CN111329508A - Heart murmur intelligent analysis method for precordial disease screening - Google Patents
Heart murmur intelligent analysis method for precordial disease screening Download PDFInfo
- Publication number
- CN111329508A CN111329508A CN202010136364.8A CN202010136364A CN111329508A CN 111329508 A CN111329508 A CN 111329508A CN 202010136364 A CN202010136364 A CN 202010136364A CN 111329508 A CN111329508 A CN 111329508A
- Authority
- CN
- China
- Prior art keywords
- heart
- screening
- analysis method
- sound
- heart sound
- 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.)
- Pending
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B7/00—Instruments for auscultation
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- Physics & Mathematics (AREA)
- Pathology (AREA)
- Cardiology (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses an intelligent analysis method for heart murmurs for screening congenital heart diseases, and belongs to the technical field of medical treatment. The heart noise intelligent analysis method for screening the heart disease comprises data processing and feature recognition, wherein the data processing is three paths of synchronously acquired heart sound, electrocardio and echocardiogram data for processing, the noise generated by the heart sound of a newborn due to cardiac organic degeneration can be effectively separated through a signal preprocessing technology, one heart sound is decomposed into a plurality of signals with different frequencies through wavelet transformation of the heart sound signals, time domain feature values and frequency domain feature values of the signals with different frequencies are calculated, the calculated time domain special diagnosis values and frequency domain special diagnosis values are used as input, and classification is carried out through a model trained by a vector machine, so that the heart noise features can be more accurately extracted, more accurate, low-cost and risk-free judgment basis can be provided for medical workers, and popularization of screening the heart disease is facilitated.
Description
Technical Field
The invention relates to the technical field of medical treatment, in particular to an intelligent analysis method for heart murmurs for screening congenital heart diseases.
Background
Congenital heart disease (called congenital heart disease for short) is a cardiac organic lesion which seriously affects the normal development of a fetus and an infant, and congenital heart disease refers to a condition that in the process of embryonic development, due to anatomical structure abnormality caused by formation obstacle or abnormal development of a heart and a large blood vessel, a blood flow channel which is automatically closed after the birth of the fetus cannot be closed or is abnormally narrow, so that blood supply of all parts of a body is insufficient, the normal development of the fetus is seriously affected, and even the fetus fails. Congenital heart disease is the most common one of congenital malformations, accounting for about 28% of various congenital malformations and about 0.4% -1% of infants in life.
Heart murmurs refer to abnormal sounds produced by the vibration of the walls, valves or vessels of the heart or blood vessels due to the turbulence of blood generated within the heart or blood vessels during systole or diastole, in addition to heart sounds and extra heart sounds. More than two levels of noise are usually indicative of organic cardiovascular diseases such as valvular heart disease, congenital heart disease, fever, etc.
In the field of fetal congenital heart disease diagnosis, more complex and larger expensive instruments are used currently; whether the fetus has the congenital heart disease can be seen by performing the four-dimensional ultrasonic examination outside the abdomen, the birth canal or the intestine and the nuclear magnetic resonance examination during the 18-20 weeks gestational period examination, but the examination has low sensitivity and high cost, and the instruments adopted by the examination have the defects of high risk, complex operation, portability and the like in use.
Disclosure of Invention
1. Technical problem to be solved
The invention aims to provide an intelligent analysis method of heart murmurs for screening congenital heart diseases, which can effectively separate the murs of newborn heart sounds generated by cardiac organic degeneration by a signal preprocessing technology, decompose one heart sound into a plurality of signals with different frequencies by performing wavelet transformation on the heart sound signals, calculate time domain characteristic values and frequency domain characteristic values of the signals with different frequencies, then use the calculated time domain characteristic value and frequency domain characteristic value as input, classify by a model trained by a vector machine, can more accurately extract the heart murs characteristics, can provide more accurate, low-cost and risk-free judgment basis for medical workers, and is favorable for popularization of screening congenital heart diseases.
2. Technical scheme
In order to solve the above problems, the present invention adopts the following technical solutions.
The heart murmur intelligent analysis method for the antecedent heart disease screening comprises data processing and feature recognition, wherein the data processing comprises the steps of synchronously acquiring three paths of data of heart sounds, electrocardio and echocardiograms, and the feature recognition comprises the following steps:
step one, performing wavelet transformation on a heart sound signal through an algorithm and semi-artificial participation, so as to decompose a heart sound into a plurality of signals with different frequencies, extracting two sound banks from the heart sound image data, wherein one sound bank consists of heart murmurs, the other bank consists of normal heart sounds, performing characteristic extraction to extract time domain characteristic values and frequency domain characteristic values of the heart murs and the normal heart sounds, then training by using a vector machine to obtain vector machine parameters, and comparing the heart murs with the normal heart sounds by using the vector machine parameters;
and secondly, performing signal preprocessing, including resampling and recognizing weak signals, performing band-pass filtering on the heart sound signals, calculating an intelligent threshold (each section of heart sound signal has a threshold value, finding out potential heart murmurs according to the threshold value), finding out all potential heart murs, extracting time domain characteristics and frequency domain characteristics of the potential heart sounds, and further judging whether the potential heart murs are real heart murs or normal heart sounds by using a vector machine.
Furthermore, in the second step, the characteristic extraction adopts a HHT marginal spectrum entropy mode, sliding window processing is utilized, each frame window is 64 points long, the frame is moved by 1 point, HHT marginal spectrum entropy of each frame signal is calculated, and a concept and an entropy analysis method based on the HHT marginal spectrum entropy and the energy spectrum entropy are provided based on the Hilbert-Huang transform theory and the concept of the generalized information entropy.
Further, the characteristic reference quantity in the HHT marginal spectral entropy includes kurtosis, skewness, a mean value of signal width at K1 and a mean value of signal width at K2, and K1 and K2 are two artificially set reference lines, and different adaptive thresholds are set: as shown in the formula (1) (2), wherein nfzcc is an entropy value
Min（nfzcc）+k1*（max（nfzcc）-min（nfzcc）） （1）
Min（nfzcc）+k2*（max（nfzcc）-min（nfzcc））（k1>k2） （2）
Adjusting K1, K2 can set different sensitivities: when K1 increases and K2 decreases, the abnormal heart sound signal recognition rate becomes large; when K1 decreases and K2 increases, the normal heart sound signal recognition rate becomes large; parameter selection: k1 takes 0.4, K2 takes 0.225, the classifier is selected as a support vector machine, and the kernel function is a linear kernel function.
Further, in the second step, all potential heart murmurs are found by moving the 20ms window along the heart sound signal.
Further, the data processing comprises the steps of denoising the heart sound signal and calculating and analyzing the heart sound by combining the electrocardiogram data, and in the process of denoising the heart sound signal, when a non-stationary signal is analyzed, the wavelet transformation has obvious advantages mainly due to strong adaptability, when the signal type is a low-frequency long-time signal, the frequency resolution is high, and the time resolution is low.
Further, the denoising of the heart sound signal adopts wavelet denoising, and the wavelet denoising step includes: step one, determining a wavelet to be processed, and performing N-layer wavelet decomposition according to the characteristics of a wavelet signal; secondly, quantizing the threshold value of the wavelet decomposition coefficient; and thirdly, wavelet reconstruction of the one-dimensional signal.
Further, the heart sound calculation includes BPM (BPMa calculated per minute heart beats by counting the number of times of the first heart sound S1 in 5 seconds), S1 split condition (first heart sound S1 is composed of M1, T1, M1 is the mitral valve component of S1, is the primary component in normal times, and occurs immediately after the mitral valve closes, and before the tricuspid component in normal times, T1 is the second component of S1, occurs normally after M1 (mitral valve component), immediately after the tricuspid valve closes, M1 and T1 are normally spaced 0.02S more than S1 split, S2 split condition (second heart sound S2 is composed of aortic a2 and pulmonary valve P2, a2 has stronger strength at valve closure than the right mechanical motion P2, normal condition a 2P 6959P 2, P2 is prior to comparison, and auscultatory sounds are auscultated by auscultatory in the region 2, which auscultatory region is sharp in the second heart valve region of the aortic valve 2, the pulmonary valve is located in the second intercostal space at the left margin of the sternum. The P2 hyperemia represents the second heart sound hyperfunction or enhancement of the pulmonary artery region, mainly caused by the pressure increase of the pulmonary artery, and most of the pressure hyperfunction can be caused by mitral stenosis, congenital heart disease of left-to-right shunting or pulmonary heart disease patients, wherein the pressure increase in the pulmonary artery can cause the P2 hyperfunction. Defining an hyperfunction standard for quantification of an energy spectrum of P2, realizing automatic identification of P2 hyperfunction) and a systolic phase-diastolic phase ratio (a systolic phase is between S1 and S2, a diastolic phase is between S2 and S1, peaks of S1 and S2 are taken as reference points for calculation, and 5 complete heartbeat cycles are taken for average calculation), and identifying the first heart sound S1 and the second heart sound S2 and the common heart sound calculation analysis and the murmur characteristic identification through an electrocardiogram QRS wave characteristic.
Further, the S2 divisions include physiological divisions (deep inspiration end appears, common in adolescents) (increased thoracic negative pressure-increased right cardiotomy volume-increased right heart ejection time-prolonged right heart ejection time-delayed pulmonary valve closure)), common divisions (most common, complete right bundle branch conduction block, pulmonary valve stenosis, mitral valve stenosis (prolonged right ventricular ejection time-delayed pulmonary valve closure), mitral valve insufficiency, ventricular septal defect (shortened left ventricular ejection time-advanced aortic valve closure)), fixed divisions (division is not affected by breathing, the time interval of division is fixed, atrial septal defect (left-to-right shunting)) and abnormal divisions (reverse division) (pathological signs; aortic valve closure is delayed in pulmonary valve, division is narrowed in inspiration, and is widened in expiration; severe hypertension, aortic valve stenosis, complete left bundle branch conduction block), when the interval between A2 and P2 is less than or equal to 0.03 seconds, physiological cleavage is considered, and when the interval is greater than 0.03 seconds, pathological cleavage is considered.
Further, the nature of the heart murmurs is related to their frequencies, including blowing-like (mainly high frequency), coarse (medium frequency or mixture of high and low frequencies) and rumbling-like (mainly low frequency), which may occur between S1 and S2 (systolic phase), or between S2 and S1 (diastolic phase), and some lesions may be heard in both systolic and diastolic phases.
3. Advantageous effects
Compared with the prior art, the invention has the advantages that:
(1) the scheme can effectively separate the noise generated by heart organic degeneration of the heart sound of the newborn by a signal preprocessing technology, decompose one heart sound into a plurality of signals with different frequencies by performing wavelet transformation on the heart sound signals, calculate the time domain characteristic value and the frequency domain characteristic value of the signals with different frequencies, take the calculated time domain special diagnosis value and the frequency domain special diagnosis value as input, classify by a model trained by a vector machine, can accurately extract the heart noise characteristics, can provide more accurate, low-cost and risk-free judgment basis for medical workers, and is favorable for popularization of prior heart disease screening.
(2) And in the second step, the characteristic extraction adopts a HHT marginal spectrum entropy mode, sliding window processing is utilized, each frame window is 64 points long, frames are moved by 1 point, HHT marginal spectrum entropy of each frame signal is calculated, and a concept and an entropy analysis method based on the HHT marginal spectrum entropy and the energy spectrum entropy are provided based on the Hilbert-Huang transform theory and the concept of generalized information entropy.
(3) The characteristic reference quantity in the HHT marginal spectrum entropy comprises kurtosis, skewness, a signal width average value at K1 and a signal width average value at K2, wherein K1 and K2 are two artificially set reference lines, and different adaptive thresholds are set: as shown in the formula (1) (2), wherein nfzcc is an entropy value
Min（nfzcc）+k1*（max（nfzcc）-min（nfzcc）） （1）
Min（nfzcc）+k2*（max（nfzcc）-min（nfzcc））（k1>k2） （2）
Adjusting K1, K2 can set different sensitivities: when K1 increases and K2 decreases, the abnormal heart sound signal recognition rate becomes large; when K1 decreases and K2 increases, the normal heart sound signal recognition rate becomes large; parameter selection: k1 takes 0.4, K2 takes 0.225, the classifier is selected as a support vector machine, and the kernel function is a linear kernel function.
(4) In step two, all potential heart murmurs are found by moving the 20ms window along the heart sound signal.
(5) The data processing comprises the steps of denoising the heart sound signal and calculating and analyzing the heart sound by combining electrocardiogram data, wherein in the process of denoising the heart sound signal, when a non-stationary signal is analyzed, the advantage of wavelet transformation is very obvious, mainly because the adaptability is very strong, when the signal type is a low-frequency long-term signal, the frequency resolution ratio is high, and the time resolution ratio is low.
(6) The method for denoising the heart sound signal adopts wavelet denoising, and comprises the following steps: step one, determining a wavelet to be processed, and performing N-layer wavelet decomposition according to the characteristics of a wavelet signal; secondly, quantizing the threshold value of the wavelet decomposition coefficient; and thirdly, wavelet reconstruction of the one-dimensional signal.
(7) The heart sound calculation includes BPM (BPMa per minute heart beat calculated by counting the number of first heart sounds S1 within 5 seconds), S1 split cases (first heart sounds S1 consisting of M1, T1, M1 being the mitral valve component of S1, being the primary component in normal times, and occurring immediately after the mitral valve closes, normally before the tricuspid valve component, T1 being the second component of S1, normal M1 occurring (mitral valve component), M1 and T1 being normally spaced 0.02S immediately after the tricuspid valve closes, over which is S1 split), S2 split cases (second heart sounds S2 consisting of aortic a2 and pulmonary P2, a2 having a stronger force at valve closure, and thus stronger than the right mechanical motion P2, normal cases a 2P 2, P2 comparing the soft 2 with the auscultatory sounds P2, which are most clearly auscultatory in the region of the second artery, the pulmonary valve is located in the second intercostal space at the left margin of the sternum. The P2 hyperemia represents the second heart sound hyperfunction or enhancement of the pulmonary artery region, mainly caused by the pressure increase of the pulmonary artery, and most of the pressure hyperfunction can be caused by mitral stenosis, congenital heart disease of left-to-right shunting or pulmonary heart disease patients, wherein the pressure increase in the pulmonary artery can cause the P2 hyperfunction. Defining the standard of hyperfunction by quantifying the energy spectrum of P2, realizing the automatic identification of P2 hyperfunction) and the ratio of systole to diastole (systole between S1 and S2, diastole between S2 and S1, calculating by taking the peak values of S1 and S2 as reference points, taking the average value of 5 complete heartbeat cycles for calculation), and identifying the first heart sound S1 and the second heart sound S2 and the computational analysis of the common heart sound and the characteristic identification of the murmurs by the QRS wave characteristics of the electrocardiogram.
(8) The S2 divisions include physiological divisions (occurrence of deep inspiration end, common in adolescents) (negative pressure increase in thoracic cavity-increase in right heart return blood volume-increase in right heart bleeding time-delay in closing of pulmonary valve)), common divisions (most common, complete right bundle branch block, pulmonary valve stenosis, mitral valve stenosis (extension of right ventricular bleeding time-delay in closing of pulmonary valve), mitral valve insufficiency, ventricular septal defect (shortening of left ventricular ejection time-delay in closing of aortic valve)), fixed divisions (division is not affected by breathing, the time interval of division is fixed, atrial septal defect of prior heart disease (left-to-right shunting)) and abnormal divisions (reverse division) (pathological signs; closing of aortic valve is delayed in pulmonary valve, division is narrowed in inspiration, widening in expiration; is seen in severe hypertension, aortic valve stenosis, complete left bundle branch block), when the interval between A2 and P2 is less than or equal to 0.03 seconds, physiological cleavage is considered, and when the interval is greater than 0.03 seconds, pathological cleavage is considered.
(9) The nature of the heart murmurs is related to their frequencies, which include bloat-like (mainly high frequency), rough (medium frequency or mixture of high and low frequencies) and carina-like (mainly low frequency), and the murmurs may occur between S1 and S2 (systolic phase), or between S2 and S1 (diastolic phase), and some lesions may be heard in both systolic and diastolic phases.
Drawings
FIG. 1 is a flow chart of the vector machine training in the first step of feature recognition of the present invention;
FIG. 2 is a block diagram of a heart murmur feature extraction technique flow diagram of the present invention;
FIG. 3 is a schematic structural diagram of a heart murmur feature parameter according to the present invention;
FIG. 4 is a schematic structural diagram of a rough relationship of the range of heart murmur frequencies indicating a lesion according to the present invention.
Detailed Description
Referring to fig. 1-4, the method for intelligently analyzing heart murmurs for screening heart disease first includes data processing and feature recognition, the data processing is performed by synchronously acquiring three-way data of heart sounds, electrocardio and echocardiograms, and the feature recognition includes the following steps:
step one, performing wavelet transformation on a heart sound signal through an algorithm and semi-artificial participation, so as to decompose a heart sound into a plurality of signals with different frequencies, extracting two sound banks from the heart sound image data, wherein one sound bank consists of heart murmurs, the other bank consists of normal heart sounds, performing characteristic extraction to extract time domain characteristic values and frequency domain characteristic values of the heart murs and the normal heart sounds, then training by using a vector machine to obtain vector machine parameters, and comparing the heart murs with the normal heart sounds by using the vector machine parameters;
and secondly, performing signal preprocessing, including resampling and recognizing weak signals, performing band-pass filtering on the heart sound signals, calculating an intelligent threshold (each section of heart sound signal has a threshold value, finding out potential heart murmurs according to the threshold value), finding out all potential heart murs, extracting time domain characteristics and frequency domain characteristics of the potential heart sounds, and further judging whether the potential heart murs are real heart murs or normal heart sounds by using a vector machine.
And in the second step, the characteristic extraction adopts a HHT marginal spectrum entropy mode, sliding window processing is utilized, each frame window is 64 points long, frames are moved by 1 point, HHT marginal spectrum entropy of each frame signal is calculated, and a concept and an entropy analysis method based on the HHT marginal spectrum entropy and the energy spectrum entropy are provided based on the Hilbert-Huang transform theory and the concept of generalized information entropy.
The characteristic reference quantity in the HHT marginal spectrum entropy comprises kurtosis, skewness, a signal width average value at K1 and a signal width average value at K2, wherein K1 and K2 are two artificially set reference lines, and different adaptive thresholds are set: as shown in the formula (1) (2), wherein nfzcc is an entropy value
Min（nfzcc）+k1*（max（nfzcc）-min（nfzcc）） （1）
Min（nfzcc）+k2*（max（nfzcc）-min（nfzcc））（k1>k2） （2）
Adjusting K1, K2 can set different sensitivities: when K1 increases and K2 decreases, the abnormal heart sound signal recognition rate becomes large; when K1 decreases and K2 increases, the normal heart sound signal recognition rate becomes large; parameter selection: k1 takes 0.4, K2 takes 0.225, the classifier is selected as a support vector machine, and the kernel function is a linear kernel function.
In step two, all potential heart murmurs are found by moving the 20ms window along the heart sound signal.
The data processing comprises the steps of denoising the heart sound signal and calculating and analyzing the heart sound by combining electrocardiogram data, wherein in the process of denoising the heart sound signal, when a non-stationary signal is analyzed, the advantage of wavelet transformation is very obvious, mainly because the adaptability is very strong, when the signal type is a low-frequency long-term signal, the frequency resolution ratio is high, and the time resolution ratio is low.
The method for denoising the heart sound signal adopts wavelet denoising, and comprises the following steps: step one, determining a wavelet to be processed, and performing N-layer wavelet decomposition according to the characteristics of a wavelet signal; secondly, quantizing the threshold value of the wavelet decomposition coefficient; and thirdly, wavelet reconstruction of the one-dimensional signal.
The heart sound calculation includes BPM (BPMa per minute heart beat calculated by counting the number of first heart sounds S1 within 5 seconds), S1 split cases (first heart sounds S1 consisting of M1, T1, M1 being the mitral valve component of S1, being the primary component in normal times, and occurring immediately after the mitral valve closes, normally before the tricuspid valve component, T1 being the second component of S1, normal M1 occurring (mitral valve component), M1 and T1 being normally spaced 0.02S immediately after the tricuspid valve closes, over which is S1 split), S2 split cases (second heart sounds S2 consisting of aortic a2 and pulmonary P2, a2 having a stronger force at valve closure, and thus stronger than the right mechanical motion P2, normal cases a 2P 2, P2 comparing the soft 2 with the auscultatory sounds P2, which are most clearly auscultatory in the region of the second artery, the pulmonary valve is located in the second intercostal space at the left margin of the sternum. The P2 hyperemia represents the second heart sound hyperfunction or enhancement of the pulmonary artery region, mainly caused by the pressure increase of the pulmonary artery, and most of the pressure hyperfunction can be caused by mitral stenosis, congenital heart disease of left-to-right shunting or pulmonary heart disease patients, wherein the pressure increase in the pulmonary artery can cause the P2 hyperfunction. Defining the standard of hyperfunction by quantifying the energy spectrum of P2, realizing the automatic identification of P2 hyperfunction) and the ratio of systole to diastole (systole between S1 and S2, diastole between S2 and S1, calculating by taking the peak values of S1 and S2 as reference points, taking the average value of 5 complete heartbeat cycles for calculation), and identifying the first heart sound S1 and the second heart sound S2 and the computational analysis of the common heart sound and the characteristic identification of the murmurs by the QRS wave characteristics of the electrocardiogram.
The S2 divisions include physiological divisions (occurrence of deep inspiration end, common in adolescents) (negative pressure increase in thoracic cavity-increase in right heart return blood volume-increase in right heart bleeding time-delay in closing of pulmonary valve)), common divisions (most common, complete right bundle branch block, pulmonary valve stenosis, mitral valve stenosis (extension of right ventricular bleeding time-delay in closing of pulmonary valve), mitral valve insufficiency, ventricular septal defect (shortening of left ventricular ejection time-delay in closing of aortic valve)), fixed divisions (division is not affected by breathing, the time interval of division is fixed, atrial septal defect of prior heart disease (left-to-right shunting)) and abnormal divisions (reverse division) (pathological signs; closing of aortic valve is delayed in pulmonary valve, division is narrowed in inspiration, widening in expiration; is seen in severe hypertension, aortic valve stenosis, complete left bundle branch block), when the interval between A2 and P2 is less than or equal to 0.03 seconds, physiological cleavage is considered, and when the interval is greater than 0.03 seconds, pathological cleavage is considered.
The nature of the heart murmurs is related to their frequencies, which include bloat-like (mainly high frequency), rough (medium frequency or mixture of high and low frequencies) and carina-like (mainly low frequency), and the murmurs may occur between S1 and S2 (systolic phase), or between S2 and S1 (diastolic phase), and some lesions may be heard in both systolic and diastolic phases.
The method can effectively separate the noise generated by heart organic degeneration of the heart sound of the newborn by a signal preprocessing technology, decompose one heart sound into a plurality of signals with different frequencies by performing wavelet transformation on the heart sound signals, calculate the time domain characteristic value and the frequency domain characteristic value of the signals with different frequencies, take the calculated time domain special diagnosis value and the calculated frequency domain special diagnosis value as input, classify the signals by a model trained by a vector machine, can accurately extract the heart noise characteristics, can provide more accurate, low-cost and risk-free judgment basis for medical workers, and is favorable for popularization of screening of the prior heart disease.
The above; but are merely preferred embodiments of the invention; the scope of the invention is not limited thereto; any person skilled in the art is within the technical scope of the present disclosure; the technical scheme and the improved concept of the invention are equally replaced or changed; are intended to be covered by the scope of the present invention.
Claims (9)
1. The heart murmur intelligent analysis method for the congenital heart disease screening is characterized by comprising the following steps: the method comprises data processing and feature recognition, wherein the data processing comprises the steps of synchronously acquiring three paths of data of heart sound, electrocardio and echocardiogram, and processing the data, and the feature recognition comprises the following steps:
step one, performing wavelet transformation on a heart sound signal through an algorithm and semi-artificial participation, so as to decompose a heart sound into a plurality of signals with different frequencies, extracting two sound banks from the heart sound image data, wherein one sound bank consists of heart murmurs, the other bank consists of normal heart sounds, performing characteristic extraction to extract time domain characteristic values and frequency domain characteristic values of the heart murs and the normal heart sounds, then training by using a vector machine to obtain vector machine parameters, and comparing the heart murs with the normal heart sounds by using the vector machine parameters;
and secondly, performing signal preprocessing, including resampling and recognizing weak signals, performing band-pass filtering on the heart sound signals, calculating an intelligent threshold value, finding out all potential heart murmurs, extracting time domain characteristics and frequency domain characteristics of the potential heart sounds, and further judging whether the potential heart murs are real heart murs or normal heart sounds by using a vector machine.
2. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 1, characterized in that: and in the second step, the HHT marginal spectrum entropy mode is adopted for feature extraction, sliding window processing is utilized, each frame window is 64 points long, the frame is shifted by 1 point, and the HHT marginal spectrum entropy of each frame signal is calculated.
3. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 2, characterized in that: the characteristic reference quantity in the HHT marginal spectrum entropy comprises kurtosis, skewness, a signal width average value at K1 and a signal width average value at K2, and K1 and K2 are two artificially set reference lines.
4. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 2, characterized in that: in the second step, all potential heart murmurs are found by moving the 20ms window along the heart sound signal.
5. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 1, characterized in that: the data processing includes heart sound signal noise reduction and computational analysis of heart sounds in conjunction with electrocardiographic data.
6. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 5, characterized in that: the heart sound signal denoising adopts wavelet denoising, and the wavelet denoising step comprises the following steps: step one, determining a wavelet to be processed, and performing N-layer wavelet decomposition according to the characteristics of a wavelet signal; secondly, quantizing the threshold value of the wavelet decomposition coefficient; and thirdly, wavelet reconstruction of the one-dimensional signal.
7. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 1, characterized in that: the heart sound calculation comprises BPM, S1 splitting condition, S2 splitting condition, P2 hyperfunction condition and ratio of systole to diastole, the first heart sound S1 and the second heart sound S2 are identified through characteristics of electrocardiogram QRS wave, and the concentric sound calculation analysis and the noise characteristic identification are carried out.
8. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 7, characterized in that: the S2 disruption includes physiological disruption, normal disruption, fixed disruption, and abnormal disruption.
9. The intelligent cardiac murmur analysis method for screening for congenital heart diseases according to claim 1, characterized in that: the nature of the heart murmur is related to its frequency, which includes blooms, roughness, and rumons.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010136364.8A CN111329508A (en) | 2020-03-02 | 2020-03-02 | Heart murmur intelligent analysis method for precordial disease screening |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010136364.8A CN111329508A (en) | 2020-03-02 | 2020-03-02 | Heart murmur intelligent analysis method for precordial disease screening |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111329508A true CN111329508A (en) | 2020-06-26 |
Family
ID=71173917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010136364.8A Pending CN111329508A (en) | 2020-03-02 | 2020-03-02 | Heart murmur intelligent analysis method for precordial disease screening |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111329508A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112137611A (en) * | 2020-10-14 | 2020-12-29 | 山东平伟医疗技术有限公司 | Detection research method of electrocardiosignal waveform |
| CN114831595A (en) * | 2022-03-16 | 2022-08-02 | 复旦大学附属儿科医院 | Big data-based neonatal congenital heart disease intelligent screening algorithm and automatic upgrading system |
| WO2024007152A1 (en) * | 2022-07-05 | 2024-01-11 | 张福伟 | Method for diagnosing pediatric cardiovascular diseases based on electrocardiographic and phonocardiographic signals |
-
2020
- 2020-03-02 CN CN202010136364.8A patent/CN111329508A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112137611A (en) * | 2020-10-14 | 2020-12-29 | 山东平伟医疗技术有限公司 | Detection research method of electrocardiosignal waveform |
| CN114831595A (en) * | 2022-03-16 | 2022-08-02 | 复旦大学附属儿科医院 | Big data-based neonatal congenital heart disease intelligent screening algorithm and automatic upgrading system |
| WO2024007152A1 (en) * | 2022-07-05 | 2024-01-11 | 张福伟 | Method for diagnosing pediatric cardiovascular diseases based on electrocardiographic and phonocardiographic signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5576116B2 (en) | Multi-parameter classification of cardiovascular sounds | |
| Sepehri et al. | A novel method for pediatric heart sound segmentation without using the ECG | |
| US9125574B2 (en) | System and method for acoustic detection of coronary artery disease and automated editing of heart sound data | |
| US20120071767A1 (en) | Pulmonary artery pressure estimator | |
| CN111329508A (en) | Heart murmur intelligent analysis method for precordial disease screening | |
| JP2012513858A (en) | Method and system for processing heart sound signals | |
| Haghighi-Mood et al. | A sub-band energy tracking algorithm for heart sound segmentation | |
| Lee et al. | Comparison between short time Fourier and wavelet transform for feature extraction of heart sound | |
| Wang et al. | Intelligent diagnosis of heart murmurs in children with congenital heart disease | |
| Sedighian et al. | Pediatric heart sound segmentation using Hidden Markov Model | |
| Argha et al. | A novel automated blood pressure estimation algorithm using sequences of Korotkoff sounds | |
| Milani et al. | A critical review of heart sound signal segmentation algorithms | |
| Balogh et al. | Application of phonocardiography on preterm infants with patent ductus arteriosus | |
| Di Maria et al. | An algorithm for the analysis of fetal ECGs from 4-channel non-invasive abdominal recordings | |
| Hossain et al. | Wavelet and spectral analysis of normal and abnormal heart sound for diagnosing cardiac disorders | |
| Choudhary et al. | Delineation and analysis of seismocardiographic systole and diastole profiles | |
| Niida et al. | Fetal Heart Rate Detection Using First Derivative of ECG Waveform and Multiple Weighting Functions | |
| White et al. | Analysing heart murmurs using time-frequency methods | |
| Golpaygani et al. | Detection and identification of S1 and S2 heart sounds using wavelet decomposition method | |
| Hikmah et al. | A signal processing framework for multimodal cardiac analysis | |
| Ning et al. | Identification and features extraction of systolic and diastolic murmurs | |
| Samanta et al. | Identification of coronary artery diseased subjects using spectral featuries | |
| Kumar | Automatic heart sound analysis for cardiovascular disease assessment | |
| Zhidong | Noninvasive diagnosis of coronary artery disease based on instantaneous frequency of diastolic murmurs and SVM | |
| Vullings et al. | Artifact reduction in maternal abdominal ECG recordings for fetal ECG estimation |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200626 |