A method and system for identifying critical respiratory states in early childhood

Abstract

The present invention relates to the technical field of respiratory monitoring, and specifically relates to a method and system for identifying critical respiratory states in early childhood. The present invention first obtains historical respiratory signals and segments them to obtain respiratory segments; further, according to the differences in respiratory amplitude and signal components between the respiratory segments, the respiratory patterns are divided; further, the respiratory frequency and respiratory amplitude range of each respiratory pattern are obtained, and the normal respiratory pattern is screened out; further, the real-time respiratory signal of the current patient is obtained and segmented in real time; according to the difference characteristics between the current respiratory segment and the normal respiratory pattern at the current moment, combined with the respiratory pattern of the current moment and the respiratory patterns of adjacent respiratory segments, and the difference in the number of respiratory segments, the respiratory abnormality coefficient is obtained; finally, based on the respiratory abnormality coefficient, it is determined whether to issue an early warning, reducing the influence of modal mixing, reducing the influence of physiological differences among different age groups or individuals, and improving the accuracy of respiratory state identification.

Description

Technical Field

[0001] The present invention relates to the technical field of respiratory monitoring, and particularly to a method and system for identifying critical respiratory states in early childhood. Background Art

[0002] Non-contact respiration detection based on microwave radar detects the human body according to the certain correlation between the micro-motion of the human body surface and the phase of the reflected echo signal, so as to achieve the purpose of respiration detection. Then, based on Empirical Mode Decomposition (EMD), the respiration signal detected by microwave can be decomposed into a series of Intrinsic Mode Function (IMF) components with frequencies decreasing from high to low, and then the respiratory state is identified according to the frequencies of the IMF components.

[0003] Pediatric critical illnesses have the problems of strong concealment, rapid changes, and being easily overlooked. Moreover, due to age differences, there are also differences in respiration. Therefore, accurate detection of the respiration of pediatric patients is extremely important. However, when the existing method decomposes the respiration signal by empirical mode decomposition for respiratory state identification, the frequencies of the IMF components are not completely concentrated, and there is a problem of mode mixing, resulting in poor detection effects and unsatisfactory accuracy of respiratory state identification. Summary of the Invention

[0004] In order to solve the technical problem of the low accuracy of the existing method based on microwave detection and signal decomposition for identifying the respiratory state of children, the purpose of the present invention is to provide a method and system for identifying critical respiratory states in early childhood, and the specific technical solutions adopted are as follows:

[0005] A method for identifying critical respiratory states in early childhood, the method comprising:

[0006] Obtain the historical respiration signal of the current patient within a preset time interval; slide a preset window on the historical respiration signal, perform signal decomposition on the historical respiration signal within the preset window after sliding based on the EMD algorithm, iteratively adjust the window length according to the frequency concentration characteristics of the signal components, segment the historical respiration signal, and obtain respiration segments;

[0007] Obtain the clustering distance between any two respiration segments according to the difference in respiration amplitude and the difference in signal components between the two respiration segments; perform clustering on the respiration segments based on the clustering distance, and each clustering cluster corresponds to a respiration pattern;

[0008] Obtain the breathing frequency and breathing amplitude range of each breathing pattern according to the overall characteristics of the breathing amplitude ranges of all the breathing segments within each clustering cluster and the distribution characteristics of the frequencies of all the first signal components; obtain the normal breathing pattern according to the length characteristics of the breathing segments within each breathing pattern.

[0009] Obtain the real-time breathing signal of the current patient and segment it in real time; according to the difference characteristics of the breathing amplitude range between the breathing segment to which the current moment belongs and the normal breathing pattern, and the difference characteristics of the breathing frequency, and combining the number difference of the breathing segments of the breathing pattern to which the current moment belongs and the breathing patterns of adjacent breathing segments, obtain the breathing abnormality coefficient at the current moment; determine whether to issue a warning based on the breathing abnormality coefficient at the current moment.

[0010] Further, the method for obtaining the breathing segments includes:

[0011] The preset window slides starting from the starting point of the historical breathing signal, and the window length is iteratively adjusted after each slide. The window length determined after each slide is used as the sliding step length for the next slide.

[0012] For the historical breathing signal with the latest window length after any slide, select any one of the signal components as the target signal component; obtain the spectrogram of the target signal component, translate the straight line parallel to the frequency axis from the maximum value of the spectral amplitude to the minimum value, and obtain the intersection length of the straight line corresponding to each spectral amplitude and the spectral line; according to the variation characteristics of the intersection length of the target signal component with the spectral amplitude, obtain the frequency concentration sub-coefficient of the target signal component.

[0013] According to the mean value of the frequency concentration sub-coefficients of all the signal components, and combining the difference between the frequency concentration sub-coefficient of each signal component and the mean value of the frequency concentration sub-coefficients, obtain the frequency concentration coefficient of the current latest window length; the mean value of the frequency concentration sub-coefficients is positively correlated with the frequency concentration coefficient; the difference between the frequency concentration sub-coefficient and the mean value of the frequency concentration sub-coefficients is negatively correlated with the frequency concentration coefficient.

[0014] If the frequency concentration coefficient of the current latest window length is less than the first preset concentration threshold, reduce the window length by a preset adjustment step length to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the second preset concentration threshold, increase the window length by a preset adjustment step length to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the first preset concentration threshold and less than the second preset concentration threshold, terminate the iteration, and use the historical breathing signal corresponding to the current latest window length as a breathing segment.

[0015] Further, the method for obtaining the frequency concentration sub-coefficient includes:

[0016] Normalize the sum value of the product of the negative correlation mapping values of all the intersection lengths of the target signal component and the corresponding spectral amplitudes, and use the normalized result as the frequency concentration sub-coefficient of the target signal component.

[0017] Further, the method for obtaining the clustering distance includes:

[0018] Record the frequency at the maximum value of the spectral amplitude in the spectrogram of each signal component as the representative frequency; obtain any two of the respiratory segments to form a target binary group. For each signal component within any one of the respiratory segments in the target binary group, obtain the signal component with the closest representative frequency within the other respiratory segment as the matching signal component.

[0019] The calculation formula of the clustering distance includes:

[0020] ;

[0021] where u and v are the serial numbers of the respiratory segments; represents the clustering distance between the u-th respiratory segment and the v-th respiratory segment; represents linear normalization; represents the absolute value of the difference between the maximum respiratory amplitudes of the u-th respiratory segment and the v-th respiratory segment; represents the absolute value of the difference between the minimum respiratory amplitudes of the u-th respiratory segment and the v-th respiratory segment; i represents the serial number of the signal component of the u-th respiratory segment; k represents the serial number of the signal component of the v-th respiratory segment; represents the number of signal components of the u-th respiratory segment; represents the number of signal components of the v-th respiratory segment; represents the representative frequency of the i-th signal component of the u-th respiratory segment; the representative frequency of the matching signal component corresponding to the i-th signal component of the u-th respiratory segment within the v-th respiratory segment; represents taking the absolute value; represents the representative frequency of the k-th signal component of the v-th respiratory segment; the representative frequency of the matching signal component corresponding to the k-th signal component of the v-th respiratory segment within the u-th respiratory segment.

[0022] Further, the method for obtaining the respiratory frequency and the range of respiratory amplitude of each respiratory pattern includes:

[0023] Take the mean of the representative frequencies of the first signal component of all the respiratory segments within each cluster as the respiratory frequency of the corresponding respiratory pattern;

[0024] Take the mean of the maximum respiratory amplitudes of all the respiratory segments within each cluster as the maximum value of the respiratory amplitude range of the corresponding respiratory pattern; take the mean of the minimum respiratory amplitudes of all the respiratory segments within each cluster as the minimum value of the respiratory amplitude range of the corresponding respiratory pattern.

[0025] Further, the method for obtaining the normal respiratory pattern includes:

[0026] Take the product of the mean of the time-domain lengths of all the respiratory segments of each respiratory pattern and the sum value of the time-domain lengths of all the respiratory segments as the normal possibility coefficient of each respiratory pattern;

[0027] Select the respiratory pattern with the largest normal possibility coefficient as the normal respiratory pattern.

[0028] Further, the calculation formula of the respiratory abnormality coefficient includes:

[0029] ;

[0030] where L represents the respiratory abnormality coefficient at the current moment; represents the exponential function with the natural constant e as the base; h is the serial number of the respiratory segment to which the current moment belongs; represents the respiratory amplitude range of the respiratory segment at the current moment; represents the respiratory amplitude range of the normal respiratory pattern; represents the intersection of the respiratory amplitude range of the respiratory segment at the current moment and the respiratory amplitude range of the normal respiratory pattern; represents the union of the respiratory amplitude range of the respiratory segment at the current moment and the respiratory amplitude range of the normal respiratory pattern; represents the linear normalization function; represents taking the absolute value; represents the representative frequency of the respiratory segment at the current moment; represents the respiratory frequency of the normal respiratory pattern; represents the number of respiratory segments of the respiratory pattern adjacent to the respiratory segment at the current moment; represents the number of respiratory segments of the respiratory pattern corresponding to the current moment.

[0031] Further, the method for determining whether to issue an early warning based on the respiratory abnormality coefficient at the current moment includes:

[0032] When the respiratory abnormality coefficient is greater than the preset abnormality threshold, issue an early warning.

[0033] Further, the preset adjustment step size is 1 second.

[0034] The present invention also provides a system for identifying critical respiratory states in early childhood, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any one of the methods for identifying critical respiratory states in early childhood.

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

[0036] The present invention first obtains historical respiratory signals to provide a data basis for determining different respiratory patterns of a patient; further, a preset window slides on the historical respiratory signals, and the historical respiratory signals within the preset window after sliding are decomposed based on the EMD algorithm. The window length is iteratively adjusted according to the frequency concentration characteristics of the signal components, and the historical respiratory signals are segmented to obtain respiratory segments, reducing the influence of mode mixing. At the same time, ensuring that longer respiratory segments can contain more complete respiratory cycles is more conducive to analyzing different respiratory patterns; further dividing the respiratory patterns, determining the respiratory frequency and respiratory amplitude range of each respiratory pattern, clarifying the parameter boundaries of the respiratory patterns, and distinguishing different respiratory patterns; further, according to the length characteristics of the respiratory segments within each respiratory pattern, obtaining the normal respiratory pattern as the reference pattern, and the normal respiratory pattern determined based on the historical respiratory signals of the current patient can better represent the normal respiratory pattern of the current patient, reducing the influence of physiological differences between different age groups or individuals; further, according to the differences in the respiratory amplitude range and respiratory frequency between the current respiratory segment and the normal respiratory pattern at the current moment, combined with the respiratory patterns of the current respiratory segment and adjacent respiratory segments, and the difference in the number of respiratory segments, the respiratory abnormality coefficient at the current moment is obtained, which characterizes the degree of abnormality of the respiratory state of the current child patient at the current moment, providing a basis for subsequent determination of whether to give an alarm; finally, it is determined whether to give an alarm based on the respiratory abnormality coefficient. By analyzing the mode mixing situation of signal decomposition during signal segmentation, the present invention reduces the influence of mode mixing by adjusting the window length; by using the normal respiratory pattern determined based on the historical respiratory signals of the current patient, it reduces the influence of physiological differences between different age groups or individuals and improves the accuracy of respiratory state recognition. 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 following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 Flow chart of a method for identifying critical respiratory states in early childhood provided by an embodiment of the present invention;

[0039] Figure 2 Flow chart of a method for obtaining respiratory segmentation provided by an embodiment of the present invention;

[0040] Figure 3 Schematic diagram of an intersection length provided by an embodiment of the present invention;

[0041] Figure 4 Schematic diagram of the respiratory amplitude range of all respiratory patterns provided by an embodiment of the present invention;

[0042] Figure 5 Schematic diagram of the respiratory frequency of all respiratory patterns provided by an embodiment of the present invention. Detailed implementation manners

[0043] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a method and system for identifying critical respiratory states in early childhood according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0045] The following specifically describes the specific solutions of a method and system for identifying critical respiratory states in early childhood provided by the present invention with reference to the accompanying drawings.

[0046] Please refer to Figure 1 , which shows a flow chart of a method for identifying critical respiratory states in early childhood provided by an embodiment of the present invention, specifically including:

[0047] Step S1: Obtain the historical respiratory signal of the current patient within a preset time interval; slide a preset window over the historical respiratory signal, decompose the historical respiratory signal within the slid preset window based on the EMD algorithm, iteratively adjust the window length according to the frequency concentration characteristics of the signal components, segment the historical respiratory signal, and obtain respiratory segmentation.

[0048] In an embodiment of the present invention, a microwave radar is used to detect the respiration of a patient to obtain a respiration signal; first, the historical respiration signals of the current patient in a preset time interval are obtained, and the historical respiration signals are analyzed to determine different respiration patterns of the patient, so as to identify the real-time respiration pattern of the patient.

[0049] It should be noted that using a microwave radar to detect the respiration of a patient and using the EMD algorithm for signal decomposition are both prior arts, and reference can be made to the literature "Design and Research of a Microwave Non-contact Respiration Detection and Analysis System" by Liang Song; in an embodiment of the present invention, the length of the preset time interval is 24 hours, and the historical respiration signals of the current patient in the 24 hours of the day adjacent to the current day are obtained. The preset time interval is from the start to the end of a day, and the historical respiration signals are updated at 0:00 every day to update the normal respiration pattern; during the update period, the respiration state recognition still uses the data before the update, and the implementer can set it according to the implementation scenario.

[0050] There are differences in the amplitude and frequency of each respiration of the patient, and the differences between different respiration patterns only lie in the magnitudes of the differences in amplitude and frequency. At the same time, since respiration exists continuously, analyzing respiration in units of single respiration will result in too many respiration patterns, which is not conducive to state recognition. Therefore, first, in the form of windowing, continuous respirations with similar states can be divided into the same respiration segment. Since the maintenance time of the respiration segments is different, the window size can be adaptively determined by analyzing the performance of the respiration signals within the window.

[0051] To reduce the influence of mode mixing, the frequencies of the signal components within a respiration segment cannot be too discrete, and at the same time, it is ensured that a longer respiration segment can contain more complete respiration cycles, which is more conducive to analyzing different respiration patterns. Therefore, a preset window slides on the historical respiration signals, and the historical respiration signals within the preset window after sliding are decomposed. The window length is iteratively adjusted according to the frequency concentration characteristics of the signal components. With each sliding and iteratively adjusted window, the historical respiration signals are segmented to obtain respiration segments, which is convenient for subsequent division of respiration patterns.

[0052] Preferably, in an embodiment of the present invention, the method for obtaining respiration segments includes:

[0053] First, the sliding process is described: the preset window starts sliding from the starting point of the historical respiration signals, and the window length is iteratively adjusted after each sliding. The window length determined after each sliding is used as the sliding step length for the next sliding; at the same time, the window length determined after each sliding is also the time domain length of a respiration segment.

[0054] As an example, the preset window is 5 seconds, the preset adjustment step is 1 second, the starting point of the historical breathing signal is set to 0, the initial position of the preset window is at 0 - 5 seconds. Suppose after iterative adjustment, the window length is finally adjusted to 20 seconds, then the first breathing segment is 0 - 20, the preset window slides correspondingly by 20 seconds, and iterative adjustment is performed again at 20 - 25 seconds; suppose after iterative adjustment, the window length is finally adjusted to 4 seconds, then the second breathing segment is 20 - 24, the preset window slides correspondingly by 4 seconds, and iterative adjustment is performed again at 24 - 29 seconds; and so on, to obtain all breathing segments of the historical breathing signal.

[0055] It should be noted that the time domain length of the preset window is the base value of the window length after each slide and remains unchanged during the sliding process.

[0056] Please refer to Figure 2 , which shows a flowchart of a method for obtaining breathing segments provided by an embodiment of the present invention, specifically including:

[0057] Step S101: For the historical breathing signal with the latest window length after any slide, select any signal component as the target signal component; obtain the spectrogram of the target signal component, translate the straight line parallel to the frequency axis from the maximum value of the spectral amplitude to the minimum value, and obtain the intersection length of the straight line corresponding to each spectral amplitude and the spectral line; according to the variation characteristics of the intersection length of the target signal component with the spectral amplitude, obtain the frequency concentration sub - coefficient of the target signal component.

[0058] Since the processing method for the latest window length after each slide is the same, and at the same time the analysis method for each signal component of the historical breathing signal with the latest window length is the same, one example is selected here as the target signal component for description, and no repeated description will be given.

[0059] Considering that the spectrogram is more conducive to analyzing the concentration characteristics of frequency, when the frequency is more concentrated, the waveform in the spectrogram is sharper, showing a higher amplitude and a narrower waveform. Therefore, the straight line parallel to the frequency axis is translated from the maximum value of the spectrogram to the minimum value to obtain the intersection length of the straight line corresponding to each spectral amplitude and the spectral line.

[0060] It should be noted that the straight - line segment where the straight line corresponding to the spectral amplitude intersects the spectral line and is between the spectral line and the horizontal axis is an intersection segment; when there are multiple intersection segments where the straight line corresponding to the spectral amplitude intersects the spectral line, the sum value of all intersection segments is used as the intersection length. Please refer to Figure 3 , which shows a schematic diagram of an intersection length provided by an embodiment of the present invention; Figure 3 In, the horizontal axis is frequency, the vertical axis is spectral amplitude, and the waveform function is set as (For illustration only, not representing actual spectral lines), Y1 corresponds to , Y2 corresponds to , Y1 is the straight line corresponding to the maximum value of the spectral amplitude. Translating downward from Y1, if Y2 is obtained by translation, there are two intersecting line segments between the straight line Y2 and the spectral line, and . The sum of the lengths of these two line segments is the intersection length corresponding to the spectral amplitude of Y2.

[0061] The shorter the intersection length at the higher spectral amplitude, the more concentrated the frequency. Therefore, according to the variation characteristics of the intersection length of the target signal component with the spectral amplitude, the frequency concentration sub-coefficient of the target signal component is obtained.

[0062] In an embodiment of the present invention, after normalizing the sum of the products of the negative correlation mapping values of all the intersection lengths of the target signal component and the corresponding spectral amplitudes, the normalized result is used as the frequency concentration sub-coefficient of the target signal component.

[0063] As an example: The calculation formula of the frequency concentration sub-coefficient includes:

[0064] ;

[0065] where, i represents the serial number of the target signal component; represents the frequency concentration sub-coefficient of the i-th target signal component; represents the linear normalization function; y represents the serial number of the spectral amplitude; represents the data value of the y-th spectral amplitude of the i-th target signal component; represents the number of spectral amplitudes; represents the exponential function with the natural constant e as the base; represents the intersection length of the y-th spectral amplitude of the i-th target signal component; represents the negative correlation mapping value of the intersection length of the y-th spectral amplitude of the i-th target signal component.

[0066] In the calculation formula of the frequency concentration sub-coefficient, through the exp(-x) function, the intersection length is used as the independent variable x for negative correlation mapping to obtain the negative correlation mapping value. The spectral amplitude and the intersection length are fused by multiplication, and then through summation, represents the variation characteristics of the intersection length with the spectral amplitude. The shorter the intersection length at the higher spectral amplitude, the more concentrated the frequency, and the larger the frequency concentration sub-coefficient.

[0067] It should be noted that when the straight line parallel to the frequency axis is translated, the translation step size is set to 1 to obtain the number of spectral amplitudes. The implementer can also adjust the translation step size by himself.

[0068] Step S102: Based on the mean of the frequency concentration sub - coefficients of all signal components, and combining the difference between the frequency concentration sub - coefficient of each signal component and the mean of the frequency concentration sub - coefficients, obtain the frequency concentration coefficient of the current latest window length.

[0069] Considering that there are multiple signal components in the historical breathing signal of the latest window length, it is necessary to further synthesize all signal components, evaluate the overall frequency concentration characteristics, and obtain the frequency concentration coefficient. Considering that the larger the mean of the frequency concentration sub - coefficients, the more concentrated the frequencies of all signal components are, so the mean of the frequency concentration sub - coefficients is positively correlated with the frequency concentration coefficient. At the same time, the smaller the difference between the frequency concentration sub - coefficient of each signal component and the mean of the frequency concentration sub - coefficients, the lower the degree of dispersion and the better the overall signal decomposition effect, so the difference between the frequency concentration sub - coefficient and the mean of the frequency concentration sub - coefficients is negatively correlated with the frequency concentration coefficient.

[0070] As an example, the calculation formula of the frequency concentration coefficient includes:

[0071] ;

[0072] where, represents the frequency concentration coefficient corresponding to the latest window length; represents the mean of the frequency concentration sub - coefficients of all signal components of the historical breathing signal of the latest window length; I represents the number of target signal components; i represents the serial number of the target signal component; represents the frequency concentration sub - coefficient of the i - th target signal component; represents taking the absolute value.

[0073] In the calculation formula of the frequency concentration coefficient, the difference between the frequency concentration sub - coefficient and the mean of the frequency concentration sub - coefficients is represented by the absolute value of the difference. Through the function for negative - correlation mapping to adjust the logical relationship, and then through multiplying and fusing and , the frequency concentration characteristics of the signal components of the latest window length are represented.

[0074] Step S103: If the frequency concentration coefficient of the current latest window length is less than the first preset concentration threshold, reduce the window length by a preset adjustment step to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the second preset concentration threshold, increase the window length by a preset adjustment step to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the first preset concentration threshold and less than the second preset concentration threshold, terminate the iteration, and use the historical breathing signal corresponding to the current latest window length as a breathing segment.

[0075] As an example, the threshold value in the first preset set is 0.5, and the threshold value in the second preset set is 0.7. For instance, the current latest window length is 5 seconds, the frequency concentration coefficient is 0.9, and the window length is adjusted to 6 seconds; the latest window length is 6 seconds, the frequency concentration coefficient is 0.85, and the window length is adjusted to 7 seconds; the latest window length is 7 seconds, the frequency concentration coefficient is 0.8, and the window length is adjusted to 8 seconds; the latest window length is 8 seconds, the frequency concentration coefficient is 0.77, and the window length is adjusted to 9 seconds; the latest window length is 9 seconds, the frequency concentration coefficient is 0.68, the iteration is terminated, the current latest window length is 9 seconds, and the historical respiration signal within the window corresponding to this window length is used as a respiration segment.

[0076] It should be noted that the implementer can adjust the threshold values in the first preset set and the second preset set by himself; when the adjustment directions of two consecutive iterative adjustments are opposite, the window length with the shorter length is taken. For example, the frequency concentration feature corresponding to a window length of 20 seconds is 0.75. After an increase adjustment, the frequency concentration feature corresponding to the new window length of 21 seconds is 0.48. At this time, the window length of 20 seconds is taken. For example, the frequency concentration feature corresponding to a window length of 20 seconds is 0.4. After a decrease adjustment, the frequency concentration feature corresponding to the new window length of 19 seconds is 0.72. At this time, the window length of 19 seconds is taken.

[0077] Step S2: Obtain the clustering distance between any two respiration segments according to the difference in respiration amplitude and the difference in signal components between the two respiration segments; based on the clustering distance, cluster the respiration segments, and each cluster corresponds to a respiration pattern.

[0078] After segmenting the respiration signal, it is necessary to analyze the differences between the respiration segments and divide the respiration patterns to prepare for the subsequent real-time identification of the respiration state of child patients.

[0079] Considering that the differences between different respiration patterns are mainly reflected in respiration amplitude and frequency, and EMD decomposition can decompose the signal into signal components in different frequency ranges, including the respiration frequency characteristics of the respiration segments, so the clustering distance between any two respiration segments is obtained according to the difference in respiration amplitude and the difference in signal components between the two respiration segments; based on the clustering distance, cluster the respiration segments, and each cluster corresponds to a respiration pattern. By means of clustering, similar respiration segments are grouped into one category to distinguish different respiration patterns.

[0080] Preferably, in an embodiment of the present invention, considering that the frequency at the maximum spectral amplitude is the main frequency of the signal component, the frequency at the maximum spectral amplitude in the spectrogram of each signal component is recorded as the representative frequency to simplify the frequency identification and facilitate the comparison of the frequency differences of different signal components;

[0081] Obtain any two respiratory segments to form a target binary group. For each signal component within any one of the respiratory segments in the target binary group, obtain the signal component within the other respiratory segment with the closest representative frequency as the matching signal component.

[0082] For example, the respiratory segment numbered Z1 has 5 signal components, and the respiratory segment numbered Z2 has 6 signal components. The representative frequency of the first signal component in Z1 is closest to that of the second signal component in Z2, that is, the absolute value of the difference in representative frequencies is the smallest. Then, the second signal component in Z2 is the matching signal component of the first signal component in Z1.

[0083] It should be noted that the same signal component can be the matching signal component of multiple signal components. For example, when the representative frequency of the second signal component in Z1 is also closest to that of the second signal component in Z2, the second signal component in Z2 is also the matching signal component of the first signal component in Z1.

[0084] The calculation formula for the clustering distance includes:

[0085] ;

[0086] where u and v are the serial numbers of the respiratory segments; represents the clustering distance between the u-th respiratory segment and the v-th respiratory segment; represents linear normalization; represents the absolute value of the difference between the maximum respiratory amplitudes of the u-th respiratory segment and the v-th respiratory segment; represents the absolute value of the difference between the minimum respiratory amplitudes of the u-th respiratory segment and the v-th respiratory segment; i represents the serial number of the signal component of the u-th respiratory segment; k represents the serial number of the signal component of the v-th respiratory segment; represents the number of signal components of the u-th respiratory segment; represents the number of signal components of the v-th respiratory segment; represents the representative frequency of the i-th signal component of the u-th respiratory segment; the representative frequency of the matching signal component corresponding to the i-th signal component of the u-th respiratory segment within the v-th respiratory segment; represents taking the absolute value; represents the representative frequency of the k-th signal component of the v-th respiratory segment; the representative frequency of the matching signal component corresponding to the k-th signal component of the v-th respiratory segment within the u-th respiratory segment.

[0087] In the calculation formula for the clustering distance, the difference characteristics of the data are shown through the absolute value of the difference, and through the and , respectively reflecting the differences in the maximum respiratory amplitude and the minimum respiratory amplitude. After multiplying and fusing, it represents the difference in respiratory amplitude between two respiratory segments; through it represents the frequency difference feature between the signal component of the u-th respiratory segment and the signal component with the closest representative frequency within the v-th respiratory segment. Through it represents the frequency difference feature between the signal component of the v-th respiratory segment and the signal component with the closest representative frequency within the u-th respiratory segment. After multiplying and fusing, it represents the difference in signal components between two respiratory segments; fusing the respiratory amplitude difference and the respiratory frequency difference reflected by the signal components measures the difference between two respiratory segments and serves as a clustering distance for subsequent clustering to divide respiratory patterns.

[0088] Through the calculation formula of the clustering distance, the clustering distances of all target binary groups are obtained.

[0089] As a preferred embodiment, the neighborhood radius is set to 0.1 and the minimum number of points is 3. The DBSCAN algorithm is used to obtain several clustering clusters, and the respiratory segments in each clustering cluster are recorded as a respiratory pattern, obtaining several respiratory patterns.

[0090] It should be noted that the DBSCAN algorithm is already an existing technology, and the implementer can also use the K-means clustering algorithm, which will not be elaborated here.

[0091] Step S3: According to the overall characteristics of the respiratory amplitude ranges of all respiratory segments within each clustering cluster and the distribution characteristics of the frequencies of all first signal components, obtain the respiratory frequency and respiratory amplitude range of each respiratory pattern; according to the length characteristics of the respiratory segments within each respiratory pattern, obtain the normal respiratory pattern.

[0092] After dividing the respiratory patterns, it is necessary to quantify the physiological characteristics of each pattern to provide a reference benchmark for subsequent abnormal respiratory state recognition. Considering that the overall characteristics of the respiratory amplitude ranges of all respiratory segments within a clustering cluster represent the variation range and law of a respiratory pattern in terms of the depth of breathing, and the frequency of the first signal component obtained by EMD decomposition represents the main respiratory frequency in the respiratory signal, so according to the overall characteristics of the respiratory amplitude ranges of all respiratory segments within each clustering cluster and the distribution characteristics of the frequencies of all first signal components, obtain the respiratory frequency and respiratory amplitude range of each respiratory pattern, clarify the parameter boundaries of the respiratory patterns, and enhance the ability to distinguish respiratory patterns.

[0093] Preferably, in an embodiment of the present invention, the mean value of the representative frequencies of the first signal components of all respiratory segments within each clustering cluster is used as the respiratory frequency of the corresponding respiratory pattern, and the distribution characteristics of the frequencies are represented by the mean value of the representative frequencies;

[0094] The mean value of the maximum respiratory amplitude of all respiratory segments within each cluster is used as the maximum value of the respiratory amplitude range corresponding to the respiratory pattern; the mean value of the minimum respiratory amplitude of all respiratory segments within each cluster is used as the minimum value of the respiratory amplitude range corresponding to the respiratory pattern; the overall characteristics of the respiratory amplitude range are represented by the mean value of the maximum respiratory amplitude and the mean value of the minimum respiratory amplitude.

[0095] Please refer to Figure 4 , which shows a schematic diagram of the respiratory amplitude range of all respiratory patterns provided by an embodiment of the present invention; Figure 4 In, the horizontal axis is the serial number of the respiratory pattern, the vertical axis is the respiratory amplitude, and each vertical line corresponds to a respiratory amplitude range.

[0096] Please refer to Figure 5 , which shows a schematic diagram of the respiratory frequency of all respiratory patterns provided by an embodiment of the present invention; Figure 5 In, the horizontal axis is the serial number of the respiratory pattern, the vertical axis is the respiratory frequency, each dot corresponds to a respiratory frequency, and the curve is the change curve of the respiratory frequency with the respiratory pattern.

[0097] Furthermore, considering that the length of the respiratory segment reflects the stable duration of respiration, and the normal respiratory pattern is more stable than other respiratory patterns, so according to the length characteristics of the respiratory segments within each respiratory pattern, the normal respiratory pattern can be effectively screened and obtained as the reference pattern, which is convenient for subsequent real-time judgment of whether the respiratory state is abnormal; at the same time, the normal respiratory pattern judged based on the historical respiratory signals of the current patient can better represent the normal respiratory pattern of the current patient and reduce the influence of physiological differences among different age groups or individuals.

[0098] Preferably, in an embodiment of the present invention, considering that for any respiratory pattern, if its window is larger and the number of respiratory segments is more, it means that the possibility of this respiratory pattern being a normal pattern is greater, so the product of the mean value of the time domain lengths of all respiratory segments of each respiratory pattern and the sum value of the time domain lengths of all respiratory segments is used as the normal possibility coefficient of each respiratory pattern; the length characteristics of the respiratory segments within the respiratory pattern are represented by the mean value and the sum value of the time domain lengths of all respiratory segments of the respiratory pattern;

[0099] The respiratory pattern with the largest normal possibility coefficient is selected as the normal respiratory pattern.

[0100] It should be noted that pediatric patients may have an abnormal respiratory pattern for a long time, and its corresponding normal possibility coefficient is the largest. For this patient, this abnormal respiratory pattern is the respiratory pattern that maintains the stability of the patient's vital signs or the normal operation of the compensation mechanism, so it should be regarded as the individualized normal respiratory pattern of this patient.

[0101] Step S4: Obtain the real-time respiratory signal of the current patient and segment it in real time; Based on the difference characteristics of the respiratory amplitude range between the current respiratory segment to which the current moment belongs and the normal respiratory pattern, as well as the difference characteristics of the respiratory frequency, and combining the respiratory pattern of the current moment to which the current moment belongs with the respiratory patterns of adjacent respiratory segments, and the difference in the number of respiratory segments, obtain the respiratory abnormality coefficient at the current moment; Determine whether to issue a warning based on the respiratory abnormality coefficient at the current moment.

[0102] After obtaining the respiratory frequency and the respiratory amplitude range of the normal respiratory pattern of the current pediatric patient through Steps S1 - S3, the real-time respiratory signal of the current patient can be obtained and segmented in real time, so as to identify whether the current pediatric patient is in an abnormal respiratory state and issue a warning in a timely manner to ensure the safety of the pediatric patient.

[0103] It should be noted that the method of segmenting in real time is the same as the segmentation method of the respiratory segments in Step S1 and will not be elaborated here; Considering that the current moment may be in a new respiratory segment, and the new respiratory segment is relatively short and difficult to show the respiratory characteristics of the patient, so if the current moment is in the preset initial stage of the new respiratory segment, it is determined that the current moment belongs to the previous respiratory segment in the time domain. As an example, the preset initial stage is the first 3 seconds of a respiratory segment, and the implementer can adjust it by himself.

[0104] Considering that the difference characteristics of the respiratory amplitude range and the respiratory frequency between the current respiratory segment and the normal respiratory pattern reflect the current degree of respiratory abnormality of the patient; At the same time, considering that there are differences in the number of respiratory segments corresponding to the normal respiratory pattern and the abnormal respiratory pattern, therefore, the difference in the number of respiratory segments between the respiratory pattern to which the current moment belongs and the respiratory patterns of adjacent respiratory segments can show the current degree of respiratory abnormality from the perspective of the pattern change of the respiratory pattern;

[0105] Therefore, based on the difference characteristics of the respiratory amplitude range between the current respiratory segment to which the current moment belongs and the normal respiratory pattern, as well as the difference characteristics of the respiratory frequency, and combining the respiratory pattern of the current moment to which the current moment belongs with the respiratory patterns of adjacent respiratory segments, and the difference in the number of respiratory segments, obtain the respiratory abnormality coefficient at the current moment, which characterizes the degree of abnormality of the respiratory state of the current pediatric patient at the current moment and provides a basis for subsequent determination of whether to give a warning.

[0106] Preferably, in an embodiment of the present invention, the calculation formula of the respiratory abnormality coefficient includes:

[0107] ;

[0108] wherein, L represents the respiratory abnormality coefficient at the current moment; represents the exponential function with the natural constant e as the base; h is the serial number of the respiratory segment to which the current moment belongs; represents the respiratory amplitude range of the respiratory segment at the current moment; represents the respiratory amplitude range of the normal breathing mode; represents the intersection of the respiratory amplitude range of the respiratory segment at the current moment and the respiratory amplitude range of the normal breathing mode; represents the union of the respiratory amplitude range of the respiratory segment at the current moment and the respiratory amplitude range of the normal breathing mode; represents the linear normalization function; represents taking the absolute value; represents the representative frequency of the respiratory segment at the current moment; represents the respiratory frequency of the normal breathing mode; represents the number of respiratory segments of the respiratory mode adjacent to the current moment; represents the number of respiratory segments of the respiratory mode corresponding to the current moment.

[0109] In the calculation formula of the respiratory abnormality coefficient, through the intersection-union ratio method, represents the difference characteristics of the respiratory amplitude range, The larger it is, the smaller the difference between the respiratory segment at the current moment and the respiratory amplitude range of the normal breathing mode. From the perspective of the respiratory amplitude range, it indicates that the current moment is less likely to be abnormal. Through the exp(-x) function for negative correlation mapping to adjust the logic, the smaller the respiratory abnormality coefficient; through the absolute value of the difference method, represents the difference characteristics of the respiratory frequency, The larger it is, the more likely the current moment is to be abnormal from the perspective of the respiratory frequency, and the larger the respiratory abnormality coefficient; through the ratio method, represents the difference in the number of respiratory segments between the respiratory mode to which the current moment belongs and the respiratory mode of the adjacent respiratory segment. Since the number of segments in the normal breathing mode is the largest and the number of segments in the abnormal breathing mode is smaller, The larger it is, the more likely it is to change from the normal breathing mode to the abnormal breathing mode, and the more likely the current moment is to be abnormal, and the larger the respiratory abnormality coefficient.

[0110] It should be noted that in an embodiment of the present invention, the frequency of early warning determination is once per second, and the implementer can adjust the frequency of early warning determination in combination with the implementation scenario.

[0111] When obtaining the breathing pattern of the breathing segment to which the current moment belongs, as a preferred embodiment, the implementer can select several breathing segments closest to the center of each clustering cluster in the historical breathing signal, such as selecting 5 breathing segments, as the representative segments of each clustering cluster, calculate the average value of the clustering distances between the breathing segment to which the current moment belongs and the representative segments of each clustering cluster, and select the breathing pattern of the clustering cluster with the smallest average value of the clustering distances as the breathing pattern at the current moment;

[0112] In another embodiment of the present invention, it is also possible to obtain the matching coefficient between the breathing segment and each breathing pattern by the difference between the breathing amplitude range of the breathing segment to which the current moment belongs and the breathing amplitude range of each breathing pattern, and combine with the difference in breathing frequency, and select the most matching breathing pattern; the calculation formula of the matching coefficient includes:

[0113] ;

[0114] where s is the serial number of the breathing pattern; represents the matching coefficient between the hth breathing segment and the sth breathing pattern; represents the breathing amplitude range of the sth breathing pattern.

[0115] In the calculation formula of the matching coefficient, the larger the intersection-over-union ratio of the breathing amplitude range, the greater the matching degree between the breathing segment at the current moment and the corresponding breathing pattern, and the larger the matching coefficient; the smaller the [specific condition not clear in the original, seems to be a missing part], the smaller the difference in breathing frequency, the greater the matching degree, and the larger the matching coefficient. Finally, the breathing pattern with the largest matching coefficient is selected as the breathing pattern of the breathing segment to which the current moment belongs.

[0116] In still another embodiment of the present invention, it is also possible to adopt a real-time clustering method to divide the breathing pattern of the breathing segment at the current moment.

[0117] After obtaining the breathing abnormality coefficient characterizing the degree of abnormality of the breathing state of the current child patient at the current moment, it is possible to determine whether to issue a warning based on the breathing abnormality coefficient at the current moment.

[0118] Preferably, in an embodiment of the present invention, when the breathing abnormality coefficient is greater than a preset abnormality threshold, a warning is issued.

[0119] As an example, the preset abnormality threshold is 0.5.

[0120] An embodiment of the present invention further provides a system for identifying critical respiratory states in early childhood, which 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 a method for identifying critical respiratory states in early childhood described in steps S1 - S4.

[0121] In summary, in view of the technical problem that the existing method based on microwave detection and signal decomposition has low accuracy in identifying children's respiratory states, the present invention proposes a method and a system for identifying critical respiratory states in early childhood. The present invention first obtains the historical respiratory signals of the current patient in a preset time interval, segments the historical respiratory signals to obtain respiratory segments; further divides the respiratory patterns according to the differences in respiratory amplitudes and signal components between the respiratory segments; further obtains the respiratory frequency and respiratory amplitude range of each respiratory pattern, and screens out the normal respiratory pattern; further obtains the real - time respiratory signals of the current patient and segments them in real - time; according to the difference characteristics between the current respiratory segment and the normal respiratory pattern, and combining the number difference of respiratory segments between the respiratory pattern to which the current moment belongs and the adjacent respiratory segments, obtains a respiratory abnormality coefficient; finally, determines whether to issue an alarm based on the respiratory abnormality coefficient, reduces the influence of modal mixing, reduces the influence of physiological differences among different age groups or individuals, and improves the accuracy of respiratory state identification.

[0122] It should be noted that: the above - mentioned sequence of embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for identifying a child's early critical respiratory state implemented by a child's early critical respiratory state identification system, characterized in that: The method comprises: obtaining a historical respiratory signal of a current patient within a preset time interval; sliding a preset window on the historical respiratory signal, performing signal decomposition on the historical respiratory signal within the preset window after sliding based on an EMD algorithm, iteratively adjusting the window length according to the frequency concentration characteristics of the signal components, segmenting the historical respiratory signal, and obtaining respiratory segments; According to the difference in breathing amplitude and signal component between any two breathing segments, a cluster distance between the corresponding two breathing segments is obtained; based on the cluster distance, the breathing segments are clustered, and each cluster cluster corresponds to a breathing pattern; According to the overall characteristics of the respiratory amplitude range of all respiratory segments in each cluster and the distribution characteristics of the frequencies of all first signal components, the respiratory frequency and respiratory amplitude range of each respiratory pattern are obtained; according to the length characteristics of the respiratory segments in each respiratory pattern, the normal respiratory pattern is obtained; The real-time breathing signal of the current patient is obtained and segmented in real time; based on the difference characteristics of the breathing amplitude range between the current breathing segment and the normal breathing pattern, as well as the difference characteristics of the breathing frequency, combined with the difference in the number of breathing segments between the current breathing pattern and the breathing pattern of the adjacent breathing segment, the breathing abnormality coefficient at the current moment is obtained; based on the breathing abnormality coefficient at the current moment, it is determined whether to issue an early warning.

2. The method for identifying early critical respiratory status in children according to claim 1, characterized in that: The method for obtaining the breathing segment comprises: The preset window starts to slide from the starting point of the historical respiratory signal, and the window length is iteratively adjusted after each sliding, and the window length determined after each sliding is used as the sliding step length for the next sliding; For the historical breathing signal of the latest window length after any sliding, select any signal component as the target signal component; obtain the spectrum of the target signal component, translate the straight line parallel to the frequency axis from the maximum value to the minimum value of the spectrum amplitude, and obtain the intersection length of the straight line corresponding to each spectrum amplitude and the spectrum line; according to the variation characteristics of the intersection length of the target signal component with the spectrum amplitude, obtain the frequency concentration sub-coefficient of the target signal component; According to the mean of the frequency concentration sub-coefficients of all the signal components, combined with the difference between the frequency concentration sub-coefficient of each signal component and the mean of the frequency concentration sub-coefficients, the frequency concentration coefficient of the latest window length is obtained; the mean of the frequency concentration sub-coefficients is positively correlated with the frequency concentration coefficient; the difference between the frequency concentration sub-coefficients and the mean of the frequency concentration sub-coefficients is negatively correlated with the frequency concentration coefficient; If the frequency concentration coefficient of the current latest window length is less than the first preset concentration threshold, the window length is reduced by the preset adjustment step to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the second preset concentration threshold, the window length is increased by the preset adjustment step to obtain a new window length; if the frequency concentration coefficient of the current latest window length is greater than or equal to the first preset concentration threshold and less than the second preset concentration threshold, the iteration is terminated and the historical breathing signal corresponding to the current latest window length is taken as a breathing segment.

3. The method for identifying the early critical respiratory state of children according to claim 2, characterized in that: The method for obtaining the sub-coefficients in the frequency concentration includes: After normalizing the sum of the products of all the negative correlation mapping values ​​of the intersection lengths of the target signal component and the corresponding spectrum amplitudes, the normalized result is used as the frequency concentrated sub-coefficient of the target signal component.

4. The method for identifying critical respiratory status in early childhood according to claim 1, characterized in that: The method for obtaining the cluster distance includes: The frequency at the maximum value of the spectrum amplitude in the spectrum diagram of each signal component is recorded as the representative frequency; any two of the breathing segments are obtained to form a target binary group, and for each of the signal components in any of the breathing segments in the target binary group, the signal component with the closest representative frequency in the other breathing segment is obtained as the matching signal component; The calculation formula for cluster distance includes: ； Among them, u and v are the serial numbers of the breathing segments; represents the clustering distance between the u-th breathing segment and the v-th breathing segment; represents linear normalization; represents the absolute value of the difference between the maximum breathing amplitude of the u-th breathing segment and the v-th breathing segment; represents the absolute value of the difference between the minimum breathing amplitude of the u-th breathing segment and the v-th breathing segment; i represents the sequence number of the signal component of the u-th breathing segment; k represents the sequence number of the signal component of the v-th breathing segment; represents the number of signal components of the u-th respiratory segment; represents the number of signal components of the vth respiratory segment; represents the representative frequency of the i-th signal component of the u-th respiratory segment; represents the representative frequency of the i-th signal component of the u-th breathing segment and the corresponding matching signal component in the v-th breathing segment; Indicates taking the absolute value; represents the representative frequency of the kth signal component of the vth respiratory segment; Represents the kth signal component of the vth respiratory segment and the representative frequency of the corresponding matching signal component in the uth respiratory segment.

5. The method for identifying the early critical respiratory state of children according to claim 4, characterized in that: The method for obtaining the respiratory frequency and respiratory amplitude range of each respiratory mode includes: Taking the mean of the representative frequencies of the first signal components of all the respiratory segments in each cluster as the respiratory frequency of the corresponding respiratory pattern; The average of the maximum values ​​of the breathing amplitude of all the breathing segments in each cluster is used as the maximum value of the breathing amplitude range of the corresponding breathing pattern; the average of the minimum values ​​of the breathing amplitude of all the breathing segments in each cluster is used as the minimum value of the breathing amplitude range of the corresponding breathing pattern.

6. The method for identifying early critical respiratory status in children according to claim 1, characterized in that: The method for acquiring the normal breathing pattern includes: The product of the mean value of the time domain lengths of all the breathing segments of each breathing pattern and the sum of the time domain lengths of all the breathing segments is used as the normal possible coefficient of each breathing pattern; The breathing pattern with the largest normal possible coefficient is selected as the normal breathing pattern.

7. The method for identifying early critical respiratory status in children according to claim 5, characterized in that: The calculation formula of the abnormal breathing coefficient includes: ； Wherein, L represents the respiratory abnormality coefficient at the current moment; represents an exponential function with the natural constant e as the base; h is the serial number of the breathing segment to which the current moment belongs; Indicates the breathing amplitude range of the breathing segment at the current moment; Respiratory amplitude range representing normal breathing pattern; Indicates the intersection of the breathing amplitude range of the breathing segment at the current moment and the breathing amplitude range of the normal breathing pattern; The union of the breathing amplitude range of the breathing segment at the current moment and the breathing amplitude range of the normal breathing pattern; represents the linear normalization function; Indicates taking the absolute value; Indicates the representative frequency of the breathing segment at the current moment; Respiratory rate, which indicates a normal breathing pattern; The number of breathing segments representing the breathing pattern of adjacent breathing segments at the current moment; Indicates the number of breathing segments corresponding to the breathing pattern at the current moment.

8. The method for identifying the early critical respiratory state of children according to claim 7, characterized in that: The method for determining whether to issue an early warning based on the abnormal breathing coefficient at the current moment includes: When the abnormal breathing coefficient is greater than a preset abnormal threshold, an early warning is issued.

9. The method for identifying the early critical respiratory state of children according to claim 2, characterized in that: The preset adjustment step is 1 second.

10. A system for identifying early critical respiratory states in children, 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 method for identifying early critical respiratory state of children as described in any one of claims 1 to 9 are implemented.

Images