KR20190058858A - Method for providing diagnostic information on cardiovascular diseases using a smart device and heart sound application for the same - Google Patents

Abstract

Provided are a method for providing diagnostic information, to enable a normal person to use a smart device to conveniently and autonomously diagnose cardiovascular diseases and a heart sound application thereof. According to the present invention, the method for providing diagnostic information comprises the following steps: converting pieces of sample heart sound data received from the outside into image data visualizing the amplitude of a heart sound in accordance with changes in time and frequency; using a convolutional neural network (CNN) to train an image object corresponding to the heat sound among the pieces of image data of the sample heart sound data with any one of a plurality of predefined heart sound patterns; directly recording a heart sound from a patient who receives a diagnosis; converting recorded heart sound data into image data visualizing the amplitude of the heart sound in accordance with changes in time and frequency; using the CNN to classify an image object corresponding to the heat sound from the image data of the heart sound data of the patient into any one of the heart sound patterns.

Description

TECHNICAL FIELD [0001] The present invention relates to a method for providing diagnostic information of a cardiovascular disease using a smart device,

The present invention relates to a method for providing diagnostic information on cardiovascular diseases and a cardiac sound application for the same, and more particularly, to a method for providing diagnostic information on cardiovascular diseases using a smart device and a cardiac sound application therefor.

Cardiovascular disease is one of the most common causes of death, accounting for 31.5% of the world's deaths. In 2015, 92.1 million adults in the United States have cardiovascular disease, and by 2030, 43.9% of the adult population is expected to have cardiovascular disease. The stethoscope invented by Rene Laennec in 1816 has played an important role in the physical examination of heart disease patients. The opening and closing of the heart valve and the blood flow through the valve give rise to specific vibrations, which are detected through a stethoscope. Heart stethoscopy using a stethoscope is used as a diagnostic and diagnostic tool in cardiology for a long time because it can perform a hemodynamic evaluation according to heart valves or defects in the heart.

In recent years, high-tech devices such as ultrasonic devices have partially replaced the role of stethoscopes, but the usefulness of stethoscopes remains valid because of their ease of use and storage and accurate diagnosis. As such, the stethoscope is a very useful tool, but it is virtually impossible for the general public without expertise to use it for self-diagnosis.

On the other hand, smartphones are widely used in modern people. By 2015, 64% of Americans and 88% of Koreans are known to own smartphones. With the spread of smartphones, a variety of healthcare applications have been developed to date, including heart rate or calorie consumption measurement applications. However, there are still no self-diagnostic applications that require specialized medical knowledge. The biggest reason is that it is difficult to diagnose accurately with smartphone applications without the doctor's opinion.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a diagnostic information providing method that allows a general person to conveniently diagnose a cardiovascular disease by using a smart device.

Another problem to be solved by the present invention is to provide a cardiac sound application for this purpose.

The problems to be solved by the present invention are not limited to the above-mentioned problems, and other matters not mentioned can be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided a method for providing diagnostic information on cardiovascular disease using a smart device equipped with cardiac sound application software, And converting image data corresponding to heart sound into image data of a plurality of predefined images using the convolutional neural network (CNN). The image processing apparatus according to claim 1, Cardiac sound data is recorded directly from a patient to be diagnosed, and the recorded patient heart sound data is converted into image data in which the cardiac amplitude according to the change of time and frequency is visualized, and the image data of the patient heart sound data The image object corresponding to the heart sound is convoluted Using a neural network design and a step of classification by any one of the heart rate pattern.

The method of claim 1, further comprising the steps of: measuring the sound pressure of the patient's heart sound data for a predetermined time window from the patient's heart sound data, and defining a time window corresponding to a lower predetermined ratio of the sound pressure as thermal noise, have.

Wherein recording of the patient's heart sounds comprises displaying a measurement screen on the screen of the smart device (a plurality of heart sound measurement sites are displayed on the measurement screen), and the input means of the smart device displays the heart sound measurement site The measurement screen is switched to the recording screen. When the smart device records cardiac sound close to the body of the patient corresponding to the cardiac sound measurement site, it is determined whether or not the recorded heart sound can be interpreted And generating an alarm event based on the determination.

The predefined plurality of cardiac tone patterns may include a normal heart sound, a third heart sound S3, a fourth heart sound S4, a systolic murmur, a diastolic murmur, or a combination thereof .

According to another aspect of the present invention, there is provided a cardiac sound application, which is combined with a smart device that provides diagnostic information on a cardiovascular disease, And an image object corresponding to the heart sound is learned by using a convolutional neural network (CNN) in one of a plurality of pre-defined cardiac sound patterns in the image data of the sampled cardiac sound data, The cardiac sound data is directly recorded from the patient, the recorded patient heart sound data is converted into image data in which the heart sound amplitude is visualized according to the change of the time and frequency, and the image object corresponding to the heart sound in the image data of the patient heart sound data is converted into convolution neural network The cardiac sound patterns are classified into one And stored in the medium to execute the system.

Image data corresponding to one cycle of the heart sound among the sample heart sound data or the patient's heart sound data may be extracted and the image data of each cycle may be machine-learned using the convolution neural network.

The details of other embodiments are included in the detailed description and drawings.

As described above, according to the diagnostic information providing method of cardiovascular disease and the cardiac sound application according to the present invention, it is possible to diagnose cardiovascular disease with high accuracy even if only the built-in microphone of the smart device is used without using an external microphone. In the present invention, in order to obtain a reliable diagnostic result from a cardiac sound recorded with a built-in microphone, first, a thermal noise generated in the smart device itself is removed using a spectral subtraction and a fast Fourier transform algorithm Second, we classify cardiac patterns by machine learning based on Convolutional Neural Network (CNN) for classical heart sound data. Third, when using convolution neural network, heart sound data (ie, audio data) And 2 - dimensional image data with visualization of the heartbeat amplitude according to the frequency change.

Furthermore, even a general person can easily perform a self-diagnosis of cardiovascular disease even if he or she does not have any special medical knowledge by using his / her smart device, for example, a smart phone. In order to enhance user convenience, the present invention provides a special user interface, which provides a measurement screen that clearly displays the heart sound measurement site and a recording screen that displays the recording state of the heart sound. It also determines whether the recorded heart sound is interpretable and provides a notification event separately.

As described above, the cardiac sound application of the present invention can be used not only for the self-diagnosis of the general public but also as a tool for selecting patients at the hospital admission stage.

1 is a diagram illustrating an embodiment of use of a smart device according to an embodiment of the present invention.
Fig. 2 is a diagram schematically showing a configuration of the smart device of Fig. 1;
FIG. 3 is a flowchart sequentially illustrating a diagnostic information providing method of a cardiovascular disease according to an embodiment of the present invention.
4 is a diagram illustrating a user interface of a cardiac sound application according to an exemplary embodiment of the present invention.
5A to 5C show the measured patient heart sound data and the converted image data.
FIG. 6 is a diagram showing diagnostic performance using a smart device in the present experimental example.

BRIEF DESCRIPTION OF THE DRAWINGS The advantages and features of the present invention and the manner of achieving them will become apparent with reference to the embodiments described in detail below with reference to the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Is provided to fully convey the scope of the invention to those skilled in the art, and the invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

Hereinafter, a smart device according to an embodiment of the present invention will be described in detail with reference to FIG. 1 and FIG. 1 is a diagram illustrating an embodiment of a smart device according to an embodiment of the present invention, and FIG. 2 is a schematic view of a configuration of the smart device of FIG.

The smart device 20 of the present invention is an electronic device that records heart sound directly from a patient or human body 10 and provides diagnosis information on cardiovascular diseases. The smart device 20 may be an electronic device capable of performing a communication function through wireless communication while the user moves and may be a mobile phone including a smart phone, a tablet computer, a PDA, a wearable device, a smart watch, And the like.

The smart device 20 includes a processor 220, input means 210, output means 230 and memory 240. The input unit 210 includes a software or a hardware input unit. For example, the hardware input unit includes a button installed outside the smart device 20, a built-in microphone for receiving audio information, a communication device for exchanging data through a wired / . ≪ / RTI > The output means 230 includes a speaker, a display, and the like. The display may include a user interface as a means for detecting a user's touch input in UI / UX of operating system software and UI / UX of application software. The display may be a means for outputting a screen and a touch screen for simultaneously executing a function of an input means for sensing a touch event of a user.

The memory 240 of the smart device 20 typically provides a location for storing computer code and data used in the device. The memory 240 of the smart device 20 in accordance with an embodiment of the present invention includes basic input and output systems, operating systems, various programs, applications, or devices Firmware for any device, including user interface functions, processor functions, and the like, that are executed in the device.

The cardiac sound application 250 installed in the smart device 20 provides diagnostic information of cardiovascular disease from cardiac sound recorded from the patient or the human body 10 using a Convolutional Neural Network (CNN) model. The cardiac sound application 250 may be application software downloaded by a user using an external platform to his or her smart device 20. [ In addition, the cardiac sound application 250 of the present invention may be software installed in the smart device 20 by default in conjunction with the operating system software of the smart device 20. [ The cardiac sound application 250 will be described in detail later.

The processor 220 of the smart device 20 executes the computer code with the operating system and executes the operations of generating and using the data. The processor 220 may also use a series of instructions to receive and process input and output data between the components of the smart device 20. The processor 220 also serves as a control unit for executing functions of operating system software and various application software installed in the smart device 20. [

Additional or conventional components such as a power supply of the smart device 20, a hardware module such as a communication modem, a GPS, an I / O device, and a camera module are not shown in the drawing, Lt; RTI ID = 0.0 > and / or < / RTI > The smart device 20 may also include hardware components (including circuitry), software components (including computer code stored on a computer readable medium), or a combination of the two.

The cardiac sound application 250 of the present invention functionally includes a heart sound-image converting unit 252, a learning unit 254, a cardiac sound classifying unit 256 and a noise removing unit 258.

When the input means 210 directly records heart sounds in proximity to the human body 10 of a patient who is to diagnose a cardiovascular disease, the heart sound-image converting unit 252 converts the heart sound data (i.e., audio data) And the amplitude of the heart sound according to the change of the frequency is converted into image data that is visualized in two dimensions. Also, the heart sound-image converting unit 252 converts a plurality of sampled heart sound data (i.e., audio data) supplied from the outside into image data visualized in two dimensions. The sample heart sound data can be obtained from a public database (eGeneralmedical, UCLA Auscultation assistant, Welchallyn, Michigan Medical School, Medicine and Teaching Hospital Motol Department of Applied Informatics, 3M stethoscope, Teachingheartauscultation, etc.). The heart sound-image converting unit 252 uses two-dimensional image data in which the time, frequency, and heart sound amplitude are all visualized in order to accurately record cardiac sound having a low frequency of 1 to 200 Hz and analyze its characteristics. Specifically, changes in heart sound amplitude are classified into color differences based on the time axis and the frequency axis, and are displayed as image data.

The learning unit 254 uses a plurality of sample cardiac sound data provided from outside as input data for the learning model of the heart sound pattern. Specifically, when the sampled heart sound data is input through the input unit 210, the heart sound-image converting unit 252 visualizes the sampled heart sound data (i.e., audio data) and converts the sampled heart sound data into image data, and the learning unit 254 The image data of the sample heart sound data is used as the input data of the convolution neural network model. The learning unit 254 uses the convolution neural network model to learn an image object corresponding to the heart sound among the image data using any one of a plurality of predefined heart sound patterns. The learning unit 254 may extract image data corresponding to one cycle of the heart sound among the sampled heart sound data, and then machine-learn image data of each cycle using the convolution neural network model. Here, the cardiac cycle or cardiac cycle refers to the period from when the heart starts to contract, to when the heart starts to systole again through the diastole.

The machine learning process by the convolution neural network model is a method in which a plurality of convolution layers and a pooling layer are repeatedly constructed to extract abstract features from two-dimensional image data, The process is repeated to get a generalized result as it goes to the upper layer. On the plurality of convolution and pulling layers, there is a fully-connected layer. The function for each layer includes a sigmoid function, a hyperbolic tangent (tanh) function, a ReLU (Rectified Linear Unit ) Function can be used.

In the present embodiment, heart sound patterns are defined in advance in a plurality of categories according to normal and abnormal states. For example, normal heart sounds, third heart sounds S3, fourth heart sounds S4, systolic murmurs, It can be classified into five patterns of diastolic murmur.

Normal heart sounds represent heart sounds without cardiovascular disease, and S1 (first heart sound) and S2 (second heart sound) are observed during one cardiac cycle. S1 is a long, dull low-pitched sound caused by the blood colliding with the valve wall during the occlusion of the tricuspid and mitral valve in the early phase of ventricular systole. S2 is a vibration sound generated by the blood colliding with the valve wall when the aortic valve and the pulmonary valve are closed.

In addition, S3 (third heart sound), S4 (fourth heart sound), systolic noise, and diastolic noise are all heart murmurs caused by cardiovascular disease. Specifically, S3 (third heart sound) is a heart murmur which occurs when blood is filled in the ventricle rapidly at a heart rate of 0.12-0.16 seconds after S2. It is anemic, hyperthyroidism, aortic insufficiency, mitral (or tricuspid) ), Cardiac septal defect, cor pulmonale, and the like. S4 (fourth heart sound) is a heart murmur which occurs when the ventricular enlargement or damage of the ventricular wall is filled with blood. It occurs as a cause of aortic stenosis, coronary artery disease, ischemic heart disease, arrhythmia, heart failure. Systolic noise is a heart murmur occurring between S1 and S2, which is caused by aortic stenosis, mitral regurgitation, ventricular septal defect, and pulmonary valve stenosis. Dilatant noise is a heart murmur that occurs in the ventricular diastole and is caused by aortic regurgitation, mitral (or tricuspid) stenosis, and aortic valve (or pulmonary valve) obstruction.

The heart sound classifying unit 256 classifies the patient's heart sound into one of the heart sound patterns based on the learning information of the learning unit 254 by the convolution neural network model. Specifically, when the heart sound-image converting unit 252 converts the recorded patient heart sound data into image data obtained by visualizing the heart sound amplitude according to the change of the time and the frequency in two dimensions, the heart sound classifying unit 256 classifies the convolution The neural network is used to classify the image object corresponding to the heart sound among the image data of the patient's heart sound data into one of the cardiac sound patterns to derive a result value.

The noise removing unit 258 removes the thermal noise contained in the patient cardiac sound data recorded through the input unit 210. Unwanted noise generated in cardiac stuttering and cardiac sound recording using the smart device 20 is roughly divided into two causes. One is environmental noise, and the other is the self noise of the smart device 20, i.e., thermal noise. Environmental noise refers to the ambient noise generated outside the smart device 20, such as a voice, music sound, traffic noise, etc., which is negligible noise in a quiet recording environment. However, thermal noise is one of the main causes of poor accuracy in heart sound analysis due to the noise generated by the heat transfer of charged electrons in the circuit of the smart device 20. In the present invention, thermal noise generated in the smart device itself is removed using spectral subtraction and fast Fourier transform (FFT) algorithm. Specifically, the sound pressure data is measured for every predetermined time window (for example, 250 ms) with respect to the recorded patient heart sound data, and then a predetermined lower rate (for example, lower 10% The corresponding time window is defined as thermal noise and removed. That is, the average spectrum of the thermal noise is calculated using the fast Fourier transform, the inverse FFT is performed on all time windows, and the noise is removed using the inverse fast Fourier transform (IFFT).

In the present embodiment, for convenience of explanation, the smart device 20 is described as an independent electronic device, but the present invention is not limited thereto. Specifically, the smart device may be configured as a combination of a mobile device and an external device (e.g., one or more server devices and / or databases) that are networked or physically separated from the mobile device. In this case, the individual configuration of the smart device may belong to a mobile device or an external device. For example, the input means, the processor, and the output means may belong to the mobile device and the memory where the cardiac sound application is installed may belong to the external device. Further, some of the cardiac sound applications are handled by the mobile device, You may. Thus, even if the smart device is composed of a mobile device and an external device physically separated from each other, it should be regarded as an electronic device functionally.

Hereinafter, a diagnostic information providing method of cardiovascular disease according to an embodiment of the present invention will be described in detail with reference to FIG. 3 and FIG. FIG. 3 is a flowchart sequentially illustrating a diagnostic information providing method of a cardiovascular disease according to an embodiment of the present invention. 4 is a diagram illustrating a user interface of a cardiac sound application according to an exemplary embodiment of the present invention.

First, a plurality of sample heart sound data provided from the outside are input to the smart device 20 (S10). The sample heart sound data is audio data and can be obtained from a public database.

The smart device 20 converts the sampled heart sound data (i.e., audio data) into image data that is two-dimensionally visualized (S20). In the present invention, two-dimensional image data in which time, frequency, and cardiac amplitude are both visualized is used in order to accurately record heart sound having a low frequency and analyze the characteristics thereof.

Next, the smart device 20 uses the image data of the sampled heart sound data as input data of the convolutional neural network model (S30). Specifically, the learning unit 254 uses the convolution neural network model to learn the image object corresponding to the heart sound among the image data using any one of a plurality of pre-defined heart sound patterns.

Next, the smart device 20 is brought close to the patient who is to be diagnosed as a cardiovascular disease to directly record the heart sound. The present invention provides a special user interface for easy self-diagnosis of cardiovascular diseases even for the general public. 4, when a user or a patient executes a cardiac sound application of the smart device 20, a measurement screen 30 is displayed on the screen of the smart device 20. On the measurement screen 30, a plurality of cardiac sound measurement sites 32 are displayed. In the present embodiment, five cardiac sound measurement sites 32 are illustratively displayed on the measurement screen 30. The measurement screen 30 is switched to the recording screen 34 when the input means 210 of the smart device 20 generates a selection event for any one of the cardiac sound measurement sites 32. [ The user records the heart sound by bringing the built-in microphone of the smart device 20 close to the body corresponding to the selected heart sound measuring part 32. [ The smart device 20 displays the recording situation on the recording screen 34. [ Upon completion of the recording, the smart device 20 determines whether or not the recorded heart sound can be analyzed and generates an alarm event on the screen. If the external noise is larger than the reference value or the smart device 20 is not accurately placed in the selected heart sound measuring part 32, if the recorded heart sound is smaller than the reference, heart sound analysis is impossible and an alarm event for requesting re-recording is generated , And generates an alarm event indicating that recording is completed when the heart sound analysis is possible.

When the interpretable cardiac sound is recorded, the smart device 20 converts the amplitude of the heart sound according to the change of time and the frequency of the recorded patient heart sound data (i.e., audio data) into image data that is two-dimensionally visualized S40).

Then, based on the learning information by the convolution neural network model, the smart device 20 classifies the image object corresponding to the heart sound among the image data of the patient heart sound data into one of the heart sound patterns (S50).

<Experimental Example>

In this experiment, a total of 46 subjects were enrolled. First, a cardiologist diagnosed each subject with a cardiovascular disease by a skilled cardiologist. Table 1 shows the characteristics of the subjects. Of the total subjects, 20 had systolic murmur, 5 had diastolic murmur, 1 had fourth heart sound (S4), and 20 had normal heart sounds.

As a result of measuring the subject's heart sound with the smart device of the present invention and removing the thermal noise from the patient's heart sound data, it was confirmed that 30 heart sounds (65.2%) out of 46 heart sounds were interpretable. In cases where heart sound analysis was impossible, it was mainly recognized as background noise and lack of cooperation of subjects.

5A to 5C show the measured patient heart sound data and the converted image data. 5A is a view showing recorded sampled heart sound data (steady state) and image data converted therefrom, FIG. 5B is a graph showing recorded sampled heart sound data (aortic stenosis) in the present experimental example, And FIG. 5C is a diagram showing recorded sample heart sound data (mitral regurgitation) and image data converted therefrom in the present experimental example. Referring to FIGS. 5A to 5C, image data visualized in two dimensions are displayed by dividing a change in heart sound amplitude with respect to a time axis and a frequency axis by color differences. Specifically, in the normal state as shown in FIG. 5A, image objects for S1 and S2 heart sounds were periodically observed in a constant pattern. As shown in FIGS. 5B and 5C, in the case of cardiovascular disease, the image object for the murmur was periodically observed in a constant pattern.

FIG. 6 is a diagram showing diagnostic performance using a smart device in the present experimental example. In this experiment, the diagnostic performance was measured by installing a heart sound application on Samsung Galaxy S5, Samsung Galaxy S6 and LG G3 smartphones. In FIG. 6, TP indicates True Positive, TN indicates True Negative, FP indicates false positive, and FN indicates false negative. Samsung Galaxy S5 is 90%, Samsung Galaxy S6 is 87%, and LG G3 is 90%. Table 2 is a table specifically showing the diagnostic performance of the present experimental example.

As described above, it has been verified that the smart device and the cardiac sound application of the present invention can conveniently record cardiac sounds and diagnose cardiovascular diseases with high accuracy by the convolution neural network model.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, You will understand. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive.

10: Human body
20: Smart Devices
30: Measurement screen
32: Heart sound measurement site
34: Recording Screen
210: input means
220: Processor
230: output means
240: Memory
250: Cardiac application
252: heart sound-image conversion unit
254:
256:
258: Noise elimination

Claims (8)

A method for providing diagnosis information of a cardiovascular disease using a smart device in which heart sound application software is installed,
A plurality of sampled heart sound data received from outside are converted into image data having visualized heartbeat amplitude according to a change in time and frequency, and an image object corresponding to heart sound is converted into convolutional neural network (CNN) in the image data of the sampled heart sound data A plurality of heart sound patterns defined in advance,
The cardiac sound data is directly recorded from the patient to be diagnosed, and the recorded patient heart sound data is converted into image data in which the amplitude of the heart sound according to the change of time and frequency is visualized. Then, in the image data of the patient heart sound data, And classifying the heartbeat pattern into one of the cardiac sound patterns using a convolution neural network. The method according to claim 1,
Wherein a thermal noise by the smart device is removed from the patient's cardiac sound data, a time window corresponding to a lower predetermined ratio of the sound pressure is defined as thermal noise after the sound pressure is measured in the predetermined time window in the patient's heart sound data A method for providing diagnostic information of a cardiovascular disease. The method according to claim 1,
Wherein recording of the patient's heart sounds comprises displaying a measurement screen on the screen of the smart device (a plurality of heart sound measurement sites are displayed on the measurement screen), and the input means of the smart device displays the heart sound measurement site The measurement screen is switched to the recording screen. When the smart device records cardiac sound close to the body of the patient corresponding to the cardiac sound measurement site, it is determined whether or not the recorded heart sound can be interpreted And generating an alarm event based on the determination result. The method according to claim 1,
The predefined plurality of cardiac tone patterns may be selected from the group consisting of normal heart sounds, third heart sounds (S3), fourth heart sounds (S4), systolic murmur, diastolic murmur, Of the diagnostic information. Combined with smart devices that provide diagnostic information on cardiovascular disease,
A plurality of sampled heart sound data received from outside are converted into image data having visualized heartbeat amplitude according to a change in time and frequency, and an image object corresponding to heart sound is converted into convolutional neural network (CNN) in the image data of the sampled heart sound data A plurality of heart sound patterns defined in advance,
The cardiac sound data is directly recorded from the patient to be diagnosed, and the recorded patient heart sound data is converted into image data in which the amplitude of the heart sound according to the change of time and frequency is visualized. Then, in the image data of the patient heart sound data, A cardiac rhythm application stored on a medium for performing the step of classifying it into one of the cardiac rhythm patterns using a convolution neural network. 6. The method of claim 5,
Wherein a thermal noise by the smart device is removed from the patient's cardiac sound data, a time window corresponding to a lower predetermined ratio of the sound pressure is defined as thermal noise after the sound pressure is measured in the predetermined time window in the patient's heart sound data Features heart sound application. 6. The method of claim 5,
Wherein recording of the patient's heart sounds comprises displaying a measurement screen on the screen of the smart device (a plurality of heart sound measurement sites are displayed on the measurement screen), and the input means of the smart device displays the heart sound measurement site The measurement screen is switched to the recording screen. When the smart device records cardiac sound close to the body of the patient corresponding to the cardiac sound measurement site, it is determined whether or not the recorded heart sound can be interpreted And generating an alarm event based on the determination. 6. The method of claim 5,
The predefined plurality of cardiac tone patterns may include at least one of a heart sound application, a third heart sound (S3), a fourth heart sound (S4), a systolic murmur, a diastolic murmur, .

Images