ventilation type, and control mode; the respiratory cycle division is based on the maximum and minimum value search algorithm to identify the extreme values ​​of the monitoring data waveform curve and divide the respiratory interval; the duration of the monitoring data is not less than 60s;

[0090] The multi-condition interval division refers to establishing a multi-model integrated architecture and dividing the system according to triggering method, ventilation type, and control mode:

[0091] S1: Based on the triggering method classification rules, the machine triggering identification module further classifies the monitoring data into machine triggering and patient triggering;

[0092] S2: Based on the ventilation type classification rules, the control ventilation identification module further classifies the monitoring data into control ventilation, assisted ventilation and spontaneous ventilation;

[0093] S3: Based on the control mode classification rules, the control identification module further classifies the monitoring data into capacity control type and pressure control type;

[0094] Based on the volume waveform curve, a maximum and minimum value search algorithm is executed, and the positions of the maximum and minimum values ​​are used as the basis for interval division. The time series of pressure, volume, and flow rate within a unit monitoring time are divided into respiratory cycle segments. The resulting pressure time series are denoted as P1, P2, ..., P... t-2 P t-1 P t The pressure waveform data of the t-th respiratory cycle is denoted as in Let V1 be the pressure value corresponding to the nth data sampling point in the tth respiratory cycle; the divided volume time series are denoted as V1, V2, ..., V3. t-2 V t-1 V t The volume waveform data of the t-th respiratory cycle is denoted as in The volume value corresponding to the nth data sampling point in the t-th respiratory cycle; flow rate time series F1, F2, ..., F t-2 F t-1 F t The flow rate waveform data for the t-th respiratory cycle is denoted as in Insp represents the flow rate value corresponding to the nth data sampling point in the tth respiratory cycle; t Let t be the inspiratory time of the t-th respiratory cycle, Exp t Let t be the exhalation time of the t-th respiratory cycle.

[0095] S103: Based on the monitoring data after the multi-condition interval division, establish a multi-level decision model for automatic monitoring of abnormal life safety events, abnormal mechanical ventilation quality events, and abnormal human-machine asynchrony events; based on specific sub-conditions after the multi-condition interval division, monitor respiratory system mechanical changes in parallel according to the feature sequence of linear fitting.

[0096] The multi-level decision-making model consists of a life safety abnormal event decision tree as the first layer, a mechanical ventilation quality abnormal event decision tree as the second layer, and a human-machine asynchrony abnormal event decision tree as the third layer. The decision tree selects waveform data for preferred monitoring, judges function characteristics through mathematical calculation formulas and fitting waveform curves, and establishes complete logical judgment rules based on rule thresholds. It establishes a correspondence between waveform morphology characteristics and change patterns and the performance characteristics of abnormal events, thereby identifying and classifying specific abnormal events.

[0097] The life safety anomalies include air leakage, airway circuit obstruction, and gas trapping; mechanical ventilation quality anomalies include condensation in the tubing, suctioning, and CPR; periodic asynchrony anomalies include flow starvation, premature switching, delayed switching, and over-firing; frequency asynchrony anomalies include automatic triggering, invalid triggering, reverse triggering, and repeated triggering; the monitoring of internal events in these three categories of anomalies adopts a parallel monitoring method.

[0098] S104: Set alarm levels according to life safety abnormal events, mechanical ventilation quality abnormal events, and human-machine asynchrony abnormal events to push early warning information, and push early warning information for respiratory system mechanical changes abnormal events in parallel.

[0099] The monitoring of gas leakage events and gas trapping events in the aforementioned life safety anomalies specifically includes the following steps:

[0100] Select capacity time series and flow rate time series for abnormal event identification:

[0101] S1: First, based on the flow velocity time series P1, P2, ..., P... t-2 P t-1 P t The judgment is based on F. t Does the flow rate return to zero at the end of expiration?

[0102] S2: If the flow rate value reaches zero at the end of expiration, then proceed to step S4;

[0103] S3: Conversely, if the flow rate does not return to zero at the end of expiration, continue according to Exp. t =0 to determine: if Exp t If 0 is true, then proceed to step S4; otherwise, proceed to step S5.

[0104] S4: Then, based on the capacity time series V1, V2, ..., V t-2 V t-1 V t The judgment is based on V. t Does it exhibit a vertical descent trend at the end of exhalation, and does it meet the following criteria? If both conditions are met, proceed to step S6; otherwise, terminate monitoring.

[0105] S5: At least three respiratory cycles F t-2 F t-1 F t The end-expiratory flow rate did not return to zero. At this time, a gas trapping event may occur during mechanical ventilation, which will be judged as an abnormal life safety event.

[0106] S6: At least three respiratory cycles V t-2 V t-1 V t If both conditions in S4 are met, an air leak may occur during mechanical ventilation, which will be classified as a life-threatening event.

[0107] Monitoring of pipeline condensate events during mechanical ventilation quality anomalies includes the following steps:

[0108] Select flow velocity time series for abnormal event identification:

[0109] First, based on the flow velocity time series F1, F2, ..., F... t-2 F t-1 F t Make a judgment: Calculate F during the expiratory phase t First-order difference F t ′ and second-order difference F t ";in Second-order difference The location and number of local maxima and local minima in the expiratory phase are determined by the first-order and second-order differences.

[0110] The flow velocity time series F1, F2, ..., F t-2 F t-1 F t Based on the fluctuation pattern of rising and then falling during the exhalation phase, the duration of each fluctuation is obtained by subtracting the adjacent minimum and maximum values. If the frequency of oscillation is greater than a preset rule threshold and the duration of each oscillation is less than the preset rule threshold, then there may be a condensation event in the mechanical ventilation equipment pipeline, which is determined to be an abnormal mechanical ventilation quality event.

[0111] The specific steps for determining human-machine asynchrony anomalies are as follows:

[0112] S1: According to the definition rules of mechanical ventilation cycle and frequency, the mechanical ventilation cycle and frequency module further divides the monitoring data into cycle asynchrony or frequency asynchrony. If it is cycle asynchrony, the cycle asynchrony sub-decision tree is used to identify and classify cycle asynchrony abnormal events; if it is frequency asynchrony, the frequency asynchrony sub-decision tree is used to identify and classify frequency asynchrony abnormal events.

[0113] S2: A human-machine asynchrony index (AI) exceeding a set threshold (10%) is considered a serious human-machine asynchrony event.

[0114]

[0115] The total number of respiratory cycles during the monitoring period is statistically analyzed, including the total number of mechanical and spontaneous respiratory cycles, the number of respiratory cycles with human-machine asynchrony events, and the total number of respiratory cycles with invalid trigger events. Based on the above formula, if the human-machine asynchrony index exceeds a set threshold, a relevant alarm will be triggered.

[0116] The monitoring of flow rate starvation anomaly and repetitive triggering anomaly events in the aforementioned human-machine asynchrony anomaly events specifically includes the following steps:

[0117] Selecting pressure time series for abnormal event identification:

[0118] S1: Sub-decision tree with asynchronous execution cycles for flow rate starvation abnormal events; S2: Sub-decision tree with asynchronous execution frequencies for repeated triggering abnormal events; S5: Sub-decision tree with asynchronous execution frequencies for repeated triggering abnormal events.

[0119] S2: First, based on the pressure time series P1, P2, ..., P... t-2 P t-1 P t The judgment is based on the "concave" state of the pressure-time series during the intake phase. The waveform characteristics are converted into a mathematical expression: the P in the intake phase is calculated. t First-order difference P t ′ and second-order difference P t ";in Second-order difference

[0120] S3: If the second difference of the inspiratory phase pressure time series is non-negative and the pressure waveform fitting curve approximately satisfies the characteristics of a concave function, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6.

[0121] S4: If there are outlier points in the second difference of the inspiratory phase pressure time series, it indicates that the rate of increase of the inspiratory phase pressure value has changed. Using the outlier points as the dividing points, if the fitting curves of the two waveform segments approximately satisfy the convex function characteristics, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchronous abnormal event, and the process will proceed to step S6.

[0122] S5: If Exp t =0 is true or Exp t ≤0.5*mean(Insp1,...,Insp t If the expiratory time is less than half of the average inspiratory time, then a repeated triggering event may occur during mechanical ventilation. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6.

[0123] S6: The human-machine asynchrony index is calculated based on the AI ​​formula. When the value exceeds the set threshold, a relevant alarm is pushed.

[0124] The linearly fitted feature sequences are used to monitor respiratory system mechanical changes in parallel. This monitoring of respiratory system mechanical changes is independent and parallel to monitoring life-threatening events, mechanical ventilation quality issues, and patient-ventilator asynchrony events. The main operational steps include:

[0125] S1: Select the triggering method based on the multi-condition interval division of the multi-model integration architecture;

[0126] S2: Select the ventilation type for control ventilation based on the multi-condition interval division of the multi-model integrated architecture;

[0127] S3: Based on the multi-condition interval division of the multi-model integrated architecture, determine the capacity control type and pressure control type;

[0128] S4: Perform adaptive change trend identification:

[0129] S4_1: In capacity-controlled ventilation mode, the characteristic sequence representing compliance trend is denoted as:

[0130]

[0131] Plateau pressure during the m-th respiratory cycle Based on the position at the beginning of the expiratory phase, the average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop is taken as the plateau pressure value for this respiratory cycle.

[0132] Positive end-expiratory pressure (PEP) in the m-th respiratory cycle mThe peep value for this respiratory cycle is the average of all pressure sampling points within the stable trend range of the pressure waveform before the end of the expiratory phase, based on the position at which the expiratory phase ends.

[0133] The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.1.

[0134] P Δ2 It is inversely proportional to the compliance value, and P Δ2 After linear fitting of the feature sequences, the fitting curve can reflect the changing trend of compliance;

[0135] S4_2: In pressure-controlled ventilation mode, the characteristic sequence representing compliance trend, the effective inspiratory time sequence is denoted as:

[0136] Δt=[Δt1,Δt2,...,Δt m , ...Δt z ];

[0137] The sequence of peak velocity angles is denoted as:

[0138] θ = [θ1, θ2, ..., θ m , ..., θ z ];

[0139] Δt and θ are positively correlated with compliance values. After linearly fitting the Δt feature sequence and the θ feature sequence respectively, the fitted curve can reflect the changing trend of compliance.

[0140] S5: Perform airway resistance change trend identification:

[0141] S5_1: In volume-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance is denoted as:

[0142]

[0143] Peak pressure during the m-th respiratory cycle The peak pressure of this respiratory cycle is defined as the maximum value of all pressure sampling data points from the beginning of the inspiratory phase to the end of the expiratory phase.

[0144] Plateau pressure during the m-th respiratory cycle The average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop at the beginning of the expiratory phase is taken as the plateau pressure value for this respiratory cycle.

[0145] The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.1.

[0146] P Δ1It is directly proportional to the airway resistance value, and P Δ1 After linear fitting of the characteristic sequence, the fitting curve can reflect the changing trend of airway resistance.

[0147] S5_2: In pressure-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance, with the peak flow rate sequence denoted as:

[0148]

[0149] Peak flow rate in the m-th respiratory cycle The peak flow rate for this respiratory cycle is defined as the maximum value of all flow rate sampling points from the beginning of the inspiratory phase to the end of the expiratory phase.

[0150] f peak It is inversely proportional to the airway resistance value, f peak After linear fitting of the characteristic sequence, the fitted curve can be used to reflect the changing trend of airway resistance;

[0151] S6: Perform dynamic changes analysis of the respiratory system. If the trend curve shows the following, it is determined to be an abnormal event of respiratory system mechanical changes:

[0152] In capacity-controlled ventilation mode, when the curve reflecting the change in compliance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment:

[0153] In pressure-controlled ventilation mode, when the two trend curves reflecting compliance change per unit time show a continuous downward trend at a certain moment, and the decrease is 4% of the corresponding value at that moment:

[0154] In volume-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment;

[0155] In pressure-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous downward trend at a certain moment, and the decrease is 6% of the corresponding value at that moment.

[0156] In the second embodiment, as Figure 2 As shown, a mechanical ventilation analysis and early warning system is characterized by comprising:

[0157] The data acquisition module 201 is used to collect and analyze the monitoring data of mechanical ventilation in real time and perform data quality analysis.

[0158] Data preprocessing module 202, under the condition that the monitoring data quality is qualified, is used to divide the respiratory cycle, triggering method, ventilation type and control mode into multiple condition intervals;

[0159] The event monitoring module 203 is used to establish a multi-level decision model based on the monitoring data after the multi-condition interval division to automatically monitor abnormal events of life safety, abnormal events of mechanical ventilation quality, and abnormal events of human-machine asynchrony; and to monitor the mechanical changes of the respiratory system in parallel based on specific sub-conditions after the multi-condition interval division and the feature sequence of linear fitting.

[0160] The event warning module 204 is used to push warning information according to the alarm level settings of abnormal events of life safety, abnormal events of mechanical ventilation quality, and abnormal events of human-machine asynchrony, and at the same time push warning information of abnormal events of respiratory system mechanical changes in parallel.

[0161] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A mechanical ventilation analysis and early warning method, characterized in that: Includes the following steps: Step 1: Collect and analyze mechanical ventilation monitoring data in real time and perform data quality analysis. Step 2: If the monitoring data quality is qualified, then divide the respiratory cycle and divide the multi-condition intervals of triggering method, ventilation type, and control mode; Step 3: Based on the monitoring data after the multi-condition interval division, establish a multi-level decision model for automatic monitoring of life safety abnormal events, mechanical ventilation quality abnormal events, and human-machine asynchrony abnormal events; based on specific sub-conditions after the multi-condition interval division, monitor respiratory system mechanical changes in parallel according to the feature sequence of linear fitting. Step 4: Set up and push early warning information according to the alarm level of abnormal life safety events, abnormal mechanical ventilation quality events, and abnormal human-machine asynchrony events, and push early warning information in parallel for abnormal respiratory system mechanical changes events; The multi-condition interval division refers to establishing a multi-model integrated architecture and dividing the system according to triggering method, ventilation type, and control mode: S1: Based on the triggering method classification rules, the machine triggering identification module further classifies the monitoring data into machine triggering and patient triggering; S2: Based on the ventilation type classification rules, the control ventilation identification module further classifies the monitoring data into control ventilation, assisted ventilation and spontaneous ventilation; S3: Based on the control mode classification rules, the control identification module further classifies the monitoring data into capacity control type and pressure control type; The monitoring data includes pressure waveform data, volume waveform data, and flow velocity waveform data; Based on the volume waveform curve, a maximum and minimum value search algorithm is executed, and the positions of the maximum and minimum values ​​are used as the basis for interval division. The time series of pressure, volume, and flow rate within a unit monitoring time are divided into respiratory cycle segments. The resulting pressure time series are denoted as P1, P2, ..., P... t-2 P t-1 P t The pressure waveform data of the t-th respiratory cycle is denoted as in Let V1 be the pressure value corresponding to the nth data sampling point in the tth respiratory cycle; the divided volume time series are denoted as V1, V2, ..., V3. t-2 V t-1 V t The volume waveform data of the t-th respiratory cycle is denoted as in The volume value corresponding to the nth data sampling point in the t-th respiratory cycle; flow rate time series F1, F2, ..., F t-2 F t-1 F t The flow rate waveform data for the t-th respiratory cycle is denoted as in Insp represents the flow rate value corresponding to the nth data sampling point in the tth respiratory cycle; t Let t be the inspiratory time of the t-th respiratory cycle, Exp t Let t be the exhalation time of the t-th respiratory cycle; The multi-level decision-making model consists of a life safety abnormal event decision tree as the first layer, a mechanical ventilation quality abnormal event decision tree as the second layer, and a human-machine asynchrony abnormal event decision tree as the third layer. Within the decision tree, waveform data is selected for preferred monitoring. Through mathematical calculation formulas and fitting waveform curves, the function characteristics are judged. A complete logical judgment rule is established based on the rule threshold method. The waveform morphology characteristics and change patterns are correlated with the performance characteristics of abnormal events, thereby identifying and classifying specific abnormal events.

2. The mechanical ventilation analysis and early warning method according to claim 1, characterized in that: While acquiring monitoring data, data quality analysis is performed, including whether there is drift, movement or interference in the waveform data; whether there are abnormalities in the waveform data values; if there are data quality problems, an early warning information on data quality problems is directly pushed out and monitoring data continues to be collected and analyzed.

3. The mechanical ventilation analysis and early warning method according to claim 2, characterized in that: The respiratory cycle division is based on a maximum and minimum value search algorithm to identify the extreme values ​​of the monitoring data waveform curve and divide the respiratory intervals; the duration of the monitoring data is not less than 60 seconds.

4. The mechanical ventilation analysis and early warning method according to claim 3, characterized in that: The life safety abnormal events include air leakage events, airway circuit obstruction events, and gas trapping events; mechanical ventilation quality abnormal events include condensation in the tubing, suctioning, and CPR; periodic asynchrony events of human-machine asynchrony abnormal events include flow starvation, premature switching, delayed switching, and over-firing; frequency asynchrony events of human-machine asynchrony abnormal events include automatic triggering, invalid triggering, reverse triggering, and repeated triggering; the monitoring of internal events of the three major categories of abnormal events adopts a parallel monitoring method.

5. The mechanical ventilation analysis and early warning method according to claim 4, characterized in that: The monitoring of gas leakage events and gas trapping events in the aforementioned life safety anomalies specifically includes the following steps: Select capacity time series and flow rate time series for abnormal event identification: S1: First, based on the flow velocity time series F1, F2, ..., F... t-2 F t-1 F t The judgment is based on F. t Does the flow rate return to zero at the end of expiration? ; S2: If the flow rate value reaches zero at the end of expiration, then proceed to step S4; S3: Conversely, if the flow rate does not return to zero at the end of expiration, continue according to Exp. t =0 to determine: if Exp t If 0 is true, then proceed to step S4; otherwise, proceed to step S5. S4: Then, based on the capacity time series V1, V2, ..., V t-2 V t-1 V t The judgment is based on V. t Does it exhibit a vertical descent trend at the end of exhalation, and does it meet the following criteria? If both conditions are met, proceed to step S6; otherwise, terminate monitoring. S5: At least three respiratory cycles F t-2 F t-1 F t The end-expiratory flow rate did not return to zero. If this occurs, a gas trapping event may occur during mechanical ventilation, which will be classified as a life-threatening event. S6: At least three respiratory cycles V t-2 V t-1 V t If both conditions in S4 are met, an air leak may occur during mechanical ventilation, which will be classified as a life-threatening event.

6. The mechanical ventilation analysis and early warning method according to claim 4, characterized in that: The monitoring of pipeline condensate events in the aforementioned mechanical ventilation quality abnormality events specifically includes the following steps: Select flow velocity time series for abnormal event identification: First, based on the flow velocity time series F1, F2, ..., F... t-2 F t-1 F t Make a judgment: Calculate F during the expiratory phase t First-order difference F t ′ and second-order difference F t ";in , Second-order difference , The location and number of local maxima and local minima in the expiratory phase are determined based on the first-order and second-order differences. The flow velocity time series F1, F2, ..., F t-2 F t-1 F t Based on the fluctuation pattern of rising and then falling during the exhalation phase, the duration of each fluctuation is obtained by subtracting the adjacent minimum and maximum values. If the frequency of oscillation is greater than a preset rule threshold and the duration of each oscillation is less than the preset rule threshold, then there may be a condensation event in the mechanical ventilation equipment pipeline, which is determined to be an abnormal mechanical ventilation quality event.

7. The mechanical ventilation analysis and early warning method according to claim 4, characterized in that: The specific steps for determining human-machine asynchrony anomalies are as follows: S11: According to the definition rules of mechanical ventilation cycle and frequency, the mechanical ventilation cycle and frequency module further divides the monitoring data into cycle asynchrony or frequency asynchrony. If it is cycle asynchrony, the cycle asynchrony sub-decision tree is used to identify and classify cycle asynchrony abnormal events; if it is frequency asynchrony, the frequency asynchrony sub-decision tree is used to identify and classify frequency asynchrony abnormal events. S12: When the human-machine asynchrony index (AI) exceeds a set threshold, it is considered a serious human-machine asynchrony event. The system calculates the total number of respiratory cycles per unit time, including the total number of mechanical and spontaneous respiratory cycles, as well as the number of respiratory cycles with abnormal human-machine asynchrony events and the total number of respiratory cycles with invalid trigger events per unit time. When the human-machine asynchrony index calculated by the above formula exceeds the set threshold, a relevant alarm will be pushed. The monitoring of flow rate starvation anomaly and repetitive triggering anomaly events in the aforementioned human-machine asynchrony anomaly events specifically includes the following steps: Selecting pressure time series for abnormal event identification: S1: The sub-decision tree for the asynchronous execution cycle of the flow rate starvation abnormal event, i.e., step S2; the sub-decision tree for the asynchronous execution frequency of the repeated triggering abnormal event, i.e., step S5; S2: First, based on the pressure time series P1, P2, ..., P... t-2 P t-1 P t The judgment is based on the "concave" state of the pressure-time series during the intake phase. The waveform characteristics are converted into a mathematical expression: the P in the intake phase is calculated. t First-order difference P t ′ and second-order difference P t ";in , Second-order difference , ; S3: If the second difference of the inspiratory phase pressure time series is non-negative and the pressure waveform fitting curve approximately satisfies the characteristics of a concave function, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6. S4: If there are outlier points in the second difference of the inspiratory phase pressure time series, it indicates that the rate of increase of the inspiratory phase pressure value has changed. Using the outlier points as the dividing points, if the fitting curves of the two waveform segments approximately satisfy the convex function characteristics, then a flow starvation event may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchronous abnormal event, and the process will proceed to step S6. S5: If the expiratory time is less than half of the average inspiratory time, then repeated triggering events may occur during mechanical ventilation treatment. This will be judged as a human-machine asynchrony abnormality event, and the process will proceed to step S6. S6: The human-machine asynchrony index is calculated based on the AI ​​formula. When the value exceeds the set threshold, a relevant alarm is pushed.

8. The mechanical ventilation analysis and early warning method according to claim 4, characterized in that: The linearly fitted feature sequences are used to monitor respiratory system mechanical changes in parallel. This monitoring of respiratory system mechanical changes will be conducted independently and in parallel with abnormal life-threatening events, abnormal mechanical ventilation quality events, and abnormal patient-ventilator asynchrony events. The process includes the following steps: S1: Select the triggering method based on the multi-condition interval division of the multi-model integration architecture; S2: Select the ventilation type for control ventilation based on the multi-condition interval division of the multi-model integrated architecture; S3: Based on the multi-condition interval division of the multi-model integrated architecture, determine the capacity control type and pressure control type; S4: Perform adaptive change trend identification: S4_1: In capacity-controlled ventilation mode, the characteristic sequence representing compliance trend is denoted as: Plateau pressure during the m-th respiratory cycle Based on the position at the beginning of the expiratory phase, the average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop is taken as the plateau pressure value for this respiratory cycle. Positive end-expiratory pressure (PEP) in the m-th respiratory cycle m The peep value for this respiratory cycle is the average of all pressure sampling points within the stable trend range of the pressure waveform before the end of the expiratory phase, based on the position at which the expiratory phase ends. The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.

1. P Δ2 It is inversely proportional to the compliance value, and P Δ2 After linear fitting of the feature sequences, the fitting curve can reflect the changing trend of compliance; S4_2: In pressure-controlled ventilation mode, the characteristic sequence representing compliance trend, the effective inspiratory time sequence is denoted as: Δt=[Δt1,Δt2,...,Δt m ,...,Δt z ]; The sequence of peak velocity angles is denoted as: θ=[θ1,θ2,...,θ m ,...,θ z ]; Δt and θ are positively correlated with compliance values. After linearly fitting the Δt feature sequence and the θ feature sequence respectively, the fitted curve can reflect the changing trend of compliance. S5: Perform airway resistance change trend identification: S5_1: In volume-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance is denoted as: Peak pressure during the m-th respiratory cycle The peak pressure of this respiratory cycle is defined as the maximum value of all pressure sampling data points from the beginning of the inspiratory phase to the end of the expiratory phase. Plateau pressure during the m-th respiratory cycle The average value of all pressure sampling points within the range where the pressure waveform stabilizes after a sharp drop at the beginning of the expiratory phase is taken as the plateau pressure value for this respiratory cycle. The aforementioned stable trend interval is a continuous interval in the first-order difference value where the fluctuation range does not exceed 0.

1. P Δ1 It is directly proportional to the airway resistance value, and P Δ1 After linear fitting of the characteristic sequence, the fitting curve can reflect the changing trend of airway resistance. S5_2: In pressure-controlled ventilation mode, the characteristic sequence representing the trend of airway resistance, with the peak flow rate sequence denoted as: ； Peak flow rate in the m-th respiratory cycle The peak flow rate for this respiratory cycle is defined as the maximum value of all flow rate sampling points from the beginning of the inspiratory phase to the end of the expiratory phase. f peak It is inversely proportional to the airway resistance value, f peak After linear fitting of the characteristic sequence, the fitted curve can be used to reflect the changing trend of airway resistance; S6: Perform dynamic changes analysis of the respiratory system. If the trend curve shows the following, it is determined to be an abnormal event of respiratory system mechanical changes: In capacity-controlled ventilation mode, when the curve reflecting the change trend of compliance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment; In pressure-controlled ventilation mode, when the two trend curves reflecting compliance change per unit time show a continuous downward trend at a certain moment, and the decrease is 4% of the corresponding value at that moment; In volume-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous upward trend at a certain moment, and the increase is 8% of the corresponding value at that moment; In pressure-controlled ventilation mode, when the curve reflecting the change trend of airway resistance per unit time shows a continuous downward trend at a certain moment, and the decrease is 6% of the corresponding value at that moment.

9. An early warning system for implementing the mechanical ventilation analysis and early warning method according to any one of claims 1-8, characterized in that: include: The data acquisition module is used to collect and analyze mechanical ventilation monitoring data in real time and perform data quality analysis. The data preprocessing module, provided that the monitoring data quality is qualified, is used to divide the respiratory cycle and divide the multi-condition intervals of triggering method, ventilation type, and control mode. The event monitoring module is used to establish a multi-level decision model based on the monitoring data after the multi-condition interval division to automatically monitor abnormal events of life safety, abnormal events of mechanical ventilation quality, and abnormal events of human-machine asynchrony. Based on specific sub-conditions after multi-condition interval division, the mechanical changes of the respiratory system are monitored in parallel according to the feature sequence of linear fitting. The event warning module is used to push warning information according to the alarm level settings of abnormal events of life safety, abnormal events of mechanical ventilation quality, and abnormal events of human-machine asynchrony, and at the same time push warning information of abnormal events of respiratory system mechanical changes in parallel.

Images