CN115721317B - Physiological state monitoring method and monitor based on heart and lung information - Google Patents
Physiological state monitoring method and monitor based on heart and lung information Download PDFInfo
- Publication number
- CN115721317B CN115721317B CN202211287888.2A CN202211287888A CN115721317B CN 115721317 B CN115721317 B CN 115721317B CN 202211287888 A CN202211287888 A CN 202211287888A CN 115721317 B CN115721317 B CN 115721317B
- Authority
- CN
- China
- Prior art keywords
- frequency
- signal
- respiratory
- fusion
- modal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000012544 monitoring process Methods 0.000 title claims abstract description 24
- 230000035790 physiological processes and functions Effects 0.000 title claims abstract description 19
- 210000004072 lung Anatomy 0.000 title claims description 4
- 230000000241 respiratory effect Effects 0.000 claims abstract description 70
- 230000004927 fusion Effects 0.000 claims abstract description 51
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 47
- 230000002612 cardiopulmonary effect Effects 0.000 claims abstract description 22
- 238000007637 random forest analysis Methods 0.000 claims abstract description 21
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 16
- 230000003183 myoelectrical effect Effects 0.000 claims abstract description 15
- 238000001914 filtration Methods 0.000 claims abstract description 13
- 208000019622 heart disease Diseases 0.000 claims abstract description 9
- 230000006872 improvement Effects 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims description 16
- 241000255581 Drosophila <fruit fly, genus> Species 0.000 claims description 14
- 230000000694 effects Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000009286 beneficial effect Effects 0.000 claims description 3
- 238000003066 decision tree Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000036391 respiratory frequency Effects 0.000 claims description 3
- 230000004962 physiological condition Effects 0.000 claims 3
- 230000006870 function Effects 0.000 description 17
- 238000001228 spectrum Methods 0.000 description 12
- 238000005070 sampling Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- 230000003595 spectral effect Effects 0.000 description 6
- 230000008878 coupling Effects 0.000 description 5
- 238000010168 coupling process Methods 0.000 description 5
- 238000005859 coupling reaction Methods 0.000 description 5
- 241000255588 Tephritidae Species 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 210000002345 respiratory system Anatomy 0.000 description 4
- 238000003786 synthesis reaction Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 210000000748 cardiovascular system Anatomy 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 230000001808 coupling effect Effects 0.000 description 2
- 230000002526 effect on cardiovascular system Effects 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 230000003434 inspiratory effect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 230000001131 transforming effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 210000003403 autonomic nervous system Anatomy 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 238000009966 trimming Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The application discloses a physiological state monitoring method based on cardiopulmonary information, which comprises the following steps: acquiring a high-frequency electrocardiosignal, and filtering an myoelectric interference signal based on a wavelet threshold method combined with variation modal decomposition and improvement; acquiring a respiratory flow signal, and removing high-frequency noise interference based on a second-order active low-pass filter; extracting time domain, frequency domain and differential entropy characteristics of signals, carrying out characteristic fusion on the characteristics through an improved random forest algorithm, and carrying out decision fusion based on a D-S evidence theory method and a weighting method to obtain a physiological state monitoring result. The application also provides a physiological state monitor based on cardiopulmonary information. The application realizes the effective fusion of cardiopulmonary information, accurately monitors the physiological state and ensures the monitoring of hidden heart diseases.
Description
Technical Field
The application relates to the field of medical equipment, in particular to a physiological state monitoring method and a physiological state monitoring instrument based on cardiopulmonary information.
Background
There is an inherent coordination mechanism between the cardiovascular system and the respiratory system, and the interaction between them is called cardiopulmonary coupling and therefore also called cardiopulmonary interaction. Compared with a single study on electrocardiosignals or respiratory signals to explain the balance and health condition of the organism, the electrocardiosignals and the respiratory signals are put together for coupling study to obtain the coupling action state between the cardiovascular circulatory system and the respiratory system, so that the cardiovascular system and the respiratory system have higher accuracy and stability.
The research shows that the coupling research based on the body surface electrocardiosignal and the respiratory signal has clinical value of diagnosis, and compared with the traditional single analysis of the electrocardiosignal or the respiratory signal, the analysis result of the cardiopulmonary coupling research has higher consistency, and can well reflect the influence of the autonomic nervous system on the coupling effect of the cardiovascular system and the respiratory system in different physiological states.
Therefore, the following problems exist in the prior art: the cardiopulmonary information cannot be effectively fused, and some hidden heart diseases cannot be found based on conventional electrocardiosignals, so that the health state monitoring cannot achieve a good effect.
Disclosure of Invention
(one) solving the technical problems
In order to solve the technical problems, the application provides a physiological state monitoring method and a physiological state monitoring instrument based on cardiopulmonary information, which realize more accurate health state monitoring by combining feature fusion and decision fusion.
(II) technical scheme
In order to solve the technical problems and achieve the aim of the application, the application is realized by the following technical scheme:
a physiological state monitoring method based on cardiopulmonary information, comprising the steps of:
s1: acquiring high-frequency electrocardiosignals and preprocessing
The high-frequency electrocardiograph is adopted to collect high-frequency electrocardiosignals, the heart and lung information fusion is carried out based on the high-frequency electrocardiosignals, and the hidden heart disease can be recognized based on the characteristic information of the high-frequency signals.
S2: acquiring respiratory signals and preprocessing
The respiratory signal is acquired based on the respiratory flow sensor, and accurate respiratory signals including information such as respiratory frequency and respiratory flow rate can be acquired based on the respiratory flow sensor.
S3: extracting signal features
The method comprises the step of normalizing signals, and extracting features through three aspects, including a time domain, a frequency domain and differential entropy.
S4: signal fusion of high-frequency electrocardiosignals and respiratory signals
And carrying out signal fusion on the high-frequency electrocardiosignals and respiratory signals based on two data fusion modes of deep learning combined feature fusion and decision fusion. The method comprises the following steps: firstly, fusing time domain features and frequency domain features of high-frequency electrocardiosignals, and further fusing the time domain features and differential entropy features; secondly, fusing the time domain features and the frequency domain features of respiration, and further fusing the time domain features and the differential entropy features; and finally, fusing the fusion result of the high-frequency electrocardiosignal and the fusion result of the respiratory signal to obtain a final result.
Further, step S1 further includes:
aiming at power frequency interference:
the power frequency interference of 50hz is filtered by adopting a software method, and based on a second-order IIR filter, the system function is as follows:
wherein,for notch number digital frequency f 0 For notch frequency, f s Is the sampling frequency.
For myoelectric interference:
the application filters myoelectric interference signals based on a wavelet threshold method combining variation modal decomposition and improvement. The method comprises the following steps:
(1) The high-frequency electrocardiosignal is adaptively decomposed into k modal components through variation modal decomposition.
Further, the k value is determined by analyzing the distribution of the center frequencies of the modal components under the continuously varying k value.
(2) The nature of the modal component is determined and is classified as either signal dominant or noise dominant. And the correlation coefficient of the modal component and the high-frequency electrocardiosignal is used as a characteristic quantity, a threshold M is set to distinguish the property of the modal component, and when the correlation coefficient is larger than the threshold, the modal component is identified as signal dominant, otherwise, the modal component is noise dominant.
(3) Wavelet threshold denoising is performed on the noise-dominated modal components.
(4) Reconstructing the modal component of the signal dominant and the modal component after denoising of the wavelet threshold to obtain the denoised electrocardiosignal.
The improved wavelet threshold denoising is as follows:
based on the modified threshold function, the following formula is shown:
wherein w' is the denoised signal, w is the original signal, and T is the threshold.
Further, step S2 includes:
preprocessing the respiratory flow signal includes:
the tidal volume was calculated as follows:
wherein C is tidal volume, n is data volume acquired in one inspiration process, x i Is the acquired inspiratory data value.
Filtering the respiratory flow from the flow waveform by adopting a low-pass filter, only reserving the air leakage, and further subtracting the obtained air leakage from the respiratory flow signal to obtain a respiratory flow compensated result; and a second-order active low-pass filter is adopted to remove high-frequency noise interference.
Further, the features in step S3 specifically include:
(1) Time domain
The time domain information includes: standard deviation, variance, root mean square, first and second differential means of the original and normalized signals, hjorth activity, mobility, complexity.
The first order differential mean value is calculated as follows:
the second order differential mean value is calculated as follows:
kurtosis and skewness of response signal fluctuations:
hjorth activity, mobility, and complexity are calculated as follows:
wherein s is the original signal and has a length of N, mu s Sigma, which is the average value of the original signal s Is the standard deviation of the original signal,s 'is the standard deviation of the original signal and s' is the first derivative of the original signal.
(2) Frequency domain
The extracted frequency domain features include energy spectral density, power spectral intensity, and differential entropy.
Obtaining frequency domain information by:
transforming the n-section time domain signals into frequency domain signals X (n) through FFT; dividing the frequency domain signal into K frequency bands, and extracting the characteristics of each frequency band. The frequency domain features include the following:
a. energy spectral density
Wherein ESD represents an energy spectrum characteristic,f k a lower band representing a kth band, f k+1 Upper band representing the kth band, +.>Representing the quotient of the lower frequency band and the sampling frequency, +.>Representing the quotient of the up-band and the sampling frequency.
b. Power spectral density
Wherein PSD represents the power spectrum characteristic,
c. power spectrum intensity
The power spectrum intensity represents the sum of the amplitudes of the frequency domains, and the calculation formula is as follows:
PSI＝∑|X i |
wherein PSI represents the power spectrum intensity,
(3) Differential entropy
For signals subject to gaussian distribution, the differential entropy is calculated as follows:
further, step S4 includes:
feature fusion is carried out based on an improved random forest algorithm, decision fusion is carried out based on a D-S evidence theory method, and finally the fusion is based on a weighting algorithm.
Further, the improved random forest algorithm is specifically:
a. optimizing the number of decision trees and the initial value of the maximum depth in a random forest algorithm based on a drosophila algorithm; the method comprises the following specific steps:
a1, initializing the position of the fruit fly and the distance between the fruit fly and food:
wherein X is 0 、Y 0 Respectively, the positions of the drosophila are respectively, and i and j are respectively random values.
a2: distance D is calculated and taste concentration determination value S is calculated:
a3 calculating taste concentration value S i :
S i ＝f(S)
Wherein f-function is a taste concentration function.
and a4, obtaining a concentration maximum value and a position optimal value of the individual.
a5, updating the coordinate position and calculating the concentration maximum value of the new position;
a6, comparing the concentration of the new position with the concentration of the original position, if the concentration of the new position is better than the concentration of the original position, executing a5, otherwise executing a1-a4.
a7. And determining whether an end condition is reached, if so, obtaining a final result, and exiting the program.
b. Constructing a random forest classifier based on the optimal initial value to classify, and forming a first-stage classifier;
and the initial parameters in the random forest algorithm are searched for optimal values through the initialized Drosophila algorithm, so that parameter uncertainty caused by the fact that the initial parameters in the algorithm training model are set by adopting artificial experience values can be avoided. The root mean square error of the random forest algorithm training model is selected as the fitness function of the particle swarm algorithm, the smaller the error value is, the more accurate the prediction capability of the model is, the minimum value optimization is carried out on the fitness function by using the drosophila algorithm, and the parameters corresponding to the optimal condition of the random forest model can be determined.
The classification is formed based on the time domain features, the frequency domain features and the differential entropy features respectively, namely a first class classifier, and the classification result is the physiological state grade of the monitored object and is classified into excellent, good, medium, bad and critical.
Further, the decision fusion includes:
decision fusion is carried out based on D-S evidence theory method, and a evidence reasoning system E is assumed i ,Θ＝{θ 1 ,θ 2 …θ n A power set of 2 Θ Where Θ is the set of n mutually exclusive systems, for a subset A of Θ, m () is the base probability of A, if m (A)>0, then A is focal.
The method comprises the following specific steps of fusing a plurality of classifiers based on a D-S evidence theory decision algorithm:
a. calculate evidence E i Degree of conflict k with other evidence ij Form conflict vector K i ＝{k i1 ,k i2 ,…k in And performing standardization processing.
b. Calculating entropy of normalized collision vector
c. Calculate evidence E i Weight coefficient:
d. according to the weight coefficient distribution m and the conflict k, the evidence synthesis mode is formed as follows:
m′(A)＝p(A)+k′×q(A)
wherein,q (A) is the average degree of support of A by evidence, < >>
e. And making a decision according to the synthesis result to obtain a classification result.
The output results of the time domain and frequency domain primary classifier are fused through a D-S evidence theory decision algorithm to form a secondary classification result, the secondary classification result is fused with the primary classifier of the differential entropy through a D-S evidence theory decision algorithm to form a tertiary classification result, and the tertiary classification result of the high-frequency electrocardiosignal and the tertiary classification result of the respiratory signal are further fused in a weighting mode respectively.
The application also provides a physiological state monitor based on cardiopulmonary information, which specifically comprises:
the high-frequency electrocardiosignal acquisition and processing module is used for acquiring and preprocessing the high-frequency electrocardiosignal and aiming at power frequency interference:
the software method is adopted to filter the power frequency interference of 50hz based on a second-order IIR filter, and the system function is as follows:
wherein,for notch number digital frequency f 0 For notch frequency, f s Is the sampling frequency;
for myoelectric interference: filtering myoelectric interference signals based on a wavelet threshold method combining variation modal decomposition and improvement;
the respiratory signal acquisition and processing module is used for acquiring the respiratory signal based on the respiratory flow sensor, and preprocessing the respiratory flow signal comprises the following steps:
filtering the respiratory flow from the flow waveform by adopting a low-pass filter, only reserving the air leakage, and further subtracting the obtained air leakage from the respiratory flow signal to obtain a respiratory flow compensated result; and a second-order active low-pass filter is adopted to remove high-frequency noise interference.
And the feature extraction module is used for extracting signal features, including time domain, frequency domain and differential entropy.
The feature fusion module is used for carrying out signal fusion on the high-frequency electrocardiosignals and respiratory signals based on two data fusion modes of deep learning combined feature fusion and decision fusion, and specifically comprises the following steps: feature fusion is carried out based on a random forest algorithm optimized by combining a drosophila algorithm, and decision fusion is carried out based on a D-S evidence theory method.
(III) beneficial effects
The application acquires the high-frequency electrocardiosignals of the monitored object through the high-frequency electrocardiosignals, and accurately acquires and judges the hidden heart diseases of the monitored object; the myoelectric interference signals are filtered through combining the variation modal decomposition and an improved wavelet threshold method, so that the high-frequency electrocardiosignals are more accurately restored; by combining the signal fusion algorithm of feature fusion and decision fusion, the accurate judgment of the monitoring state is realized.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:
FIG. 1 is a flow chart of a monitoring method according to an embodiment of the application;
fig. 2 is a schematic diagram of a signal fusion algorithm according to an embodiment of the application.
Detailed Description
Embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
Other advantages and effects of the present disclosure will become readily apparent to those skilled in the art from the following disclosure, which describes embodiments of the present disclosure by way of specific examples. It will be apparent that the described embodiments are merely some, but not all embodiments of the present disclosure. The disclosure may be embodied or practiced in other different specific embodiments, and details within the subject specification may be modified or changed from various points of view and applications without departing from the spirit of the disclosure. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.
It should also be noted that the illustrations provided in the following embodiments merely illustrate the basic concepts of the disclosure by way of illustration, and only the components related to the disclosure are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
Referring to fig. 1, the physiological state monitoring method based on cardiopulmonary information includes:
s1: acquiring high-frequency electrocardiosignals and preprocessing
The conventional electrocardiograph with a frequency response range of about 40Hz is not more than 100Hz, namely, only the low-frequency signal of the electrocardio can be traced, and the high-frequency signal above 100Hz can not be filtered out and can not be reacted, so that a lot of information with diagnostic value on heart diseases is lost, the electrocardio high-frequency information refers to the electrocardio components with the frequency of more than 100Hz, the diagnostic value of the electrocardio high-frequency components on heart diseases, especially on coronary heart diseases, is gradually brought to the attention of cardiovascular clinical workers, and the high-frequency electrocardiograph is gradually becoming a new method for detecting the non-invasive heart diseases and is applied to clinic.
Therefore, the application adopts the high-frequency electrocardiograph to collect the high-frequency electrocardiosignals, carries out cardiopulmonary coupling based on the high-frequency electrocardiosignals, and is beneficial to identifying the existence of hidden diseases based on the characteristic information of the high-frequency signals.
The preprocessing of the high-frequency electrocardiosignals differs from the preprocessing of the low-frequency signals in that the interference signals cannot be filtered out simply by low-pass filtering.
Aiming at power frequency interference:
the power frequency interference of 50hz is filtered by adopting a software method, and based on a second-order IIR filter, the system function is as follows:
wherein,for notch number digital frequency f 0 For notch frequency, f s Is the sampling frequency.
For myoelectric interference:
the application filters myoelectric interference signals based on a wavelet threshold method combining variation modal decomposition and improvement. The method comprises the following steps:
(1) The high-frequency electrocardiosignal is adaptively decomposed into k modal components through variation modal decomposition.
k is the number of decomposition layers, and the selection of k values is significant for the decomposition effect. If the k value is too small, under-decomposition can be caused, the signal cannot be effectively decomposed, and modal aliasing is easy to occur; too large a k value can lead to over-decomposition, generating spurious components, affecting accurate analysis and reconstruction of the signal. Thus, the present application determines the k value by analyzing the distribution of the center frequencies of the modal components at continuously varying k values.
(2) The nature of the modal component is determined and is classified as either signal dominant or noise dominant. And the correlation coefficient of the modal component and the high-frequency electrocardiosignal is used as a characteristic quantity, a threshold M is set to distinguish the property of the modal component, and when the correlation coefficient is larger than the threshold, the modal component is identified as signal dominant, otherwise, the modal component is noise dominant.
(3) Wavelet threshold denoising is performed on the noise-dominated modal components.
(4) Reconstructing the modal component of the signal dominant and the modal component after denoising of the wavelet threshold to obtain the denoised electrocardiosignal.
The improved wavelet threshold denoising is as follows:
the present application provides an improved threshold function, as shown in the following formula:
wherein w' is the denoised signal, w is the original signal, and T is the threshold.
The threshold function has a continuity, as w approaches T, as follows:
w→T + in the time-course of which the first and second contact surfaces,
w→T - when w' =0.
Thus, the function is continuous at T, solving the problem of the hard threshold being discontinuous at T.
When w & gtto 1, therefore, the asymptote of the threshold function curve is w' =w, the problem of constant deviation of the soft threshold function is solved, and the distortion of the signal after denoising is smaller.
In actual denoising, the results of denoising evaluation indexes are combined, and parameters p and q are adjusted according to the noise interference degree so as to achieve the best denoising effect. Firstly, the integral parameter q is regulated, the q value is not too small, otherwise, oscillation is easy to occur, and is not too large, otherwise, signal distortion is easy to occur. After q is completed, the threshold function is trimmed by adjusting the trimming parameter p until the expected effect is achieved.
S2: acquiring respiratory signals and preprocessing
The respiratory signal is acquired based on the respiratory flow sensor, and accurate respiratory signals including information such as respiratory frequency and respiratory flow rate can be acquired based on the respiratory flow sensor.
Preprocessing the respiratory flow signal includes:
calculating respiratory tidal volume, judging whether the current respiratory state is inspiration or expiration by detecting flow data, storing the flow data into an inspiration storage space if the current respiratory state is inspiration, then reading next data, adding the flow data and total flow data of the inspiration storage space if the current respiratory state is inspiration, and storing the total inspiration storage space, calculating total inspiration volume, outputting the flow stored in the inspiration storage space as tidal volume if the next data is expiration, and clearing the storage space data to wait for the next inspiration process. The calculation mode is as follows:
wherein C is tidal volume, n is data volume acquired in one inspiration process, x i Is the acquired inspiratory data value.
Filtering the respiratory flow from the flow waveform by adopting a low-pass filter, only reserving the air leakage, and further subtracting the obtained air leakage from the respiratory flow signal to obtain a respiratory flow compensated result; and a second-order active low-pass filter is adopted to remove high-frequency noise interference.
S3: extracting signal features
Because the dimensions of the signals are different, the signals are required to be normalized, the signals are normalized into dimensionless values, the calculation complexity is reduced, and the calculation speed is improved. The normalization calculation mode is as follows:
where x' is the result after normalization, σ is the standard deviation, and μ is the mean.
Features are extracted in three ways, including time domain, frequency domain, and differential entropy.
(1) Time domain
The time domain information can show the time variation of the signal, and the application is based on the time domain characteristics of statistics, which comprises the following steps: standard deviation, variance, root mean square, first and second differential means of the original and normalized signals, hjorth activity, mobility, complexity.
The first order differential mean value is calculated as follows:
the second order differential mean value is calculated as follows:
kurtosis and skewness of response signal fluctuations:
hjorth activity, mobility, and complexity are calculated as follows:
wherein s is the original signal and has a length of N, mu s Sigma, which is the average value of the original signal s Is the standard deviation of the original signal,s 'is the standard deviation of the original signal and s' is the first derivative of the original signal.
(2) Frequency domain
The time domain information of the signal is visual, and the frequency domain characteristics of the signal can reflect greater details of the signal, so that the time domain signal is converted into the frequency domain, and the characteristics of the frequency domain are extracted, including energy spectrum density, power spectrum intensity and differential entropy.
Obtaining frequency domain information by:
transforming the n-section time domain signals into frequency domain signals X (n) through FFT; dividing the frequency domain signal into K frequency bands, and extracting the characteristics of each frequency band. The frequency domain features include the following:
a. energy spectral density
Wherein ESD represents an energy spectrum characteristic,f k a lower band representing a kth band, f k+1 Upper band representing the kth band, +.>Representing the quotient of the lower frequency band and the sampling frequency, +.>Representing the quotient of the up-band and the sampling frequency.
b. Power spectral density
Wherein PSD represents the power spectrum characteristic,
c. power spectrum intensity
The power spectrum intensity represents the sum of the amplitudes of the frequency domains, and the calculation formula is as follows:
PSI＝∑|X i |
wherein PSI represents the power spectrum intensity,
(3) Differential entropy
For signals subject to gaussian distribution, the differential entropy is calculated as follows:
s4: signal fusion of high-frequency electrocardiosignals and respiratory signals
Based on two data fusion modes of deep learning and feature fusion and decision fusion, the high-frequency electrocardiosignals and respiratory signals are subjected to signal fusion, and a specific fusion method is shown in fig. 2:
(1) Feature fusion
And classifying the random forest algorithm with the improved time domain characteristics, frequency domain characteristics and differential entropy characteristics respectively to form a first-stage classifier. The improved random forest algorithm is optimized by combining a drosophila algorithm, and the specific steps are as follows:
a. optimizing the number of decision trees and the initial value of the maximum depth in a random forest algorithm based on a drosophila algorithm; the method comprises the following specific steps:
a1, initializing the position of the fruit fly and the distance between the fruit fly and food:
wherein X is 0 、Y 0 Respectively, the positions of the drosophila are respectively, and i and j are respectively random values.
a2: distance D is calculated and taste concentration determination value S is calculated:
a3 calculating taste concentration value S i :
S i ＝f(S)
Wherein f-function is a taste concentration function.
and a4, obtaining a concentration maximum value and a position optimal value of the individual.
a5, updating the coordinate position and calculating the concentration maximum value of the new position;
a6, comparing the concentration of the new position with the concentration of the original position, if the concentration of the new position is better than the concentration of the original position, executing a5, otherwise executing a1-a4.
a7. And determining whether an end condition is reached, if so, obtaining a final result, and exiting the program.
b. Constructing a random forest classifier based on the optimal initial value to classify, and forming a first-stage classifier;
and the initial parameters in the random forest algorithm are searched for optimal values through the initialized Drosophila algorithm, so that parameter uncertainty caused by the fact that the initial parameters in the algorithm training model are set by adopting artificial experience values can be avoided. The root mean square error of the random forest algorithm training model is selected as the fitness function of the particle swarm algorithm, the smaller the error value is, the more accurate the prediction capability of the model is, the minimum value optimization is carried out on the fitness function by using the drosophila algorithm, and the parameters corresponding to the optimal condition of the random forest model can be determined.
Classification results are the physiological state grades of the monitored objects and are classified into excellent, good, medium, bad and critical.
(2) Decision fusion
D-S evidence theory method is abbreviated as D-S rule, and a evidence reasoning system E is assumed i ,Θ＝{θ 1 ,θ 2 …θ n A power set of 2 Θ Where Θ is the set of n mutually exclusive systems, for a subset A of Θ, m () is the base probability of A, if m (A)>0, then A is focal.
The method comprises the following specific steps of fusing a plurality of classifiers based on a D-S evidence theory decision algorithm:
a. calculate evidence E i Degree of conflict k with other evidence ij Form conflict vector K i ＝{k i1 ,k i2 ,…k in And performing standardization processing.
b. Calculating entropy of normalized collision vector
c. Calculate evidence E i Weight coefficient:
d. according to the weight coefficient distribution m and the conflict k, the evidence synthesis mode is formed as follows:
m′(A)＝p(A)+k′×q(A)
wherein,q (A) is the average degree of support of A by evidence, < >>
e. And making a decision according to the synthesis result to obtain a classification result.
The output results of the time domain and frequency domain primary classifier are fused through a D-S evidence theory decision algorithm to form a secondary classification result, the secondary classification result is fused with the primary classifier of the differential entropy through a D-S evidence theory decision algorithm to form a tertiary classification result, and the tertiary classification result of the high-frequency electrocardiosignal and the tertiary classification result of the respiratory signal are further fused in a weighting mode respectively.
In the embodiment, the high-frequency electrocardiosignals of the monitored object are used for acquiring the high-frequency electrocardiosignals, so that the hidden heart diseases of the monitored object are accurately acquired and judged; the myoelectric interference signals are filtered through combining the variation modal decomposition and an improved wavelet threshold method, so that the high-frequency electrocardiosignals are more accurately restored; by combining the signal fusion algorithm of feature fusion and decision fusion, the accurate judgment of the monitoring state is realized.
The embodiment of the application also provides a physiological state monitor based on cardiopulmonary information, which specifically comprises:
the high-frequency electrocardiosignal acquisition and processing module is used for acquiring and preprocessing the high-frequency electrocardiosignal and aiming at power frequency interference:
the software method is adopted to filter the power frequency interference of 50hz based on a second-order IIR filter, and the system function is as follows:
wherein,for notch number digital frequency f 0 For notch frequency, f s Is the sampling frequency;
for myoelectric interference: filtering myoelectric interference signals based on a wavelet threshold method combining variation modal decomposition and improvement;
the respiratory signal acquisition and processing module is used for acquiring the respiratory signal based on the respiratory flow sensor, and preprocessing the respiratory flow signal comprises the following steps:
filtering the respiratory flow from the flow waveform by adopting a low-pass filter, only reserving the air leakage, and further subtracting the obtained air leakage from the respiratory flow signal to obtain a respiratory flow compensated result; and a second-order active low-pass filter is adopted to remove high-frequency noise interference.
And the feature extraction module is used for extracting signal features, including time domain, frequency domain and differential entropy.
The feature fusion module is used for carrying out signal fusion on the high-frequency electrocardiosignals and respiratory signals based on two data fusion modes of deep learning combined feature fusion and decision fusion, and specifically comprises the following steps: feature fusion is carried out based on a random forest algorithm optimized by combining a drosophila algorithm, and decision fusion is carried out based on a D-S evidence theory method.
The above examples are only illustrative of the preferred embodiments of the present application and are not intended to limit the scope of the present application, and various modifications and improvements made by those skilled in the art to the technical solution of the present application should fall within the scope of protection defined by the claims of the present application without departing from the spirit of the present application.
Claims (9)
1. A physiological condition monitoring method based on cardiopulmonary information, comprising the steps of:
s1: acquiring high-frequency electrocardiosignals and preprocessing;
collecting high-frequency electrocardiosignals by using a high-frequency electrocardiograph, carrying out heart and lung information fusion based on the high-frequency electrocardiosignals, and being beneficial to identifying the existence of hidden heart diseases based on characteristic information of the high-frequency signals;
s2: acquiring a respiratory signal and preprocessing;
acquiring a respiratory signal based on a respiratory flow sensor, wherein the respiratory flow sensor can acquire an accurate respiratory signal, and the respiratory signal comprises a respiratory frequency and a respiratory flow rate;
s3: extracting signal characteristics;
the method comprises the steps of carrying out normalization processing on signals, extracting features through three aspects, including time domain, frequency domain and differential entropy;
s4: carrying out signal fusion on the high-frequency electrocardiosignal and the respiratory signal;
the method comprises the following steps: classifying the time domain features, the frequency domain features and the differential entropy features through a random forest algorithm improved based on a drosophila algorithm respectively to form a first-stage classifier, fusing the output results of the first-stage classifier of the time domain and the frequency domain through a D-S evidence theory decision algorithm to form a second-stage classification result, fusing the second-stage classification result with the first-stage classifier of the differential entropy through a D-S evidence decision algorithm to form a third-stage classification result, and further fusing the third-stage classification result of the high-frequency electrocardiosignal with the third-stage classification result of the respiratory signal in a weighted mode to obtain a final result.
2. The method for monitoring physiological status based on cardiopulmonary information according to claim 1, wherein the step S1 includes filtering out myoelectric interference signals based on a wavelet thresholding method combined with a variational modal decomposition.
3. The method for monitoring physiological status based on cardiopulmonary information according to claim 2, wherein the myoelectric interference signal filtering method specifically comprises:
the high-frequency electrocardiosignal is adaptively decomposed into k modal components through variation modal decomposition;
determining the properties of the modal components, and classifying them as signal dominant or noise dominant; the correlation coefficient of the modal component and the high-frequency electrocardiosignal is used as a characteristic quantity, a threshold M is set to distinguish the property of the modal component, when the correlation coefficient is larger than the threshold, the modal component is identified as signal dominant, otherwise, the modal component is noise dominant;
performing wavelet threshold denoising on the modal component dominated by noise;
reconstructing the modal component of the signal dominant and the modal component after denoising of the wavelet threshold to obtain the denoised electrocardiosignal.
4. A physiological condition monitoring method based on cardiopulmonary information according to claim 3, wherein the k-value is determined by analyzing a distribution of center frequencies of modal components at continuously varying k-values.
5. The method of claim 2, wherein the improved wavelet threshold denoising is as follows:
based on the modified threshold function, the following formula is shown:
；
wherein,for denoised signal, ++>As the original signal->Is a threshold value.
6. The method for monitoring physiological status based on cardiopulmonary information according to claim 1, wherein the time domain information in step S3 includes: standard deviation, variance, root mean square, first and second differential means of the original and normalized signals, hjorth activity, mobility, complexity.
7. The method for monitoring physiological states based on cardiopulmonary information according to claim 1, wherein the modified random forest algorithm optimizes the number of decision trees and the initial value of the maximum depth in the random forest algorithm based on a drosophila algorithm.
8. The method for monitoring physiological states based on cardiopulmonary information according to claim 1, wherein decision fusion is performed based on a D-S evidence theory method.
9. A monitor for performing the method for monitoring physiological conditions based on cardiopulmonary information according to any of claims 1-8, wherein the monitor comprises:
the high-frequency electrocardiosignal acquisition and processing module is used for acquiring and preprocessing the high-frequency electrocardiosignal and aiming at power frequency interference: filtering the power frequency interference of 50hz based on a second-order IIR filter by adopting a software method;
for myoelectric interference: filtering myoelectric interference signals based on a wavelet threshold method combining variation modal decomposition and improvement;
the respiratory signal acquisition and processing module is used for acquiring the respiratory signal based on the respiratory flow sensor, and preprocessing the respiratory flow signal comprises the following steps:
filtering the respiratory flow from the flow waveform by adopting a low-pass filter, only reserving the air leakage, and further subtracting the obtained air leakage from the respiratory flow signal to obtain a respiratory flow compensated result; removing high-frequency noise interference by adopting a second-order active low-pass filter;
the characteristic extraction module is used for extracting signal characteristics, including time domain, frequency domain and differential entropy;
the feature fusion module is used for carrying out signal fusion on the high-frequency electrocardiosignal and the respiratory signal, and specifically comprises the following steps: classifying the time domain features, the frequency domain features and the differential entropy features through a random forest algorithm improved based on a drosophila algorithm respectively to form a first-stage classifier, fusing the output results of the first-stage classifier of the time domain and the frequency domain through a D-S evidence theory decision algorithm to form a second-stage classification result, fusing the second-stage classification result with the first-stage classifier of the differential entropy through a D-S evidence decision algorithm to form a third-stage classification result, and further fusing the third-stage classification result of the high-frequency electrocardiosignal with the third-stage classification result of the respiratory signal in a weighted mode to obtain a final result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211287888.2A CN115721317B (en) | 2022-10-20 | 2022-10-20 | Physiological state monitoring method and monitor based on heart and lung information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211287888.2A CN115721317B (en) | 2022-10-20 | 2022-10-20 | Physiological state monitoring method and monitor based on heart and lung information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115721317A CN115721317A (en) | 2023-03-03 |
| CN115721317B true CN115721317B (en) | 2023-11-21 |
Family
ID=85293922
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211287888.2A Active CN115721317B (en) | 2022-10-20 | 2022-10-20 | Physiological state monitoring method and monitor based on heart and lung information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115721317B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108113657A (en) * | 2016-11-30 | 2018-06-05 | 天津职业技术师范大学 | A kind of Driver Fatigue Detection based on D-S evidence theory |
| CN109770892A (en) * | 2019-02-01 | 2019-05-21 | 中国科学院电子学研究所 | A kind of sleep stage method based on electrocardiosignal |
| CN110151169A (en) * | 2019-07-04 | 2019-08-23 | 中山大学 | A kind of sleep state method for identifying and classifying based on electrocardiogram (ECG) data |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105792741A (en) * | 2013-12-06 | 2016-07-20 | 心脏起搏器股份公司 | Heart failure event prediction using classifier fusion |
| AU2020221866A1 (en) * | 2019-02-12 | 2021-10-07 | Sleep Number Corporation | Systems and methods for generating synthetic cardio-respiratory signals |
-
2022
- 2022-10-20 CN CN202211287888.2A patent/CN115721317B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108113657A (en) * | 2016-11-30 | 2018-06-05 | 天津职业技术师范大学 | A kind of Driver Fatigue Detection based on D-S evidence theory |
| CN109770892A (en) * | 2019-02-01 | 2019-05-21 | 中国科学院电子学研究所 | A kind of sleep stage method based on electrocardiosignal |
| CN110151169A (en) * | 2019-07-04 | 2019-08-23 | 中山大学 | A kind of sleep state method for identifying and classifying based on electrocardiogram (ECG) data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115721317A (en) | 2023-03-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Schmidt et al. | Acoustic features for the identification of coronary artery disease | |
| Altan et al. | Deep learning with 3D-second order difference plot on respiratory sounds | |
| Al-Angari et al. | Use of sample entropy approach to study heart rate variability in obstructive sleep apnea syndrome | |
| Belgacem et al. | ECG based human authentication using wavelets and random forests | |
| Emanet et al. | A comparative analysis of machine learning methods for classification type decision problems in healthcare | |
| Lin et al. | Automatic wheezing detection based on signal processing of spectrogram and back-propagation neural network | |
| Diker et al. | Intelligent system based on Genetic Algorithm and support vector machine for detection of myocardial infarction from ECG signals | |
| CN106805965A (en) | A kind of electrocardiosignal sorting technique and device | |
| CN110811609A (en) | Intelligent epileptic spike detection method based on fusion of adaptive template matching and machine learning algorithm | |
| Deng et al. | A robust heart sound segmentation and classification algorithm using wavelet decomposition and spectrogram | |
| CN108256453A (en) | A kind of method based on one-dimensional ECG signal extraction two dimension CNN features | |
| CN108090509A (en) | A kind of adaptive electrocardiogram sorting technique of data length | |
| Oud et al. | Asthmatic airways obstruction assessment based on detailed analysis of respiratory sound spectra | |
| RU2016146176A (en) | METHOD AND SYSTEM OF NON-INVASIVE SCREENING ASSESSMENT OF PHYSIOLOGICAL PARAMETERS AND PATHOLOGIES | |
| Kahya et al. | Classifying respiratory sounds with different feature sets | |
| Bashar et al. | Extraction of heart rate from PPG Signal: a machine learning approach using decision tree regression algorithm | |
| CN114569096B (en) | Non-contact continuous blood pressure measurement method and system based on video stream | |
| CN115721317B (en) | Physiological state monitoring method and monitor based on heart and lung information | |
| Jiang et al. | Topological data analysis approach to extract the persistent homology features of ballistocardiogram signal in unobstructive atrial fibrillation detection | |
| Saraswat et al. | Classification of ECG signals using cross-recurrence quantification analysis and probabilistic neural network classifier for ventricular tachycardia patients | |
| SenthilPandi et al. | Multi-Level Interpretable and Adaptive Representation of EEG Signals for Sleep Scoring Using Ensemble Learning Multi Classifiers | |
| CN114864095A (en) | Analysis method for blood circulation change of narrow coronary artery under combination of multiple exercise strengths | |
| CN114569116A (en) | Three-channel image and transfer learning-based ballistocardiogram ventricular fibrillation auxiliary diagnosis system | |
| Tamanna et al. | Controlling body sources of noise generated by niddle electrogram machines: A new emg idea for skipping traditional approaches | |
| Sezgin et al. | Classification of respiratory sounds by using an artificial neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |