applied sciences Article
Classification of Heart Sound Signal Using Multiple Features Yaseen, Gui-Young Son and Soonil Kwon * Department of Digital Contents, Sejong University, Seoul 05006, Korea;
[email protected] (Y.);
[email protected] (G.-Y.S.) * Correspondence:
[email protected]; Tel.: +82-2-3408-3847 Received: 10 October 2018; Accepted: 20 November 2018; Published: 22 November 2018
Abstract: Cardiac disorders are critical and must be diagnosed in the early stage using routine auscultation examination with high precision. Cardiac auscultation is a technique to analyze and listen to heart sound using electronic stethoscope, an electronic stethoscope is a device which provides the digital recording of the heart sound called phonocardiogram (PCG). This PCG signal carries useful information about the functionality and status of the heart and hence several signal processing and machine learning technique can be applied to study and diagnose heart disorders. Based on PCG signal, the heart sound signal can be classified to two main categories i.e., normal and abnormal categories. We have created database of 5 categories of heart sound signal (PCG signals) from various sources which contains one normal and 4 are abnormal categories. This study proposes an improved, automatic classification algorithm for cardiac disorder by heart sound signal. We extract features from phonocardiogram signal and then process those features using machine learning techniques for classification. In features extraction, we have used Mel Frequency Cepstral Coefficient (MFCCs) and Discrete Wavelets Transform (DWT) features from the heart sound signal, and for learning and classification we have used support vector machine (SVM), deep neural network (DNN) and centroid displacement based k nearest neighbor. To improve the results and classification accuracy, we have combined MFCCs and DWT features for training and classification using SVM and DWT. From our experiments it has been clear that results can be greatly improved when Mel Frequency Cepstral Coefficient and Discrete Wavelets Transform features are fused together and used for classification via support vector machine, deep neural network and k-neareast neighbor(KNN). The methodology discussed in this paper can be used to diagnose heart disorders in patients up to 97% accuracy. The code and dataset can be accessed at “https://github.com/yaseen21khan/Classification-of-HeartSound-Signal-Using-Multiple-Features-/blob/master/README.md”. Keywords: heart sound signal classification; discrete wavelets transform; mel frequency cepstral coefficient
1. Introduction Cardiovascular system is a perpetual source of data that sanctions soothsaying or distinguishing among cardiovascular diseases. External constrains can lead to cardiac diseases that can cause sudden heart failure [1]. Cardiovascular diseases are censorious and must be detected with no time delay [2]. Heart diseases may be identified by elucidating the cardiac sound data. The heart sound signal characteristics may vary with respect to different kinds of heart diseases. A huge difference in the pattern can be found between a normal heart sound signal and abnormal heart sound signal as their PCG signal varies from each other with respect to time, amplitude, intensity, homogeneity, spectral content, etc [3].
Appl. Sci. 2018, 8, 2344; doi:10.3390/app8122344
www.mdpi.com/journal/applsci
Appl. Sci. 2018, 8, 2344 Appl. Sci. 2018, 8, x FOR PEER REVIEW
2 of 14 2 of 14
Cardiovascular auscultation auscultation is is the the principal principal yet yet simple simple diagnosing diagnosing method method used used to to assess assess and and Cardiovascular analyzethe theoperation operationand andfunctionality functionalityof ofthe theheart. heart. ItIt is is aa technique technique of of listening listening to to heart heart sound sound with with analyze stethoscope. The The main main source source of of the the generation generation of of heart heart sound sound is is due due to to unsteady unsteady moment moment of of blood blood stethoscope. called blood blood turbulence. turbulence. The opening opening and and closing closing of of atrioventricular atrioventricular values, values, mitral mitral and and tricuspid tricuspid called valuescauses causesdifferential differentialblood bloodpressure pressureand andhigh highacceleration accelerationand andretardation retardationofofblood bloodflow flow[4]. The values [4]. The auscultation of cardiac disorders is carried out using an electronic stethoscope which is indeed cost auscultation of cardiac disorders is carried out using an electronic stethoscope which is indeed cost effectiveand andnon-invasive non-invasive approach digital recording ofsound heartwith sound the help of effective approach [3]. [3]. The The digital recording of heart the with help of electronic electronic stethoscope is called PCG [4]. stethoscope is called PCG [4]. Once the the heart heart sound sound is is obtained, obtained, itit can can be be classified classified via via computer computer aided aided software software techniques, techniques, Once these techniques techniques need need more more accurately accurately defined defined heart heart sound sound cycle for feature feature extraction process [3]. these Manydifferent differentautomated automated classifying approaches been including used, including neural Many classifying approaches have have been used, artificial artificial neural network networkand (ANN) and hiddenmodel Markov modelMultilayer (HMM). Multilayer perceptron-back (ANN) hidden Markov (HMM). perceptron-back propagationpropagation (MLP-BP), (MLP-BP), Wavelet and coefficients and neural used forof classification of heart Wavelet coefficients neural networks werenetworks used for were classification heart sound signal. Butsound they signal. But they have often resulted in lower accuracy due to segmentation error [2]. have often resulted in lower accuracy due to segmentation error [2]. Toachieve achievebetter betteraccuracy, accuracy, this work, propose SVM, centroid displacement based To in in this work, wewe propose SVM, centroid displacement based KNNKNN and and DNN based classification algorithm using discrete wavelets transform (DWT) and (MFCC) DNN based classification algorithm using discrete wavelets transform (DWT) and (MFCC) features. features. The recognition accuracy can be increased for both clearsignals, speechusing signals, using The recognition accuracy can be increased for both noisy andnoisy clearand speech discrete discrete wavelets transform fused together with MFCCs features [5,6]. We extracted MFCCs features wavelets transform fused together with MFCCs features [5,6]. We extracted MFCCs features and and discrete wavelets transform features for heart sound signals (normal and abnormal) and discrete wavelets transform features for heart sound signals (normal and abnormal) and classified classified using support vectordeep machine, neural centroid displacement them usingthem support vector machine, neuraldeep network andnetwork centroid and displacement based KNN, based KNN, the highest accuracy we achieved so far is 97.9%. the highest accuracy we achieved so far is 97.9%. The beginning beginning of of this this paper paper describes describes some some background background knowledge knowledge about about heart heart sound sound signal signal The processing and and categories categories of of heart heart disorders disorders in in brief briefdetail, detail,afterwards afterwards the themethodology methodology of of features features processing (MFCCs and DWT) and classifiers we used is discussed. In the experiment section we talk about (MFCCs and DWT) and classifiers we used is discussed. In the experiment section we talk about classificationof offeatures, features,detail detailabout aboutclassifiers, classifiers,the thetools toolswe weused, used,and and database database in in detail. detail. In In the the end, end, classification wehave havediscussed discussedour ourresults results and performance evaluation using accuracy averaged F1 score we and performance evaluation using accuracy and and averaged F1 score and and the last paragraph concludes the paper. the last paragraph concludes the paper. 2. 2. Background Background The chambers. Two of them areare called atrias andand theythey make the Thehuman humanheart heartisiscomprised comprisedofoffour four chambers. Two of them called atrias make upper portion of theofheart whilewhile remaining two chambers are called ventricles and they the upper portion the heart remaining two chambers are called ventricles andmake theylower make portion of heart, blood heart atrias and exit through ventricles. Normal heart lower portion ofand heart, and enters blood the enters thethrough heart through atrias and exit through ventricles. Normal sounds (Figure(Figure 1a is an1aexample of normal heart sound generated by closing opening heart sounds is an example of normal heartsignal) soundare signal) are generated byand closing and of the valves of valves heart. The heartThe sound signals are directly opening and closing opening of the of heart. heart soundproduced, signals produced, are related directlyto related to opening and of the valves, flow andflow viscosity. ThereforeTherefore during exercise other activities increases closing of theblood valves, blood and viscosity. duringorexercise or other which activities which the heart rate, bloodrate, flowblood through thethrough valves is increased hence heart soundheart signal’s intensity is increases the heart flow the valves isand increased and hence sound signal’s increased and during and shock (low shock blood (low flow)blood the sound intensity decreased [2]. Figure[2]. 1f intensity is increased during flow) signal’s the sound signal’sisintensity is decreased also shows the shows spectrum a PCG signal. Figure 1 f also theof spectrum of a PCG signal.
(a)
(b) Figure 1. Cont.
Appl. Sci. 2018, 8, 2344 Appl. Sci. 2018, 8, x FOR PEER REVIEW
3 of 14 3 of 14
(c)
(d)
(e)
(f)
Figure 1. sound signal; (b)(b) Murmur in systole (MVP); (c) Mitral Regurgitation (MR); 1. (a) (a)AAnormal normalheart heart sound signal; Murmur in systole (MVP); (c) Mitral Regurgitation (d) Mitral Stenosis (MS); (e) Aortic Stenosis (AS); (f) Spectrum of a PCG (MR); (d) Mitral Stenosis(MS); (e) Aortic Stenosis (AS); (f) Spectrum of asignal. PCG signal.
The of heart valves generate generate audible audible sounds sounds of of frequency frequency range range lower lower than than 22 kHz, kHz, The movements movements of heart valves commonly TheThe “lub” is the in the in heart is denoted commonly refer refertotoasas“lub-dub” “lub-dub” “lub” is first the portion first portion thesound heart signal soundand signal and is by (S1), it is generated from the closing of mitral and tricuspid valve. One complete cardiac cycle is the denoted by (S1), it is generated from the closing of mitral and tricuspid valve. One complete cardiac heart sound signal wave which starts from S1 and ends to the start of next S1 and it is described cycle is the heart sound signal wave which starts from S1 and ends to the start of next S1 and it as is one heartbeat. The duration,The pitch, and shape ofand the heart us detail about different described as one heartbeat. duration, pitch, shapesound of the shows heart sound shows us the detail about conditions of conditions heart [4]. Closing of[4]. mitral valves followed by closing of tricuspid valve usually the the different of heart Closing ofismitral valves is followed by closing of and tricuspid valve delay between operation is this 20 tooperation 30 ms. Due toto the contraction first,contraction the mitral valve and usually thethis delay between is 20 30left ms.ventricle Due to the left ventricle first, component occurs first followed by tricuspid valve component in the signal. If the duration between the mitral valve component occurs first followed by tricuspid valve component in the signal. If the these twobetween sound components is in between 100 is to in 200 ms, it is100 called a split and its frequency range duration these two sound components between to 200 ms, it is called a split and its lies from 40range to 200 Hzfrom and40 it is considered if is theconsidered delay is above msdelay [7]. “Dub”, is generated frequency lies Hz to 200 Hz fatal and it fatal 30 if the is above 30 ms [7]. (the second heart sound denoted “S2”), when the aortic vales and pulmonary are “Dub”, is generated (thecomponent second heart soundby component denoted by “S2”), when the aortic valves vales and closed. S1 is usually of longer time period and lower frequency than S2 which has shorter duration pulmonary valves are closed. S1 is usually of longer time period and lower frequency than S2 which but ranges (50 to 250 Hz).ranges A2 is another heart sound component generated has higher shorterfrequency duration but higher frequency (50 to 250 Hz). A2 issignal another heart sound signal due to movement of Aortic valve and P2 is another heart sound signal component generated due to component generated due to movement of Aortic valve and P2 is another heart sound signal movement pulmonary Aortic pressure is highervalve. compared pulmonary pressure hence the componentof generated duevalve. to movement of pulmonary Aortictopressure is higher compared to A2 component appears before P2 in the heart sound signal. The presence of cardiac diseases such as pulmonary pressure hence the A2 component appears before P2 in the heart sound signal. The pulmonary and atrial septal defect can be identified by analyzing closely between presence of stenosis cardiac diseases such as pulmonary stenosis and atrial septal defect canthe besplit identified by aortic component A2 and component P2 and intensities. component During deep analyzing closely the split pulmonary between aortic component A2 their and pulmonary P2emotional and their and excitement state, the emotional duration between A2 and state, P2 may longer than usual,A2 causing intensities. During deep and excitement thebeduration between and P2long mayand be wide in between thelong components. state isin considered if the delay between these longersplitting than usual, causing and wideThis splitting between fatal the components. This state is component is longer 30 between ms [7]. these component is longer than 30 ms [7]. considered fatal if thethan delay Beside “lub-dub”, some some noisy noisy signals signals may may be be present present in the heart sounds called called murmurs. murmurs. Murmurs, areare continuous vibrations produced due due to thetoirregular flow of blood Murmurs, both bothnormal normaland andfatal, fatal, continuous vibrations produced the irregular flow of in cardiovascular system. system. Murmurs can be ofcan twobe types “normal and “abnormal murmurs”, blood in cardiovascular Murmurs of two typesmurmurs” “normal murmurs” and “abnormal normal murmurs are murmurs usually present in heartpresent sound signal component of infants, children of andinfants, adults murmurs”, normal are usually in heart sound signal component during also during in women (during this type of murmurthis cantype be detected in the childrenexercise and adults exercise alsopregnancy), in women (during pregnancy), of murmur canfirst be heart sound. Abnormal murmurs are usually present in heart patience which indicate a heart valve detected in the first heart sound. Abnormal murmurs are usually present in heart patience which defect called stenosis (squeezed heart stenosis valves) and regurgitation heart valves) [7]. Murmurs indicate a heart valve defect called (squeezed heart (bleeding valves) and regurgitation (bleeding
heart valves) [7]. Murmurs are the unusual sound present in heart sound cycle of an abnormal heart sound which indicates some abnormalities. Based on the murmur position in the heart cycle, they
Appl. Sci. 2018, 8, 2344
4 of 14
are the unusual sound present in heart sound cycle of an abnormal heart sound which indicates some abnormalities. Based on the murmur position in the heart cycle, they can be called systolic murmurs, diastolic murmurs and continuous murmurs [4]. These murmurs sounds and clicks are a key points to identify cardiovascular disease [7]. Murmurs in the heart sound can be discovered using stethoscope, echocardiography or phonocardiography. Murmurs are classified as continuous murmurs, systolic murmurs and diastolic murmurs. Systolic murmurs are present during systole and they are generated during the ventricles contraction (ventricular ejection). In the heart sound component they are present between S1 and S2 hence are called as systolic murmurs. Based on their types, these systolic murmurs can be called as either ejection murmurs (atrial septal defect, pulmonary stenosis, or aortic stenosis) Figure 1e, or regurgitant murmurs (ventricular septal defect, tricuspid regurgitation, mitral regurgitation or mitral valve prolapse) Figure 1c. Diastolic murmurs are created during diastole (after systole), when the ventricles relax. Diastolic murmurs are present between the second and first heart sound portion, this type of murmur is usually due to mitral stenosis (MS) or aortic regurgitation (AR) by as shown in Figure 1d. Mitral valve prolapse (MVP) is a disease where the murmur sound is present in between the systole section as shown in Figure 1b. The murmur of AR is high pitched and the murmur of AS is low pitched. In mitral regurgitation (MR), systolic component S1 is either soft or buried or absent, and the diastolic component S2 is widely split. In mitral stenosis, the murmur is low pitched and rumbling and is present along the diastolic component. In mitral valve prolapse, the murmur can be found throughout S1compnent. The sound signal of VSD and MR are mostly alike [2,7]. Summarizing the above discussion, we can see that heart sound signal can be acquired using and electronic stethoscope, in each complete cycle of heart sound signal we have S1–S4 intervals, S3 and S4 are rare heart sounds and are not normally audible but can be shown on the graphical recording i.e., phonocardiogram [4]. S3 and S4 intervals are called murmur sound and the heart sound signal which carries murmurs are called abnormal heart sounds and can be classified according to murmur position in the signal. In this study automatic classification of heart sound signal is carried out using five categories, one normal category and four abnormal categories. These abnormal categories are aortic stenosis (AS), mitral regurgitation (MR), mitral stenosis (MS) and MVP (murmur exist in the systole interval). Figure 1 shows graphical representation of theses heart sound signals categorically. The classification of heart sound signal can be carried out by several machine learning classifiers available for biomedical signal processing. One of them is SVM which can be used for classification of heart sounds. SVM is machine learning, classification and recognition technique that totally works on statistical learning and theorems, the classification accuracy of SVM is much more efficient than conventional classification techniques [8]. Another classifier that recently gained attention is DNN, DNN acoustic models have high performance in speech processing as well as other bio medical signal processing [9]. In this paper, our study shows that these classifiers (SVM and DNN also centroid displacement based KNN) have good performance for heart sound signal classification. The performance of Convolutional Neural Network is optimized for image classification, so this method is not suitable for the feature set we used in this paper. We also used Recurrent Neural Network in the pilot test, but our DNN method showed better results [10,11]. 3. Methodology The heart sound signal contains useful data about the working and health status of the heart, the heart sound signal can be processed using signal processing approach to diagnose various heart diseases before the condition of the heart get worse. Hence, various signal processing techniques can be applied to process the PCG signal. The stages involved in processing and diagnosing the heart sound signal are, the acquisition of heart sound signal, noise removal, sampling the PCG signal at a specific frequency rate, feature extraction, training and classification. As shown in the following Figure 2.
Appl. Sci. 2018, 8, 2344 Appl. Sci. 2018, 8, x FOR PEER REVIEW
5 of 14 5 of 14
(a)
(b)
Figure Figure 2. 2. (a) Proposed heart sound sound signal signal classification classification algorithm algorithm using SVM; (b) (b) Proposed Proposed heart heart sound sound signal signal classification classification algorithm algorithm using using DNN. DNN.
From heart sound sound signal signal we we extract extracttwo twodifferent differenttypes typesof offeatures: features: MFCCs DWT. From heart MFCCs andand DWT. To To classify these features we used SVM, centroid displacement based KNN and DNN as classifiers. classify these features we used SVM, centroid displacement based KNN and DNN as classifiers. For For performance evaluation of features using these classifiers,we wecarried carriedout out our our experiment performance evaluation of features using these classifiers, experiment and and evaluated the averaged F1 score and accuracy for all of the three features (MFCCs, DWT and evaluated the averaged F1 score and accuracy for all of the three features (MFCCs, DWT and a acombination combinationofofthese) these)using usingthe thethree threeclassifies classifies(SVM, (SVM,DNN DNNand andcentroid centroidbased basedKNN). KNN). 3.1. 3.1. Mel Mel Frequency Frequency Cepstral Cepstral Coefficients Coefficients (MFCCs) (MFCCs) We extracted features as a set of measured, non-redundant and derived valuesvalues from heart We extracted features as a set of measured, non-redundant and derived fromsignal heart which has been used in our study. As we used two types of features, MFCCs, DWT and the combination signal which has been used in our study. As we used two types of features, MFCCs, DWT and the of both, MFCCs features are enormously used in signal processing and processing recognition.and They were first combination of both, MFCCs features are enormously used in signal recognition. calculated and used in speech analysis by Davis and Mermelstein in the 1980’s. Human can recognize They were first calculated and used in speech analysis by Davis and Mermelstein in the 1980’s. small changes in pitch of sound signalinatpitch loweroffrequencies and MFCC scale is linear forMFCC those Human can recognize small changes sound signal atthe lower frequencies and the frequencies having range
