Automatic flow feedback adjusting method and system for breathing machine

Abstract

The invention relates to the technical field of flow regulation of breathing machines, in particular to an automatic flow feedback regulation method and system for a breathing machine. The method comprises the following steps: firstly, in to-be-analyzed historical data, carrying out linear fitting on data of target positions of different respiratory cycles, and obtaining a respiratory characteristic value and credibility of each data point based on a fitting result; further obtaining the correction credibility of the corresponding data point according to the correlation characteristics of the respiration intensity sequence and the distance sequence in combination with the credibility; further obtaining a distinguishing capability coefficient of the target position by combining correction credibility according to difference characteristics of the data points of the target position in different historical respiration data respiration characteristic values and distribution difference characteristics of the respiration characteristic values of the data points in different ventilation modes; further combining the distinguishing capability coefficient and the distinguishing degree to obtain a final principal component of the target mode; and finally, determining the current ventilation mode based on the final principal components of all the ventilation modes.

Description

Technical Field

[0001] The present invention relates to the technical field of ventilator flow rate regulation, and in particular to an automatic flow rate feedback regulation method and system for a ventilator. Background Art

[0002] The appropriate ventilator is a kind of effective ventilation for the patient's lungs in an appropriate way, which not only guarantees the patient's life needs, but also reduces complications as much as possible, and is safe and comfortable. The automatic feedback adjustment of the flow control of the ventilator is an intelligent ventilation mode, which realizes the set tidal volume by automatically adjusting the inspiratory pressure of the ventilator, and at the same time reduces the airway pressure as much as possible to reduce the pressure injury of positive pressure mechanical ventilation.

[0003] Some ventilators can automatically determine the appropriate ventilation mode based on the patient's respiratory data by extracting respiratory data features through historical data training. However, due to the normal differences in respiratory data corresponding to different patients and different monitoring time periods, the principal components obtained by principal component analysis cannot truly reflect the characteristics of respiratory data, and the obtained principal components cannot be used as the main basis for distinguishing ventilation modes, which in turn leads to inaccurate automatic feedback adjustment of the ventilator's flow rate. Summary of the invention

[0004] In order to solve the technical problem that the automatic feedback adjustment of the flow rate of a ventilator is inaccurate due to the inaccurate selection of the main component, the purpose of the present invention is to provide a method and system for automatic feedback adjustment of the flow rate of a ventilator. The technical solution adopted is as follows:

[0005] A method for automatic flow feedback regulation of a ventilator, the method comprising:

[0006] Acquire current respiratory data and historical respiratory data; the historical respiratory data contains a corresponding ventilation mode; all respiratory data contain the same number of respiratory cycles, and the number of data points in each respiratory cycle is the same; select any of the historical respiratory data as the historical data to be analyzed; the ventilation mode corresponding to the historical data to be analyzed is the target mode; select any position in the respiratory cycle as the target position;

[0007] In the historical data to be analyzed, the data of the target position of different respiratory cycles are fitted with a straight line to obtain a distance sequence composed of the distance between each data point and the fitting straight line; according to the violent fluctuation characteristics of the data in the distance sequence, the credibility of each data point is obtained; based on the parameters of the fitting straight line, the respiratory characteristic value of each data point is obtained; according to the peak value in each respiratory cycle and the data change trend on both sides of the adjacent peak value in the historical data to be analyzed, the respiratory intensity of each respiratory cycle is obtained; according to the correlation characteristics of the respiratory intensity sequence composed of the respiratory intensity of the respiratory cycle and the distance sequence, combined with the credibility of the corresponding data point, the corrected credibility of the corresponding data point is obtained;

[0008] In the target mode, according to the difference characteristics of the respiratory characteristic values ​​of the data points of the target position of different historical respiratory data, combined with the corrected credibility, the discrimination degree of the target position is obtained; according to the distribution difference characteristics of the respiratory characteristic values ​​of the target position under different ventilation modes, combined with the discrimination degree, the discrimination ability coefficient of the target position is obtained;

[0009] Perform principal component analysis on the historical breathing data under the target mode, and obtain the final principal component of the target mode by combining the discrimination coefficient and the discrimination degree; determine the current ventilation mode based on the difference characteristics between the principal component of the current breathing data and the final principal component of each ventilation mode.

[0010] Furthermore, the method for obtaining the credibility includes:

[0011] The fluctuation intensity coefficient is obtained at least according to the variance and the average value of the elements in the distance sequence; the variance and the average value of the elements in the distance sequence are both positively correlated with the fluctuation intensity coefficient;

[0012] The fluctuation intensity coefficient is negatively correlated and normalized to serve as the credibility of each data point at the target position in the historical data to be analyzed.

[0013] Furthermore, the method for obtaining the breathing characteristic value includes:

[0014] The Euclidean norm of the slope and intercept of the fitting straight line is used as the respiratory characteristic value of each data point at the target position in the historical data to be analyzed.

[0015] Furthermore, the method for obtaining the breathing intensity includes:

[0016] In each of the respiratory cycles in the historical data to be analyzed, the peak value of the respiratory data is extended to both sides to obtain extended data, and the duration of the sequence formed by the extended data and the peak value is used as the peak duration; the extended data is greater than a preset ratio of the peak value of the respiratory data;

[0017] The respiratory intensity of each respiratory cycle is obtained according to the peak value of the respiratory data in each respiratory cycle and the peak duration; the peak value of the respiratory data and the peak duration are both positively correlated with the respiratory intensity.

[0018] Furthermore, the method for obtaining the modified credibility includes:

[0019] At the target position in the historical data to be analyzed, the correlation coefficient between the respiratory intensity sequence and the distance sequence is obtained; the sum of the correlation coefficient and 1 is used as a correction coefficient, and the product of the correction coefficient and the credibility is used as the corrected credibility of the corresponding data point at the target position.

[0020] Furthermore, the method for obtaining the discrimination degree includes:

[0021] In the target mode, any two of the historical breathing data are taken as a historical data group; the absolute value of the difference between the breathing characteristic values ​​of the data points at the target position in the historical data group is taken as the first numerator, the sum of the corrected credibility is taken as the first denominator, and the ratio of the first numerator to the first denominator is taken as the breathing difference parameter corresponding to the historical data group;

[0022] According to the overall characteristics of the breathing difference parameters of all the historical data groups of the target pattern, the discrimination degree of the target position of the target pattern is obtained; the overall characteristics of the breathing difference parameters and the discrimination degree are negatively correlated.

[0023] Furthermore, the method for obtaining the discrimination ability coefficient includes:

[0024] When the discrimination degree of the target position is less than or equal to a preset discrimination threshold, the discrimination ability coefficient of the target position is set to zero;

[0025] When the discrimination degree of the target position is greater than a preset discrimination threshold: based on the statistical characteristics of the respiratory characteristic values ​​of all the historical respiratory data under each ventilation mode, obtaining the empirical distribution function of each ventilation mode; selecting any other ventilation mode except the target mode as a target comparison mode; and forming a mode binary with the target mode and the target comparison mode;

[0026] Obtain the extreme value of the respiratory characteristic value of the data point at the target position for all the historical respiratory data in each ventilation mode in the pattern binary group; use the union of the ranges of the extreme values ​​corresponding to the two ventilation modes in the pattern binary group as the comparison range of the pattern binary group at the target position; within the comparison range at the target position, normalize the area between the empirical distribution functions of the two ventilation modes in the pattern binary group, and use the normalized result as the discrimination sub-coefficient of the pattern binary group;

[0027] The minimum value of all the discrimination sub-coefficients corresponding to the target position of the target pattern is used as the discrimination capability coefficient of the target position of the target pattern.

[0028] Furthermore, the method for obtaining the final principal component includes:

[0029] Determine all principal components according to the result of the principal component analysis, and determine the weight of each element in the eigenvector corresponding to each principal component in the entire eigenvector; in the target mode, take the sum of the weight of each element in the eigenvector corresponding to each principal component, the discrimination degree of the original data point corresponding to the element in the eigenvector and the product of the discrimination coefficient as the principal component importance of each principal component;

[0030] The principal component with the greatest importance is selected according to the preset variance contribution rate to obtain the final principal component of the target pattern.

[0031] Furthermore, the method for obtaining the current ventilation mode includes:

[0032] Obtaining a transformation matrix of the final principal component of each ventilation mode; obtaining the principal component of the current respiratory data under the target mode based on the transformation matrix of the target mode and the current respiratory data; taking the maximum Euclidean distance of data points between the final principal component of the target mode and the principal component of the current respiratory data under the target mode as the matching distance between the current respiratory data and the target mode;

[0033] The ventilation mode having the smallest matching distance between the current respiratory data and all the ventilation modes is selected as the current ventilation mode.

[0034] The present invention also proposes an automatic flow feedback regulation system for a ventilator, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements any one of the steps of the automatic flow feedback regulation method for a ventilator when executing the computer program.

[0035] The present invention has the following beneficial effects:

[0036] The present invention first obtains current respiratory data and historical respiratory data to provide a basis for subsequent data analysis; further utilizes a straight line fitting method to quantify the fluctuation characteristics of data at the same position in different respiratory cycles, obtains the credibility and respiratory characteristic value of each data point at the target position, characterizes the respiratory characteristics under the ventilation mode and evaluates the credibility of the characterization, providing a basis for the subsequent calculation of the ability of the target position to distinguish the ventilation mode; further quantifies the respiratory intensity of each respiratory cycle, corrects the credibility of the data point according to the correlation characteristics of the respiratory intensity sequence and the distance sequence, obtains the corrected credibility, eliminates the interference of the patient's respiratory intensity, and improves the accuracy of the corrected credibility; further, in the target mode, according to the difference characteristics of the respiratory characteristic values ​​of the data points at the target position of different historical respiratory data, combined with the corrected credibility, the consistency of the single ventilation mode is obtained. From the perspective of reliability and credibility, the discrimination degree of the target position is obtained to characterize the ability of the data at the target position to distinguish and discriminate the target mode; further, according to the distribution difference characteristics of the respiratory characteristic values ​​of the target position under different ventilation modes, combined with the discrimination degree, the discrimination coefficient of the target position is obtained, and multiple ventilation modes are combined for analysis to avoid the generalization problem of the target position characteristics, which is convenient for optimizing the principal component analysis results; further, the historical respiratory data under the target mode is subjected to principal component analysis, and the final principal component of the target mode is obtained by combining the discrimination coefficient and the discrimination degree. The extracted final principal component contains the most important characteristic information of the target mode, which provides more reliable data support for the subsequent classification and matching of the ventilation mode of the current respiratory data; finally, the current ventilation mode is determined according to the difference characteristics of the principal component of the current respiratory data and the final principal component of each ventilation mode. The present invention optimizes the principal component by analyzing the discrimination ability of the historical respiratory data of each ventilation mode, so that the extracted final principal component contains the most important characteristic information of the target mode, and the current respiratory data can be accurately matched to the most similar ventilation mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 A flow chart of a method for automatic flow feedback regulation of a ventilator provided by one embodiment of the present invention;

[0039] Figure 2A flowchart of a method for obtaining a differentiation capability coefficient provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the flow automatic feedback adjustment method and system for a ventilator proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0041] Unless defined otherwise, 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 belongs.

[0042] The following is a detailed description of a specific scheme of an automatic flow feedback adjustment method and system for a ventilator provided by the present invention in conjunction with the accompanying drawings.

[0043] See also Figure 1 , which shows a flow chart of a method for automatic flow feedback regulation of a ventilator provided by an embodiment of the present invention, specifically comprising:

[0044] Step S1: Acquire current respiratory data and historical respiratory data; the historical respiratory data contains the corresponding ventilation mode; all respiratory data contain the same number of respiratory cycles, and the same number of data points in each respiratory cycle; select any historical respiratory data as the historical data to be analyzed; the ventilation mode corresponding to the historical data to be analyzed is the target mode; select any position in the respiratory cycle as the target position.

[0045] The embodiment of the present invention analyzes the historical respiratory data of each ventilation mode and extracts the data features of each ventilation mode, thereby matching the current respiratory data with each ventilation mode and determining the current ventilation mode. First, the current respiratory data and historical respiratory data are obtained to provide a basis for data analysis. In order to classify and analyze the historical respiratory data, the historical respiratory data contains the corresponding ventilation modes, and at the same time, the format of the respiratory data is unified and standardized. All respiratory data contain the same number of respiratory cycles, and the number of data points in each respiratory cycle is the same.

[0046] In one embodiment of the present invention, the ventilation modes include: assisted controlled ventilation mode (A / C or ACV), continuous positive airway pressure mode (CPAP), pressure support ventilation mode (PSV), pressure controlled ventilation mode (PCV), enhanced pressure controlled ventilation mode (APCV), synchronized intermittent mandatory ventilation mode (SIMV), pressure-regulated volume controlled ventilation mode (PRVC), bidirectional positive airway pressure mode (Bi-PAP or Bi-level), and volume controlled ventilation mode (VCV).

[0047] It should be noted that the ventilation mode can also be called the breathing mode; in order to accurately identify the current breathing data and determine the current ventilation mode, it is necessary to analyze the historical breathing data of all ventilation modes, so the historical breathing data needs to include all ventilation modes; the specific data format standard of each breathing data, that is, the number of breathing cycles in each breathing data, the number of data points in each breathing cycle, and the collection frequency can all be set by the implementer; in one embodiment of the present invention, the collection is carried out at a frequency of 10 times per second, 3 minutes is 1 cycle, and 10 cycles are one breathing data.

[0048] The ventilator mainly determines the ventilation mode according to the actual breathing condition of the patient. The respiratory data under the same ventilation mode have the same characteristics, which is manifested in that the data points of these respiratory data have a certain stability or trend. For example, in VCV and PSV, it is mainly determined based on the characteristics of the plateau data at the end of the respiratory stage in the patient's respiratory data, or it is easier to judge the ventilation mode based on the data at the plateau position. Therefore, not every respiratory data point can be used as a feature point for auxiliary ventilation mode judgment. Therefore, it is necessary to determine the expression and distinction ability of the data point at each position for respiratory characteristics. Therefore, any historical respiratory data is selected as the historical data to be analyzed; the ventilation mode corresponding to the historical data to be analyzed is the target mode; and any position in the respiratory cycle is selected as the target position, so that it is easy to analyze one by one.

[0049] It should be noted that the analysis method for all positions of all ventilation modes is consistent and will not be described one by one; data collection is authorized by the user, does not violate relevant laws and regulations, and does not violate public order and good customs.

[0050] Step S2: In the historical data to be analyzed, the data of the target position in different respiratory cycles are fitted with a straight line to obtain a distance sequence consisting of the distance between each data point and the fitting straight line; according to the violent fluctuation characteristics of the data in the distance sequence, the credibility of each data point is obtained; based on the parameters of the fitting straight line, the respiratory characteristic value of each data point is obtained; according to the peak value in each respiratory cycle in the historical data to be analyzed and the data change trend on both sides, the respiratory intensity of each respiratory cycle is obtained; according to the correlation characteristics of the respiratory intensity sequence formed by the respiratory intensity of the respiratory cycle and the distance sequence, combined with the credibility of the corresponding data point, the corrected credibility of the corresponding data point is obtained.

[0051] Considering that the data at the target position can be used as a characteristic identifier of respiratory data, when distinguishing different ventilation modes, the more regular the data at the target position is, the more accurate the data is. Since straight line fitting can intuitively analyze the degree of deviation between data points and the overall trend, in the historical data to be analyzed, the data at the target position in different respiratory cycles are fitted with a straight line to obtain a distance sequence consisting of the distance between each data point and the fitting straight line; considering that the violent fluctuation characteristics of the data in the distance sequence reflect the regular characteristics of the data at the target position, thereby reflecting the credibility of each data point at the target position as an identifier of the historical data to be analyzed, the credibility of each data point is obtained according to the violent fluctuation characteristics of the data in the distance sequence, and the credibility of each data point at the target position is quantified, providing a basis for the subsequent analysis of the ability to distinguish data at the target position.

[0052] Preferably, in one embodiment of the present invention, considering that the larger the variance and the larger the mean of the elements in the distance sequence, the farther the distance between the data point at the target position and the fitting straight line is, the more unstable it is, and the more obvious the violent fluctuation characteristics are, so the violent fluctuation coefficient is obtained at least according to the variance and the mean value of the distance sequence; the variance and the mean value of the distance sequence are both positively correlated with the violent fluctuation coefficient;

[0053] The more drastic the distance fluctuation is, the worse the regularity and stability of the data points involved in the fitting straight line are, and the more difficult it is to use them as identification data for respiratory data. It is more difficult to distinguish different ventilation modes through these data. Therefore, the fluctuation intensity coefficient is negatively correlated and normalized, and used as the credibility of each data point at the target position in the historical data to be analyzed.

[0054] As an example, the product of the variance and the mean of the elements in the distance sequence is taken as the volatility coefficient, the volatility coefficient is linearly normalized, and the difference between the constant 1 and the normalized volatility coefficient is taken as the credibility of each data point at the target position in the historical data to be analyzed.

[0055] In other embodiments of the present invention, the implementer may also fuse the variance and mean of the elements in the distance sequence by a positive correlation method of addition or weighted summation; considering that the absolute mean deviation may also reflect the violent fluctuation characteristics of the data, the maximum value of the elements in the distance sequence may represent the maximum deviation of the distance, so the implementer may also fuse the variance, mean, absolute mean deviation and maximum value of the elements in the distance sequence by addition or multiplication to obtain the violent fluctuation coefficient; negative correlation mapping may also be performed and normalized by an exponential function exp(-x) with the natural constant e as the base, where x represents the independent variable.

[0056] Considering that the fitting straight line represents the characteristic change trend of the data points at the target position, the respiratory characteristic values ​​extracted from it can characterize the respiratory characteristics under different ventilation modes. Therefore, the respiratory characteristic values ​​of each data point are obtained based on the parameters of the fitting straight line, which provides a basis for the subsequent calculation of the target position's ability to distinguish ventilation modes, and finally extracts the principal components with more distinguishing ability, thereby preparing for improving the recognition accuracy of different ventilation modes.

[0057] Preferably, in one embodiment of the present invention, considering that the slope and intercept of the fitted straight line are important parameters of the straight line, the fused slope and intercept can express the characteristics of the fitted straight line, thereby reflecting the characteristics of the data points involved in the straight line fitting, so the Euclidean norm of the slope and intercept of the fitted straight line is used as the respiratory characteristic value of each data point at the target position in the historical data to be analyzed.

[0058] It should be noted that straight-line fitting is a technical means well known to those skilled in the art. In one embodiment of the present invention, the least square method is used for straight-line fitting, which will not be described in detail. The data points selected at the target position are sorted according to the chronological order, and the corresponding distance sequence also adopts the same order. The breathing intensity sequence obtained subsequently is also sorted in the chronological order.

[0059] Taking into account that the patient's unique breathing characteristics will interfere with the respiratory data and cause corresponding fluctuations in the respiratory data, the peak value of the data within the respiratory cycle and the changing trend of the data on both sides of the adjacent data reflect the patient's respiratory intensity state. Therefore, according to the peak value of each respiratory cycle in the historical data to be analyzed and the changing trend of the data on both sides of the adjacent data, the respiratory intensity of each respiratory cycle is obtained, and the respiratory intensity characteristics of each respiratory cycle are measured, so as to prepare for the subsequent correction of the credibility of the data points in combination with the changes in the patient's respiratory intensity, and ultimately improve the accuracy of the final principal component.

[0060] Preferably, in one embodiment of the present invention, considering that the larger the peak value of the data in the respiratory cycle, the longer the duration of the larger data, and the greater the respiratory intensity, in each respiratory cycle in the historical data to be analyzed, the peak value of the respiratory data is extended to both sides to obtain extended data, and the duration of the sequence composed of the extended data and the peak value is taken as the peak duration; the extended data is greater than a preset ratio of the peak value of the respiratory data;

[0061] The respiratory intensity of each respiratory cycle is obtained according to the peak value and peak duration of the respiratory data in each respiratory cycle; the peak value and peak duration of the respiratory data are both positively correlated with the respiratory intensity.

[0062] As an example, the peak value of the data in a certain respiratory cycle of the historical data to be analyzed is 60, and the respiratory data sequence adjacent to the peak is [36, 39, 45, 49, 52, 60, 53, 47, 40, 38, 35]. The preset ratio is set to 2 / 3, then 60 can extend to the left to [45, 49, 52], and to the right to [53, 47]. The combined peak value itself is [45, 49, 52, 60, 53, 47], and the duration corresponding to [45, 49, 52, 60, 53, 47] is the peak duration; the product of the peak value of the respiratory data in each respiratory cycle and the peak duration is normalized and used as the respiratory intensity of each respiratory cycle.

[0063] It should be noted that when there are multiple peak values ​​in a respiratory cycle, multiple respiratory intensities will be obtained, and the maximum respiratory intensity is selected as the respiratory intensity of the corresponding respiratory cycle. In other embodiments of the present invention, the implementer can also fuse the peak duration and the peak value by adding or weighted summing to obtain the respiratory intensity.

[0064] Considering that the greater the breathing intensity, the greater the change in the respiratory data within the respiratory cycle, which causes fluctuations in the data points at the target position and the data in the distance sequence, the sharp fluctuations in the data in the distance sequence may be normal fluctuations caused by the breathing intensity. Therefore, based on the relevant characteristics of the respiratory intensity sequence and the distance sequence constructed by the breathing intensity of the respiratory cycle, combined with the credibility of the corresponding data points, the corrected credibility of the corresponding data points is obtained, and the credibility is corrected from the perspective of breathing intensity to eliminate the interference of the patient's breathing intensity and improve the accuracy of the corrected credibility.

[0065] Preferably, in one embodiment of the present invention, when the changes of the breathing intensity sequence and the distance sequence are more consistent, the larger the correlation coefficient is, indicating that the fluctuation of the distance sequence is a normal fluctuation caused by the patient's breathing, the credibility is low, and the credibility needs to be increased; on the contrary, the smaller the correlation coefficient is, the higher the credibility is, and the credibility needs to be reduced;

[0066] Based on this, the correlation coefficient between the breathing intensity sequence and the distance sequence is obtained at the target position in the historical data to be analyzed; the sum of the correlation coefficient and 1 is used as the correction coefficient, and the product of the correction coefficient and the credibility is used as the corrected credibility of the corresponding data point at the target position.

[0067] Among them, when the correlation coefficient is positive, it means that the changes in the respiratory intensity sequence and the distance sequence are positively correlated, and the fluctuation of the distance sequence is a normal fluctuation caused by the patient's breathing. The correction coefficient is greater than 1, and the credibility can be increased and corrected; when the correlation coefficient is negative, the credibility can be reduced and corrected; when the correlation coefficient is 0, the respiratory intensity sequence and the distance sequence have no linear correlation, and the credibility cannot be corrected by the respiratory intensity. The correction coefficient is 1, and the credibility and corrected credibility are equal.

[0068] It should be noted that, in one embodiment of the present invention, the correlation coefficient is specifically the Pearson correlation coefficient; in other embodiments of the present invention, the implementer may also use cosine similarity as the correlation coefficient, which is already existing technology and will not be described in detail.

[0069] Step S3: In the target mode, the discrimination degree of the target position is obtained according to the difference characteristics of the respiratory characteristic values ​​of the data points at the target position of different historical respiratory data, combined with the corrected credibility; the discrimination ability coefficient of the target position is obtained according to the distribution difference characteristics of the respiratory characteristic values ​​of the target position under different ventilation modes, combined with the discrimination degree.

[0070] Considering that the data at the target position can be used as the basis for distinguishing the target pattern, the respiratory characteristic values ​​of all historical respiratory data of the target pattern at the target position should be highly consistent. At the same time, the corrected credibility can be used as the credibility basis of the respiratory characteristic values. Therefore, the discrimination of the target position can be obtained based on the difference characteristics of the respiratory characteristic values ​​of the data points at the target position of different historical respiratory data, combined with the corrected credibility, to characterize the ability of the data at the target position to distinguish and discriminate the target pattern, and prepare for the subsequent acquisition of accurate final principal components.

[0071] Preferably, in one embodiment of the present invention, the greater the absolute value of the difference of the breathing characteristic value, the greater the difference, the weaker the consistency, and the lower the discrimination; at the same time, the smaller the correction credibility, the lower the credibility of the breathing characteristic value of the data at the target position, the lower the credibility of the breathing characteristic value at the target position in distinguishing the target pattern, and the lower the discrimination;

[0072] Based on this, in the target mode, any two historical breathing data are taken as a historical data group; the absolute value of the difference in the breathing characteristic values ​​of the data points at the target position in the historical data group is taken as the first numerator, the sum of the corrected credibility is taken as the first denominator, and the ratio of the first numerator to the first denominator is taken as the breathing difference parameter of the corresponding historical data group;

[0073] According to the overall characteristics of the breathing difference parameters of all historical data groups of the target pattern, the discrimination of the target position of the target pattern is obtained; the overall characteristics of the breathing difference parameters and the discrimination are negatively correlated.

[0074] As an example: The calculation formula for discrimination includes:

[0075]

[0076] Where D represents the discrimination of the target position of the target pattern; norm{} represents the linear normalization function; n represents the sequence number of the historical data group of the target pattern; N represents the number of the historical data groups of the target pattern; λ n,1 represents the respiratory characteristic value of the data point at the target position of the first historical respiratory data in the nth historical data group of the target mode; p n,1 represents the corrected credibility of the data point at the target position of the first historical breathing data in the nth historical data group of the target mode; n,2 represents the respiratory characteristic value of the data point at the target position of the second historical respiratory data in the nth historical data group of the target mode; p n,2 represents the corrected credibility of the data point at the target position of the second historical respiratory data in the nth historical data group of the target mode; Represents the respiratory difference parameter of the nth historical data group of the target mode at the target position; || indicates taking the absolute value.

[0077] In the calculation formula of the discrimination, the overall characteristics of the respiratory difference parameters of all historical data groups are expressed by the mean of the respiratory difference parameters. The overall characteristics of the respiratory difference parameters are negatively correlated and normalized by subtracting the linear normalization result from 1 to obtain the discrimination. The larger the respiratory difference parameter, the lower the credibility of the respiratory characteristic value of the data at the target position, the weaker the consistency, and the smaller the discrimination.

[0078] It should be noted that in other embodiments of the present invention, the implementer may also use the mean, median and mode of the respiratory difference parameters to jointly represent the overall characteristics of the respiratory difference parameters, such as weighted summing the mean, median and mode of the respiratory difference parameters with weights of 0.5, 0.25 and 0.25, performing negative correlation mapping on the weighted summation result and normalizing it to obtain the discrimination degree.

[0079] Since the discrimination is only analyzed and obtained within a single ventilation mode, the data points of the selected target position may be common features of multiple ventilation modes and cannot accurately distinguish the ventilation modes. Therefore, it is also necessary to obtain the discrimination coefficient of the target position based on the distribution difference characteristics of the respiratory characteristic values ​​of the target position under different ventilation modes, combined with the discrimination, to avoid the generalization problem of the target position characteristics, improve the accuracy of the target position's ability to distinguish the target ventilation mode, facilitate optimization of the principal component analysis results, and ultimately select the most suitable ventilation mode for the current respiratory data.

[0080] Preferably, in one embodiment of the present invention, the method for obtaining the discrimination capability coefficient includes:

[0081] See also Figure 2 , which shows a flow chart of a method for obtaining a distinguishing ability coefficient provided by an embodiment of the present invention, specifically comprising:

[0082] Step S301: When the discrimination degree of the target position is less than or equal to a preset discrimination threshold, the discrimination ability coefficient of the target position is set to zero.

[0083] Considering that when the discrimination degree is too small, it means that the data of the target position analyzed from a single ventilation mode does not have the ability to distinguish the ventilation mode, so when the discrimination degree of the target position is less than or equal to the preset discrimination threshold, the discrimination ability coefficient of the target position is set to zero.

[0084] As an example, the preset distinction threshold is 0.5.

[0085] Step S302: When the discrimination degree of the target position is greater than the preset discrimination threshold: based on the statistical characteristics of the respiratory characteristic values ​​of all historical respiratory data under each ventilation mode, obtain the empirical distribution function of each ventilation mode; select any other ventilation mode except the target mode as the target comparison mode; and form a mode binary group with the target mode and the target comparison mode.

[0086] When the discrimination degree of the target position is greater than the preset discrimination threshold, it means that the data of the target position analyzed from a single ventilation mode has a certain discrimination ability, and further analysis is needed in combination with different ventilation modes;

[0087] Considering that the empirical distribution function can intuitively reflect the distribution of data points, the empirical distribution functions of the two data sets are very different in the same range, which means that there are significant differences in the data distribution of the two data sets in the corresponding interval. Therefore, the empirical distribution function is used to analyze the distribution difference characteristics of the respiratory characteristic values ​​at the target position under different ventilation modes.

[0088] Therefore, based on the statistical characteristics of the respiratory characteristic values ​​of all historical respiratory data in each ventilation mode, the empirical distribution function of each ventilation mode is obtained; any ventilation mode other than the target mode is selected as the target comparison mode; the target mode and the target comparison mode form a mode binary.

[0089] Step S303: Obtain the extreme values ​​of the respiratory characteristic values ​​of the data points at the target position of all historical respiratory data in each ventilation mode in the pattern binary group; use the union of the ranges of the extreme values ​​corresponding to the two ventilation modes in the pattern binary group as the comparison range of the pattern binary group at the target position; within the comparison range at the target position, normalize the area between the empirical distribution functions of the two ventilation modes in the pattern binary group, and use the normalized result as the discrimination sub-coefficient of the pattern binary group.

[0090] In order to use the empirical distribution function to analyze the distribution difference characteristics, it is necessary to determine the comparison range of the empirical distribution function. Considering that the distribution difference of the respiratory characteristic values ​​of the data points at the target position needs to be compared, the comparison range is the union of the ranges of the extreme values ​​corresponding to the two ventilation modes in the mode binary group.

[0091] Considering that the larger the area between the functions, the greater the difference in the functions, reflecting that the greater the distribution difference of the respiratory characteristic values ​​of the data of different ventilation modes at the target position, the stronger the discrimination ability, so within the comparison range at the target position, the area between the empirical distribution functions of the two ventilation modes in the mode binary group is normalized, and the normalized result is used as the discrimination sub-coefficient of the mode binary group. As an example, the calculation formula of the discrimination sub-coefficient includes:

[0092]

[0093] Wherein, k represents the sequence number of the pattern binary group of the target position of the target pattern, which is also the sequence number of the target comparison pattern; b k represents the discriminant coefficient of the target position of the target pattern in the k-th pattern binary group; norm{} represents the linear normalization function; λ min Indicates the minimum value of the comparison range, which is also the lower limit of the integral; λ max represents the maximum value of the comparison range, which is also the upper limit of the integral; || represents the absolute value; φ(λ) represents the empirical distribution function of the target mode; φ(λ) k Represents the empirical distribution function of the k-th target contrast pattern.

[0094] In the calculation formula of the discriminant sub-coefficient, the area between the empirical distribution functions of the two ventilation modes in the mode binary group is obtained by the integral method, and normalized by the normalization function to obtain the discriminant sub-coefficient of the mode binary group.

[0095] Step S304: taking the minimum value of all discrimination sub-coefficients corresponding to the target position of the target pattern as the discrimination capability coefficient of the target position of the target pattern.

[0096] Considering that the minimum value of the discriminator sub-coefficient represents the weakest discrimination ability of the target position for distinguishing the target mode from other ventilation modes, the minimum value of all discriminator sub-coefficients corresponding to the target position of the target mode is taken as the discrimination ability coefficient of the target position of the target mode.

[0097] Step S4: Perform principal component analysis on the historical respiratory data under the target mode, and obtain the final principal component of the target mode by combining the discrimination coefficient and the discrimination degree; determine the current ventilation mode according to the difference characteristics between the principal component of the current respiratory data and the final principal component of each ventilation mode.

[0098] Both the discrimination coefficient and the discrimination degree reflect the ability of data points to distinguish ventilation patterns. Therefore, principal component analysis is performed on the historical respiratory data under the target pattern. The final principal components of the target pattern are obtained by combining the discrimination coefficient and the discrimination degree. The extracted final principal components contain the most important characteristic information of the target pattern, which provides more reliable data support for the subsequent classification and matching of ventilation patterns of the current respiratory data.

[0099] Preferably, in one embodiment of the present invention, considering the initial principal component directly obtained by the principal component analysis algorithm, the greater the weight of the element in the characteristic vector of the principal component, the greater the discrimination and discrimination coefficient of the original data point in the historical respiratory data corresponding to the element, indicating that the principal component is more important and can be used as the judgment identification data of the target mode;

[0100] Based on this, all principal components are determined according to the results of principal component analysis, and the weight of each element in the eigenvector corresponding to each principal component in the entire eigenvector is determined; in the target mode, the sum of the weight of each element in the eigenvector corresponding to each principal component, the discrimination degree of the original data point corresponding to the element in the eigenvector, and the product of the discrimination coefficient is used as the principal component importance of each principal component;

[0101] The principal component with the greatest importance is selected according to the preset variance contribution rate to obtain the final principal component of the target pattern.

[0102] It should be noted that the principal component analysis algorithm, obtaining the weight of each element in the eigenvector corresponding to the principal component in the entire eigenvector, and selecting the principal component using the preset variance contribution rate are all technical means well known to those skilled in the art and will not be described in detail.

[0103] It should be noted that the variance contribution rate is also called the variance explanation rate. The result of the principal component analysis retains all components as principal components without screening. As an example, the preset variance contribution rate is 0.7, and the principal components are sorted in descending order according to their importance. According to the order of descending sorting, the variance contribution rates of the principal components are accumulated until the accumulated value exceeds 0.7 for the first time. The principal component involved in the accumulation is the final principal component. In other embodiments of the present invention, the implementer can set other values; the implementer can also fuse the principal component weight, discrimination and discrimination ability coefficient of each data point by addition or weighted summation.

[0104] In another embodiment of the present invention, the implementer may also set a preset number of principal components, such as setting the preset number of principal components to 5, and selecting the first 5 principal components with the greatest importance as the final principal components.

[0105] After analyzing each ventilation mode and obtaining the final principal component of each ventilation mode, the current ventilation mode can be determined based on the difference characteristics between the principal component of the current respiratory data and the final principal component of each ventilation mode.

[0106] Preferably, in one embodiment of the present invention, in order to ensure that the current respiratory data matches the principal components of the ventilation mode, the principal components of the current respiratory data are obtained through the change matrix of the final principal components of the ventilation mode, and the matching distance is measured by means of the maximum Euclidean distance between the data points of the principal components, so as to select the ventilation mode with the smallest matching distance as the current ventilation mode;

[0107] Based on this, the transformation matrix of the final principal component of each ventilation mode is obtained; based on the transformation matrix of the target mode and the current respiratory data, the principal component of the current respiratory data in the target mode is obtained; the maximum Euclidean distance of the data points between the final principal component of the target mode and the principal component of the current respiratory data in the target mode is used as the matching distance between the current respiratory data and the target mode;

[0108] The ventilation mode with the smallest matching distance between the current respiratory data and all ventilation modes is selected as the current ventilation mode.

[0109] It should be noted that the transformation matrix in the principal component analysis algorithm, the use of the transformation matrix to reduce the dimension of data to obtain the principal components, and the Euclidean distance are all existing technologies and will not be described in detail.

[0110] It should be noted that the final main component of the ventilation mode can be set to a longer update cycle, not updated every time new respiratory data is obtained, for example, it can be set to be updated once a month.

[0111] An embodiment of the present invention also provides an automatic flow feedback regulation system for a ventilator, the system comprising a memory, a processor and a computer program, wherein the memory is used to store a corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program runs in the processor, it can implement an automatic flow feedback regulation method for a ventilator described in steps S1-S4.

[0112] In summary, in order to solve the technical problem that the automatic feedback flow adjustment of the ventilator is inaccurate due to the inaccurate selection of the principal component, the present invention provides a method and system for automatic feedback flow adjustment of the ventilator. The present invention first obtains the current respiratory data and the historical respiratory data; further, in the historical data to be analyzed, the data of the target position of different respiratory cycles are linearly fitted, and the respiratory characteristic value and credibility of each data point are obtained based on the fitting result; further, according to the correlation characteristics of the respiratory intensity sequence and the distance sequence combined with the credibility, the corrected credibility of the corresponding data point is obtained; further, according to the difference characteristics of the respiratory characteristic values ​​of the data point at the target position in different historical respiratory data, and the distribution difference characteristics of the respiratory characteristic values ​​of the data point in different ventilation modes, combined with the corrected credibility, the discrimination coefficient of the target position is obtained; further, the final principal component of the target mode is obtained by combining the discrimination coefficient and the discrimination degree; finally, based on the final principal components of all ventilation modes, the current ventilation mode is determined.

[0113] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. 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, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for automatic flow feedback regulation of a ventilator, characterized in that: The method comprises: Acquire current respiratory data and historical respiratory data; the historical respiratory data contains a corresponding ventilation mode; all respiratory data contain the same number of respiratory cycles, and the number of data points in each respiratory cycle is the same; select any of the historical respiratory data as the historical data to be analyzed; the ventilation mode corresponding to the historical data to be analyzed is the target mode; select any position in the respiratory cycle as the target position; In the historical data to be analyzed, the data of the target position of different respiratory cycles are fitted with a straight line to obtain a distance sequence composed of the distance between each data point and the fitting straight line; according to the violent fluctuation characteristics of the data in the distance sequence, the credibility of each data point is obtained; based on the parameters of the fitting straight line, the respiratory characteristic value of each data point is obtained; according to the peak value in each respiratory cycle and the data change trend on both sides of the adjacent peak value in the historical data to be analyzed, the respiratory intensity of each respiratory cycle is obtained; according to the correlation characteristics of the respiratory intensity sequence composed of the respiratory intensity of the respiratory cycle and the distance sequence, combined with the credibility of the corresponding data point, the corrected credibility of the corresponding data point is obtained; In the target mode, according to the difference characteristics of the respiratory characteristic values ​​of the data points of the target position of different historical respiratory data, combined with the corrected credibility, the discrimination degree of the target position is obtained; according to the distribution difference characteristics of the respiratory characteristic values ​​of the target position under different ventilation modes, combined with the discrimination degree, the discrimination ability coefficient of the target position is obtained; Perform principal component analysis on the historical breathing data under the target mode, and obtain the final principal component of the target mode by combining the discrimination coefficient and the discrimination degree; determine the current ventilation mode based on the difference characteristics between the principal component of the current breathing data and the final principal component of each ventilation mode.

2. The method for automatic flow feedback regulation of a ventilator according to claim 1, characterized in that: The method for obtaining the credibility includes: The fluctuation intensity coefficient is obtained at least according to the variance and the average value of the elements in the distance sequence; the variance and the average value of the elements in the distance sequence are both positively correlated with the fluctuation intensity coefficient; The fluctuation intensity coefficient is negatively correlated and normalized to serve as the credibility of each data point at the target position in the historical data to be analyzed.

3. The automatic feedback flow control method for a ventilator according to claim 1, characterized in that: The method for obtaining the respiratory characteristic value comprises: The Euclidean norm of the slope and intercept of the fitting straight line is used as the respiratory characteristic value of each data point at the target position in the historical data to be analyzed.

4. The method for automatic flow feedback regulation of a ventilator according to claim 1, characterized in that: The method for obtaining the breathing intensity comprises: In each of the respiratory cycles in the historical data to be analyzed, the peak value of the respiratory data is extended to both sides to obtain extended data, and the duration of the sequence formed by the extended data and the peak value is used as the peak duration; the extended data is greater than a preset ratio of the peak value of the respiratory data; The respiratory intensity of each respiratory cycle is obtained according to the peak value of the respiratory data in each respiratory cycle and the peak duration; the peak value of the respiratory data and the peak duration are both positively correlated with the respiratory intensity.

5. The method for automatic flow feedback regulation of a ventilator according to claim 4, characterized in that: The method for obtaining the modified credibility includes: At the target position in the historical data to be analyzed, the correlation coefficient between the respiratory intensity sequence and the distance sequence is obtained; the sum of the correlation coefficient and 1 is used as a correction coefficient, and the product of the correction coefficient and the credibility is used as the corrected credibility of the corresponding data point at the target position.

6. The method for automatic flow feedback regulation of a ventilator according to claim 1, characterized in that: The method for obtaining the discrimination degree includes: In the target mode, any two of the historical breathing data are taken as a historical data group; the absolute value of the difference between the breathing characteristic values ​​of the data points at the target position in the historical data group is taken as the first numerator, the sum of the corrected credibility is taken as the first denominator, and the ratio of the first numerator to the first denominator is taken as the breathing difference parameter corresponding to the historical data group; According to the overall characteristics of the breathing difference parameters of all the historical data groups of the target pattern, the discrimination degree of the target position of the target pattern is obtained; the overall characteristics of the breathing difference parameters and the discrimination degree are negatively correlated.

7. The method for automatic flow feedback regulation of a ventilator according to claim 6, characterized in that: The method for obtaining the discrimination capability coefficient includes: When the discrimination degree of the target position is less than or equal to a preset discrimination threshold, the discrimination ability coefficient of the target position is set to zero; When the discrimination degree of the target position is greater than a preset discrimination threshold: based on the statistical characteristics of the respiratory characteristic values ​​of all the historical respiratory data under each ventilation mode, obtaining the empirical distribution function of each ventilation mode; selecting any other ventilation mode except the target mode as a target comparison mode; and forming a mode binary with the target mode and the target comparison mode; Obtain the extreme value of the respiratory characteristic value of the data point at the target position for all the historical respiratory data in each ventilation mode in the pattern binary group; use the union of the ranges of the extreme values ​​corresponding to the two ventilation modes in the pattern binary group as the comparison range of the pattern binary group at the target position; within the comparison range at the target position, normalize the area between the empirical distribution functions of the two ventilation modes in the pattern binary group, and use the normalized result as the discrimination sub-coefficient of the pattern binary group; The minimum value of all the discrimination sub-coefficients corresponding to the target position of the target pattern is used as the discrimination capability coefficient of the target position of the target pattern.

8. The method for automatic flow feedback regulation of a ventilator according to claim 1, characterized in that: The method for obtaining the final principal component includes: Determine all principal components according to the result of the principal component analysis, and determine the weight of each element in the eigenvector corresponding to each principal component in the entire eigenvector; in the target mode, take the sum of the weight of each element in the eigenvector corresponding to each principal component, the discrimination degree of the original data point corresponding to the element in the eigenvector and the product of the discrimination coefficient as the principal component importance of each principal component; The principal component with the greatest importance is selected according to the preset variance contribution rate to obtain the final principal component of the target pattern.

9. The method for automatic flow feedback regulation of a ventilator according to claim 1, characterized in that: The method for obtaining the current ventilation mode includes: Obtaining a transformation matrix of the final principal component of each ventilation mode; obtaining the principal component of the current respiratory data under the target mode based on the transformation matrix of the target mode and the current respiratory data; taking the maximum Euclidean distance of data points between the final principal component of the target mode and the principal component of the current respiratory data under the target mode as the matching distance between the current respiratory data and the target mode; The ventilation mode having the smallest matching distance between the current respiratory data and all the ventilation modes is selected as the current ventilation mode.

10. A flow automatic feedback regulation system for a ventilator, the system 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, the steps of the automatic flow feedback adjustment method for a ventilator as claimed in any one of claims 1 to 9 are implemented.

Images