Heart valve abnormality detection device, computer program, and heart valve abnormality detection method

Abstract

[Problem] To detect, in a simple, highly precise manner, a cardiac-valve abnormality, in particular, serious aortic valve stenosis, in the form of a clear result. [Solution] Provided is a cardiac-valve abnormality detection device having: a cardiac-sound-data acquisition unit 21 that acquires cardiac sound data corresponding to cardiac sounds; a first processing unit that acquires, on the basis of the cardiac sound data, a peak frequency which is a peak value of a frequency in a frequency component of the cardiac sounds in a target range, the target range being between a first sound and a second sound of the cardiac sounds; and a second processing unit that outputs a detection signal indicating a high likelihood of serious aortic valve stenosis when the peak frequency is equal to or greater than a peak reference value.

Description

Heart valve abnormality detection device, detection method, and computer program

[0001] The present invention relates to a heart valve abnormality detection device, a detection method, and a computer program for detecting heart valve abnormalities such as aortic valve stenosis.

[0002] With the aging of society in recent years, the number of patients with valvular heart disease, particularly severe aortic stenosis (AS), has been increasing. Aortic stenosis is a heart disease in which calcification occurs in the aortic valve with age, causing the valve to harden and become difficult to open, significantly reducing the amount of blood pumped into the aorta. Because it is a serious disease that can lead to sudden death or heart failure, overlooking it poses a major medical problem. Severe AS is said to affect 3.4% of people aged 75 and over, and is estimated to affect 560,000 people in Japan.

[0003] Auscultation, which is widely used in clinical settings, is useful for screening for valvular disease, but it is said that the accuracy of diagnosing valvular disease by non-cardiologists with little experience in auscultation is low, at around 40-50%.The reason for the low diagnostic accuracy of auscultation is that the timing, location, and range of cardiac murmurs vary depending on the type and condition of valvular disease, and it is difficult to make a comprehensive judgment based on this information.

[0004] Conventionally, techniques have been proposed for detecting abnormalities in heart valves by analyzing heart sound information obtained using a stethoscope or the like.

[0005] Patent Document 1 discloses a method for diagnosing heart disease by extracting a first sound signal and a second sound signal from heart sound data based on the heart sounds of a living body, and comparing the amplitude and power of a specific frequency band of the first sound signal and the second sound signal with reference values.

[0006] Patent Document 2 discloses a method for detecting a cardiac murmur from collected heart sounds, generating an analysis signal using the section of the cardiac murmur, and classifying cases of cardiac murmur based on a comparison between a first frequency component, which is a predetermined frequency component of the analysis signal, and a second frequency component, which is a frequency component that constitutes a part of the first frequency component. Paragraph 0026 of Patent Document 2 describes that mitral valve regurgitation or aortic valve stenosis is determined based on the frequency band power ratio.

[0007] Patent Document 3 discloses diagnosing the presence or absence of heart disease based on the shape of the power spectrum in a partial frequency band of systolic heart sound data.

[0008] Furthermore, in order to detect cardiac murmurs, it has been proposed to analyze sound sources acquired by a digital stethoscope using techniques such as deep learning to detect various cardiac diseases (Non-Patent Document 1).

[0009] JP 2017-12366 A JP 2014-87543 A JP 2015-188512 A

[0010] W.Reid Thompson, et al. Artificial Intelligence-Assisted Aus cultation of Heart Murmurs: Validation by Virtual Clinical Trial. Pediatric Cardiology (2019) 40:623-629)

[0011] However, conventionally, it has not been possible to diagnose heart valve abnormalities, particularly severe AS (aortic stenosis), with high accuracy using simple procedures.

[0012] For example, in the case of Patent Document 1 mentioned above, cardiac disease is diagnosed by comparing the amplitude and power of specific frequency bands of the first sound signal and the second sound signal with reference values, so the number of detection parameters used for comparison increases, which increases the burden of calculation processing and makes it easy for errors to accumulate. The same is true for Patent Document 2, which can only simply classify cases. Furthermore, in Patent Document 3, diagnosis is made by comparing the shapes of power spectra, so the judgment is ambiguous, making it difficult to achieve high accuracy, and at best it can only determine the presence or absence of cardiac disease.

[0013] Valvular disease can progress without any obvious subjective symptoms, and if it is severe AS, the only treatment is surgery, so early detection through accurate diagnosis is important. Large hospitals and other specialized facilities use echocardiography to accurately assess the severity of valvular disease, but because it requires equipment and specialists, it is not easy to perform in small clinics, etc., from a cost-effective perspective.

[0014] In the current situation where screening for the need for echocardiography is not adequately performed by auscultation, a simple detection device that is accurate, reliable, and provides clear results is desired.

[0015] The present invention has been made in consideration of the above points, and has as its object to easily detect abnormalities of heart valves, particularly severe aortic valve stenosis, with clear results and high accuracy.

[0016] A first aspect of the heart valve abnormality detection device of the present invention includes a heart sound data acquisition unit that acquires heart sound data corresponding to heart sounds; a first processing unit that, based on the heart sound data, acquires a peak frequency that is the peak value of the frequency components of the heart sound within a target range between the first and second heart sounds; and a second processing unit that outputs a detection signal indicating a high possibility of severe aortic stenosis when the peak frequency is equal to or greater than a peak reference value.

[0017] Preferably, the heart sound data includes a diastolic duration acquisition unit that acquires a diastolic duration based on the heart sound data, and the second processing unit outputs the detection signal when a condition that the peak frequency is equal to or greater than a peak reference value and a condition that the diastolic duration is equal to or greater than an ejection reference value are satisfied.

[0018] More preferably, the device has a detection unit for detecting an ejection click occurring near the first sound based on the heart sound data, and the second processing unit outputs the detection signal when the condition that the peak frequency is equal to or greater than a peak reference value and the condition that the ejection click has been detected are satisfied.

[0019] More preferably, the first processing unit acquires the peak frequency by performing wavelet analysis on the heart sound data.

[0020] A second aspect of the cardiac valve abnormality detection device of the present invention includes a first processing unit that, based on heart sound data, acquires a peak frequency, which is a peak value of a frequency component of the heart sound that has an intensity equal to or greater than a reference intensity within a target range between the first and second heart sounds, and a second processing unit that outputs a detection signal indicating a severe cardiac valve abnormality when the peak frequency is equal to or greater than the reference peak value.

[0021] A third aspect of the cardiac valve abnormality detection device of the present invention includes a first processing unit that acquires, based on heart sound data, a peak frequency that is a peak value of a frequency component of the heart sound that has an intensity equal to or greater than a reference intensity within a target range between the first and second heart sounds; an ejection duration acquisition unit that acquires an ejection duration based on the heart sound data; and a second processing unit that outputs a detection signal that indicates the degree of abnormality of the cardiac valve based on a result of comparing the peak frequency with a set frequency range and a result of comparing the ejection duration with an ejection reference value.

[0022] According to the present invention, abnormalities of the heart valve, particularly severe aortic valve stenosis, can be detected simply and with high accuracy, producing clear results.

[0023] According to the third aspect of the present invention, the degree of abnormality of a heart valve can be detected simply and with high accuracy to provide clear results.

[0024] 1 is a diagram showing the flow from diagnosis to treatment of aortic valve stenosis. FIG. 2 is a block diagram showing an example of the configuration of a heart valve abnormality diagnosis system according to an embodiment of the present invention. FIG. 3 is a diagram showing an example of heart sounds obtained from a heart sound sensor. FIG. 4 is a diagram schematically showing an example of frequency components of heart sounds obtained by wavelet analysis. FIG. 5 is a diagram showing an example of the flow of heart sound signal processing up to wavelet analysis. FIG. 6 is a diagram conceptually showing an example of peak frequency diagnosis of severe AS. FIG. 7 is a diagram conceptually showing an example of detection of the severity of AS in peak frequency diagnosis. FIG. 8 is a diagram showing an example of the result of wavelet analysis of normal heart sounds. FIG. 9 is a diagram showing an example of the result of wavelet analysis of moderate heart sounds. FIG. 10 is a diagram showing an example of the result of wavelet analysis of severe heart sounds. FIG. 11 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 12 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 13 is a diagram for explaining the result of wavelet analysis of normal heart sounds. FIG. 14 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 15 is a diagram for explaining the result of wavelet analysis of normal heart sounds. FIG. 16 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 17 is a diagram for explaining the result of wavelet analysis of normal heart sounds. FIG. 18 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 19 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 20 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 21 is a diagram for explaining the result of wavelet analysis of normal heart sounds. FIG. 22 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 23 is a diagram for explaining the result of wavelet analysis of severe heart sounds. FIG. 24 is a FIG. 10 is a diagram showing the correlation between peak frequency and maximum aortic valve blood flow velocity Vmax.

[0025] [Principle of the Present Invention] First, the principle of the present invention will be explained.

[0026] Fig. 3 shows an example of heart sounds obtained from a heart sound sensor, showing the heart sounds S1-H of a patient with severe aortic stenosis (AS) and S1-N of a healthy person, along with an electrocardiogram. Fig. 4 is a diagram showing an example of the frequency components of heart sounds obtained by wavelet analysis, with the results of the wavelet analysis being shown schematically by the contour line WD of the scalogram.

[0027] Aortic stenosis (AS) is a disease in which calcification of the aortic valve causes the valve to become stiff and difficult to open. As a result, blood takes longer than usual to pass through the narrow valve opening. The turbulence that occurs during this process generates high-pitched noise. As the blood passes through the narrowed lumen, its flow velocity increases, generating high-pitched white noise. In other words, the more severe the condition, the higher the noise. In echocardiography, the increase in blood flow velocity at the stenotic site is used as an indicator of severity.

[0028] In the present invention, heart sounds obtained by auscultation, rather than by echocardiography, are used as the sound source, and wavelet analysis is applied to the heart sound data obtained from the heart sounds to detect frequency increases (higher harmonics) in specific intervals to detect severe AS (severe aortic stenosis).To determine whether or not a patient has severe AS, peak frequencies in specific intervals are detected and evaluated.

[0029] As shown in Figure 3, in severe AS, abnormal heart sounds occur between the first and second sounds, and this abnormal heart sound is subjected to wavelet analysis. By wavelet analysis, a peak frequency fp appears somewhere during the ejection phase, as shown in Figure 4, for example. This is detected, and the detected peak frequency fp is evaluated.

[0030] Furthermore, focusing on the fact that blood ejection time increases as AS becomes more severe, the cardiac ejection time is obtained using wavelet analysis. The obtained ejection time is evaluated and used to detect severe AS. Furthermore, the frequency of the main component of the first heart sound obtained from the heart sounds is evaluated to detect the occurrence of an ejection click. The presence or absence of an ejection click is used to detect severe AS.

[0031] In this way, the present invention utilizes specific acoustic patterns contained in heart sounds to accurately detect severe AS.

[0032] To verify the accuracy of the detection method, 85 cases of severe AS and 100 cases without AS were evaluated using peak frequency and ejection time. As a result, the detection rate of severe AS was increased to 90%.

[0033] Since the peak frequency and ejection duration were found to be positively correlated with the severity of AS, even moderate AS can be detected by wavelet analysis of heart sounds. In other words, the present invention makes it possible to diagnose the severity of AS.

[0034] Until now, no technology has been able to classify the severity of AS using auscultation alone. Using the technology of this invention, even doctors with little experience listening to heart murmurs can easily diagnose the severity of AS. Furthermore, by using a digital stethoscope at home, patients can detect AS themselves. This will enable referral and treatment at specialized facilities, and is expected to reduce the number of patients who die suddenly from AS.

[0035] In specialized facilities, patients with moderate AS are sometimes required to undergo echocardiography about once every six months for follow-up, but the technology of the present invention makes it possible to detect the progression of AS using auscultation alone, which is expected to replace echocardiography, thereby reducing testing costs and medical expenses.

[0036] The present invention can also be applied to the detection of other valvular diseases such as mitral regurgitation and aortic regurgitation.

[0037] Hereinafter, an embodiment of the present invention will be described. [Flow of diagnosis with the assistance of a cardiac valve abnormality detection system] Fig. 1 is a diagram showing the flow from diagnosis to treatment of aortic stenosis.

[0038] In Figure 1, for example, a patient with chronic heart disease who is recovering from hospital stays in the vicinity of his or her home or community (P1), and visits his or her regular clinic for a medical examination when symptoms appear (P2). The clinic is equipped with a cardiac valve abnormality detection device, and a cardiac valve abnormality diagnosis system 1 has been constructed together with its peripheral devices. The cardiac valve abnormality diagnosis detection system 1 performs peak frequency diagnosis using wavelet analysis to detect cardiac valve abnormalities, particularly severe aortic valve stenosis, with high accuracy, and outputs and displays the detection results. A doctor examines the patient using a stethoscope, including auscultation, and makes a final diagnosis of heart disease with the support of the cardiac valve abnormality diagnosis system 1 (P3).

[0039] If a doctor diagnoses the patient as highly suspected of having severe AS, the patient may go to a hospital, for example, and undergo an echocardiogram to make a definitive diagnosis (P4). If the diagnosis of severe AS is confirmed, the patient may undergo surgery such as artificial valve replacement, valvuloplasty, or catheterization for treatment (P5).

[0040] In the cardiac valve abnormality diagnosis system 1, the cardiac valve abnormality detection device 5 is connected to hospitals, specialized facilities, etc. via a network to exchange various types of information. For example, the results of detection by the cardiac valve abnormality detection device 5 are sent to and displayed in the echocardiography room and operating room of the hospital, and the results of the echocardiography and surgery are received and used for learning in the cardiac valve abnormality detection device 5 and for setting various parameters.

[0041] In this way, by constructing a cardiac valve abnormality detection system 1 centered around a cardiac valve abnormality detection device installed in a clinic or the like, severe AS can be easily detected by auscultation, and subsequent life-saving steps such as a definitive diagnosis and treatment can be reliably followed, thereby significantly improving the survival rate of AS patients. [Configuration of the cardiac valve abnormality detection system] Figure 2 shows an example of the configuration of the cardiac valve abnormality diagnosis system 1 according to an embodiment of the present invention.

[0042] In FIG. 2, the cardiac valve abnormality diagnostic system 1 includes a cardiac valve abnormality detection device 5, a heart sound sensor 6, an electrocardiograph 7, etc., and is connected to an echocardiograph 8, etc. via a network NW.

[0043] The heart sound sensor 6 is an acoustic-electrical transducer such as a microphone, piezoelectric sensor, or acceleration sensor, and outputs a heart sound signal S1 corresponding to the patient's heart sounds. The heart sound signal S1 includes signals corresponding to the first heart sound, second heart sound, heart murmur, etc. The heart sound sensor 6 may be incorporated into a stethoscope. The heart sound sensor 6 is placed, for example, in the second intercostal space on the right border of the sternum, in the Erb area, or under the nipple to acquire heart sounds.

[0044] The electrocardiograph 7 acquires an electrocardiogram of the patient and outputs an electrical signal (electrocardiogram signal) S2 containing signals of P, Q, R, S, T, U, and P waves.

[0045] The echocardiogram 18 is used to obtain images and cross-sectional images of the heart and blood vessels using ultrasound waves and to examine abnormalities in the structure and function of each part, and is installed in hospitals and specialized facilities.

[0046] The cardiac valve abnormality detection device 5 includes a heart sound data acquisition unit 21, a first processing unit 22, a second processing unit 23, an ejection duration acquisition unit 24, a detection unit 25, an output / display unit 26, and a communication unit 27.

[0047] Each unit or part of the cardiac valve abnormality detection device 5 can be configured by an information processing device such as a server or a personal computer. Each function of the cardiac valve abnormality detection device 5 described below can be realized by a CPU executing a computer program, or by cooperation between the CPU and hardware elements or devices. The computer program can be read from a recording medium or downloaded from a server and executed. The output / display unit 26 and the communication unit 2 can be realized by a display device or communication function provided in the information processing device. [Heart Sound Data Acquisition Unit] Figure 5 is a diagram showing an example of the flow of heart sound signal processing up to wavelet analysis.

[0048] The heart sound data acquisition unit 21 acquires and outputs heart sound data D1 corresponding to heart sounds, and includes a normalization unit 211, a noise removal unit 212, and an amplification unit 213. The input heart sound signal S1 is an analog signal or a digital signal. If it is an analog signal, it is converted into a digital signal by an appropriate AD converter, and the subsequent processing is performed.

[0049] The normalization unit 211 normalizes the heart sounds to accommodate differences in loudness of the heart sounds.

[0050] The noise removal unit 212 removes environmental noise, including ambient sounds unrelated to heart sounds, electrical noise, etc. Furthermore, breathing sounds may be mixed in with heart sounds as noise, and it is even better to remove these breathing sounds.

[0051] The amplifier 213 amplifies the heart sound signal S1, for example, by 16 times, so that it reaches a predetermined intensity level (amplitude level). The degree of amplification is adjustable. As a result, the heart sound data D1 output from the heart sound data acquisition unit 21 exhibits a waveform corresponding to the heart sound, and the amplitudes of various frequency components contained in the waveform are normalized. [First Processing Unit] Based on the heart sound data D1, the first processing unit 22 acquires a peak frequency fp, which is the peak value of the frequency of heart sounds having an intensity (amplitude) equal to or greater than a reference intensity within the target range, between the first and second heart sounds.

[0052] The first processing unit 22 includes a wavelet analysis unit 221, a reference intensity storage unit 222, a target range storage unit 223, a peak frequency storage unit 224, and the like.

[0053] The wavelet analysis unit 221 performs a continuous wavelet transform on the input heart sound data D1 to localize both the time and frequency components of the heart sound data D1. The continuous wavelet transform obtains time-frequency features of the heart sound data D1, i.e., the frequency on the time axis and the intensity (amplitude) of each frequency component. For example, analysis data is obtained that shows the time t, frequency f, and intensity (amplitude) dB of each frequency component of the heart sound data D1 on the x-axis, y-axis, and z-axis, respectively. This analysis data can be displayed on a two-dimensional plane using a scalogram, as shown in Figures 8 to 12 (described later). In a scalogram, the horizontal axis represents time and the vertical axis represents frequency. The brightness or color of each point indicates the intensity (amplitude) at a certain frequency at a certain point in time.

[0054] The various parameters and coefficients obtained by the wavelet analysis unit 221 can be used as feature quantities for machine learning using artificial intelligence (AI).

[0055] The reference strength storage unit 222 stores a reference strength ad for determining the peak frequency fp. The reference strength ad is a condition that, when acquiring the peak frequency fp, the strength of the frequency component must be equal to or greater than the reference strength ad. The reference strength ad is used to remove the effects of noise components, harmonic components, and the like remaining in the heart sound data D1 to obtain the correct frequency component. The reference strength ad can be determined through experiments so that the correct frequency component can be empirically obtained. If there are no noise components, harmonic components, and the like, or they are negligible, the reference strength ad can be set to a value close to 0, for example, or the reference strength ad can be excluded from the conditions.

[0056] Furthermore, if there is a signal stronger than the peak frequency obtained by wavelet analysis near a frequency half the peak frequency, the stronger signal is the fundamental wave, so this frequency is taken as the peak frequency fp and harmonics are excluded. In addition, it is possible to empirically exclude any suspicious peak frequencies fp.

[0057] In short, it is important to obtain a significant value for a frequency component whose intensity is not 0 as the peak frequency fp.

[0058] The target range storage unit 223 stores a temporal target range tr for determining the peak frequency fp. The target range tr may be set as a time range during which the aortic valve is open and blood flows therethrough. However, since the target range tr is for determining the peak frequency fp within this time range, it does not necessarily have to coincide with the time range during which blood actually flows, as long as a significant peak frequency fp is extracted.

[0059] As an example of how to set the target range tr, for example, the range from the start of the first sound to the end of the second sound, which corresponds to ventricular systole, is set as the target range tr. Furthermore, taking into consideration that the first sound is loud when the mitral valve opens and that it takes time for blood flow to actually start up, the target range tr is set to the range from time t1 after the start of the first sound. Furthermore, taking into consideration that the ejection duration may be extended and the time period in which the peak frequency fp may actually be detected, the target range tr is set to the range from time t1 after the start of the first sound but not beyond time t2. Furthermore, when the second sound is not clearly discernible, the target range tr is set to the range from time t1 after the start of the first sound but not beyond time t2. In other words, the target range tr is limited in terms of time. Furthermore, using the electrocardiogram signal S2, the target range tr is set to the range from the position of the S wave to the start or end of the second sound.

[0060] Specifically, for example, time t1 may be 100 ms and time t2 may be 500 ms, and the target range tr may be any range from 100 ms after the start of the first heart sound but not exceeding 500 ms. Other times may also be used. This setting method allows the target range tr to be set based solely on the heart sound data D1, without using the electrocardiogram signal S2. Furthermore, the target range tr can be specified even if the second heart sound is unclear.

[0061] The peak frequency storage unit 224 stores the acquired peak frequency fp. When acquiring the peak frequency fp, it is possible to acquire only one peak frequency fp, but there is a possibility that an inappropriate peak frequency will be extracted due to a large error. In other words, heart sounds vary for each pulse wave (aortic wave), and the peak frequency fp may also change. Furthermore, if there is arrhythmia or the pulse is extended, the position of the peak will shift. Therefore, it is preferable to acquire multiple peak frequencies and use their average value, etc., in order to increase the reliability of the peak frequency fp.

[0062] Therefore, to determine a representative value for the peak frequency fp, for example, the average value of multiple peak frequencies acquired from the target range tr for each of multiple pulse waves is determined as the representative value for the peak frequency fp. Alternatively, for example, five peak frequencies are acquired from the target range tr for each of five consecutive pulse waves, and the average value of these five peak frequencies is determined as the representative value for the peak frequency fp. Alternatively, the top n peak frequencies, for example, five, are selected from the multiple peak frequencies acquired from the target range tr, and the representative value for the peak frequency fp is determined by averaging the five peak frequencies. Alternatively, the maximum, minimum, and other singular values ​​are excluded from the maximum ten peak frequencies acquired from the target range tr for each of ten consecutive pulse waves, and the average value of the remaining five peak frequencies is determined as the representative value for the peak frequency fp. Instead of using the average value, the root mean square or median may be used, or an appropriate selection algorithm may be used to determine the representative value. Furthermore, if respiratory sounds are mixed into the heart sound data D1, the representative value may be determined using peak frequencies acquired only from the target range tr that do not overlap with the time of the respiratory sounds.

[0063] The first processing unit 22 outputs the peak data D2 including the peak frequency fp thus obtained to the second processing unit 23.

[0064] The ejection duration acquisition unit 24 acquires the ejection duration tk based on the heart sound data D1. The acquired ejection duration tk is stored in the ejection duration storage unit 241 and is output as ejection data D3.

[0065] The ejection time tk is the time during which blood flows out of the aortic valve, and corresponds to the ejection period of ventricular contraction. The ejection period is usually the period during ventricular contraction from the first heart sound to the second heart sound after the intraventricular pressure becomes higher than the intraaortic pressure and the aortic valve opens. Therefore, the ejection time tk can be obtained by detecting the timing of the aortic valve opening and the timing of the second heart sound based on the heart sound data D1.

[0066] The ejection time tk is about 200 ms in healthy individuals, but increases as the condition worsens. For example, in severe AS, it increases to about 300 ms to 500 ms. There is little individual variation.

[0067] Note that the ejection duration tk is for comparison with the ejection reference value ce, and therefore does not necessarily have to coincide with the actual ejection duration of the patient, and in short, it may be a value that corresponds to the ejection duration of the patient. Therefore, for example, the time from the first heart sound to the second heart sound may be acquired as the ejection duration tk, and an appropriate value for the ejection reference value ce for comparison may be set accordingly.

[0068] Based on the heart sound data D1, the detector 25 detects the main component of the first sound, that is, the frequency of a sound with a high intensity near the first sound. The detected main frequency fm is stored in the main frequency memory 251.

[0069] In addition, the I sound reference value cs is stored in the I sound reference value storage unit 252. When the main frequency fm is equal to or greater than the I sound reference value cs, click data D4 indicating the occurrence of an ejection click is output to the second processing unit 23.

[0070] The first heart sound is usually the sound made when the mitral valve closes, and the aortic valve opens immediately after it. However, if the aortic valve becomes stiff and less mobile, it does not open easily. Therefore, the aortic valve opens due to the impact of blood being pushed out by the contraction of the ventricles hitting the aortic valve. At this time, an ejection click, which is a higher-pitched sound than the normal first heart sound, occurs. In other words, the ejection click indicates that the aortic valve has become stiff and its function is deteriorating.

[0071] Examination of the main frequency components of the ejection click revealed a range of 50 to 60 Hz. Because the ejection click occurs immediately after the first sound, it sounds like a higher-pitched first sound when heard with a stethoscope, and is detected as being integrated with the first sound in the heart sound data D1. In normal heart sounds of healthy individuals, the first sound is composed of frequency components between 30 Hz and 40 Hz. Furthermore, in patients with severe AS, the early first sound is composed of frequency components of 40 Hz or higher.

[0072] Therefore, in order to detect an ejection click, the first sound reference value cs is set to, for example, 50 Hz, 55 Hz, or 60 Hz. When the first sound reference value cs is set to a lower value of 50 Hz, the detection sensitivity is increased, increasing the possibility of falsely detecting a sound that is not an ejection click, but the possibility of failing to detect an ejection click is reduced. Therefore, by ANDing this with other conditions, the accuracy of detecting severe AS is improved. When the first sound reference value cs is set to a higher value of 60 Hz, the detection sensitivity is reduced, but if click data D4 is detected, there is a high possibility that the aortic valve is sclerotic and that the patient has severe AS. Therefore, the first sound reference value cs can be determined taking these factors into consideration. Furthermore, the weight of the click data D4 in detecting severe AS may be varied depending on the value of the first sound reference value cs.

[0073] In this way, by detecting the ejection click by focusing on the fact that the main frequency of the sound component is high, cardiac valve abnormalities, including severe AS, can be detected with higher accuracy. [Second Processing Unit and Peak Frequency Diagnosis] The second processing unit 23 detects cardiac valve abnormalities, particularly severe AS, based on the peak data D2, ejection data D3, and click data D4, under the combination of the following conditions 1 to 3, depending on the selection of the determination mode. (Condition 1) When the peak frequency fp is equal to or greater than the peak reference value cp. (Condition 2) When the ejection time tk is equal to or greater than the ejection reference value ce. (Condition 3) When the main frequency fm is equal to or greater than the first sound reference value cs, that is, when an ejection click is detected.

[0074] In this embodiment, in determination mode 1, when condition 1 is satisfied, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output. In determination mode 2, when conditions 1 and 2 are satisfied simultaneously, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output. In determination mode 3, when conditions 1, 2, and 3 are satisfied simultaneously, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output.

[0075] When the detection signal D5 is output, the output / display unit 26 displays on the screen that it has been determined that there is a high possibility of severe AS, and if necessary, issues a sound such as a warning, and also transmits the information to other devices via the communication unit 27 and the network NW. For example, a message saying "Severe AS is suspected" is displayed on the screen.

[0076] This will be explained in more detail below.

[0077] The second processing unit 23 has the peak reference value cp and the ejection reference value ce stored in respective storage units 231, 232. In addition, although not shown, an operation unit, a setting unit, etc. for switching and setting the determination mode are provided.

[0078] In determination mode 1, when the peak frequency fp is equal to or greater than the peak reference value cp, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output. The peak reference value cp is set to, for example, 300 Hz. In other words, when the peak frequency fp is equal to or greater than 300 Hz, the detection signal D5 is output.

[0079] Based on past case experience, the peak frequency fp in severe AS is 250 Hz or higher, with an average value of approximately 320 Hz, with a variation of approximately ±50 Hz. Therefore, if the peak reference value cp is set to 300 Hz and severe AS is detected when the peak frequency fp is 300 Hz or higher, most severe AS cases will be detected. However, a peak frequency fp of 300 Hz or higher does not necessarily indicate severe AS, so in this regard, it is necessary to wait for combinations with other conditions or for medical examinations such as ultrasound examinations.

[0080] Thus, in determination mode 1, focusing on the fact that the peak frequency fp is high in cases of severe AS, the detected peak frequency fp is compared with the peak reference value cp. When the peak frequency fp is, for example, 300 Hz or higher, a detection signal D5 is output. This makes it possible to detect patients who are likely to have severe AS with a high degree of accuracy.

[0081] If the peak reference value cp is set to a value higher than 300 Hz, such as 320 Hz or 340 Hz, the possibility of not detecting severe AS increases, but the possibility of detecting severe AS when it is not. Therefore, it is advisable to set the peak reference value cp taking this into consideration, as well as taking into account combinations with other conditions that compensate for this.

[0082] In determination mode 2, when the peak frequency fp and the ejection time tk are equal to or greater than the ejection reference value ce, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output. The ejection reference value ce is set to, for example, 300 ms.

[0083] As mentioned above, patients with severe AS have a prolonged ejection time, empirically ranging from 300 to 460 ms. Therefore, by setting the ejection reference value ce to the lower value of 300 ms and adding the requirement that the ejection time tk be 300 ms or greater to the conditions for detecting severe AS, the likelihood that a detection signal D5 will be issued when the patient actually has severe AS increases, improving the reliability of detection.

[0084] In this way, in determination mode 2, when conditions 1 and 2 are satisfied, that is, when the peak frequency fp is, for example, 300 Hz or more and the ejection time tk is, for example, 300 ms or more, the detection signal D5 is output, thereby enabling patients who are likely to have severe AS to be detected with higher accuracy.

[0085] As mentioned above, there are errors and variations when acquiring the actual ejection duration tk, so an appropriate ejection reference value ce can be set depending on the errors and variations in acquiring the ejection duration tk and the method for acquiring the ejection duration tk.

[0086] In determination mode 3, in addition to the conditions of the peak frequency fp and the ejection time tk, when the main frequency fm is equal to or greater than the first sound reference value cs, that is, when an ejection click occurs, it is determined that there is a high possibility of severe AS, and a detection signal D5 indicating this is output.

[0087] As described above, ejection clicks indicate that the aortic valve is hardening and its function is deteriorating, and the occurrence of an ejection click suggests the possibility of severe AS. Therefore, by adding the occurrence of an ejection click as condition 3 to conditions 1 and 2, a more reliable detection signal D5 can be obtained, and patients who are likely to have severe AS can be detected with even greater accuracy.

[0088] The detection accuracy and sensitivity of ejection clicks vary depending on the value of the first sound reference value cs. Therefore, by setting the first sound reference value cs while taking into consideration its relationship with and compatibility with other conditions, it becomes possible to detect severe AS with extremely high accuracy.

[0089] In addition to the above-described determination modes 1 to 3, it is possible to set determination modes that combine various conditions 1 to 3. For example, it is possible to combine conditions 1 and 3, or conditions 2 and 3. Conditions other than these conditions 1 to 3 may also be combined.

[0090] Furthermore, the settings of the ejection reference value ce, the first sound reference value cs, the peak reference value cp, etc. under the other conditions may be changed depending on the values ​​of the peak frequency fp, the ejection duration tk, the main frequency fm, the determination results, etc. In this case, the order in which the conditions 1 to 3 are applied may be set in various ways, or one condition may be applied multiple times.

[0091] The second processing unit 23 also has a determination mode 4 for detecting the severity of AS. That is, in determination mode 4, the severity of AS is determined by comparing the detected peak frequency fp with a set frequency range cf. Here, an example will be described in which a detection signal D5 is output when the patient has severe AS and a detection assistance signal D6 is output when the patient has moderate AS.

[0092] The frequency range cf is set to a first frequency for determining severe AS and a second frequency for determining moderate AS. The first frequency may be substituted by the peak reference value cp. If the peak frequency fp is equal to or greater than the first frequency, the condition is determined to be severe AS. If the peak frequency fp is less than the first frequency and equal to or greater than the second frequency, the condition is determined to be moderate AS (Condition 4).

[0093] For example, the first frequency is set to 300 Hz and the second frequency is set to 180 Hz. In this case, the second processing unit 23 outputs a detection signal D5 indicating a high possibility of severe AS when the peak frequency fp is 300 Hz or higher, and outputs a detection assistance signal D6 indicating a high possibility of moderate AS when the peak frequency fp is 180 Hz or higher but lower than 300 Hz.

[0094] Based on past experience with cases, the peak frequency fp was in the range of 180 Hz to 300 Hz in cases of moderate AS. Furthermore, no peak waveform appeared in cases of mild AS. Therefore, if the acquired peak frequency fp was in the range of 180 Hz to 300 Hz, there was a high possibility that the patient had moderate AS.

[0095] Furthermore, in determination mode 4, it is possible to detect the severity of AS by determining whether there is a high possibility of developing severe AS in the future or whether an echocardiogram should be performed at the present time. This will be described next.

[0096] FIG. 17 is a diagram showing the peak frequencies fp detected and plotted for 189 cases (patients) including mild, moderate, and severe AS.

[0097] In Figure 17, the horizontal axis represents the peak frequency fp (Hz), and the vertical axis represents the maximum aortic valve blood flow velocity Vmax (m / sec). Here, the maximum aortic valve blood flow velocity Vmax (hereinafter referred to as "Vmax") was measured for each case using continuous wave Doppler. Vmax is an important factor used in echocardiography to evaluate the severity of AS, and is described in detail in the 2020 revised edition of the "Guidelines for the Treatment of Valvular Heart Disease" and the joint guidelines of the Japanese Circulation Society, the Japanese Association for Thoracic Surgery, the Japanese Society for Vascular Surgery, and the Japanese Society for Cardiovascular Surgery (JCS / JATS / JSVS / JSCS 2020 Guideline on the Management of Valvular Heart Disease).

[0098] According to "Table 30: Assessment of AS Severity by Echocardiography" on page 63 of Chapter 5 of the aforementioned "Guidelines for the Treatment of Valvular Heart Disease," "aortic valve sclerosis" indicates a Vmax of 2.5 (m / sec) or less, "mild AS" is 2.6-2.9, "moderate AS" is 3.0-3.9, and "severe AS" is 4.0 or more. Figure 17 shows 78 cases of "mild AS," 15 cases of "moderate AS," and 96 cases of "severe AS" based on these assessment criteria. Note that "mild AS" here includes not only the original "mild AS," but also "aortic valve sclerosis," which has milder symptoms, and a normal group.

[0099] 17, the correlation coefficient is 0.84, which indicates that there is a strong positive correlation between the peak frequency fp and Vmax. In other words, it indicates that Vmax tends to be larger in cases where the peak frequency fp is high.

[0100] In cases of "mild AS" (mild), where Vmax is less than 3.0 (m / sec), most of the peak frequencies fp are 150 Hz or less. Furthermore, using the regression line, the peak frequency fp corresponding to Vmax of 4.0 is approximately 300 Hz.

[0101] From these findings, it can be said that cases in which the peak frequency fp is 150 Hz or higher cannot be said to have "mild AS" and that an echocardiogram should be performed to determine the severity. Furthermore, it can be said that cases in which the peak frequency fp is 150 Hz or higher but less than 300 Hz are likely to develop severe AS in the future. Since there is a high possibility that surgery will be required if severe AS develops, these patients can also be said to be a target group for whom future surgery should be considered.

[0102] Therefore, when the peak frequency fp is equal to or greater than 150 Hz and less than 300 Hz, the second processing unit 23 outputs a detection assistance signal D6a indicating that there is a high possibility that the subject will develop severe aortic stenosis in the future.Furthermore, when the peak frequency fp is equal to or greater than 150 Hz, the second processing unit 23 outputs a detection assistance signal D6b indicating that an echocardiogram should be performed.

[0103] Although the detection assistance signals D6, D6a, and D6b contain overlapping content, by understanding and using the meaning of each, it becomes possible to respond quickly and accurately after the signal is displayed.

[0104] Furthermore, for the detection of these severity levels, the ejection time tk or the main frequency fm may be taken into account instead of only the peak frequency fp, as will be called next.

[0105] Furthermore, by statistically using these detection assistance signals D6, D6a, and D6b, it is possible to predict the approximate number of patients with severe AS or patients who will need surgery in the future, as well as the tendency for increases or decreases, which can be useful for hospital personnel and equipment planning, or national medical policies.

[0106] Note that the judgment in judgment mode 4 can be made by combining conditions 2 and 3 used in judgment modes 2 and 3. In this case, the parameters in conditions 2 and 3 may be adjusted so that continuous detection can be performed correctly according to the actual severity.

[0107] For example, when determination mode 2 is combined, severe AS is determined when conditions 1 and 2 are satisfied, and moderate AS is determined when conditions 4 and 2 are satisfied. When determining moderate AS, the value of the ejection reference value ce in condition 2 may be lowered compared to the case of severe AS. This is because the ejection time tk is considered to be shorter in the case of moderate AS than in the case of severe AS.

[0108] In this way, by combining conditions 1 to 4 and adjusting each parameter, the cardiac valve abnormality detection device 5 can detect the severity of AS.

[0109] Furthermore, the peak frequency fp, ejection time tk, and main frequency fm are associated not only with aortic stenosis but also with other valvular diseases such as aortic regurgitation and mitral regurgitation, as well as abnormalities in artificial valves. Therefore, they can be applied to the detection of abnormalities and deterioration of these various heart valves, including artificial valves.

[0110] Furthermore, by providing the artificial intelligence with a large amount of data such as the peak frequency fp, ejection time tk, main frequency fm, other test values, actual symptoms and progress, and allowing it to learn, it is possible to use the artificial intelligence to diagnose the possibility and severity of severe AS, and to detect various cardiac valve abnormalities or determine the degree of abnormality.

[0111] FIG. 6 conceptually shows an example of peak frequency diagnosis of severe AS, and FIG. 7 conceptually shows an example of detection of the severity of AS in peak frequency diagnosis.

[0112] In Figure 6, when condition 1 for peak frequency fp, condition 2 for ejection duration tk, and condition 3 for dominant frequency fm are simultaneously satisfied, a diagnosis of severe AS is made. This corresponds to the above-mentioned determination mode 3. In this case, signals indicating the states of a portion V1 that satisfies conditions 1 and 2 but not 3, a portion V2 that satisfies conditions 1 and 3 but not 2, and a portion V3 that satisfies condition 1 but not 2 or 3 may be included in the detection signal D5. These portions V1, V2, and V3 may be considered to be, for example, severe AS or a suspected severe AS condition.

[0113] In Figure 7, severe AS is determined when the peak frequency fp is, for example, 300 Hz or higher under condition 3, but various methods for determining moderate AS are conceivable. For example, when the peak frequency fp is in the range of 180 to 300 Hz, a portion M1 that simultaneously satisfies condition 2 for the ejection duration tk and condition 3 for the main frequency fm is determined to be moderate AS. For portions M2 to M7 that lack any of the conditions, a signal indicating that state can be included in the detection assistance signal D6. These portions M2 to M7 can be considered to be treated as moderate AS or a suspected case of moderate AS, for example.

[0114] 7 shows a case where the peak frequency fp is 180 Hz or higher, but by replacing "180 Hz" with "150 Hz," the method can be applied to detecting the severity of AS when the peak frequency fp is 150 Hz or higher. In this case, for example, it may be determined that there is a high possibility that the portions M1 and M5 to M7 will develop severe AS in the future, and that future surgery should be considered. It may also be determined that the portions are suspected of having moderate AS.

[0115] The example in Figure 7 corresponds to the above-mentioned determination mode 4, and the severity of AS can be indicated according to the portions M2 to M7. [Explanation of Scalograms by Wavelet Analysis] Next, wavelet analysis was performed on the heart sound data D1 for several cases, and the results are shown in Figures 8 to 15 as scalograms (amplitude scalograms), which will be explained.

[0116] FIG. 8 is a diagram showing an example of the results of wavelet analysis of normal heart sounds, FIG. 9 is a diagram showing an example of the results of wavelet analysis of moderate heart sounds, FIG. 10 is a diagram showing an example of the results of wavelet analysis of severe heart sounds, FIG. 11 is a diagram showing another example of the results of wavelet analysis of severe heart sounds, FIG. 12 is a diagram showing an enlarged portion of the results of wavelet analysis of severe heart sounds, FIG. 13 is a diagram for explaining the results of wavelet analysis of severe heart sounds, FIG. 14 is a diagram showing an enlarged portion of the results of wavelet analysis of normal heart sounds, FIG. 15 is a diagram showing an enlarged portion of the first sound in the results of wavelet analysis of severe heart sounds, and FIG. 16 is a diagram showing an example of the correlation between severity and peak frequency in AS.

[0117] In these figures, the horizontal axis represents time (sec) on a linear scale, and the vertical axis represents frequency (kHz) on a logarithmic scale. While intensity (amplitude) is typically represented on a color scale that changes continuously from blue to green to yellow for improved visibility, here it is represented on a gray scale in which brightness changes continuously from black to white. For example, the intensity is normalized such that black, which has the lowest brightness, is 0, and white, which has the highest brightness, is 1. The interval between black and white is scaled as 0.1, 0.2, 0.3, etc., depending on the brightness. Therefore, in the scalograms shown in the figures, brightness can be said to represent intensity (amplitude).

[0118] For example, in Figure 8, which shows a scalogram of normal heart sounds of a healthy person, the first and second sounds of a pulse wave are recorded approximately 20 times over a period of approximately 14 seconds along the horizontal axis. Areas with an intensity (amplitude) of 0 are shown in black, and as the intensity increases, the brightness increases and approaches white. The frequency of the first sound of a normal heart sound of a healthy person is between approximately 30 Hz and 40 Hz, and it appears as a white dot with an intensity of approximately 0.6 to 0.7 within this frequency range on the vertical axis.

[0119] In Figure 8, a cloud of points with an intensity of approximately 0.2 to 0.5 appears to rise like flames from near the first and second sounds upward. The peak frequency fp of normal heart sounds, i.e., the peak value of the frequency components of normal heart sounds, is in the range of approximately 90 to 160 Hz, and the highest point of these points does not exceed 160 Hz.

[0120] Note that five arrows are shown at the top of Figure 8, and five peaks of a point cloud with an intensity of approximately 0.3 can be seen below these arrows. These are respiratory sounds and are unrelated to heart sounds. Therefore, when acquiring the peak frequency fp, it is necessary to remove the respiratory sounds. Therefore, as described above, the noise removal unit 212 removes as much respiratory sound as possible, but if the noise removal unit 212 is unable to completely remove the respiratory sounds and some of the remaining respiratory sounds remain, the pulse wave that overlaps with the respiratory sounds is removed when acquiring the peak frequency fp. The period of the respiratory sounds is almost constant, and it is sufficient to identify the respiratory sounds based on their waveform.

[0121] In the scalogram of the moderate heart sound shown in FIG. 9, the white dot of the first sound has shifted slightly upward overall compared to the normal heart sound, and the peak frequency fp exceeds approximately 180 Hz.

[0122] In the scalograms for severe heart sounds shown in Figures 10 and 11, the white dot for the first sound has generally shifted further upward than in the case of moderate heart sounds, and the interval between the first sound and the second sound has widened. Following the first sound is a peak due to the frequency components of the cardiac murmur, with the highest point of the peak being the peak frequency fp, and both peak frequencies fp exceeding approximately 300 Hz.

[0123] 9 to 11, the target range tr is set between the first and second sounds. Peaks visible other than the peak frequency fp shown in the figures are either outside the target range tr or are due to respiratory sounds or the like, and are therefore excluded from the target range tr. In the example shown in FIG. 10, the top five peak frequencies were selected from the recorded heartbeats, and the representative value of the peak frequency fp was determined by averaging the five peak frequencies. In this case, the representative value of the peak frequency fp is 304.4 Hz, which is above the peak reference value cp of 300 Hz, and it is determined that there is a high possibility of severe AS.

[0124] Figures 12 and 13 show peaks due to the frequency components of four heart murmurs. The strongest point (whitest point) near the first sound indicates an ejection click. The ejection click appears immediately after the first sound or is integrated with the first sound, and its main component has a frequency of about 60 Hz. From there, a peak due to the frequency components of the heart murmur spreads out over a width of about 280 ms to 500 ms. This peak is caused by aortic valve stenosis restricting blood flow, which increases the blood flow velocity and generates a high-frequency heart murmur, with a peak frequency fp exceeding 300 Hz. At the same time, the ejection time is lengthened to about 280 ms to 500 ms, which delays the generation of the second sound.

[0125] Figure 14 shows normal heart sounds, the first and second sounds. It can be seen that the frequency of the first sound is between 30 Hz and 40 Hz. Note that while the heart rate in Figure 14 is 60 bpm, the frequency of the first sound is also between 30 Hz and 40 Hz in a scalogram for an example with a heart rate of 75 bpm.

[0126] FIG. 15 shows the first heart sound, which is a severe heart sound, and it can be seen that the frequency of the first heart sound exceeds 40 Hz and reaches almost 60 Hz.

[0127] As described above, the cardiac valve abnormality detection device 5 of this embodiment can easily and accurately detect cardiac valve abnormalities, particularly severe aortic valve stenosis, and can also easily and accurately detect the degree of cardiac valve abnormality.

[0128] In particular, since results can be obtained solely based on the heart sound data D1 obtained by auscultation, this method can be easily implemented without increasing the burden on doctors and patients in clinical settings. Severe AS can be diagnosed by comparing the peak frequency fp obtained based on the heart sound data D1 with the peak reference value cp (peak frequency diagnosis), so the diagnosis result and the reason for the diagnosis are clear. Since the peak frequency fp can be easily obtained using wavelet analysis, processing is simple, the processing time is short, and it can be implemented at a relatively low cost. Further improvement in detection accuracy can be achieved by adjusting the peak reference value cp, the ejection reference value ce, etc. based on feedback from echocardiography results, etc.

[0129] Incidentally, it has been found that the severity of AS is positively correlated with the peak frequency fp, as shown by the hypothetical curve CV1 in Figure 16. Based on this finding, it is expected that the severity of AS can be detected with even higher accuracy.

[0130] In the above-described embodiment, the configuration of the cardiac valve abnormality detection device 5 can be modified in various ways. For example, the second processing unit 23 may output not only the detection result such as "moderate AS" or "severe AS" but also the severity according to the peak frequency fp, ejection time tk, or main frequency fm, i.e., the progression of AS. Alternatively, the peak frequency fp, ejection time tk, main frequency fm, etc. may be output as they are.

[0131] In the above-described embodiment, the cardiac valve abnormality detection device 5 can be implemented as a miniaturized health device for home use or as a portable device. A computer program that causes the cardiac valve abnormality detection device 5 to function may be executed by a personal computer or a smartphone, and the personal computer or smartphone may be implemented as the cardiac valve abnormality detection device.

[0132] To implement the present invention as a cardiac valve abnormality detection method, the functions of each part of the cardiac valve abnormality detection device 5 or cardiac valve abnormality detection system 1, including the first processing unit 22, the second processing unit 23, and the ejection duration acquisition unit 24, etc., described above, may be executed.

[0133] In addition, the configuration, structure, shape, function, setting values, processing content, processing order, output format of the whole or each part of the above-mentioned heart sound data acquisition unit 21, first processing unit 22, second processing unit 23, ejection duration acquisition unit 24, detection unit 25, output / display unit 26, heart valve abnormality detection device 5, heart valve abnormality detection system 1, etc., as well as the content, signal form, output timing, etc. of the detection signal D5, detection auxiliary signal D6, etc. can be changed in various ways other than those described above.

[0134] REFERENCE SIGNS LIST 1 Heart valve abnormality detection system 5 Heart valve abnormality detection device 6 Heart sound sensor 7 Electrocardiograph 21 Heart sound data acquisition unit 22 First processing unit 23 Second processing unit 24 Diaphragmatic time acquisition unit 25 Detection unit 26 Output / display unit 211 Normalization unit 212 Noise removal unit 213 Amplification unit 213 D1 Heart sound data D5 Detection signal D6 Detection auxiliary signal fp Peak frequency cp Peak reference value tr Target range ad Reference intensity fm Main frequency cs First sound reference value tk Diaphragmatic time ce Ejection reference value cf Frequency range

Claims

1. a heart sound data acquisition unit for acquiring heart sound data corresponding to a heart sound; a first processing unit that acquires a peak frequency, which is a peak value of a frequency component of the heart sound within a target range between the first sound and the second sound of the heart sound, based on the heart sound data; an ejection duration acquisition unit that acquires an ejection duration based on the heart sound data; a second processing unit that is provided with a storage unit that stores a first frequency, which is a peak reference value for determining severe AS, a second frequency that is lower than the first frequency, and an ejection reference value, and outputs a detection signal indicating severe AS when the peak frequency is equal to or higher than the first frequency and the ejection time is equal to or higher than the ejection reference value, and outputs a detection auxiliary signal indicating moderate AS when the peak frequency is lower than the first frequency and equal to or higher than the second frequency and the ejection time is equal to or higher than the ejection reference value; 1. A heart valve abnormality detection device comprising:

2. the first frequency is 300 Hz; 2. The heart valve abnormality detection device according to claim 1.

3. the second frequency is 180 Hz; 3. The heart valve abnormality detection device according to claim 2.

4. The second processing unit outputs a detection assistance signal indicating a high possibility of developing severe aortic valve stenosis in the future when the peak frequency is equal to or higher than 150 Hz and lower than 300 Hz.

3. The heart valve abnormality detection device according to claim 2.

5. The second processing unit outputs a detection auxiliary signal indicating that an echocardiogram should be performed when the peak frequency is 150 Hz or more.

3. The heart valve abnormality detection device according to claim 1 or 2.

6. The target range is a range from 100 ms to 500 ms after the start of the I sound.

6. The heart valve abnormality detection device according to claim 1.

7. A representative value determined based on a plurality of peak frequencies obtained from the target range for each of a plurality of pulse waves is used as the first frequency.

7. The heart valve abnormality detection device according to claim 1.

8. The ejection reference value is 280 ms.

8. The heart valve abnormality detection device according to claim 1.

9. a detector for detecting an ejection click occurring near the first sound based on the heart sound data; The second processing unit outputs the detection signal when a condition that the ejection click is detected is satisfied in addition to the condition that the peak frequency is equal to or higher than the first frequency and the ejection time is equal to or higher than the ejection reference value, and outputs the detection assistance signal when a condition that the ejection click is detected is satisfied in addition to the condition that the peak frequency is lower than the first frequency and equal to or higher than the second frequency and the ejection time is equal to or higher than the ejection reference value.

9. The heart valve abnormality detection device according to claim 1.

10. the detection unit detects a main frequency that is a frequency of a main component of the first sound, and detects the ejection click when the main frequency is equal to or higher than an first sound reference value.

10. The heart valve abnormality detection device according to claim 9.

11. The I sound reference value is 50 Hz.

11. The heart valve abnormality detection device according to claim 10.

12. The first processing unit targets a frequency component of the heart sound having an intensity equal to or greater than a reference intensity within the target range when acquiring the peak frequency.

12. The heart valve abnormality detection device according to claim 1.

13. The first processing unit obtains the peak frequency by performing a wavelet analysis on the heart sound data.

13. The heart valve abnormality detection device according to claim 1.

14. The heart sound data acquisition unit includes: a normalization processor for normalizing the heart sounds; a noise removal unit that removes environmental noise; an amplifier for amplifying the signal to a predetermined intensity level; 14. The heart valve abnormality detection device according to claim 1, further comprising:

15. A computer program that causes a computer to function as the heart valve abnormality detection device according to any one of claims 1 to 14.

16. A computer readable medium having recorded thereon the computer program of claim 15.

17. 1. A method for detecting a heart valve abnormality by processing heart sound data corresponding to a heart sound, comprising: a first process for acquiring a peak frequency, which is a peak value of a frequency component of the heart sound having an intensity equal to or greater than a reference intensity within a target range between the first sound and the second sound of the heart sound based on the heart sound data; An ejection duration acquisition process for acquiring an ejection duration based on the heart sound data; a second process that uses a first frequency, which is a peak reference value for determining severe AS, a second frequency lower than the first frequency, and an ejection reference value, which are stored in a storage unit, to output a detection signal indicating severe AS when the peak frequency is equal to or higher than the first frequency and the ejection time is equal to or higher than the ejection reference value, and outputs a detection assistance signal indicating moderate AS when the peak frequency is lower than the first frequency and equal to or higher than the second frequency, and the ejection time is equal to or higher than the ejection reference value; 4. A method for detecting a heart valve abnormality, comprising: