Respiratory and lung sound auxiliary recognition method and system for clinical nursing

Abstract

The present invention relates to the technical field of lung sound signal analysis, and specifically to a respiratory lung sound auxiliary identification method and system for clinical nursing. The present invention first obtains the segmented lung sound signal, segmented signal component and segmented respiratory signal of each day; further, according to the relevant characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation law characteristics of the segmented signal component, the initial qualified component is obtained; further, according to the fluctuation difference characteristics of each qualified signal component segment and the segmented lung sound signal corresponding to the time series, combined with the lung sound law coefficient, the noise influence coefficient of each qualified signal component segment is obtained, and the low-noise period is screened out; further, according to the distribution similarity characteristics of the low-noise period in the initial qualified components of the current day and all historical days, combined with the similar characteristics of the noise influence coefficient, the credibility of each initial qualified component in the current day is obtained; finally, the optimal signal component of the current day is obtained based on the credibility screening.

Description

Technical Field

[0001] The present invention relates to the technical field of lung sound signal analysis, and in particular to a respiratory lung sound auxiliary recognition method and system for clinical nursing. Background Art

[0002] When local inflammation or bronchial stenosis in the lungs prevents air from entering the alveoli evenly, intermittent breath sounds may occur, which are common in diseases such as tuberculosis and pneumonia. For some patients, long-term lung sound monitoring is required, and some patients will be connected to ventilators and other monitoring devices, which can provide real-time respiratory data, including lung auscultation sounds.

[0003] When the patient is in pain or mentally tense, intermittent additional sounds of muscle contractions will also be collected. This intermittent additional sound has nothing to do with lung disease. In addition, during long-term monitoring, it will be affected by the noise in the environment, such as the conversations of other patients in the ward, the footsteps in the corridor, the operation of machinery and equipment, etc., which will also interfere with the collection of lung sounds. These noises will mask the lung sound signals, especially some weak abnormal lung sounds, making it difficult to distinguish the true lung sound characteristics from the collected lung sound signals. Summary of the invention

[0004] In order to solve the technical problem that noise interference exists in lung sound signals and affects the analysis of lung sound characteristics, the purpose of the present invention is to provide a method and system for assisting in the recognition of respiratory and lung sounds for clinical nursing. The technical solutions adopted are as follows:

[0005] A method for assisting the recognition of respiratory and lung sounds for clinical nursing, the method comprising:

[0006] Acquire daily lung sound signals and respiratory signals; perform signal decomposition on the lung sound signals to obtain signal components, and segment the lung sound signals, the signal components, and the respiratory signals according to a preset time domain length to obtain segmented lung sound signals, segmented signal components, and segmented respiratory signals;

[0007] According to the correlation characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation regularity characteristics of the segmented signal component, the lung sound regularity coefficient of each segmented signal component is obtained; the target signal component segment is screened out according to the lung sound regularity coefficient; according to the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient of the target signal component segment in each signal component, an initial qualified component is obtained; each segment of the initial qualified component is a qualified signal component segment;

[0008] According to the fluctuation difference characteristics of each qualified signal component segment and the segmented lung sound signal corresponding to the time series, combined with the lung sound regularity coefficient, the noise influence coefficient of each qualified signal component segment is obtained; according to the noise influence coefficient, the low-noise time period is screened out; according to the low-noise time period in each of the initial qualified components in the current day, the distribution similarity characteristics of the low-noise time period in each of the initial qualified components in all historical days, combined with the similar characteristics of the noise influence coefficient, the credibility of each of the initial qualified components in the current day is obtained;

[0009] The optimal signal component of the current day is obtained based on the confidence level screening.

[0010] Furthermore, the method for obtaining the lung sound regularity coefficient includes:

[0011] Obtain the correlation coefficient between each of the segmented signal components and the segmented breathing signal corresponding to the time series; obtain the lung sound regularity coefficient of each of the segmented signal components based on the numerical stability and interval stability of the maximum value in each of the segmented signal components and the corresponding correlation coefficient; the correlation coefficient, the numerical stability and interval stability of the maximum value are all positively correlated with the lung sound regularity coefficient.

[0012] Furthermore, the method for obtaining the initial qualified component includes:

[0013] Obtaining the overall parameter of each of the signal components at least according to the mean value of the lung sound regularity coefficient of the target signal component segment in each of the signal components;

[0014] Acquire an initial score for each of the signal components according to the number of the lung sound regularity coefficients of the target signal component segment in each of the signal components, the overall parameter, and the autocorrelation coefficient;

[0015] The signal component whose initial score is greater than a second preset threshold is marked as an initial qualified component.

[0016] Furthermore, the method for obtaining the noise-affected coefficient includes:

[0017] Obtaining a fluctuation difference parameter of each qualified signal component segment according to the distribution difference characteristics of the maximum value in the segmented lung sound signal corresponding to each qualified signal component segment in time sequence;

[0018] According to the mutual correlation coefficient between each segment of the qualified signal component segment and the segmented lung sound signal corresponding to the time series, the corresponding lung sound regularity coefficient and the fluctuation difference parameter, the noise influence coefficient of each segment of the qualified signal component segment is obtained; the mutual correlation coefficient is positively correlated with the noise influence coefficient; the fluctuation difference parameter and the lung sound regularity coefficient are both negatively correlated with the noise influence coefficient.

[0019] Furthermore, the method for obtaining the noise-affected coefficient includes:

[0020] According to the difference in the number of maximum values ​​in each qualified signal component segment and the segmented lung sound signal corresponding to the timing, combined with the sum of the maximum value in the qualified signal component segment and the minimum time interval between the maximum value in the segmented lung sound signal corresponding to the timing, the fluctuation difference parameter of each qualified signal component segment is obtained.

[0021] Furthermore, the method for obtaining the credibility includes:

[0022] One of the initial qualified components is randomly selected every day in the current day and all the historical days to form a component group; any of the initial qualified components in the current day is selected as a target qualified component; any of the component groups containing the target qualified component is selected as a target component group;

[0023] Sort the low-noise time periods according to their time sequence in each day; take the ratio of the low-noise time periods with the same sequence number in all days in the target component group to all the low-noise time periods in the current day as the first credible parameter of the target component group of the current day;

[0024] Select any historical day of the current day as the target historical day; in the low-noise time period of the current day and the target historical day with the same serial number, obtain the second credible sub-parameter of the current day and the target historical day according to the overall characteristics of the noise impact coefficient of the low-noise time period of the current day, combined with the difference characteristics of the noise impact coefficient of the low-noise time period of the current day and the noise impact coefficient of the low-noise time period of the target historical day; the overall characteristics of the noise impact coefficient of the low-noise time period of the current day and the difference characteristics of the noise impact coefficient of the low-noise time period of the current day and the noise impact coefficient of the low-noise time period of the target historical day are both positively correlated with the second credible sub-parameter; negatively correlate the means of all the second credible sub-parameters of the current day are used as the second credible parameter of the target component group of the current day;

[0025] According to the first trusted parameter and the second trusted parameter, the initial credibility of the target component group of the current day is obtained; the first trusted parameter and the second trusted parameter are both positively correlated with the initial credibility; the maximum value of all the initial credibility corresponding to the target qualified component is used as the target qualified component credibility.

[0026] Furthermore, the method for obtaining the optimal signal component includes:

[0027] The initial qualified component with the greatest credibility is selected as the optimal signal component.

[0028] Furthermore, the method for acquiring the target signal component segment includes:

[0029] The segmented signal component whose lung sound regularity coefficient is greater than a first preset threshold is marked as a target signal component segment.

[0030] Furthermore, the method for obtaining the low-noise period includes:

[0031] The qualified signal component segment whose noise impact coefficient is less than a third preset threshold is marked as a low-noise period.

[0032] The present invention also proposes a respiratory and lung sound auxiliary recognition system for clinical care, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing any one of the steps of the respiratory and lung sound auxiliary recognition method for clinical care when executing the computer program.

[0033] The present invention has the following beneficial effects:

[0034] The present invention first obtains the segmented lung sound signals, segmented signal components and segmented respiratory signals of each day, so as to facilitate a more detailed analysis of local features and provide a basis for subsequent analysis; further, according to the relevant characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation regularity characteristics of the segmented signal components, the lung sound regularity coefficient of each segmented signal component is obtained, the representativeness of the segmented signal component for the lung sound signal and its own regularity are characterized, and the target signal component segment is screened out according to the lung sound regularity coefficient, and the local features of the signal component are expressed from a local perspective; further, the overall characteristics of the entire signal component are analyzed comprehensively, and the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient of the target signal component segment in each signal component are obtained. , obtain the initial qualified component, reflecting the overall regularity; further, according to the noise influence coefficient of each qualified signal component segment, the degree of elimination of noise in the lung sound signal by the qualified signal component segment is mapped out, the influence of noise on the qualified signal component segment is characterized, and the low-noise period is screened out to prepare for the evaluation of the credibility of the entire qualified signal component; further, according to the distribution similarity characteristics of the low-noise period in each initial qualified component in the current day and all historical days, combined with the similar characteristics of the noise influence coefficient, from the interference similarity angle of fixed interference factors, the credibility of each initial qualified component in the current day is obtained, providing a basis for finally obtaining the optimal signal component; finally, the optimal signal component of the current day is obtained based on the credibility screening. The present invention decomposes and segments the lung sound signal, and screens the signal component in combination with the respiratory signal. Finally, from the interference similarity angle of fixed interference factors, the similar characteristics of the current day and the historical day are analyzed to obtain the optimal signal component of the current day, reduce the noise interference in the original lung sound signal, and provide an accurate analysis basis for relevant personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] 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.

[0036] Figure 1 A flowchart of a method for assisting in identifying respiratory and lung sounds for clinical care provided by one embodiment of the present invention;

[0037] Figure 2 A breathing signal waveform diagram provided by an embodiment of the present invention;

[0038] Figure 3 A flowchart of a method for obtaining credibility provided by an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of a low-noise period of a target component group 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 specific implementation method, structure, features and effects of a respiratory lung sound auxiliary identification method and system for clinical care proposed by the present invention in combination 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 method and system for assisting in identifying respiratory and lung sounds for clinical care 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 assisting the recognition of respiratory and lung sounds for clinical care provided by an embodiment of the present invention, specifically comprising:

[0044] Step S1: Acquire daily lung sound signals and respiratory signals; perform signal decomposition on the lung sound signals to obtain signal components, and segment the lung sound signals, signal components and respiratory signals according to a preset time domain length to obtain segmented lung sound signals, segmented signal components and segmented respiratory signals.

[0045] In the embodiment of the present invention, considering that lung activity is closely related to breathing, the lung sound signal and the breathing signal of each day are first obtained, and the breathing signal is used to provide more analysis basis for the lung sound signal, thereby providing a basis for subsequent analysis; in order to eliminate the interference of various noises in the lung sound signal, the lung sound signal is decomposed to extract the most representative optimal signal component from the complex lung sound signal to represent the lung sound signal, thereby providing an accurate and effective analysis basis; considering that the patient's activity pattern changes in different time periods, such as daytime activities and nighttime rest, which will cause different change patterns of the lung sound signal and the breathing signal, the lung sound signal, the signal component and the breathing signal are segmented according to a preset time domain length to obtain segmented lung sound signals, segmented signal components and segmented breathing signals, thereby facilitating a more detailed analysis of local features.

[0046] As an example, an acquisition frequency of 44.1 kHz and a quantization bit number of 16 bits are set to acquire lung sound signals, and a respiratory sensor is used to acquire respiratory signals at the same time. Lung sound signals and respiratory signals are acquired at intervals of days, that is, the time domain length of the lung sound signals and respiratory signals is 1 day. Signal decomposition is performed using an empirical mode decomposition (EMD) algorithm, and from 0:00 to 24:00 every day with a preset time domain length of 2 hours, the lung sound signal is equally divided into segmented lung sound signals, the signal components are equally divided into segmented signal components, and the respiratory signal is equally divided into segmented respiratory signals.

[0047] It should be noted that the EMD algorithm is already an existing technology, the acquisition frequency is generally between 44.1kHz-100kHz, and the number of quantization bits is generally 16-24 bits to ensure the accuracy and resolution of the collected lung sound signals; in other embodiments of the present invention, the implementer may also set other acquisition frequencies and quantization bits, set other preset time domain lengths for segmentation, and may also use the LMD algorithm for signal decomposition, which will not be elaborated here.

[0048] See also Figure 2 , which shows a breathing signal waveform diagram provided by an embodiment of the present invention, Figure 2 The horizontal axis is the time axis and the vertical axis is the amplitude axis.

[0049] Step S2: According to the correlation characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation regularity characteristics of the segmented signal component, the lung sound regularity coefficient of each segmented signal component is obtained; the target signal component segment is screened out according to the lung sound regularity coefficient; according to the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient of the target signal component segment in each signal component, the initial qualified component is obtained; each segment of the initial qualified component is a qualified signal component segment.

[0050] Considering that different signal components have different representations of lung sound signals and are affected by noise in different ways, the signal components are screened first. Since the periodicity of the patient's lung sound signal is highly consistent with the periodicity of the respiratory signal, for example, when the patient's respiratory rate increases, each respiratory cycle is shortened, and the time for intermittent respiratory sounds to appear in each cycle is also shortened accordingly. Therefore, the correlation characteristics of the segmented signal component and the segmented respiratory signal corresponding to the time series reflect the representativeness of the segmented signal component for the lung sound signal;

[0051] Other noises, such as the additional sound caused by muscle contraction and noise in the environment, appear randomly and may appear at any time period with unclear periodicity, such as the sudden contraction of intercostal muscles, the sound of door opening and closing, and the sound of conversation. All these noises may affect the regularity of lung sound signals. Therefore, the fluctuation regularity of the segmented signal components reflects the influence of noise on the segmented signal components.

[0052] Therefore, according to the correlation characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation law characteristics of the segmented signal component, the lung sound regularity coefficient of each segmented signal component is obtained to characterize the representativeness of the segmented signal component for the lung sound signal and its own regularity, so as to provide a basis for screening the segmented signal components and finally screening the signal components.

[0053] Preferably, in one embodiment of the present invention, considering that the correlation coefficient can directly represent the correlation between signals, the larger the correlation coefficient is, the stronger the correlation is, the more it can represent the regularity of lung sounds, and the larger the regularity coefficient of lung sounds is; the maximum value is an obvious feature point in the signal, the more stable the value of the maximum value is, the more stable the interval is, the more it can be explained that the segmented signal components are more regular, the less affected by random environmental noise is, and the larger the regularity coefficient of lung sounds is;

[0054] Based on this, the correlation coefficient between each segmented signal component and the segmented respiratory signal corresponding to the time series is obtained; according to the numerical stability and interval stability of the maximum value in each segmented signal component, combined with the corresponding correlation coefficient, the lung sound regularity coefficient of each segmented signal component is obtained; the correlation coefficient, the numerical stability of the maximum value and the interval stability are all positively correlated with the lung sound regularity coefficient.

[0055] As an example, the correlation coefficient is the Pearson correlation coefficient, and the Pearson correlation coefficient is normalized by the maximum and minimum values; the variance of the numerical value of the segmented signal component is obtained to indicate the stability of the numerical value of the maximum value. The larger the variance of the numerical value, the worse the numerical stability; the time interval between adjacent maximum value points is obtained, and the average value of the absolute value of the difference between the time intervals of two adjacent groups of adjacent maximum value points is obtained as the interval difference coefficient to indicate the interval stability of the maximum value. The larger the interval difference coefficient, the worse the interval stability; the calculation formula of the lung sound regularity coefficient includes:

[0056] ;

[0057] Wherein, a represents the serial number of the segmented signal component; represents the lung sound regularity coefficient of the a-th segment signal component; represents the correlation coefficient between the a-th segment signal component and the corresponding segment respiratory signal; It represents the exponential function with the natural constant e as the base; Represents the variance of all the maximum values ​​in the a-th segment signal component; Represents the interval difference coefficient of the segment signal component of segment a.

[0058] In the formula, through The product of the interval difference coefficient and the variance of the maximum value is negatively correlated and fused by multiplication. and ; represents the independent variable, As an independent variable The function performs negative correlation mapping.

[0059] In other embodiments of the present invention, the implementer may also obtain the variance of the time intervals between adjacent maximum value points, indicating the interval stability, as the interval difference coefficient; and may also fuse the time intervals by positive correlation methods such as addition or weighted summation. and ; You can also use the reciprocal method to Negative correlation mapping is performed; and the minimum value can be selected to replace the maximum value for analysis, and the cosine similarity or DTW similarity is used to replace the Pearson correlation coefficient as the correlation coefficient to obtain the lung sound regularity coefficient, which are all existing technologies and will not be repeated here.

[0060] By obtaining the representativeness of the segmented signal components for the lung sound signal and the degree of their own regularity, the target signal component segments can be screened out according to the lung sound regularity coefficient, thus preparing for screening the signal components.

[0061] Preferably, in one embodiment of the present invention, considering that the larger the lung sound regularity coefficient is, the more regular the segmented signal component is and the less affected by random environmental noise is, the segmented signal component whose lung sound regularity coefficient is greater than the first preset threshold is marked as a target signal component segment.

[0062] As an example, the first preset threshold is 0.6; in other embodiments of the present invention, the implementer may set other first preset thresholds.

[0063] Taking into account that the lung sound regularity coefficient can only represent the regular characteristics of local signal components, it is necessary to further analyze the overall characteristics of the entire signal component. Taking into account that the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient reflect the overall noise interference of the target signal component segment within a specific time range, reflecting the overall regularity, the initial qualified component is obtained according to the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient of the target signal component segment in each signal component; each segment of the initial qualified component is a qualified signal component segment.

[0064] Preferably, in one embodiment of the present invention, considering that the more target component segments in the signal component, the larger the lung sound regularity coefficient of the target component segment is as a whole, the more regular the signal component is as a whole, the stronger its representativeness is, and the higher the initial score is; the autocorrelation coefficient reflects the periodic consistency of the signal component, and the lung sound signal usually has a periodic feature. The higher the autocorrelation coefficient, the more significant the periodicity of the signal component is, and the better it can reflect the real lung sound features, the higher the initial score is;

[0065] Based on this, the overall parameter of each signal component is obtained according to at least the mean value of the lung sound regularity coefficient of the target signal component segment in each signal component;

[0066] Obtaining an initial score for each signal component according to the number, overall parameters, and autocorrelation coefficient of the lung sound regularity coefficients of the target signal component segment within each signal component;

[0067] The signal component whose initial score is greater than the second preset threshold is marked as an initial qualified component.

[0068] As an example, the mean of the lung sound regularity coefficients of the target signal component segment in each signal component is used as the overall parameter of each signal component; after the autocorrelation coefficient is linearly normalized, the product of the normalized autocorrelation coefficient, the number of lung sound regularity coefficients and the overall parameter is linearly normalized and used as the initial score of the corresponding signal component; the second preset threshold is 0.5, and the signal component with an initial score greater than 0.5 is marked as an initial qualified component.

[0069] As another example, the mean, median and mode of the lung sound regularity coefficient of the target signal component segment in each signal component are weighted and summed with weights of 0.6, 0.2 and 0.2, and the weighted sum result is used as the overall parameter of each signal component; then the initial score and the initial qualified component are obtained.

[0070] It should be noted that when there is only one target signal component segment in the signal component, the autocorrelation coefficient cannot be calculated, and the periodicity of the signal component cannot be analyzed by the autocorrelation coefficient. At this time, the average of the autocorrelation coefficients of all signal components is taken as the autocorrelation coefficient of this signal component to eliminate the influence of the autocorrelation coefficient.

[0071] Step S3: According to the fluctuation difference characteristics of each qualified signal component segment and the corresponding segmented lung sound signal in time series, combined with the lung sound regularity coefficient, the noise influence coefficient of each qualified signal component segment is obtained; according to the noise influence coefficient, the low-noise time period is screened out; according to the distribution similarity characteristics of the low-noise time period in each initial qualified component in the current day and the low-noise time period of each initial qualified component in all historical days, combined with the similar characteristics of the noise influence coefficient, the credibility of each initial qualified component in the current day is obtained.

[0072] There may be multiple initial qualified components screened out in step S2, and the noise contained in each initial qualified component is not completely the same, so it is necessary to further analyze the initial qualified components.

[0073] Taking into account that the undecomposed segmented lung sound signal contains noise, the less noise the qualified signal component segment contains, the more obvious the fluctuation difference characteristics between the qualified signal component segment and the segmented lung sound signal corresponding to the time series. At the same time, the lung sound regularity coefficient represents the representativeness of the lung sound signal and its own regularity. Therefore, according to the fluctuation difference characteristics between each qualified signal component segment and the segmented lung sound signal corresponding to the time series, combined with the lung sound regularity coefficient, the noise influence coefficient of each qualified signal component segment is obtained to represent the influence of noise on the qualified signal component segment, so as to prepare for the subsequent screening to determine the low-noise period and evaluate the credibility.

[0074] Preferably, in one embodiment of the present invention, considering that the maximum value point is an obvious landmark point in the signal, the greater the maximum value distribution difference is, the greater the fluctuation difference of different signals is, so according to the distribution difference characteristics of the maximum value in each qualified signal component segment and the segmented lung sound signal corresponding to the time sequence, the fluctuation difference parameter of each qualified signal component segment is obtained;

[0075] Considering that the larger the regularity coefficient of lung sound is, the more regular the qualified signal component segment is, the less affected by noise is, and the smaller the noise influence coefficient is; considering that the mutual correlation coefficient can reflect the similar characteristics between signals and can be used as a supplement to the difference in maximum value distribution, the larger the mutual correlation coefficient is, the more consistent the change trend is, indicating that the qualified signal component segment has not effectively removed noise interference, and the greater the noise influence coefficient is;

[0076] Based on this, the noise influence coefficient of each qualified signal component segment is obtained according to the mutual correlation coefficient between each qualified signal component segment and the corresponding segmented lung sound signal in time series, the corresponding lung sound regularity coefficient and the fluctuation difference parameter; the mutual correlation coefficient is positively correlated with the noise influence coefficient; the fluctuation difference parameter and the lung sound regularity coefficient are both negatively correlated with the noise influence coefficient.

[0077] As an example, the calculation formula of the noise influence coefficient includes:

[0078] ;

[0079] Wherein, b represents the serial number of the qualified signal component segment; represents the lung sound regularity coefficient of the b-th segment of the qualified signal component; Indicates the noise impact coefficient of the b-th segment of the qualified signal component; represents the linear normalization function; The cross-correlation coefficient between the qualified signal component segment b and the corresponding segmented lung sound signal in time sequence; It represents the exponential function with the natural constant e as the base; Indicates the fluctuation difference parameter corresponding to the b-th segment of the qualified signal component.

[0080] In the calculation formula of the coefficient affected by noise, the cross-correlation coefficient is normalized by a linear normalization function. The larger the value, the more similar the fluctuation of the qualified signal component segment is to the corresponding segmented lung sound signal, which means that the qualified signal component segment has not effectively removed the noise interference and the greater the coefficient affected by the noise; the lung sound regularity coefficient and the fluctuation difference parameter are fused by multiplication, and the Negative correlation mapping is performed on the product of the lung sound regularity coefficient and the fluctuation difference parameter to adjust the correlation relationship; fusion is performed by multiplication and , the fluctuation difference between the qualified signal component segment and the corresponding segmented lung sound signal is analyzed from the two perspectives of mutual correlation coefficient and maximum distribution, and the noise influence coefficient is obtained to characterize the influence of noise on the qualified signal component segment.

[0081] It should be noted that in other embodiments of the present invention, the implementer may also obtain the DTW similarity between the diaphragm signal component segment and the corresponding segmented lung sound signal, and replace the mutual correlation coefficient to represent the similar features between the signals. The DTW similarity, autocorrelation coefficient and mutual correlation coefficient have been realized by existing technologies; the implementer may also fuse the DTW similarity by positive correlation methods such as addition or weighted summation. and , I will not go into details here.

[0082] It should be noted that the length of the lung sound signals and signal components of each day is consistent, the segmentation method is consistent, and the time series correspondence is the correspondence of the start and end time in the time domain.

[0083] Preferably, in one embodiment of the present invention, considering that the greater the difference in the number of maximum values ​​in different signals, the greater the difference in time intervals, which means that the greater the difference in the distribution of the maximum values, the greater the difference in the fluctuations of the two signals, therefore, the fluctuation difference parameter of each qualified signal component segment is obtained based on the difference in the number of maximum values ​​in each qualified signal component segment and the segmented lung sound signal corresponding to the timing, combined with the sum of the minimum time intervals between the maximum values ​​in the qualified signal component segment and the maximum values ​​in the corresponding segmented lung sound signal.

[0084] As an example, the minimum time interval between the maximum value in the qualified signal component segment and the maximum value in the corresponding segmented lung sound signal is: select any maximum value in the qualified signal component segment as the component maximum value, obtain the time interval between the component maximum value and each maximum value in the corresponding segmented lung sound signal, and finally select the minimum time interval; then combine the sum of the minimum time intervals corresponding to all the maxima in the grid signal component segment and the product of the difference in the number of maxima as the fluctuation difference parameter of each qualified signal component segment.

[0085] It should be noted that in other embodiments of the present invention, the implementer may also fuse the sum of the minimum time intervals and the difference between the number of maximum values ​​by a positive correlation method such as addition or weighted summation, which will not be described in detail.

[0086] It should be noted that the implementer can select the minimum value to replace the maximum value as the analysis object to obtain the fluctuation difference parameter and the coefficient affected by noise.

[0087] After obtaining the noise influence coefficient that characterizes the influence of noise on the qualified signal component segment, the low-noise period can be screened out by the noise influence coefficient.

[0088] Preferably, in one embodiment of the present invention, considering that the smaller the noise impact coefficient is, the less the qualified signal component segment is affected by noise, the qualified signal component segment with a noise impact coefficient less than a third preset threshold is marked as a low noise period.

[0089] As an example, the third preset threshold is 0.4; in other embodiments of the present invention, the implementer may set other third preset thresholds.

[0090] Considering that there are some definite factors that cause noise, such as daily ward rounds, medication and cleaning, which are usually carried out at fixed times, the undecomposed lung sound signals are affected by noise during these time periods, and the corresponding initial qualified components will have low-noise periods. Therefore, based on the low-noise periods in each initial qualified component in the current day, and the distribution similarity characteristics of the low-noise periods of each initial qualified component in all historical days, combined with the similar characteristics of the coefficient affected by noise, from the perspective of interference similarity of fixed interference factors, the credibility of each initial qualified component in the current day is obtained, providing a basis for ultimately obtaining the optimal signal component.

[0091] Preferably, in one embodiment of the present invention, the method for obtaining credibility includes:

[0092] See also Figure 3 , which shows a flow chart of a method for obtaining credibility provided by an embodiment of the present invention, specifically comprising:

[0093] Step S301: randomly extract an initial qualified component every day in the current day and all historical days to form a component group; select any initial qualified component in the current day as the target qualified component; select any component group containing the target qualified component as the target component group.

[0094] Taking into account the differences in the signal decomposition results of lung sound signals on different days, an initial qualified component is randomly selected every day in the current day and all historical days, and a full arrangement is performed. Each arrangement constitutes a component group; any initial qualified component in the current day is selected as the target qualified component; any component group containing the target qualified component is selected as the target component group, which is convenient for analysis one by one.

[0095] Step S302: Sort the low-noise periods according to their time sequence in each day; take the ratio of the low-noise periods with the same sequence number in all days in the target component group to all the low-noise periods in the current day as the first credible parameter of the target component group of the current day.

[0096] See also Figure 4 , which shows a schematic diagram of a low-noise period of a target component group provided by an embodiment of the present invention; Figure 4 It contains the ath target qualified component of the current day, the dth initial qualified component of the first day of history, and the eth initial qualified component of the second day of history. The numbers are the serial numbers of the time periods. The time period between the two long vertical lines is the low-noise time period. For example, the time periods with serial numbers 1, 2, and 3 in the ath are low-noise time periods, and the low-noise time period with serial number 3 is the low-noise time period with the same serial number on all days in the target component group.

[0097] In the target qualified component of the current day, the more low-noise periods appear in all days and the larger the proportion, the more likely it is that it is an interference time period corresponding to a fixed influencing factor. The better the denoising effect of the target qualified component, the more it can reflect the inherent fluctuation characteristics of lung sounds, the higher the credibility, and the larger the first credibility parameter.

[0098] Step S303: Select any historical day of the current day as the target historical day; in the low-noise time periods with the same serial numbers of the current day and the target historical day, obtain the second credible sub-parameters of the current day and the target historical day according to the overall characteristics of the noise impact coefficient of the low-noise time period of the current day, combined with the difference characteristics of the noise impact coefficient of the low-noise time period of the current day and the noise impact coefficient of the low-noise time period of the target historical day; negatively correlate the means of all the second credible sub-parameters of the current day and use them as the second credible parameters of the target component group of the current day.

[0099] Considering that the smaller the difference in the interference influence coefficient between the current day and the same low-noise period in the historical days, the more consistent the interference situation is, the more likely it is that the interference is caused by fixed factors, and the higher the credibility is; at the same time, the smaller the interference influence coefficient between the current day and the same low-noise period in the historical days, the smaller the interference degree of the target qualified component is, the better the denoising effect is, the more it can reflect the inherent fluctuation characteristics of the lung sound, and the higher the credibility is, so the second credible sub-parameters of the current day and the target historical day are obtained; and considering that the target component group contains multiple historical days, the means of all the second credible sub-parameters of the current day are negatively correlated and mapped as the second credible parameters of the target component group of the current day.

[0100] As an example, the calculation formula of the second trusted parameter includes:

[0101] ;

[0102] Where x represents the serial number of the target component group of the current day; The second credible parameter representing the xth target component group of the current day; represents an exponential function with the natural constant e as the base; i represents the serial number of the historical day of the current day; I represents the number of the historical days of the current day; Indicates the average value of the noise impact coefficient of the low-noise period of the current day in the low-noise period with the same serial number as the target qualified component of the current day and the initial qualified component of the i-th historical day in the x-th target component group of the current day; Indicates the number of low-noise periods with the same sequence number between the target qualified component of the current day and the initial qualified component of the i-th historical day in the x-th target component group of the current day; Indicates the sequence number of the low-noise period with the same sequence number as the target qualified component of the current day and the initial qualified component of the i-th historical day in the x-th target component group of the current day; Indicates the difference between the target qualified component of the current day and the initial qualified component of the i-th historical day in the x-th target component group of the current day. The absolute value of the difference between the noise-affected coefficients of the low-noise time periods with the same serial number; Represents the second credible sub-parameter between the current day and the i-th historical day in the target component group.

[0103] In the calculation formula of the second credible parameter, the overall characteristics of the noise impact coefficient of the low-noise period of the current day, and the difference characteristics of the noise impact coefficient of the low-noise period of the current day and the noise impact coefficient of the low-noise period of the target historical day are both positively correlated with the second credible sub-parameter; the two are fused by multiplication to obtain the second credible sub-parameter; and The mean of the second credible sub-parameter is negatively mapped, the correlation is adjusted, and the credibility of the target qualified component of the current day is evaluated from the similarity of the noise impact coefficients in the low-noise period of the current day and the historical day, providing a basis for finally obtaining the credibility.

[0104] Step S304: According to the first credible parameter and the second credible parameter, the initial credibility of the target component group of the current day is obtained; the maximum value of all initial credibility corresponding to the target qualified components is used as the target qualified component credibility.

[0105] After obtaining the first credible parameter and the second credible parameter, the initial credibility of the target component group of the current day can be obtained based on the first credible parameter and the second credible parameter, and the credibility of the target qualified component can be comprehensively evaluated by integrating the two angles of similar distribution in the low-noise period and similar coefficient of noise influence. Considering that there are multiple component groups corresponding to the target qualified component, that is, there are multiple initial credibility, among which the maximum value represents the optimal performance of the target qualified component, the maximum value of all initial credibility corresponding to the target qualified component is taken as the credibility of the target qualified component.

[0106] As an example, the calculation formula of credibility includes:

[0107] ;

[0108] Where x represents the serial number of the target component group of the current day; Indicates the initial credibility of the target qualified component of the xth target component group of the current day; represents the first credible parameter of the xth target component group of the current day, , Indicates the duration length of the low-noise period with the same sequence number between the target qualified component of the current day and the initial qualified components of all historical days in the xth target component group of the current day; Indicates the duration of all low-noise periods of the target qualified components of the current day in the xth target component group of the current day; Represents the second credible parameter of the xth target component group of the current day.

[0109] In the calculation formula of credibility, the first credibility parameter and the second credibility parameter are both positively correlated with the initial credibility. The first credibility parameter and the second credibility parameter are integrated by multiplying each other, and the credibility of the target qualified component is comprehensively evaluated from two perspectives: the similarity of the distribution in the low-noise period and the similarity of the coefficient affected by noise, to obtain the initial credibility.

[0110] It should be noted that the current day and all historical days are monitoring of the same patient; in another embodiment of the present invention, taking into account that as the monitoring proceeds, the number of historical days gradually increases, the amount of calculation continues to increase, and the demand for computing resources is likely to be too large, the implementer can limit the total number of historical days of the current day to not exceed a preset day threshold such as 6. When the total number of historical days exceeds 6, the latest 6 historical days are selected for analysis.

[0111] Step S4: Filter and obtain the optimal signal component of the current day based on the credibility.

[0112] The signal components of the current day are screened multiple times, and the credibility of the initial qualified components is evaluated from the perspective of interference similarity of fixed interference factors to characterize the credibility of the initial qualified components. Finally, the optimal signal components of the current day can be obtained based on credibility screening, reducing noise interference in the original lung sound signal, providing an accurate analysis basis for relevant personnel.

[0113] Preferably, in one embodiment of the present invention, the initial qualified component with the greatest credibility is selected as the optimal signal component.

[0114] It should be noted that, when there are multiple initial qualified components with the highest credibility, the initial qualified components with the highest credibility are all used as optimal signal components for use by relevant personnel.

[0115] An embodiment of the present invention also provides a respiratory and lung sound auxiliary recognition system for clinical care, the system comprising a memory, a processor and a computer program, wherein the memory is used to store the corresponding computer program, the processor is used to run the corresponding computer program, and when the computer program is running in the processor, it can implement a respiratory and lung sound auxiliary recognition method for clinical care described in steps S1-S4.

[0116] In summary, the present invention aims at the technical problem that noise interference exists in lung sound signals and affects the analysis of lung sound characteristics, and proposes a respiratory lung sound auxiliary identification method and system for clinical nursing. The present invention first obtains the segmented lung sound signal, segmented signal component and segmented respiratory signal of each day; further, according to the relevant characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation law characteristics of the segmented signal component, the initial qualified component is obtained; further, according to the fluctuation difference characteristics of each qualified signal component segment and the segmented lung sound signal corresponding to the time series, combined with the lung sound law coefficient, the noise influence coefficient of each qualified signal component segment is obtained, and the low-noise period is screened out; further, according to the low-noise period in each initial qualified component in the current day, the distribution similarity characteristics of the low-noise period of each initial qualified component in all historical days, combined with the similar characteristics of the noise influence coefficient, the credibility of each initial qualified component in the current day is obtained; finally, the optimal signal component of the current day is obtained based on the credibility screening. The present invention decomposes and segments the lung sound signal, and screens the signal components in combination with the respiratory signal. Finally, from the perspective of interference similarity of fixed interference factors, the similar characteristics of the current day and the historical days are analyzed to obtain the optimal signal component of the current day, reduce the noise interference in the original lung sound signal, and provide an accurate analysis basis for relevant personnel.

[0117] 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.

[0118] 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 respiratory and lung sound auxiliary recognition method for clinical nursing, characterized in that: The method comprises: Acquire daily lung sound signals and respiratory signals; the respiratory signals are collected by a respiratory sensor; perform signal decomposition on the lung sound signals to obtain signal components, and segment the lung sound signals, the signal components and the respiratory signals according to a preset time domain length to obtain segmented lung sound signals, segmented signal components and segmented respiratory signals; According to the correlation characteristics of each segmented signal component and the segmented respiratory signal corresponding to the time series, combined with the fluctuation regularity characteristics of the segmented signal component, the lung sound regularity coefficient of each segmented signal component is obtained; the target signal component segment is screened out according to the lung sound regularity coefficient; according to the fluctuation characteristics and concentration characteristics of the lung sound regularity coefficient of the target signal component segment in each signal component, an initial qualified component is obtained; each segment of the initial qualified component is a qualified signal component segment; According to the fluctuation difference characteristics of each qualified signal component segment and the segmented lung sound signal corresponding to the time series, combined with the lung sound regularity coefficient, the noise influence coefficient of each qualified signal component segment is obtained; according to the noise influence coefficient, the low-noise time period is screened out; according to the low-noise time period in each of the initial qualified components in the current day, the distribution similarity characteristics of the low-noise time period in each of the initial qualified components in all historical days, combined with the similar characteristics of the noise influence coefficient, the credibility of each of the initial qualified components in the current day is obtained; The optimal signal component of the current day is obtained based on the confidence level screening.

2. A method for assisting the identification of respiratory and lung sounds for clinical nursing according to claim 1, characterized in that: The method for obtaining the lung sound regularity coefficient comprises: Obtain the correlation coefficient between each of the segmented signal components and the segmented breathing signal corresponding to the time series; obtain the lung sound regularity coefficient of each of the segmented signal components based on the numerical stability and interval stability of the maximum value in each of the segmented signal components and the corresponding correlation coefficient; the correlation coefficient, the numerical stability and interval stability of the maximum value are all positively correlated with the lung sound regularity coefficient.

3. The respiratory and lung sound auxiliary recognition method for clinical nursing according to claim 1, characterized in that: The method for obtaining the initial qualified component includes: Obtaining the overall parameter of each of the signal components at least according to the mean value of the lung sound regularity coefficient of the target signal component segment in each of the signal components; Acquire an initial score for each of the signal components according to the number of the lung sound regularity coefficients of the target signal component segment in each of the signal components, the overall parameter, and the autocorrelation coefficient; The signal component whose initial score is greater than a second preset threshold is marked as an initial qualified component.

4. The respiratory and lung sound auxiliary recognition method for clinical nursing according to claim 1, characterized in that: The method for obtaining the noise-affected coefficient includes: Obtaining a fluctuation difference parameter of each qualified signal component segment according to the distribution difference characteristics of the maximum value in the segmented lung sound signal corresponding to each qualified signal component segment in time sequence; According to the mutual correlation coefficient between each segment of the qualified signal component segment and the segmented lung sound signal corresponding to the time series, the corresponding lung sound regularity coefficient and the fluctuation difference parameter, the noise influence coefficient of each segment of the qualified signal component segment is obtained; the mutual correlation coefficient is positively correlated with the noise influence coefficient; the fluctuation difference parameter and the lung sound regularity coefficient are both negatively correlated with the noise influence coefficient.

5. A method for assisting the identification of respiratory and lung sounds for clinical nursing according to claim 4, characterized in that: The method for obtaining the noise-affected coefficient includes: According to the difference in the number of maximum values ​​in each qualified signal component segment and the segmented lung sound signal corresponding to the timing, combined with the sum of the maximum value in the qualified signal component segment and the minimum time interval between the maximum value in the segmented lung sound signal corresponding to the timing, the fluctuation difference parameter of each qualified signal component segment is obtained.

6. The respiratory and lung sound auxiliary recognition method for clinical nursing according to claim 1, characterized in that: The method for obtaining the credibility includes: One of the initial qualified components is randomly selected every day in the current day and all the historical days to form a component group; any of the initial qualified components in the current day is selected as a target qualified component; any of the component groups containing the target qualified component is selected as a target component group; Sort the low-noise time periods according to their time sequence in each day; take the ratio of the low-noise time periods with the same sequence number in all days in the target component group to all the low-noise time periods in the current day as the first credible parameter of the target component group of the current day; Select any historical day of the current day as the target historical day; in the low-noise time period of the current day and the target historical day with the same serial number, obtain the second credible sub-parameter of the current day and the target historical day according to the overall characteristics of the noise impact coefficient of the low-noise time period of the current day, combined with the difference characteristics of the noise impact coefficient of the low-noise time period of the current day and the noise impact coefficient of the low-noise time period of the target historical day; the overall characteristics of the noise impact coefficient of the low-noise time period of the current day and the difference characteristics of the noise impact coefficient of the low-noise time period of the current day and the noise impact coefficient of the low-noise time period of the target historical day are both positively correlated with the second credible sub-parameter; negatively correlate the means of all the second credible sub-parameters of the current day are used as the second credible parameter of the target component group of the current day; According to the first trusted parameter and the second trusted parameter, the initial credibility of the target component group of the current day is obtained; the first trusted parameter and the second trusted parameter are both positively correlated with the initial credibility; the maximum value of all the initial credibility corresponding to the target qualified component is used as the target qualified component credibility.

7. A method for assisting the identification of respiratory and lung sounds for clinical nursing according to claim 6, characterized in that: The method for obtaining the optimal signal component includes: The initial qualified component with the greatest credibility is selected as the optimal signal component.

8. The method for assisting in identifying respiratory and lung sounds for clinical nursing according to claim 2, characterized in that: The method for acquiring the target signal component segment comprises: The segmented signal component whose lung sound regularity coefficient is greater than a first preset threshold is marked as a target signal component segment.

9. The method for assisting in identifying respiratory and lung sounds for clinical nursing according to claim 4, characterized in that: The method for obtaining the low-noise period includes: The qualified signal component segment whose noise impact coefficient is less than a third preset threshold is marked as a low-noise period.

10. A respiratory and lung sound auxiliary recognition system for clinical nursing, 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 respiratory and lung sound auxiliary identification method for clinical care as described in any one of claims 1 to 9 are implemented.

Images