TWI669101B - Heart sound processing method and system with eigenvalue detection - Google Patents

Abstract

A heart sound processing method includes: cutting a heart sound signal to obtain a plurality of heart sound signal segments; performing continuous wavelet transform on the heart sound signal segment to obtain a CWT data; or performing a short time interval Fourier transform on the heart sound signal segment, Obtaining an STFT data; and calculating, by using the CWT data or the STFT data, at least one eigenvalue and at least one eigenvector, and generating a correlation value from the eigenvalue, or comparing the eigenvalue with the same eigenvalue, To produce a comparison result.

Description

Heart sound processing method and system using eigenvalue detection

The present invention relates to a heart sound processing method and system thereof using feature value detection; and more particularly to a heart sound processing method and system using cardiopathy feature value detection.

A related art of cardiac sound processing for cardiac pathological detection, for example, a discriminating system for heart sound/heart disease risk and a method thereof in the Republic of China Patent Publication No. 201244691, and a method for inventing a patent, which discloses a discrimination of heart sound/heart disease risk system. The heart sound/heart disease risk discrimination system comprises a signal receiving unit, a signal processing unit, a storage unit, an output unit and a display unit.

The signal receiving unit is configured to receive a heart sound signal A, and the heart sound signal includes a plurality of heart rate. The signal processing unit includes a first operation unit, a filtering unit, a second operation unit, and a comparison unit. The first operation unit performs a special function calculation on the heart sound signal based on the absolute value of the heart sound signal as a natural logarithm, and then multiplies the value by the product of the heart sound signal (for example, X=cAln|A'|, where c can be Any value or function value] to generate a first operational signal. When A ≠ 0, A' = A; when A = 0, A' = R [R ≧ 1, and R is a real number]. The filtering unit filters the first operational signal to generate a filtered signal. The second operation unit performs a conversion operation on the filtered signal to generate a corresponding image map data, and generates an image map data according to the corresponding image map data.

The storage unit is configured to store a predetermined heart sound spectrum data related to a heart disease, and the output unit is a wireless transmission module or a wired transmission module. The comparison unit compares the image map data with the predetermined heart sound map data and generates a comparison result. Then, the output unit outputs a signal of the comparison result to the display unit to display a physiological state and/or a subsequent processing method.

However, the heart sound processing technique of the aforementioned No. 201244691 must first generate image map data, and then compare the image map data with the predetermined heart sound map data. Therefore, there is still a need for further improvement in the heart sound processing technology of the aforementioned No. 201244691. The foregoing patent application of the Republic of China is merely a reference to the technical background of the present invention and the state of the art is not intended to limit the scope of the present invention.

In view of the above, the present invention provides a heart sound processing method and system for eigenvalue detection, which cuts a heart sound signal to obtain a plurality of heart sound signal segments, and continuously performs the heart sound signal segment. Wavelet transform to obtain a CWT data, and performing short-time Fourier transform on the heart sound signal segment to obtain an STFT data, and calculating at least one feature value and at least one feature vector by using the CWT data or the STFT data, and The eigenvalue produces a correlation value, or the eigenvalue is compared with the same eigenvalue to produce a comparison result, so as to greatly improve the convenience of cardiac pathological detection.

The main object of the present invention is to provide a method for processing heart sounds using feature value detection and a system thereof, which cuts a heart sound signal to obtain a plurality of heart sound signal segments, and performs continuous wavelet transform on the heart sound signal segments to obtain a CWT data, and performing a short-time Fourier transform on the heart sound signal segment to obtain an STFT data, using the CWT data or the STFT data to generate at least one eigenvalue and at least one eigenvector, and generating a correlation from the eigenvalue The value, or the feature value, is the same as The eigenvalues are compared and determined to produce a comparison result for the purpose of providing heart pathological detection by heart sound.

In order to achieve the above object, a heart sound processing method using feature value detection according to a preferred embodiment of the present invention includes: cutting a heart sound signal to obtain a plurality of heart sound signal segments; and performing continuous wavelet transform on the heart sound signal segment to obtain a CWT data; or performing a short-time Fourier transform on the heart sound signal segment to obtain an STFT data; and calculating at least one eigenvalue and at least one eigenvector using the CWT data or the STFT data, and generating a feature value from the eigenvalue The correlation value, or the feature value is compared with the same eigenvalue to produce a comparison result.

The preferred embodiment of the present invention uses the correlation value to determine at least one CWT data or at least one STFT data of another measured heart sound signal to generate another comparison result.

The alignment of the preferred embodiment of the invention comprises a ventricular septal defect, a comparison of the ventricle dividers, an atrial septal defect or an atrial septum.

The correlation value of the characteristic value of the preferred embodiment of the invention comprises a positive correlation value or a cross correlation value.

In the preferred embodiment of the present invention, the heart sound signal includes a first heart sound S1 signal and a second heart sound S2 signal, and the feature value is generated by the second heart sound S2 signal.

The positive correlation value and the cross-correlation value of the preferred embodiment of the present invention are used to establish a congenital ventricular separation model.

The congenital ventricular separation model of the preferred embodiment of the invention is a related numerical model.

In the preferred embodiment of the present invention, the CWT data, the STFT data and the ultrasonic data are pre-aligned to find a maximum frequency point sound, a highest amplitude position sound, and at least one sound interval interval.

The preferred embodiment of the present invention calculates the CWT data and the STFT data using the Pearson product difference correlation coefficient.

In order to achieve the above object, a heart sound processing system using feature value detection according to another preferred embodiment of the present invention includes: a heart sound receiving unit for receiving at least one heart sound signal; and a heart sound processing unit for performing the heart sound signal Cutting to obtain a plurality of heart sound signal segments, and performing continuous wavelet transform on the heart sound signal segment to obtain a CWT data, or performing short-time Fourier transform on the heart sound signal segment to obtain an STFT data; a data storage unit And the output unit is configured to output a correlation value, wherein the CWT data or the STFT data is used to generate at least one eigenvalue and at least one eigenvector, and the eigenvalue is generated by the eigenvalue A correlation value, or the eigenvalue is compared with the same eigenvalue to produce a comparison result.

The heart sound receiving unit of the preferred embodiment of the present invention comprises a first receiving unit and a second receiving unit.

The heart sound receiving unit of the preferred embodiment of the present invention is attached to a surface of a human skin.

The correlation value of the characteristic value of the preferred embodiment of the invention comprises a positive correlation value or a cross correlation value.

In the preferred embodiment of the present invention, the heart sound signal includes a first heart sound S1 signal and a second heart sound S2 signal, and the feature value is generated by the second heart sound S2 signal.

The positive correlation value and the cross-correlation value of the preferred embodiment of the present invention are used to establish a congenital ventricular separation model.

The congenital ventricular separation model of the preferred embodiment of the invention is a related numerical model.

In the preferred embodiment of the present invention, the frequency of the heart sound signal segment is selected between 1 Hz and 100 Hz.

In the preferred embodiment of the present invention, the heart sound signal segment includes a predetermined number of heart sound signals, and the predetermined number of heart sound signals of each of the heart sound signal segments is the same.

In the preferred embodiment of the present invention, when the heart sound signal segment includes a heart sound murmur signal, the heart path detection is performed by detecting the heart sound murmur signal.

S1‧‧‧ steps

S2‧‧‧ steps

S3‧‧‧ steps

S4A‧‧‧ steps

S4B‧‧‧ steps

10‧‧‧heart sound receiving unit

20‧‧‧heart sound processing unit

30‧‧‧Data storage unit

40‧‧‧Output unit

Figure 1 is a block diagram of a heart sound processing system employing feature value detection in accordance with a preferred embodiment of the present invention.

2 is a flow chart of a heart sound processing method using feature value detection according to a preferred embodiment of the present invention.

3(a) and 3(b) are diagrams showing waveforms of heart sound data and electrocardiogram data by a heart sound processing system using feature value detection according to a preferred embodiment of the present invention.

4(A) and 4(B) are diagrams showing waveforms of the first heart sound data and the second heart sound data detected by the heart rate processing system using the feature value detection in the preferred embodiment of the present invention.

FIG. 5(A) is a schematic diagram showing the waveform of a slight aortic regurgitation and a slight aortic stenosis after the continuous wavelet transform of the heart sound data of the first patient is detected by the heart sound processing system using the feature value detection in the preferred embodiment of the present invention.

Figure 5(B): Heart sound of eigenvalue detection according to a preferred embodiment of the present invention The processing system detects the heartbeat data of the second patient and displays the waveform of the continuous cardiac murmur and the open arterial catheter condition after continuous wavelet transformation.

Figure 6 (A): The heart sound processing system using the feature value detection of the preferred embodiment of the present invention detects the heart sound data of the first patient, and displays the waveform of a slight aortic regurgitation and a slight aortic stenosis after a short time interval Fourier transform schematic diagram.

Figure 6(B): The heart sound processing system using the feature value detection according to the preferred embodiment of the present invention detects the heart sound data of the second patient and displays the continuous cardiac murmur and the open arterial catheter condition after the short-time Fourier transform The waveform diagram.

FIG. 7(A) is a schematic diagram showing the waveform of the original heart sound data, CWT data and STFT data of the first patient by using the heart rate processing system with feature value detection according to a preferred embodiment of the present invention.

FIG. 7(B) is a waveform diagram showing the comparison of the original heart sound data, the CWT data and the STFT data of the second patient by the heart sound processing system using the feature value detection according to the preferred embodiment of the present invention.

In order to fully understand the present invention, the preferred embodiments of the present invention are described in detail below and are not intended to limit the invention.

The heart sound processing method and system thereof using cardiac pathological feature value detection according to a preferred embodiment of the present invention are applicable to various cardiac pathological detecting devices and related application devices, for example, various types of home care systems, medical equipment automatic A control system (eg, a medical examination system), a telemedicine system, or a medical teaching system is not intended to limit the scope of the invention.

1 is a block diagram showing a heart sound processing system using feature value detection in accordance with a preferred embodiment of the present invention. Referring to FIG. 1 , for example, a heart sound processing system for cardiac pathology detection according to a preferred embodiment of the present invention includes a heart sound receiving unit 10, a heart sound processing unit 20, and a data storage. The memory unit 30 and an output unit 40 are connected to the heart sound processing unit 20 as appropriate. For example, the heart sound receiving unit 10 includes a first receiving unit and a second receiving unit, and the heart sound receiving unit 10 is attached to a human skin surface.

Referring to FIG. 1 again, for example, the data storage unit 30 is configured to store a plurality of ultrasonic data or cardiac pathological ultrasound data samples for comparing the heart sound data of the continuous wavelet transform and the short-time Fourier transform processing. To establish a correlation sample of time and frequency, so that cardiac pathology can be performed accordingly.

FIG. 2 is a flow chart showing a heart sound processing method using feature value detection according to a preferred embodiment of the present invention, which corresponds to the heart sound processing system for cardiac pathology detection of FIG. 1. Referring to FIGS. 1 and 2, the method for processing heart sounds using feature value detection according to a preferred embodiment of the present invention includes the step S1: first, a heart sound signal or a heart sound signal segment and other signals (eg, ECG, electrocardiogram) ] After receiving it with an appropriate measuring device (for example, the measuring device of the AUDICOR brand), it is cut to obtain a plurality of heart sound signal fragments.

3(a) and 3(b) are diagrams showing waveforms of heart sound data and electrocardiogram data using a heart rate processing system using feature value detection according to a preferred embodiment of the present invention. Referring to FIG. 3(a), the heart sound processing system for cardiac pathology detection of the preferred embodiment of the present invention simultaneously receives two heart sound signals of two receiving units, so that two heart sound maps (PCG, Phonocardiogram]. In contrast, referring to FIG. 3(b), the heart sound processing system for cardiac pathology detection of the preferred embodiment of the present invention simultaneously receives corresponding electrocardiogram (ECG) data.

In general, the heart sound includes a first heart sound S1, a second heart sound S2, a third heart sound S3, and a fourth heart sound S4, and the heart sound is caused by vibration caused by myocardial contraction, heart valve closure, blood impact ventricular wall or aortic wall. The voice of life. When the heart contracts and relaxes, the first heart sound S1 and the second heart sound S2 are generated, so that it is extremely easy to detect (for example, a stethoscope or an instrument). As for the third heart sound S3, it can be detected during childhood or adolescence, while the fourth heart sound S4 is extremely difficult to detect.

4(A) and 4(B) are diagrams showing waveforms of the first heart sound data and the second heart sound data detected by the heart rate processing system using the feature value detection according to a preferred embodiment of the present invention, which are different in the area around the heart. The heart sound data detected by the patch. For example, the heart sound signal segment selection includes a predetermined number of heart sound signals (for example, 400 pens), and the predetermined number of heart sound signals of each of the heart sound signal segments is selected to be the same. In addition, the frequency of the heart sound signal segment is selected between 1 Hz and 100 Hz or other low frequency intervals.

Referring to FIG. 1 and FIG. 2 again, the heart sound processing method using feature value detection according to the preferred embodiment of the present invention includes the step S2: then, the heart sound signal segment is subjected to continuous wavelet transformation (CWT) processing. To obtain a CWT data that shows a first time and frequency of cardiac pathology waveforms.

FIG. 5(A) illustrates a heart sound processing system using feature value detection according to a preferred embodiment of the present invention for detecting heart sound data of a first patient, showing a small aortic regurgitation and a slight aorta after continuous wavelet conversion. Schematic diagram of the state of the stenosis (AS, aortic stenosis). Referring to Figure 5(A), after the continuous wavelet conversion of the heart sound data of the first patient, a first continuous wavelet transform waveform (as shown in the upper half of Fig. 5(A)) and one The second continuous wavelet transform waveform is as shown in the lower half of Fig. 5(A). The first continuous wavelet converted waveform and the second continuous wavelet converted waveform have a first maximum amplitude and a second maximum amplitude, as indicated by the two arrows.

Referring to FIG. 5(A), FIG. 5(B) discloses a heart sound processing system using feature value detection according to a preferred embodiment of the present invention to detect the heart of a second patient. After the continuous wavelet transform of the sound data, the waveforms of the continuous murmur and the open ductal artery (PDA, patent ductus arteriosum) are shown. Referring to Figure 5(B), after the continuous wavelet conversion of the heart sound data of the second patient, a first continuous wavelet transform waveform is obtained (as shown in the upper half of Fig. 5(B)) and one The second continuous wavelet transform waveform (as shown in the lower half of Fig. 5(B)), and the heart sound murmur (outside a main wave) appears, as shown by the smaller waveform. Similarly, the first continuous wavelet transform waveform and the second continuous wavelet transform waveform have a first maximum amplitude and a second maximum amplitude, as indicated by two arrows, which are located in the main wave.

Referring to FIG. 1 and FIG. 2 again, the heart sound processing method using the feature value detection according to the preferred embodiment of the present invention includes the step S3: then, the heart sound signal segment is subjected to short-time Fourier transform (short-time Fourier transformation). , STFT] processing to obtain an STFT data showing a second time and frequency of cardiac pathology waveforms.

FIG. 6(A) is a diagram showing a waveform of a heart rate processing system for detecting a first patient's heart sound data after a short time interval Fourier transform according to a preferred embodiment of the present invention, showing a slight aortic regurgitation and a slight state of aortic stenosis. schematic diagram. Referring to Figure 6(A), after the short-distance Fourier transform of the first patient's heart sound data, a first short-time Fourier transform waveform is obtained (as in the upper half of Figure 6(A)). And a second short-time Fourier transform waveform (as shown in the lower half of Figure 6(A)). The first short time interval The Fourier transform waveform and the second short time Fourier transform waveform have a first maximum amplitude and a second maximum amplitude, as indicated by the two arrows.

Referring to FIG. 6(A), FIG. 6(B) discloses a heart sound processing system using feature value detection according to a preferred embodiment of the present invention to detect a heart sound data of a second patient after a short time interval Fourier transform, and display continuous Schematic diagram of the waveform of murmur and open arterial catheter status. Referring to Figure 6(B), after the short-distance Fourier transform of the heart sound data of the second patient, a first short-time Fourier transform waveform is obtained (as in the upper half of Figure 6(B)). And a second short-time Fourier transform waveform (as shown in the lower half of Figure 6(B)), and a heart sound murmur occurs, as shown by the smaller waveform. Similarly, the first short-time Fourier transform waveform and the second short-time Fourier transform waveform have a first maximum amplitude and a second maximum amplitude, as indicated by the two arrows.

Referring to FIGS. 1 and 2, the method for processing heart sounds using feature value detection according to a preferred embodiment of the present invention includes the step S4A: then, using the CWT data, the STFT data, and an ultrasonic data to compare A time-dependent numerical state of the frequency to produce a correlation value. The at least one CWT data and the at least one STFT data of the other measured heart sound signal are determined by using the correlation judgment value to generate a determination result. The judgment result of the preferred embodiment of the present invention includes a ventricular septal defect (VSD) or atrial septal defect (ASD). When the heart sound signal segment includes a heart sound noise signal, the heart path detection is performed by detecting the heart sound noise signal.

FIG. 7(A) is a schematic diagram showing the waveforms of the original heart sound data, CWT data and STFT data of the first patient by using the heart rate processing system with feature value detection according to a preferred embodiment of the present invention. In contrast, FIG. 7(B) is a schematic diagram showing waveforms of the original heart sound data, CWT data and STFT data of the second patient using the heart rate processing system using the feature value detection according to the preferred embodiment of the present invention. Please refer to paragraphs 7(A) and 7(B) to properly adjust the original heart sound data, CWT data and STFT data of the first patient and the second patient (eg, 7(A) And the left and right sides of the 7(B) diagram, and select the waveform pattern comparison and calculation correlation between the CWT data and the STFT data respectively (as shown in Table 3), or select in the CWT Waveform pattern comparison and computational correlation between data and STFT data.

In the preferred embodiment of the present invention, the CWT data and the STFT data are calculated using a Pearson product-moment coefficient. The preferred embodiment of the present invention uses the calculation formula of the Pearson product difference correlation coefficient as follows:

Where z _x and z _y are the normalized z values of x and y , respectively, and r is the correlation coefficient.

Referring to FIG. 1 and FIG. 2 again, the heart sound processing method using the feature value detection according to the preferred embodiment of the present invention includes the step S4B: then, using the CWT data or the STFT data to calculate and generate at least one eigenvalue and At least one feature vector [eigenvector], and a correlation value is generated from the feature value, or the feature value is the same as the sample The feature value (for example, as shown in step S4A) is subjected to comparison discrimination to generate a comparison result, and the sample feature value, the correlation value or the comparison result may be optionally stored in the data storage unit 30.

For example, the correlation value of the eigenvalue according to another preferred embodiment of the present invention includes an auto-correlation value or a cross-correlation value, and the eigenvalue is determined by the second The heart sound S2 signal is generated.

Another preferred embodiment of the present invention employs a unitary correlation value, also referred to as a sequence correlation, which is a cross-correlation of a signal at its own point in time. The positive correlation value is a function of the similarity between two observations as a function of the time difference between itself, which is used to find a repeating pattern (eg, a periodic signal masked by noise or noise), or to identify hidden signals in the signal. A tool for the fundamental frequency that disappears in harmonic frequencies.

Another preferred embodiment of the present invention employs a 〝 cross-correlation value 〞, also known as a 〝 mutual covariance 〞. A cross-correlation tool used in the field of signal processing to represent the similarity between two signals, which is typically applied to find characteristics in unknown signals using a comparison with known signals, which is time-to-time between two signals. A function, sometimes called a sliding dot product.

Referring to FIGS. 1 and 2 again, for example, in the preferred embodiment of the present invention, when calculating the positive correlation of the characteristic values of the CWT data or the STFT data, the formula may be selected as follows: However, it is not intended to limit the invention.

Referring to FIGS. 1 and 2, for example, in the preferred embodiment of the present invention, when calculating the cross-correlation of the characteristic values of the CWT data or the STFT data, the formula may be selected as follows: However, it is not intended to limit the invention.

For example, the positive correlation value and the cross-correlation value of another preferred embodiment of the present invention are selected to establish a congenital septal defect (CSD), and the congenital ventricular separity model is a related value. Model or other mathematical model.

For example, the positive correlation value and the cross-correlation value of the correlation value of the heart sound signal segment A and its characteristic value in the preferred embodiment of the present invention are as follows:

Where the diagonal singular value is a positive correlation value, and the rest are cross-correlation values;

Because the polynomial is Therefore, the heart sound signal segment A can be expressed as And further can be expressed as |A-λI|=0, thus, the congenital ventricular separation model of the preferred embodiment of the present invention is expressed as follows: However, it is not intended to limit the invention.

The foregoing preferred embodiments are merely illustrative of the invention and the technical features thereof, and the techniques of the embodiments can be carried out with various substantial equivalent modifications and/or alternatives; therefore, the scope of the invention is subject to the appended claims. The scope defined by the scope shall prevail.

Claims (10)

A heart sound processing method using feature value detection, comprising: cutting a heart sound signal to obtain a plurality of heart sound signal segments; performing continuous wavelet transform on the heart sound signal segment to obtain a CWT data; or the heart sound signal Performing a short-time Fourier transform to obtain an STFT data; and calculating at least one eigenvalue and at least one eigenvector by using the CWT data or the STFT data, and generating a correlation value from the eigenvalue, or the eigenvalue is the same The eigenvalues are compared and compared to produce a comparison result. The heart sound processing method using feature value detection according to claim 1 of the patent application scope, wherein the correlation value of the feature value comprises a positive correlation value or a cross correlation value. The heart sound processing method using the feature value detection according to the first aspect of the patent application, wherein the heart sound signal comprises a first heart sound S1 signal and a second heart sound S2 signal, and the feature value is determined by the second heart sound S2 The signal is generated. The heart sound processing system using feature value detection according to claim 1 of the patent application scope, wherein the positive correlation value and the cross correlation value are used to establish a congenital ventricular separation model. The heart sound processing system using feature value detection according to item 4 of the patent application scope, wherein the congenital ventricular separation model is a related numerical model. A heart sound processing system using feature value detection, comprising: a heart sound receiving unit for receiving at least one heart sound signal; and a heart sound processing unit for cutting the heart sound signal to obtain a plurality of heart sound signal segments, and Performing continuous wavelet transform on the heart sound signal segment to obtain a CWT data, or performing short-time Fourier transform on the heart sound signal segment to obtain an STFT data; and a data storage unit for storing at least one related cardiac pathological super a sample of the sound wave data; and an output unit for outputting a correlation judgment value; The at least one eigenvalue and the at least one eigenvector are generated by using the CWT data or the STFT data, and a correlation value is generated by the eigenvalue, or the eigenvalue is compared with the same eigenvalue to generate an alignment result. . A heart sound processing system using feature value detection according to item 6 of the patent application scope, wherein the correlation value of the feature value comprises a positive correlation value or a cross correlation value. The heart sound processing system using the feature value detection according to the sixth aspect of the patent application, wherein the heart sound signal comprises a first heart sound S1 signal and a second heart sound S2 signal, and the feature value is determined by the second heart sound S2 The signal is generated. The heart sound processing system using feature value detection according to item 6 of the patent application scope, wherein the positive correlation value and the cross correlation value are used to establish a congenital ventricular separation model. The heart sound processing system using feature value detection according to claim 9 of the patent application scope, wherein the congenital ventricular separation model is a related numerical model.