WO2020151236A1

WO2020151236A1 - Method for automatically positioning and repairing clipping distortion waveforms of electronic stethoscope

Info

Publication number: WO2020151236A1
Application number: PCT/CN2019/106382
Authority: WO
Inventors: 蔡盛盛; 胡南; 徐兴国; 周宁
Original assignee: 苏州美糯爱医疗科技有限公司
Priority date: 2019-01-23
Filing date: 2019-09-18
Publication date: 2020-07-30
Also published as: CN109893163A; CN109893163B

Abstract

A method for automatically positioning and repairing clipping distortion waveforms of an electronic stethoscope, comprising the following steps: reading an auscultation signal; obtaining a differential of a signal x to obtain a differential signal x_d; determining, by means of threshold comparison and endpoint pairing, K non-overlapping time intervals where clipping distortion occurs, sequencing the time intervals in time order, and forming a stack in a manner of first entering and then exiting; if K=0, determining that there is no clipping distortion area and repairing is no longer needed; if K≥1, determining a data time point set Ψ_K interpolated on the stack; by using the time point set Ψ_K and a corresponding value x(Ψ_K) on the x at each time point of the time point set, fitting a value in a Kth section of clipping distortion area by means of interpolation to replace an original clipping distortion value on the x; removing the time interval at the top of the stack; and outputting the auscultation signal x without clipping distortion. According to the method, the clipping distortion area of a stethoscope signal can be automatically detected and positioned, and the signal of the clipping distortion area and a plurality of close clipping distortion areas can be automatically repaired.

Description

Method for automatically positioning and repairing clipping and distorted waveform of electronic stethoscope

Technical field

The invention belongs to the technical field of stethoscopes, and relates to a method for automatically positioning and repairing the clipping and distorted waveforms of an electronic stethoscope.

Background technique

Auscultation is one of the most important technologies for physicians to understand the patient’s condition at the first time. However, this technology has long been restricted by factors such as the location of the clinic and the level of medical skills. With the rapid development of Internet of Things (IoT) technology, various types of The continuous emergence of electronic stethoscopes has brought the possibility of real-time monitoring, automatic transcription, cloud diagnosis and treatment, and intelligent diagnosis of patients' heart and lung sound data.

The electronic stethoscope uses a transducer (microphone) to convert weak physiological sound signals (such as heart sounds, lung sounds, etc.) into electrical signals. Currently, electronic stethoscopes generally use piezoelectric microphones with high sensitivity and excellent frequency characteristics. The thin film on the cavity collects sound vibrations and causes the crystal deformation to be converted into electric current. However, in practical applications, the output current of the electronic stethoscope quickly reaches saturation due to excessive pressing during auscultation, which causes clipping and distortion of the recorded auscultation signal. The phenomenon shown in Figure 1 occurs from time to time: such as the pediatric clinic due to infant patients The degree of cooperation is low, and excessive compression may occur during the auscultation process; another example is when the patient uses an electronic stethoscope to upload the auscultation data for cloud diagnosis and treatment, it is easy to cause clipping distortion and clipping distortion due to improper operation The emergence of on the one hand makes the experience of using the electronic stethoscope worse, on the other hand it affects the signal quality and then affects the subsequent heart sound localization and automatic diagnosis of heart and lung sounds.

Regarding the problem of signal clipping and distortion caused by pressing, the shortcomings of the prior art are embodied in the following aspects: a. There is currently no automatic positioning method for signal clipping and distortion regions; b. There is currently no automatic repair method for signal clipping and distortion regions. Perfection: It is mainly an important prerequisite to solve the problem-the lack of automatic positioning of the clipping distortion area, and how to better recover the missing part of the signal caused by clipping according to the context signal content is still a problem to be considered; c At present, there is no technical solution for repairing multiple clipping distortion regions in a short distance.

In particular, it can automatically locate and repair the clipping distortion in the auscultation signal for the purpose of realizing high-precision real-time monitoring of the fetus, intelligent assessment of heart/lung function, automatic diagnosis of heart/lung diseases and other electronic stethoscopes. , Is one of the important prerequisites for the realization of these artificial intelligence functions. Therefore, clipping and distortion of electronic stethoscope signals is a common problem that electronic stethoscopes urgently need to solve.

The focus of existing patents related to electronic stethoscopes includes hardware acquisition system, transmission system, appearance, signal preprocessing (including noise reduction, heart sound localization, heart and lung sound separation, etc.), signal intelligent analysis (fetal heart monitoring, Intelligent diagnosis of heart disease based on heart sound, intelligent diagnosis of respiratory disease based on lung sound, etc., but there is no patent to propose an automatic repair solution for signal clipping and distortion.

And the paper "Emmanouilidou D, McCollum E D, Park D E, et al. Computerized Lung Sound Screening for Pediatric Auscultation in Noisy Field Environment[J]. IEEE Transactions on Biomedical Engineering, 2018, 65(7): 1564-1574 A method for clipping distortion repair using cubic spline interpolation is proposed in. However, the method in this document has the following problems: (1) No method for automatically detecting and positioning the clipping distortion area is proposed. Therefore, the premise of this method is to manually select the clipping distortion area, which does not meet the requirements of practical applications. ; (2) The curve fitted by the cubic spline interpolation used in this method can only guarantee the continuity at the connection point, but cannot guarantee its smoothness; the entire fitted curve is controlled by all interpolation points, if there is any interpolation point One change will affect the entire curve, which is not in line with its actual characteristics for the restoration of auscultation data. Especially for the auscultation of respiratory diseases, the duration of the specific breathing additional sounds is often very short ( For example, the duration of a single wet rale is about 20 milliseconds), in this case the cubic spline interpolation fitting will bring a larger error; (3) It is not proposed when there are two or more segments that are close together. The repair strategy for the distorted area, and if they are directly repaired at the same time, it may cause underfitting due to too many points to be fitted.

Summary of the invention

The purpose of the present invention is to provide a method for automatically locating and repairing the clipping distortion waveform of an electronic stethoscope, which solves the problem that the prior art cannot automatically detect and locate the clipping distortion area of the stethoscope, and cannot automatically repair the clipping distortion area. Fix the problem of clipping the signal in the distortion area.

The purpose of the present invention can be achieved through the following technical solutions:

A method for automatically locating and repairing the clipped and distorted waveform of an electronic stethoscope has the following steps:

Step 1. Read the auscultation signal sampling sequence x(n) with a duration of N in the buffer, n=1, 2,...,N, and express it as a vector form x;

Step 2. Differentiate the original signal x to obtain the differential signal x _d . The expression for calculating the differential signal of the original signal is x _d (n) = x(n)-x(n-1), where x _d (1)=0;

Step 3. According to the differential signal x _d , through threshold comparison and end-point pairing, determine the K non-overlapping time intervals (K≥0) where clipping distortion occurs, which are in chronological order: [n _K,begin ,n _K,end ],[n _K-1,begin ,n _K-1,end ],...,[n _1,begin ,n _1,end ], the clipping distortion interval that appears after pressing enters first and then exits Way to form a stack;

Step 4. If K=0, there is no clipping distortion area in this segment of data and no longer needs to be repaired. Go directly to step 7. If K≥1, determine the earliest clipping distortion area in the stack [n _K,begin ,n _K,end ] The set of interpolation data time points Ψ _K required for repairing;

Step 5. Use the set of time points Ψ _K and the value x (Ψ _K ) on x corresponding to each time point, through Hermite interpolation, to fit the value in the K-th clipping distortion region to replace the original value in x The clipping distortion value of, thereby updating x;

Step 6. Remove the time interval [n _K,begin ,n _K,end ] at the top of the stack, set K=K-1, and return to step 4.

Step 7. Output the auscultation signal x without clipping distortion.

Further, forming a stack in the step 3 specifically includes the following steps:

S1. Find the time point corresponding to the value of the difference signal x _d whose absolute value is greater than the threshold α·max(|x|) to form a set {n ₁ ,n ₂ ,...,n _L }, where 0＜ α<1 is a preset constant;

S2. From the set {n ₁ ,n ₂ ,...,n _L }, find the time points with opposite signs on adjacent x _d and pair them, and determine that they are the beginning of a certain clipping distortion interval Start point n _{k, begin} and end point n _{k, end} ; if a single time point that fails to be matched is found near the beginning or end of the entire data, it will be paired additionally with the previous or subsequent data;

S3. According to the pairwise pairing results, if a total of K non-overlapping time intervals are obtained in time sequence, they are [n _K,begin ,n _K,end ],[n _K-1,begin ,n _K-1,end ],...,[n _1,begin ,n _1,end ], the stack is formed in the way that the clipping distortion interval that appears later enters the stack first.

Further, the method for acquiring the interpolation data time point set Ψ _K required in step 4 specifically includes the following steps:

H1. Starting from the time point n _{K, begin} , find the time point n _{K, left3} corresponding to the third zero-crossing point on x starting from n _{K, begin} against the direction of the time _axis , if the corresponding point on x is not equal to 0 , Then take the time point corresponding to the value closest to 0;

H2. Starting from the time point n _K,end , follow the time axis to find the time point n _K,right1 corresponding to the first zero-crossing point on x starting from n _K,end , if the corresponding point on x is not exactly equal to 0 , Then take the time point corresponding to the value closest to 0;

H3. Determine the data time point set Ψ _K required to repair the clipping distortion area [n _K,begin ,n _K,end ] is [n _K,left3 ,n _K,begin -P]∪[n _K,end +P ,n _K,right1 ], where P is the number of reserved points set in consideration of the conversion time of the stethoscope output to the saturated state when over-pressing, and the value is generally a non-negative integer from 1 to 10.

Further, the method for updating x in step 5 specifically includes the following steps:

F1, if there are T _k +1 time points in the time set Ψ _K :

Estimate the derivative x′(n _i ) of the data point x(n _i ) at each time point, i=0,1,...,T _k ;

F2. For any point x(m) that needs to be fitted and repaired, where m∈[n _K,begin -P+1,n _K,end +P-1] is the time point on the area to be fitted, use Ψ _K Data point x(n _i ) and its derivative at T _k +1 time point

Find its Hermite interpolation:

F3. Update the value of the K-th to-be-repaired interval in x: x(m)=x _H (m), m∈[n _K,begin -P+1,n _K,end +P-1].

The beneficial effects of the present invention:

The method for automatically locating and repairing the clipping distortion waveform of an electronic stethoscope provided by the present invention realizes the automatic detection and positioning of the clipping distortion area of the stethoscope signal through the difference of the signal, the comparison with the threshold, and the pairing; The area’s automatic positioning combined with the zero-crossing point selects the set of interpolation data time points required for repair, and the signal in the clipping distortion area can be automatically repaired through Hermite interpolation; multiple clipping can be realized in sequence by adopting the "first in, last out" feature of the stack Automatically repair the distortion area, and can solve the repair problem when the clipping distortion area is very close. It has the characteristics of fast positioning and repair and high accuracy.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the following will briefly introduce the drawings used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained from these drawings without creative work.

Fig. 1 is a schematic diagram of clipping distortion of auscultation signal in the prior art;

2 is a schematic diagram of a method for automatically locating and repairing the clipping and distorted waveform of an electronic stethoscope in the present invention;

Fig. 3 is a schematic diagram of repair in embodiment 1 of the present invention;

4 is a schematic diagram of repair in Embodiment 2 of the present invention;

detailed description

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

As shown in Figure 1, a method for automatically positioning and repairing the clipped and distorted waveform of an electronic stethoscope has the following steps:

The formation of a stack in step 3 includes the following steps:

S1. Find the time point corresponding to the value of the difference signal x _d whose absolute value is greater than the threshold α·max(|x|) to form a set {n ₁ ,n ₂ ,...,n _L }, where 0<α<1 is a preset constant;

In step 4, if K≥1, determine the interpolation data time point set Ψ _K acquisition method required to repair the earliest clipping distortion area [n _K,begin ,n _K,end ] in the stack, which specifically includes the following steps:

Step 5. Use the set of time points Ψ _K and the value x (Ψ _K ) on x corresponding to each time point, through Hermite interpolation, fit the value in the K-th band of clipping distortion area to replace the original value in x The clipping distortion value of, thereby updating x;

The method of updating x includes the following steps:

F1, if there are T _k +1 time points in the time set Ψ _K :

F2. For any point x(m) that needs to be fitted and repaired, where m∈[n _K,begin -P+1,n _K,end +P-1] is the time point on the area to be fitted, use Ψ _K Among T _k +1 data points x(n _i ) and its derivative x′(n _i ),i=0,1,...,T _k , obtain its Hermite interpolation:

Step 7. Output the auscultation signal x without clipping distortion.

Example 1:

Read a piece of lung sound auscultation data x of a patient with pediatric pneumonia, which contains two clipping distortion regions, and automatically locate and repair the clipping distortion regions. The data sampling rate is 4KHz and the duration is 2.5 seconds.

First of all, the difference of this piece of data, the difference signal x _d as shown in Figure 3 (a), it can be seen that at the beginning and end of the two clipping distortion regions, there are a pair of very large peaks with opposite polarities.

Then, find the time point corresponding to the value of the difference signal x _d whose absolute value is greater than the threshold 0.8×max(|x|), and find the time point of the opposite sign of the corresponding value on the adjacent x _d for pairing. Detect two clipping distortion areas, and determine their respective starting and ending points, as shown in Figure 3(b);

Finally, put the two positioned clipping distortion intervals into the stack in reverse time order, use the stack operation to determine the set of interpolation points in turn, and repair them in turn through Hermite interpolation, and complete the signal update. The final repaired signal is as As shown in Fig. 3(c), the number of reserved points P=5 is set in consideration of the conversion time of the stethoscope output to the saturated state when over-pressing.

Example 2:

Read a piece of normal human heart sound auscultation data x. There are 5 clipping distortion areas in the data. Among them, there are two clipping distortion areas that are very close, 0.0136 seconds apart, and automatically locate and repair the clipping distortion area. , The data sampling rate is 4KHz, and the duration is 1.5 seconds.

First of all, the difference between the data, the difference signal x _d as shown in Figure 4 (a), it can be seen that there are a pair of very large peaks with opposite polarities at the beginning and the end of the clipping distortion area.

Then, find the time point corresponding to the value of the difference signal x _d whose absolute value is greater than the threshold 0.8×max(|x|), and find the time point of the opposite sign of the corresponding value on the adjacent x _d for pairing. Detect all 5 clipping distortion areas, and determine their respective starting and ending points, as shown in Figure 4(b);

Finally, put the 5 positioned clipping distortion intervals into the stack in reverse time order, use the stack operation to determine the set of interpolation points in turn, and repair them in turn through Hermite interpolation, and complete the signal update. The final repaired signal is shown in the figure As shown in 4(c), the number of reserved points P=5 is set in consideration of the transition time for the stethoscope output to switch to the saturated state when over-pressing. It can be seen that even if there are two clipping distortion areas that are very close, the use of the present invention The proposed method can still achieve a more accurate repair.

The above content is merely an example and description of the concept of the present invention. Those skilled in the art make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention. Or beyond the scope defined by the claims, all should belong to the protection scope of the present invention.

Claims

A method for automatically locating and repairing the clipped and distorted waveform of an electronic stethoscope is characterized in that it has the following steps:

Step 1. Read the auscultation signal sampling sequence x(n) with a duration of N in the buffer, n=1, 2,...,N, and express it as a vector form x;

Step 2. Differentiate the original signal x to obtain the differential signal x d . The expression for calculating the differential signal of the original signal is x d (n) = x(n)-x(n-1), where x d (1)=0;

Step 3. According to the differential signal x d , through threshold comparison and end-point pairing, determine the K non-overlapping time intervals (K≥0) where clipping distortion occurs, which are in chronological order: [n K,begin ,n K,end ],[n K-1,begin ,n K-1,end ],...,[n 1,begin ,n 1,end ], the clipping distortion interval that appears after pressing enters first and then exits Way to form a stack;

Step 4. If K=0, there is no clipping distortion area in this segment of data and no longer needs to be repaired. Go directly to step 7. If K≥1, determine the earliest clipping distortion area in the stack [n K,begin ,n K,end ] The set of interpolation data time points Ψ K required for repairing;

Step 5. Use the set of time points Ψ K and the value x (Ψ K ) on x corresponding to each time point, through Hermite interpolation, to fit the value in the K-th clipping distortion region to replace the original value in x The clipping distortion value of, thereby updating x;

Step 6. Remove the time interval [n K,begin ,n K,end ] at the top of the stack, set K=K-1, and return to step 4.

Step 7. Output the auscultation signal x without clipping distortion.
The method for automatically locating and repairing the clipped and distorted waveform of an electronic stethoscope according to claim 1, wherein the forming of the stack in step 3 specifically includes the following steps:

S1. Find the time point corresponding to the value of the difference signal x d whose absolute value is greater than the threshold α·max(|x|) to form a set {n 1 ,n 2 ,...,n L }, where 0＜ α<1 is a preset constant;

S2. From the set {n 1 ,n 2 ,...,n L }, find the time points with opposite signs on adjacent x d and pair them, and determine that they are the beginning of a certain clipping distortion interval Start point n k, begin and end point n k, end ; if a single time point that fails to be matched is found near the beginning or end of the entire data, it will be paired additionally with the previous or subsequent data;

S3. According to the pairwise pairing results, if a total of K non-overlapping time intervals are obtained in time sequence, they are [n K,begin ,n K,end ],[n K-1,begin ,n K-1,end ],...,[n 1,begin ,n 1,end ], the stack is formed in the way that the clipping distortion interval that appears later enters the stack first.
The method for automatically locating and repairing the clipped and distorted waveform of an electronic stethoscope according to claim 1, wherein the method for obtaining the set of interpolated data time points Ψ K required in step 4 specifically includes the following steps:

H1. Starting from the time point n K, begin , find the time point n K, left3 corresponding to the third zero-crossing point on x starting from n K, begin against the direction of the time axis , if the corresponding point on x is not equal to 0 , Then take the time point corresponding to the value closest to 0;

H2. Starting from the time point n K,end , follow the time axis to find the time point n K,right1 corresponding to the first zero-crossing point on x starting from n K,end , if the corresponding point on x is not exactly equal to 0 , Then take the time point corresponding to the value closest to 0;

H3. Determine the data time point set Ψ K required to repair the clipping distortion area [n K,begin ,n K,end ] is [n K,left3 ,n K,begin -P]∪[n K,end +P ,n K,right1 ], where P is the number of reserved points set in consideration of the conversion time of the stethoscope output to the saturated state when over-pressing, and the value is generally a non-negative integer from 1 to 10.
The method for automatically locating and repairing the clipping and distorted waveform of an electronic stethoscope according to claim 1, wherein the method for updating x in step 5 specifically includes the following steps:
F1, if there are T k +1 time points in the time set Ψ K :
Estimate the derivative x′(n j ) at each time point, j=0,1,...,T k ;

F1, if there are T k +1 time points in the time set Ψ K :
Estimate the derivative x′(n i ) of the data point x(n i ) at each time point, i=0,1,...,T k ;

F2. For any point x(m) that needs to be fitted and repaired, where m∈[n K,begin -P+1,n K,end +P-1] is the time point on the area to be fitted, use Ψ K The data point x(n i ) and its derivative x′(n i ), i=0,1,...T k , at the time point of T k +1 in T k +1, obtain its Hermite interpolation:

F3. Update the value of the K-th to-be-repaired interval in x: x(m)=x H (m), m∈[n K,begin -P+1,n K,end +P-1].