A big data wisdom health respiratory dynamic monitoring method, system and terminal

Abstract

The present application relates to the technical field of respiratory monitoring, in particular to a big data intelligent health respiratory dynamic monitoring method, system and terminal. According to the change characteristics of the respiratory pressure signal, the time domain interval of each respiratory cycle is obtained; further, the work amount corresponding to each respiratory cycle is obtained, and the respiratory mode is divided; further, the latest seasonal cycle of the respiratory pressure signal is extracted, according to the change characteristics of the work amount of the time series adjacent respiratory cycles in the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles of the respiratory mode, the respiratory stability coefficient is obtained; further, when the respiratory stability coefficient is greater than or equal to the preset stability threshold, the predicted respiratory pressure signal is obtained and verified, and finally the predicted respiratory mode is obtained according to the final predicted respiratory pressure signal, the hysteresis of real-time processing is reduced, the response speed of respiratory support is improved, and the comfort is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of respiratory monitoring, in particular to a big data intelligent health respiratory dynamic monitoring method, system and terminal. BACKGROUND

[0002] In the prior art, respiratory behavior is usually achieved by monitoring key parameters such as respiratory frequency, tidal volume, airway pressure, minute ventilation, etc. Airflow pressure is generated during the respiratory process, which can reflect the respiratory attributes of the user such as respiratory state and respiratory mode. Respiratory monitoring masks are usually equipped with gas pressure sensors, which can detect the airway pressure during the patient's inhalation or exhalation process through the pressure probe installed in the airway.

[0003] In order to realize dynamic monitoring of respiratory behavior, real-time analysis of the current respiratory state of the user, and thus cater to the current respiratory state of the user and maintain the user's healthy breathing, the prior art generally relies on physiological indicators of the patient and monitoring respiratory work to identify respiratory patterns, such as blood oxygen saturation. However, physiological indicators such as blood oxygen saturation often show corresponding changes only after a period of respiratory state changes, and monitoring respiratory work requires a certain amount of respiratory data to be accumulated to determine respiratory pattern conversion, resulting in a lag in the respiratory monitoring mask determining respiratory patterns through monitoring respiratory work data. SUMMARY

[0004] In order to solve the technical problem of the lag in the respiratory monitoring mask determining respiratory patterns through monitoring respiratory work data, the purpose of the present application is to provide a big data intelligent health respiratory dynamic monitoring method, system and terminal, and the technical solution adopted is as follows:

[0005] A big data intelligent health respiratory dynamic monitoring method, the method comprising:

[0006] obtaining a respiratory pressure signal and a respiratory pressure value of a current patient; obtaining a time domain interval of each respiratory cycle of the respiratory pressure signal according to the change characteristics of the respiratory pressure signal;

[0007] obtaining the work amount corresponding to each respiratory cycle based on all the pressure values in each time domain interval; dividing the respiratory mode of each respiratory cycle based on the work amount; extracting the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm; obtaining a respiratory stability coefficient according to the change characteristics of the work amount of the time series adjacent respiratory cycles in the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles of the respiratory mode;

[0008] When the respiratory stability coefficient is greater than or equal to a preset stability threshold, a preset prediction model is used to obtain a predicted respiratory pressure signal; according to a difference feature of the respiratory pressure signal and the predicted respiratory pressure signal in the seasonal period, in combination with a difference of the respiratory stability coefficient of an adjacent seasonal period, it is determined whether to re-predict; when it is determined not to re-predict, a predicted respiratory mode is obtained according to the final predicted respiratory pressure signal.

[0009] Further, the method for obtaining the time domain interval comprises:

[0010] A trend curve of the respiratory pressure signal is extracted; a time domain between two adjacent minimum value points in the trend curve is divided into a time domain interval of one respiratory cycle.

[0011] Further, the method for obtaining the respiratory mode comprises:

[0012] When the work amount is in a first preset work amount interval, the corresponding respiratory cycle is marked as rapid shallow breathing, and the identification serial number is set as 01; when the work amount is in a second preset work amount interval, the corresponding respiratory cycle is marked as normal resting breathing, and the identification serial number is set as 02; when the work amount is in a third preset work amount interval, the corresponding respiratory cycle is marked as deep breathing, and the identification serial number is set as 03; when the work amount is in a fourth preset work amount interval, the corresponding respiratory cycle is marked as forced breathing, and the identification serial number is set as 04.

[0013] Further, the method for obtaining the respiratory stability coefficient comprises:

[0014] In the latest seasonal period, when the number of the respiratory cycles corresponding to a certain respiratory mode is the largest, the respiratory mode is marked as a main respiratory mode, and other respiratory modes are marked as secondary respiratory modes.

[0015] According to the number of the respiratory cycles corresponding to all the secondary respiratory modes in the seasonal period, in combination with the absolute value of the difference of the work amount of the time-sequentially adjacent respiratory cycles, a respiratory stability coefficient of the latest seasonal period is obtained; the number of the respiratory cycles corresponding to all the secondary respiratory modes and the absolute value of the difference of the work amount of the time-sequentially adjacent respiratory cycles are negatively correlated with the respiratory stability coefficient.

[0016] Further, the method for determining whether to re-predict comprises:

[0017] Obtain the time domain interval of each respiratory cycle of the predicted respiratory pressure signal; take the number of respiratory cycles in the latest seasonal cycle as the number of comparison cycles; extract the predicted respiratory pressure signal of the first number of comparison cycles in the time domain as the predicted comparison signal, and obtain the respiratory pattern of each respiratory cycle in the predicted comparison signal.

[0018] In the predicted comparison signal and the respiratory pressure signal within the seasonal cycle, a preset number of respiratory cycles with the same time sequence number are extracted to form a comparison binary; based on the difference in the identifier number of the respiratory pattern of the two respiratory cycles in each comparison binary, combined with the difference in the respiratory stability coefficient of adjacent seasonal cycles, the prediction confidence is obtained; the difference in the identifier number is negatively correlated with the prediction confidence.

[0019] When the prediction confidence is less than a preset confidence threshold, a re-prediction is determined and a new determination is made after the re-prediction; when the prediction confidence is greater than or equal to the preset confidence threshold, a re-prediction is not determined.

[0020] Furthermore, the method for obtaining the predicted breathing pattern includes:

[0021] In the final predicted respiratory pressure signal, the proportion of the number of respiratory cycles of each respiratory pattern to the total number of all respiratory cycles is taken as the weight of each respiratory pattern; the respiratory pattern with the largest weight is selected as the predicted respiratory pattern.

[0022] Furthermore, the method for obtaining the amount of work done includes:

[0023] The pressure value within each time interval is integrated, and the integrated value is used as the work done in the corresponding respiratory cycle.

[0024] Furthermore, the preset prediction model is the ARIMA model.

[0025] This invention also proposes a big data-driven intelligent health respiratory dynamic monitoring system, the system comprising:

[0026] Data acquisition module: acquires the current patient's respiratory pressure signal and respiratory pressure value; based on the variation characteristics of the respiratory pressure signal, acquires the time domain interval of each respiratory cycle of the respiratory pressure signal;

[0027] Respiratory analysis module: Based on all pressure values ​​within each time domain interval, obtain the work done for each respiratory cycle; classify the respiratory cycle into respiratory patterns based on the work done; extract the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm; obtain the respiratory stability coefficient based on the variation characteristics of the work done in the temporally adjacent respiratory cycles within the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles in the respiratory pattern.

[0028] Respiratory prediction module: When the respiratory stability coefficient is greater than or equal to a preset stability threshold, a predicted respiratory pressure signal is obtained using a preset prediction model; based on the difference characteristics between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, and combined with the difference in respiratory stability coefficients of adjacent seasonal cycles, it is determined whether to re-predict; when it is determined not to re-predict, a predicted respiratory pattern is obtained based on the final predicted respiratory pressure signal.

[0029] The present invention also proposes a big data-driven smart health respiratory dynamic monitoring terminal, the terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the steps of the big data-driven smart health respiratory dynamic monitoring method.

[0030] The present invention has the following beneficial effects:

[0031] This invention first acquires the current patient's respiratory pressure signal and respiratory pressure value, providing a data foundation. It then acquires the time-domain interval of each respiratory cycle of the respiratory pressure signal, dividing the signal into different respiratory cycles, which is more conducive to analyzing local signal features and identifying respiratory patterns. Furthermore, it acquires the work done in each respiratory cycle, providing a basis for classifying respiratory patterns. Finally, it classifies respiratory patterns based on work done in each respiratory cycle, facilitating a more comprehensive understanding of the patient's respiratory status. Finally, it extracts the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm, which can reflect the current state of the respiratory signal in real time, providing a basis for subsequent respiratory pattern analysis and stabilization. The invention provides a foundation for assessment and prepares for predictive signals. It further acquires a respiratory stability coefficient, characterizing the patient's respiratory stability within the seasonal cycle, providing a basis for determining whether subsequent respiratory prediction is possible. When the respiratory stability coefficient is greater than or equal to a preset stability threshold, a preset predictive model is used to obtain a predicted respiratory pressure signal, providing a data basis for reducing the lag in respiratory pattern determination by the respiratory monitoring mask, while ensuring nursing effectiveness. It further determines whether to re-predict, ensuring the model adapts to the current actual state at any time, improving the accuracy and reliability of prediction. Finally, when it is determined not to re-predict, the predicted respiratory pattern is obtained based on the final predicted respiratory pressure signal. This invention identifies the respiratory pattern of a single respiratory cycle, analyzes the stable characteristics of the patient's breathing within the seasonal cycle, and then obtains a predicted respiratory pressure signal to obtain a predicted respiratory pattern, reducing the lag in real-time processing, improving the response speed of respiratory support, and enhancing comfort. Attached Figure Description

[0032] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A flowchart illustrating a big data-driven smart health respiratory dynamic monitoring method provided in one embodiment of the present invention;

[0034] Figure 2 A pressure value waveform diagram provided in one embodiment of the present invention;

[0035] Figure 3 This is a flowchart of a method for determining whether to re-predict, provided as an embodiment of the present invention. Detailed Implementation

[0036] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a big data-driven intelligent health respiratory dynamic monitoring method, system, and terminal proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] The following description, in conjunction with the accompanying drawings, details the specific solution of the big data-driven intelligent health respiratory dynamic monitoring method, system, and terminal provided by this invention.

[0039] Please see Figure 1 The diagram illustrates a flowchart of a big data-driven smart health respiratory dynamic monitoring method according to an embodiment of the present invention, specifically including:

[0040] In this embodiment of the invention, the respiratory pressure signal of the patient is acquired in real time by a respiratory monitoring mask equipped with a piezoelectric sensor. The respiratory pressure signal is then divided into respiratory cycles and respiratory patterns. Based on the existing respiratory pressure signal, a predicted respiratory pressure signal is provided to the respiratory monitoring mask to predict the respiratory pattern, avoiding the lag of the existing determination method and adapting to the patient's respiratory changes in a timely manner.

[0041] Step S1: Obtain the current patient's respiratory pressure signal and respiratory pressure value; based on the characteristics of the respiratory pressure signal change, obtain the time domain interval of the respiratory pressure signal for each respiratory cycle.

[0042] In one embodiment of the present invention, a piezoelectric sensor is typically installed in the tubing connecting the airway and the respiratory monitoring mask to accurately monitor pressure fluctuations within the airway. The sensor's sensitivity parameters are: 1-10 pC / Pa (picoulons / Pascal); operating pressure range: 0-100 cmH2O (cm water column); frequency response: 0.1-500 Hz. It detects pressure changes within the airway in real time and converts them into electrical signals to obtain the current respiratory pressure signal and respiratory pressure value of the patient, providing a data basis.

[0043] Please see Figure 2 It shows a pressure value waveform provided by an embodiment of the present invention. Figure 2 The horizontal axis represents time, and the vertical axis represents pressure values.

[0044] Considering the obvious periodicity of human respiration, which causes significant periodic changes in respiratory pressure signals, obtaining the time domain interval of each respiratory cycle based on the changing characteristics of the respiratory pressure signal and dividing the respiratory pressure signal into different respiratory cycles is more conducive to analyzing the local features of the signal and identifying the breathing pattern.

[0045] Preferably, in one embodiment of the present invention, considering that the original pressure value waveform curve has a local sawtooth shape, which is not conducive to analyzing the periodic characteristics of the respiratory pressure signal, the trend curve of the respiratory pressure signal is extracted first, and the change characteristics of the respiratory pressure signal are analyzed by means of the trend curve; considering that each breath will cause the trend curve to rise or fall once, the time domain between two adjacent minimum points in the trend curve is divided into the time domain interval of a respiratory cycle. A respiratory cycle includes one exhalation and one inhalation. The change characteristics of the trend curve are reflected by the minimum value, which reflects the change characteristics of the respiratory pressure signal.

[0046] As an example, the trend curve of the respiratory pressure signal is extracted by the Singular Spectrum Analysis (SSA) algorithm, and then the minimum point is obtained to divide the time domain interval.

[0047] It should be noted that in other embodiments of the present invention, the implementer may also obtain the trend curve by polynomial fitting, which, along with the SSA algorithm and the minimum value acquisition method, are existing technologies and will not be described in detail here.

[0048] Step S2: Based on all pressure values ​​within each time domain interval, obtain the work done for each respiratory cycle; classify the respiratory pattern for each respiratory cycle based on the work done; extract the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm; obtain the respiratory stability coefficient based on the variation characteristics of the work done in temporally adjacent respiratory cycles within the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles of the respiratory pattern.

[0049] Considering the differences in respiratory characteristics across different respiratory cycles, and to gain a more comprehensive understanding of a patient's respiratory status, it is necessary to divide the respiratory cycle into respiratory modes. The amount of work done by all pressure values ​​within a respiratory cycle represents the work done by the respiratory monitoring mask to maintain normal ventilation for the patient. Therefore, based on all pressure values ​​within each time interval, the amount of work done for each respiratory cycle is obtained, providing a basis for dividing the respiratory modes.

[0050] Preferably, in one embodiment of the present invention, the pressure value in each time domain interval is integrated, and the integrated value is used as the work done in the corresponding respiratory cycle.

[0051] It should be noted that the integration method is already existing technology and will not be elaborated further.

[0052] Increased work is usually associated with the body needing more oxygen, increased airway resistance, or impaired lung function, while decreased work indicates shallow and rapid breathing, which is common in cases of anxiety or mild hypoxia. Therefore, breathing patterns can be classified for each respiratory cycle based on the amount of work done.

[0053] Preferably, in one embodiment of the present invention, when the amount of work done is within a first preset work interval, the corresponding respiratory cycle is marked as rapid shallow breathing and the identification number is set to 01; when the amount of work done is within a second preset work interval, the corresponding respiratory cycle is marked as normal resting breathing and the identification number is set to 02; when the amount of work done is within a third preset work interval, the corresponding respiratory cycle is marked as deep breathing and the identification number is set to 03; when the amount of work done is within a fourth preset work interval, the corresponding respiratory cycle is marked as forced breathing and the identification number is set to 04.

[0054] As an example, the average work done under normal breathing is set to 0.5 to obtain a normalized baseline value. The work done in each respiratory cycle is linearly normalized, and work values ​​exceeding twice the average work done under normal breathing are set to 1. The first preset work interval is set to [0, 0.4], the second preset work interval to (0.4, 0.6], the third preset work interval to (0.6, 0.8], and the fourth preset work interval to (0.8, 1). The identifier number is the symbol for the breathing pattern, which is convenient for subsequent analysis of changes in the breathing pattern.

[0055] Normal breathing refers to breathing performed by the patient at normal rest. Relevant personnel pre-determine and measure multiple respiratory cycles under normal breathing conditions, such as 100 cycles (which may involve multiple patients, such as 10 cycles per patient, involving 10 patients), and calculate the average amount of work done per respiratory cycle.

[0056] Respiratory pressure signals may exhibit non-stationary characteristics due to factors such as patient condition, external environment, and equipment changes. The STL (Seasonal-Trend decomposition using Loess) algorithm has good tolerance for noise and outliers, and can extract stable seasonal cycles from the signal, avoiding the influence of short-term interference data.

[0057] Extracting the latest seasonal cycle of respiratory pressure signals based on the STL algorithm can reflect the current state of respiratory signals in real time, providing a basis for subsequent respiratory pattern analysis and stability assessment, and preparing for signal prediction.

[0058] In another embodiment of the present invention, the implementer may also set the length of the seasonal cycle with a preset prediction time interval, for example, the preset prediction time interval is 10 minutes, and the length of the seasonal cycle is fixed at 10 minutes.

[0059] Considering that the changes in the amount of work done in adjacent respiratory cycles can accurately reflect short-term dynamic fluctuations, and that the quantitative distribution characteristics of respiratory patterns can comprehensively assess long-term trends and respiratory patterns, a respiratory stability coefficient is obtained based on the changes in the amount of work done in temporally adjacent respiratory cycles within the seasonal cycle, combined with the quantitative distribution characteristics of respiratory cycles in the respiratory pattern. This coefficient characterizes the respiratory stability characteristics of patients within the seasonal cycle and provides a basis for determining whether respiratory prediction can be performed subsequently.

[0060] Preferably, in one embodiment of the present invention, considering that the breathing pattern that appears most frequently in the latest seasonal cycle usually represents the patient's normal breathing state or the most common breathing behavior, in the latest seasonal cycle, when a certain breathing pattern corresponds to the most breathing cycles, this breathing pattern is marked as the primary breathing pattern, and other breathing patterns are marked as secondary breathing patterns.

[0061] Based on the number of respiratory cycles corresponding to all secondary respiratory patterns within the seasonal cycle, and combined with the absolute value of the difference in the amount of work done by temporally adjacent respiratory cycles, the latest respiratory stability coefficient for the seasonal cycle is obtained.

[0062] Considering that the number of secondary breathing patterns directly reflects the degree of unstable breathing, and the difference in work done between breathing cycles reflects the dynamic fluctuation of breathing behavior and also reflects the degree of unstable breathing, the absolute value of the difference between the number of breathing cycles corresponding to all secondary breathing patterns and the work done by sequentially adjacent breathing cycles is negatively correlated with the respiratory stability coefficient.

[0063] As an example, the formula for calculating the respiratory stability coefficient includes:

[0064] ;

[0065] Where Sta represents the respiratory stability coefficient of the latest seasonal cycle; norm{} represents the linear normalization function; r represents the number of respiratory cycles corresponding to all minor respiratory patterns within the seasonal cycle; C0 represents the positive parameter divided by zero, in this example C0=0.001; M represents the number of respiratory cycles within the seasonal cycle; m represents the sequence number of the respiratory cycle within the seasonal cycle; w m w represents the amount of work done in the m-th respiratory cycle; m+1 This represents the amount of work done in the (m+1)th respiratory cycle; || represents taking the absolute value.

[0066] In the formula for calculating the respiratory stability coefficient, the reciprocal of r and |w is used. m -w m+1| Perform negative correlation mapping; represent the distribution characteristics of the number of respiratory cycles of the respiratory patterns by the number of respiratory cycles corresponding to all secondary respiratory patterns. The larger the r, the more respiratory cycles of the secondary respiratory patterns, reflecting more switching of respiratory patterns and stronger respiratory instability, and a smaller Sta; represent the change characteristics of work done by the absolute value of the difference, |w m -w m+1 The larger the value of Sta, the greater the change in the amount of work done in adjacent respiratory cycles, the more unstable the respiratory work, the stronger the respiratory instability, and the smaller Sta is.

[0067] Meanwhile, the patient's respiratory work over multiple cycles actually reflects the level of the patient's respiratory compensation ability. m -w m+1 The smaller the value, the higher the patient's respiratory compensation ability. m -w m+1 The larger the value, the lower the patient's respiratory compensation ability.

[0068] It should be noted that in other embodiments of the present invention, the implementer may also perform negative correlation mapping using a negative correlation mapping function such as an exponential function exp(-x) with the natural constant e as the base, where x represents the independent variable; when the number of respiratory cycles of two or more respiratory modes is the highest at the same time, the respiratory mode with the highest proportion of time is selected as the primary respiratory mode.

[0069] Step S3: When the respiratory stability coefficient is greater than or equal to the preset stability threshold, the predicted respiratory pressure signal is obtained using the preset prediction model; based on the difference characteristics between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, and combined with the difference in the respiratory stability coefficient of adjacent seasonal cycles, it is determined whether to re-predict; when it is determined not to re-predict, the predicted respiratory pattern is obtained based on the final predicted respiratory pressure signal.

[0070] Considering that a larger respiratory stability coefficient indicates a more stable respiratory state and that the changes in respiratory pressure signals have a stronger regularity, the prediction model can more accurately fit these regularities and output reliable prediction results. Therefore, when the respiratory stability coefficient is greater than or equal to the preset stability threshold, the preset prediction model is used to obtain the predicted respiratory pressure signal.

[0071] As an example, the preset stability threshold is 0.8, and the preset prediction model is the ARIMA model. The existing respiratory pressure signal data of the [1,M] respiratory cycle is input into the ARIMA model, and the predicted respiratory pressure signal is output.

[0072] It should be noted that when the respiratory stability coefficient is less than the preset stability threshold, it indicates that the patient's current respiratory status is volatile and requires timely professional monitoring and management by medical personnel. Therefore, when the respiratory stability coefficient is less than 0.8, feedback is sent to relevant personnel, and no prediction is made to ensure safety and treatment effectiveness.

[0073] It should be noted that the embodiments of the present invention do not directly predict the work done by breathing because the work done is obtained by integrating the pressure signal in each respiratory cycle. The respiratory pressure signal contains more details about respiratory behavior, and the respiratory pressure signal usually has obvious periodicity and regularity, making it more suitable for modeling and prediction using time series analysis methods.

[0074] Using the ARIMA model to predict time-series signals is an existing technology. In other embodiments of this invention, the implementer may also use other methods such as training a long short-term memory network for prediction, which will not be elaborated further.

[0075] Considering that respiratory status is relatively stable during seasonal cycles, the predicted respiratory pressure signal should be similar to the actual respiratory pressure signal. Therefore, the difference between the respiratory pressure signal and the predicted respiratory pressure signal within a seasonal cycle reflects the accuracy of the prediction. The difference in respiratory stability coefficient between adjacent seasonal cycles also reflects the changes in the patient's respiratory compensation ability, which also reflects the accuracy of the prediction. Therefore, based on the difference between the respiratory pressure signal and the predicted respiratory pressure signal within a seasonal cycle, combined with the difference in respiratory stability coefficient between adjacent seasonal cycles, it is determined whether to re-predict, ensuring that the model adapts to the current actual state at any time, avoiding the accumulation of errors, and improving the accuracy and reliability of the prediction.

[0076] Preferably, in one embodiment of the present invention, the method for determining whether to re-predict includes:

[0077] Please see Figure 3 The diagram illustrates a flowchart of a method for determining whether to re-predict, provided by an embodiment of the present invention, specifically including:

[0078] Step S301: Obtain the time domain interval of each respiratory cycle of the predicted respiratory pressure signal; take the number of respiratory cycles in the latest seasonal cycle as the number of comparison cycles; extract the predicted respiratory pressure signal of the first number of comparison cycles in the time domain as the predicted comparison signal, and obtain the respiratory pattern of each respiratory cycle in the predicted comparison signal.

[0079] Considering that when the difference between the respiratory pressure signal and the predicted respiratory pressure signal is small within the seasonal cycle, the respiratory patterns of respiratory cycles with the same time sequence are relatively consistent. Therefore, the predicted respiratory pressure signal corresponding to the number of respiratory cycles before the time domain comparison cycle is extracted within the predicted respiratory pressure signal, and the respiratory pattern of each respiratory cycle is obtained. Thus, based on the respiratory pressure signal of the existing respiratory cycle with time sequence number [1,M], the predicted respiratory pressure signal of the respiratory cycle with time sequence number [1,M] is obtained, which is convenient for subsequent comparison.

[0080] Meanwhile, the predicted respiratory pressure signal is the current predicted signal, so the predicted comparison signal can also be regarded as the respiratory pressure signal of the current [M,2M] respiratory cycle.

[0081] It should be noted that the time domain interval and breathing pattern of the predicted respiratory pressure signal are obtained in the same way as those of the actual respiratory pressure signal, and will not be repeated here.

[0082] Step S302: From the predicted comparison signal and the respiratory pressure signal within the seasonal cycle, extract a preset number of respiratory cycles with the same time sequence number to form a comparison binary; based on the difference in the identification sequence number of the respiratory pattern of the two respiratory cycles in each comparison binary, and combined with the difference in the respiratory stability coefficient of adjacent seasonal cycles, obtain the prediction confidence.

[0083] Considering the large computational load of comparing respiratory patterns of all respiratory cycles with the same time sequence number, this embodiment of the invention selects a portion for comparison. As an example, the preset quantity is 10. Using random sampling, 10 respiratory cycles are randomly selected from the latest seasonal respiratory pressure signal. At the same time, according to the time sequence number of the selected respiratory cycles, 10 respiratory cycles are selected from the predicted comparison signal. Respiratory cycles with the same time sequence number are formed into comparison pairs, resulting in 10 comparison pairs.

[0084] While classifying respiratory patterns in the respiratory cycle, a serial number was assigned to each pattern, allowing for direct comparison of differences in respiratory patterns. Furthermore, considering that patients' respiratory function can vary, the greater the difference between the respiratory stability coefficient of the latest seasonal cycle and that of the adjacent historical seasonal cycle, the greater the change in the patient's respiratory stability and the greater the change in respiratory compensation capacity, resulting in lower predictive reliability. Therefore, this method was used to obtain predictive reliability.

[0085] As an example, the formula for calculating prediction confidence includes:

[0086] ;

[0087] Where Tr represents the prediction confidence of the latest prediction; norm{} represents the linear normalization function; K represents the preset number, which is also the number of comparison pairs; k is the index of the comparison pair; c k This indicates the identifier number of the respiratory pattern of the respiratory cycle in the k-th comparison pair; represents the identifier number of the respiratory pattern of the predicted contrast signal in the k-th contrast pair; || represents taking the absolute value; ΔSta represents the absolute value of the difference between the respiratory stability coefficient of the latest seasonal cycle and the respiratory stability coefficient of the adjacent historical seasonal cycle.

[0088] The formula for calculating prediction confidence uses the absolute value of the difference to represent the difference in the identifiers of the respiratory patterns between two respiratory cycles in the comparison pair. Then, the summation of all differences across the comparison pairs reflects the differences between the predicted and predicted respiratory pressure signals. The larger the value, the greater the difference between the respiratory pressure signal and the predicted respiratory pressure signal, the less ideal the prediction effect, and the lower the prediction reliability; the larger the value, the greater the patient's respiratory variation, and the lower the prediction reliability.

[0089] It should be noted that random sampling is an existing technology. In other embodiments of the present invention, the implementer may also use sampling methods such as equidistant sampling; all respiratory cycles with the same time sequence number may also be compared, which will not be elaborated further.

[0090] Step S303: When the prediction confidence is less than the preset confidence threshold, determine to re-predict and re-determine after re-prediction; when the prediction confidence is greater than or equal to the preset confidence threshold, determine not to re-predict.

[0091] As an example, the preset confidence threshold is 0.3. The prediction is repeated iteratively until the confidence of the prediction corresponding to the latest prediction is greater than 0.3. At this time, it is determined not to re-predict. The final predicted respiratory pressure signal is obtained, which provides a basis for the respiratory monitoring mask to pre-determine the respiratory mode and reduces the lag in the respiratory monitoring mask's determination of the respiratory mode.

[0092] It should be noted that when prediction is performed continuously, the current prediction interval is the time domain length of the prediction comparison signal of the final predicted respiratory pressure signal. For example, if the time domain length of the corresponding prediction comparison signal is 5 minutes for the first prediction, the second prediction will be performed after 5 minutes; if the time domain length of the corresponding prediction comparison signal is 4 minutes for the second prediction, the third prediction will be performed after 4 minutes. When prediction is interrupted, i.e., when the respiratory stability coefficient is less than the preset stability threshold, the prediction interval is set to 10 minutes. After 10 minutes, the latest seasonal cycle is obtained again, and the respiratory stability coefficient is used to determine whether to perform prediction.

[0093] Finally, when it is determined not to re-predict, the respiratory monitoring mask determines the breathing pattern based on the final predicted respiratory pressure signal, thereby obtaining the predicted breathing pattern and reducing the lag in the respiratory monitoring mask's determination of the breathing pattern.

[0094] Preferably, in one embodiment of the present invention, in the prediction comparison signal of the final predicted respiratory pressure signal, the proportion of the number of respiratory cycles of each respiratory mode to the total number of all respiratory cycles is taken as the weight of each respiratory mode; the respiratory mode with the largest weight is selected as the predicted respiratory mode.

[0095] In another embodiment of the present invention, after obtaining the predicted breathing pattern, the ventilator is adjusted using a prediction contrast signal feedback, specifically including:

[0096] In the final predicted respiratory pressure signal, the proportion of the number of respiratory cycles of each respiratory mode to the total number of respiratory cycles is taken as the weight of each respiratory mode.

[0097] When the primary breathing mode is rapid shallow breathing or normal resting breathing, and the combined proportion of deep breathing and forced breathing is less than the first preset proportion threshold, the ventilator is set to S mode.

[0098] When the primary breathing mode is rapid shallow breathing or normal resting breathing, and the combined proportion of deep breathing and forced breathing is greater than or equal to the first preset proportion threshold, the ventilator is set to S / T mode.

[0099] When the primary breathing mode is deep breathing and the proportion of forced breathing is less than the second preset threshold, the ventilator is set to S / T mode.

[0100] When the primary breathing mode is deep breathing and the proportion of forced breathing is greater than or equal to the second preset threshold, the ventilator is set to CACP mode.

[0101] When the primary breathing mode is forced breathing and the proportion of forced breathing is less than the third preset proportion threshold, the ventilator is set to CACP mode.

[0102] When the primary breathing mode is forced breathing and the proportion of forced breathing is greater than or equal to the third preset proportion threshold, the ventilator is set to T mode.

[0103] In other embodiments of the present invention, the implementer may also discard the third preset weight threshold and set the ventilator to T mode when the main breathing mode is forced breathing.

[0104] As an example, the first preset weight threshold is 15%, the second preset threshold is 30%, and the third preset weight threshold is 50%.

[0105] It should be noted that the S mode, S / T mode, CACP mode and T mode of the ventilator are existing technologies, and the implementer can adjust the first preset weight threshold, the second preset weight threshold and the third preset weight threshold by himself.

[0106] It should be noted that when using a ventilator with an adaptive operating mode that can adjust its working mode based on the respiratory pressure signal, the final predicted respiratory pressure signal can also be input, allowing the ventilator to acquire the respiratory pressure signal in advance and make adjustments, reducing the lag in real-time processing, improving the response speed of respiratory support, and enhancing comfort.

[0107] One embodiment of the present invention also provides a big data-driven intelligent health respiratory dynamic monitoring system, specifically comprising:

[0108] Data acquisition module: acquires the current patient's respiratory pressure signal and respiratory pressure value; based on the characteristics of respiratory pressure signal changes, acquires the time domain interval of the respiratory pressure signal for each respiratory cycle;

[0109] Respiratory Analysis Module: Based on all pressure values ​​within each time domain interval, obtain the work done for each respiratory cycle; classify the respiratory pattern for each respiratory cycle based on the work done; extract the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm; obtain the respiratory stability coefficient based on the variation characteristics of the work done in temporally adjacent respiratory cycles within the seasonal cycle, combined with the number distribution characteristics of respiratory cycles in the respiratory pattern.

[0110] Respiratory prediction module: When the respiratory stability coefficient is greater than or equal to the preset stability threshold, the preset prediction model is used to obtain the predicted respiratory pressure signal; based on the difference characteristics between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, and combined with the difference in respiratory stability coefficient between adjacent seasonal cycles, it is determined whether to re-predict; when it is determined not to re-predict, the predicted respiratory pattern is obtained based on the final predicted respiratory pressure signal.

[0111] An embodiment of the present invention also provides a big data-driven smart health respiratory dynamic monitoring terminal. The module includes a memory, a processor, and a computer program. The memory is used to store the corresponding computer program, and the processor is used to run the corresponding computer program. When the computer program runs in the processor, it can implement the big data-driven smart health respiratory dynamic monitoring method described in steps S1-S3.

[0112] In summary, addressing the technical problem of existing respiratory monitoring masks relying on monitoring respiratory work data to determine breathing patterns with lag, this invention obtains the time domain interval of each respiratory cycle based on the changing characteristics of the respiratory pressure signal; further, it obtains the work done for each respiratory cycle to classify the breathing pattern; further, it extracts the latest seasonal cycle of the respiratory pressure signal, and obtains a respiratory stability coefficient based on the changing characteristics of the work done in temporally adjacent respiratory cycles within the seasonal cycle, combined with the distribution characteristics of the number of respiratory cycles in the breathing pattern; further, when the respiratory stability coefficient is greater than or equal to a preset stability threshold, it uses a preset prediction model to obtain a predicted respiratory pressure signal; based on the difference between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, it determines whether to re-predict; finally, when it is determined not to re-predict, the predicted breathing pattern is obtained based on the final predicted respiratory pressure signal.

[0113] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0114] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A big data-driven intelligent health respiratory dynamic monitoring method, characterized in that, The method includes: Acquire the current patient's respiratory pressure signal and respiratory pressure value; based on the variation characteristics of the respiratory pressure signal, obtain the time domain interval of the respiratory pressure signal for each respiratory cycle; Based on all pressure values ​​within each time domain interval, the work done for each respiratory cycle is obtained; based on the work done, each respiratory cycle is divided into respiratory patterns; the latest seasonal cycle of the respiratory pressure signal is extracted based on the STL algorithm; and based on the variation characteristics of the work done in the time-adjacent respiratory cycles within the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles in the respiratory pattern, a respiratory stability coefficient is obtained. When the respiratory stability coefficient is greater than or equal to a preset stability threshold, a predicted respiratory pressure signal is obtained using a preset prediction model; based on the difference characteristics between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, and combined with the difference in respiratory stability coefficients of adjacent seasonal cycles, it is determined whether to re-predict; when it is determined not to re-predict, a predicted respiratory pattern is obtained based on the final predicted respiratory pressure signal.

2. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 1, characterized in that, The method for obtaining the time domain interval includes: Extract the trend curve of the respiratory pressure signal; divide the time domain between two adjacent minimum points in the trend curve into a time domain interval of one respiratory cycle.

3. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 1, characterized in that, The method for obtaining the breathing pattern includes: When the amount of work performed is within the first preset work range, the corresponding respiratory cycle is marked as rapid shallow breathing, and the identifier number is set to 01; when the amount of work performed is within the second preset work range, the corresponding respiratory cycle is marked as normal resting breathing, and the identifier number is set to 02; when the amount of work performed is within the third preset work range, the corresponding respiratory cycle is marked as deep breathing, and the identifier number is set to 03; when the amount of work performed is within the fourth preset work range, the corresponding respiratory cycle is marked as forced breathing, and the identifier number is set to 04.

4. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 3, characterized in that, The method for obtaining the respiratory stability coefficient includes: Within the latest seasonal cycle, when the number of respiratory cycles corresponding to a certain breathing pattern is the largest, this breathing pattern is marked as the primary breathing pattern, and the other breathing patterns are marked as secondary breathing patterns. The latest respiratory stability coefficient for the seasonal cycle is obtained by combining the number of respiratory cycles corresponding to all secondary respiratory patterns within the seasonal cycle with the absolute value of the difference in the amount of work done by the time-adjacent respiratory cycles. The number of respiratory cycles corresponding to all secondary respiratory patterns and the absolute value of the difference in the amount of work done by the time-adjacent respiratory cycles are both negatively correlated with the respiratory stability coefficient.

5. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 4, characterized in that, The method for determining whether to re-predict includes: Obtain the time domain interval of each respiratory cycle of the predicted respiratory pressure signal; take the number of respiratory cycles in the latest seasonal cycle as the number of comparison cycles; extract the predicted respiratory pressure signal of the first number of comparison cycles in the time domain as the predicted comparison signal, and obtain the respiratory pattern of each respiratory cycle in the predicted comparison signal. In the predicted comparison signal and the respiratory pressure signal within the seasonal cycle, a preset number of respiratory cycles with the same time sequence number are extracted to form a comparison binary; based on the difference in the identifier number of the respiratory pattern of the two respiratory cycles in each comparison binary, combined with the difference in the respiratory stability coefficient of adjacent seasonal cycles, the prediction confidence is obtained; the difference in the identifier number is negatively correlated with the prediction confidence. When the prediction confidence is less than a preset confidence threshold, a re-prediction is determined and a new determination is made after the re-prediction; when the prediction confidence is greater than or equal to the preset confidence threshold, a re-prediction is not determined.

6. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 5, characterized in that, The method for obtaining the predicted breathing pattern includes: In the final predicted respiratory pressure signal, the proportion of the number of respiratory cycles of each respiratory pattern to the total number of all respiratory cycles is taken as the weight of each respiratory pattern; the respiratory pattern with the largest weight is selected as the predicted respiratory pattern.

7. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 1, characterized in that, The method for obtaining the amount of work done includes: The pressure value within each time interval is integrated, and the integrated value is used as the work done in the corresponding respiratory cycle.

8. The method for dynamic respiratory monitoring in big data-driven smart health according to claim 1, characterized in that, The preset prediction model is the ARIMA model.

9. A big data-driven intelligent health respiratory dynamic monitoring system, the system comprising: Data acquisition module: acquires the current patient's respiratory pressure signal and respiratory pressure value; based on the changing characteristics of the respiratory pressure signal, acquires the time domain interval of each respiratory cycle of the respiratory pressure signal; Respiratory analysis module: Based on all pressure values ​​within each time domain interval, obtain the work done for each respiratory cycle; classify the respiratory cycle into respiratory patterns based on the work done; extract the latest seasonal cycle of the respiratory pressure signal based on the STL algorithm; obtain the respiratory stability coefficient based on the variation characteristics of the work done in the temporally adjacent respiratory cycles within the seasonal cycle, combined with the number distribution characteristics of the respiratory cycles in the respiratory pattern. Respiratory prediction module: When the respiratory stability coefficient is greater than or equal to a preset stability threshold, a predicted respiratory pressure signal is obtained using a preset prediction model; based on the difference characteristics between the respiratory pressure signal and the predicted respiratory pressure signal within the seasonal cycle, and combined with the difference in respiratory stability coefficients between adjacent seasonal cycles, it is determined whether to re-predict. When it is determined not to re-predict, the predicted breathing pattern is obtained based on the final predicted breathing pressure signal.

10. A big data-driven intelligent health respiratory dynamic monitoring terminal, the terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the big data smart health respiratory dynamic monitoring method as described in any one of claims 1 to 8.

Images