Hardware Implementation with Phonocardiographic

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 9, Number 21 (2014) pp. 11265-11280 © Research India Publications http://www.ripublication.com

Hardware Implementation with Phonocardiographic Signal for Heart Murmur Process 1

P.S. Rajakumar, 2S. Ravi and 3R.M. Suresh

1

Research Scholar, Department of CSE, JNTU, Hyderabad. [email protected] 2 Professor and Head, Department of ECE, Dr. M.G.R. Educational and Research Institute University, Chennai [email protected] 3 Principal, Sri Muthukumaran Institute of Technology, Chennai. [email protected]

Abstract The heart sounds based diagnosis in cardiac field poses a great challenge to non-invasively assess the physiological changes occurring in the heart. An added constraint is that, heart murmurs are usually weak, non stationary signals and can be distorted by noise and artifacts. Thus processing of heart murmur signal need to be performed by taking into account their spatiotemporal correlation and mutual statistical dependence.The analysis are integrated to identify the normal and pathological heart sound patterns that once effectively interpreted are useful for clinical interpretation along existing with cardiac diagnosis like ECG. Keywords— Heart Murmur, Phonocardiogram, PCA, SVM Classifier, wavelet.

INTRODUCTION A novel algorithm for dynamic assessment of cardiac acoustics signal, such as PCG but not limited to, improve the associated researchers for better understanding of phonocardiographic signals (PCG). The signal processing techniques can be employed to improve the accuracy of diagnosis. Phonocardiography is the registration of sound vibrations of heart and blood flow that has a capability of presenting various heart valve disorders. The dynamic events are associated with the contraction and relaxation of the atria and ventricles, valve movements and blood flow for heart sound. It can be heard from the chest through a stethoscope, a device commonly used

Paper Code: 27521 - IJAER

11266

P.S. Rajakumar et al

for screening and diagnosis in primary health care. The cardiac auscultation evaluates the acoustic properties of heart sounds and murmurs that include the intensity, frequency, duration, number and quality of the sounds. Psychoacoustics is the science in which the human perception of sounds is quantified. Each of the sound characteristics has a corresponding perceptual variable. The aim is to derive a quantitative model that matches the results of auditory experiments that can be contrived. Typical artefacts for heart murmurs include gastrointestinal acoustic phenomena, lung sounds, transplanted heart signals etc. The cardiac cycle has two phases, a) Diastolic Phase and b) Systolic Phase The audible sound is produced due to the opening and closing of the heart values, which pump the blood. So, flow of blood in the heart causes the contraction and expansion of heart muscles. Basically, this work, measure the acoustic feature of the cardiac cycle. In this research, treating murmurs as audio signals and using these audio signals to extract useful features is explored, so that it can be used as indicators of abnormal sounds. The abnormal sound is a direct measure/ representative of the valve problems or sounds made by artificial pace makers. Once the behavior of the murmur is properly understood it will be very useful to directly map with valve problems. The valves necessary to quantify the sound includes, (i) Tricuspid valve: best heard at the Rt half the lower end of the sternum body. (ii) Mitral valve: best heard at the Apex of the heart (Lt 5th intercostal space at the mid-clavicular line). (iii) Pulmonary valves: best heard at the Lt medial 2nd intercostal space (iv) Aortic valve: best heard in the medial 2nd Rt inetercostal space. The sensor placement for spatial processing (i.e. location of the probes) renders unique information as shown in Figure 1.

Fig. 1 Sensor Placement

Hardware Implementation with Phonocardiographic Signal

11267

Inferring heart function with audio input Powerful ultrasound system can give accurate details of cardiac anomalies. The technologies are available to do intravascular ultrasound, 3D echo cardiogram and intra cardiac echo. A stethoscope can assess heart function, obstruct and leakage of valves. It can predict and follow up the effects of treatment of heart failure. Most of the birth defects involving the heart can be diagnosed fairly well by the use of the stethoscope, besides ECGS, X-rays and Echocardiogram. Doctors can detect cardiac complications such as papillary muscle rupture and ventricular septal rupture with the stethoscope. It helps to accurately evaluate valve functions, assess lungs and lungs diseases and the functioning of artificial valves. An anesthetist with a stethoscope can know if a patient endotracheal tube is in the wind pipe (in the food pipe) and check if both lungs is proper. An ultrasound device cannot replace a stethoscope. The most important part of the stethoscope is the medical professional’s brain situated between two ear pieces. It gives immense confidence to both patient and the doctor and serve the purpose for which it is invented by clinician with great clinical knowledge. Factors Influencing Analysis of Heart Murmur The influencing analysis of heart murmur includes; (i) Turbulent blood flow (due to the pressure difference between adjacent cardiac structures) (ii) Can be normal (physiologic, benign, flow, transitional, etc.) or abnormal (pathologic) (iii) Most (80%) children will have soft murmurs in the perinatal period Function of the valves The valves functions are, (i) Valves prevent the back flow of blood. (ii) The papillary muscles will not close the valves, but will maintain the closure of the valves. (iii) The importance of chordeatendinei attached to the papillary muscles is because during ventricular contraction the ventricle size decreases and the papillary muscle must contract to shorten the chordeatendinei to prevent the leakage of valves. Phonocardiogram Signal (PCG) based Heart Murmur Processing The phonocardiography refers to the tracing technique of heart sounds and the recording of cardiac acoustics vibration by means of microphone-transducer. Therefore, the signal provide a competent tool for further analysis and processing to enhance and optimize cardiac clinical diagnostic approach that assist in identifying and obtaining useful clinical and physiological information. Although these PCG acquisition techniques are plain, noninvasive, low-cost, and precise for assessing a wide range of heart diseases, diagnosis by auscultation requires good experience and considerable observation ability. The PCG signal is analyzed by morphological properties in time domain, by spectral properties in the frequency domain and a combination of both in time-frequency domain.

11268


Heart Murmur’s (Sound) Recordings Amplifiers are used to record heart sounds, using microphones and pick up not only the sounds /and murmurs at the body surface, but also all extraneous noises and vibrations in the immediate vicinity of the patient. Many sources of noise can contribute to the overall background level in a room. In addition, some soundproofing measures can be taken, if necessary. Any noise making apparatus within the room, such as a room air conditioner or fan should be disconnected during the examination. To check whether the environment is adequate for phonocardiography, place a microphone connected to the instrument on a table near the recording area. Set the amplifier controls at typical sensitivity settings and at the various filter positions. Room temperature must not be allowed to become too cool, for this can induce shivering in the patient, and the resultant muscle tremors, could interfere with good sound.

PREVIOUS WORK Kwak. C and Kwon.O-W [2012] proposed a new algorithm for cardiac disorder classification by heart sound signals that consists of three steps: segmentation, likelihood computation and classification. Kiran Kumari Patil, et al., [2011] proposed the phonocardiography of recording and interpreting of heart sounds using latest digital technology and in particular valvular diseases. Kiran Kumari Patil, et al., [2010] presented the theoretical foundations, psychoacoustic principles and presented mathematical models for the psychoacoustic features such as loudness, sharpness, intensity, strength, roughness, tonality etc. for a set of heart sounds and murmurs. P.Stein [1981] proposed a physical and physiological basis for the interpretation of cardiac auscultation, evaluations based primarily on the second sound and ejection murmurs. Adrian D.C. Chan, et al., [2008] presented an evaluation of a new biometric based on electrocardiogram (ECG) waveforms. ECG data were collected from 50 subjects during three data-recording sessions on different days using a simple user interface, where subjects held two electrodes on the pads of their thumbs using their thumb and index fingers. Lee, Y.M. [2002] presented a prototype of a simple and noninvasive system to remotely monitor the real-time heart rate of patients or individuals based on phonocardiography. M.R. Homaeinezhad, et al., [2010] proposed discrete wavelet-aided delineation of PCG signal events via analysis of an area curve length-based decision statistic. Muthalif, et al., [2013] presented a software tool to predict cardiac abnormalities that can be identified using heart sounds. M.R. Homaeinezhad, et al., [2011] proposed optimal delineation of PCG sounds via falsealarm bounded segmentation of a wavelet-based principal components analyzed metric. C. Ahlstrom, et al., [2006] proposed distinguishing innocent murmurs from murmurs caused by aortic stenosis by recurrence quantification analysis. Yildiz, et al., [2014] presented the heart sound data taken from different patients and recorded with the help of an electronic stethoscope are classified by using hidden markov model. S. Choi, et al., [2011] selection of wavelet packetmeasures for insufficiency murmur identification is proposed. S. Choi [2008] proposed detection of valvular heart disorders using wavelet packetdecomposition and support vector machine (SVM).


11269

Mostafa, et al., [2013] proposed the ability of rough set methodology to classify heart sound diseases without applying feature selection.S. Babaeiet al., [2009] proposed heart sound reproduction based onneural network classification of cardiac valve disorders using wavelet transforms PCG signals. Fatemeh Safara, et al., [2013] extracted a new set of features to distinguish categories of heartsounds, including normal, mitral stenosis, mitral regurgitation, aortic stenosis and aortic regurgitation. H. Naseri and M.R. Homaeinezhad [2013] proposed detection and boundary identification of phonocardiogram sounds using an expert frequency-energy based metric. S. Choi and Z. Jiang [2010] proposed cardiac sound murmurs classification with autoregressive spectral analysis and multi-support vector machine technique.

PHYSIOLOGY OF THE HEART SOUND During the systolic and the diastolic phase of the cardiac cycle, audible sounds are produced from the opening and the closing of the heart valves, the flow of blood in the heart, and the vibration of heart muscles. Usually, four heart sounds are generated in a cardiac cycle. The first heart sound and the second heart sound can be easily heard in a normal heart through a stethoscope placed on a proper area on the chest. The normal third heart sound is audible in children and adolescents but not in most adults. The fourth heart sound is seldom audible in normal individuals through the conventional mechanical stethoscopes but can be detected by sensors with high sensitivity, such as electronic stethoscopes and phonocardiography systems. Sounds other than these four, called murmurs, are abnormal sounds resulting from valve problems, or sounds made by artificial pacemakers or prosthetic valves. The murmurs during the early systolic phase are common in children and normally heard in nearly all adults after exercise. Abnormal murmurs may be caused by stenosis and insufficiencies (leaks) at the aortic, pulmonary or mitral valves. It is important from a diagnostic point of view (to note the time) and the location of murmurs. The identification of murmurs may assist the diagnosis of heat defections like aortic, stenosis, mitral and tricuspid regurgitation, etc.

BEHAVIOUR OF AORTIC VALVE Energy level associated will change depending on the position of the probes and where the probes are placed play a role. The recorded energy level varies that gives the requirement to have special processing of the cardiac cycle. The basic four valves that are studied include A.V, P.V, B.V, T.V, the lowest energy denoted by Red, where AV violet is highest energy level. Heart murmur caused by Valvular Lesions The murmurs of the aortic stenosis are narrowing of the aorta increases resistance to ejection of blood. As a result, severe turbulence of blood at the root of the aorta causes intense vibration resulting in a loud systemic murmur (after 1st heart sound).

11270


Murmur of the aortic regurgitation In aortic regurgitation, the aortic valves don’t close which is essential during diastole, therefore in aortic regurgitation blood backflow into the ventricles causing diastolic murmurs (after the 2nd heart sound). Murmurs of Mitral Stenosis In mitral stenosis, narrowing of the mitral valve increase resistance of blood flow to the ventricles. After 1/3 of diastole when enough blood fills the ventricle, it causes vibration which occurs as diastolic murmur. The murmur is often not heard but could be felt as thrill at the apex of the heart. Murmurs of Mitral regurgitation In Mitral regurgitation, Mitral valves are unable to close which is essential during systole  therefore blood flows back to the atrium causing a systolic murmur. As a result vibration effects will take place, in turn generates systemic murmur at a loud micro scale. It is not audible (usually negligible by stethoscope / ECG) at greater than micro scale. Machinery murmur of Patent Ductus Arteriosis In PDA blood flows from the aorta to the pulmonary artery resulting in a murmur during systole and diastole. The murmurs during systole are much denser than in diastole because the pressure in aorta is higher during systole than diastole. The fourth heart sound occurs in late diastole and before first heart sound produced by vibrations in expanding ventricles when atria contract that is rarely heard in a normal heart. The results are from the reduced distensibility of one or both ventricles, the stiff ventricles, the force of atrial contraction increases, causing sharp movement of the ventricular wall. (i) Heart sound classification and analysis play an important role in the auscultative diagnosis. (ii) Human response variables are not linearly proportional to the value of the corresponding stimulus variables. The sample murmur used for study in this work is listed in Table 1. Table 1 Heart Murmur Samples Description Sample Murmur Description Mitral Blood flows between different chambers of the heart must flow Regurgitation through a valve. The valve between the two chambers on the left (MR) side of heart is called the mitral valve Pulmonary Stenosis The pulmonary valve is too tight, so that the flow of blood from (PS) the right ventricle of the heart into the pulmonary artery is impeded.


11271

Mitral Stenosis (MS)

It restricts the flow of blood through the valve. Back pressure which builds up behind the narrowed valve can cause various problems and symptoms. Benign Murmur Due to physiologic conditions outside the heart, as opposed to (BM) structural defects in the heart itself. Ventricular Septal Hole between the right and left pumping chambers of the heart. Defect (VSD) Patent Ductus The ductusarteriosus is a blood vessel that allows blood to go Arteriosus (PDA) around the baby's lungs before birth. Tricuspid It is heard at the left lower sternal border and radiates to the right Regurgitation (TR) lower sternal border. Midsystolic It begins after the S1 heart sound and terminates before P2 heart Murmur (MM) sound, so S1 and S2 are audible. Atrial Septal Defect It defects due to increased flow through the pulmonic valves (ASD) similar to pulmonic stenosis.

CONTINUOUS WAVELET TRANSFORM AND WAVELET SERIES The continuous wavelet transform (CWT) is provided by equation where x(t) is the signal to be analyzed. ψ(t) is the basis function and the wavelet functions used in the transformation are derived from the mother wavelet through translation and scaling (compression) is given in eqn. [1], ( , )= … (1) ∫ ( ). ∗ | |

where, the translation parameter ‘τ’ relates to the location of the wavelet function as shifted through the signal. The scale parameter‘S’ is defined as corresponds to frequency information and performs the convolution operation of the signal and the high frequencies.

IMPLEMENTATION OF SVM CLASSIFIER WITH HEART MURMUR SVM analyzes data and recognize patterns, used for classification and regression analysis. SVM classifier can predict the input data into two distinct classes and used as multiclass classifiers by treating a K-class classification problem as K two-class problems. It depends on the kernel and essentially is a linear learning machine.

PRINCIPAL COMPONENT ANALYSIS The heart murmur signal needs to be extracted without knowing apriori information. This makes PCA a good choice, since most relevant information about the murmur can be restored. PCA extracts décorrelated heart murmurs according to the decreasing order of their variance. PCA is a statistical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables.It measures a number of observed

11272


variables and creates a smaller number of artificial variables from the observed variables. PCA defines a linear combination of optimally-weighted observed variables. The murmurs in heart sounds are accomplished by means of time-frequency representations (TFR) which deals with non-stationary signals.PCA is used in this paper toderive a relatively lesser number of decorrelated linear combinations called principal components from a set of random zero-mean heart murmurs. The objective of PCA includes (i) retaining as much of the information even after rejecting minor components (ii) perform dimensionality reduction. In this work, PCA is converted to the Eigen valve problem of the covariance matrix of x. PCA Algorithm 1 Compute the covariance matrix from the i/p from the heart murmur signal 2 Compute the eigenvectors and associated Eigen values. To facilitate easy computation of (1) and (2) adaptive learning algorithms are used in this work. This eliminates the need to handle very large covariance matrix. The identification of the Principal Components is done based on the mean square error of approximation. This is achieved by sorting the eigenvectors in descending order such that retaining the components with largest Eigen values leads to the minimal mean square error, i.e, the PC’s are extracted sequentially as long as the Eigen values λ1 exceed a set threshold. Implementation 1. The first principal component is extracted with Eigenvalue λ1, such that λ1 = E[y12] y1 is the PCA o/p transformation on the input vector. 2. The second PC is extracted with Eigen value λ2 where λ2 = E[y22] where y2 is a transformation on the residual error rather than on the i/p vector

SYSTOLIC SAMPLE Systolic heart murmurs are heard during systole the murmur begins and ends during the second heart sound. It involves stenosis of the semilunar valves or regurgitation of the atrioventricular valves. The innocent subclavian murmur (supraclavicular or brachiocephalic systolic murmur) is detected frequently in children and adolescents. It has a crescendo-decrescendo configuration, an abrupt onset and brief duration in early to midsystole, and is medium to high pitched. Early systolic murmurs begin with the first heart sound and extend to middle or late systole. Midsystolic murmurs begin a murmur-free interval in early systole and end with a murmur-free interval (of variable duration) in late systole. The late systolic murmur and second heart sounds patterns are shown in figure 2 and.figure 3 respectively.


Fig. 2 Late Systolic Murmur

Fig. 3 Second heart sounds as (S1 and S2)

The systolic sample of a child heart sound is shown in Figure 4.

Fig. 4 Systolic (Child) Sample

11273

11274


HARDWARE IMPLEMENTATION Psychoacoustic Model The novelty is in the Psychoacoustic model that is suited for clinical decision support (Clustering/Classification Process) (refer Figure 5).

Fig. 5 Classification Process based Psychoacoustic model

The hardware implementation use SVM classifiers implemented with FL2440 board processor in Linux kernel. The arm tool chain and boot loader are used with network capability. Built in compiler already exist in tool chains facility that can be easily interpreted of a compilation part. The user space program consists of a program for removing noise, and extract feature from cleaned murmur signal using wavelets and perform data reduction with PCA.

RESULTS AND DISCUSSION The Linux hardware acquires heart murmur and the heart murmur signal is extracted. Different type of heart murmur is given to test the algorithms namely systolic, MT, AC and soft murmur (occur in child). The hardware interface output is shown in Figure 6 and Figure 7.


Fig. 6 Hardware interface with FL2440

Fig. 7 Hardware Interface Output

Classification Accuracy of Systolic Samples The classification accuracy of systolic is shown in Figure 8(a) to Figure 8(i).

11275

11276


Fig. 8 (a) PS Accuracy increased of systolic samples

Fig. 8 (b) MR Accuracy increased of systolic samples

Fig. 8 (c) TR Accuracy increased of systolic samples


Fig. 8 (d) BM Accuracy increased of systolic samples

Fig. 8 (e) MM Accuracy increased of systolic samples

Fig. 8 (f) VSD Accuracy increased of systolic samples

11277

11278


Fig. 8 (g) PDA Accuracy increased of systolic samples

Fig. 8 (h) MS Accuracy increased of systolic samples

Fig. 8 (i) ASD Accuracy increased of systolic samples


11279

CONCLUSION The artifacts in the acquired heart murmur samples were filtered using wavelet tool and the filtered heart sound signals (wave signals) were classified using SVM technique. The heart sound murmurs were recorded from different subjects along with other measured signals. From these observations, different features were applied to a PCA network for dimension reduction.In this work, a hardware setup perform the signal analysis of heart sounds and murmurs and distinguish innocent murmurs from murmurs caused by Aortic Stenosis using recurrence quantification analysis.

ACKNOWLEDGEMENT My sincere thanks to the Management, Dr. M.G.R. Educational and Research Institute University, Chennai for providing the lab facilities in carrying out this study.

REFERENCES [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

Adrian D.C. Chan, Mohyeldin M. Hamdy, ArminBadre and VesalBadee, “Wavelet Distance Measure for Person Identification Using Electrocardiograms”, IEEE Transactions on Instrumentation and Measurement, Vol. 57, No. 2, February 2008. Ahlstrom, C., K. Höglund and P. Hult, “Distinguishing innocent murmurs from murmurs caused by aortic stenosis by recurrence quantification analysis”,International Journal of Biological and Medical Sciences, Vol. 1, No. 3, pp. 213-218, 2006. Babaei, S and A. Geranmayeh, “Heart sound reproduction based onneural network classification of cardiac valve disorders using wavelet transforms of PCG signals”,Computers in biology and medicine, Vol.39, No. 1, pp. 8-15, Jan. 2009. Choi, S., Y. Shin and H.K. Park, “Selection of wavelet packetmeasures for insufficiency murmur identification”,Expert Systemswith Applications, Vol. 38, No. 4, pp. 4264-4271, Apr. 2011. Choi, S., “Detection of valvular heart disorders using wavelet packetdecomposition and support vector machine”,Expert Systems withApplications, Vol. 35, No. 4, pp. 1679-1687, Nov. 2008. Choi, S. and Z. Jiang, “Cardiac sound murmurs classification withautoregressive spectral analysis and multi-support vector machinetechnique,”Computers in Biology and Medicine, Vol. 40, No. 1, pp.8-20, Jan. 2010. Fatemeh Safara, Shyamala Doraisamy, Azreen Azman, Azrul Jantanand Sri Ranga, “Diagnosis of Heart Valve Disorders through TrapezoidalFeatures and Hybrid Classifier”, International Journal of Bioscience, Biochemistry and Bioinformatics, Vol. 3, No. 6, November 2013. Homaeinezhad, M.R. S.A. Atyabi, E. Deneshvar, A. Ghaffari and M.Tahmasebi, “Optimal delineation of PCG sounds via false-alarmbounded segmentation of a wavelet-based principal components analyzed metric”,International Journal for

11280

[9]

[10]

[11]

[12]

[13] [14]

[15]

[16]

[17]

[18]


Numerical Methods inBiomedical Engineering, Vol. 27, No. 11, pp. 1711-1739, 2011. Homaeinezhad, M.R., S.A. Atyabi, E. Daneshvar, A. Ghaffari and M.Tahmasebi, “Discrete wavelet-aided delineation of PCG signal eventsvia analysis of an area curve length-based decision statistic”, Cardiovascular Engineering, Vol. 10, No. 4, pp. 218-34, Dec. 2010. Kwak. C. and Kwon.O.W., “Cardiac disorder classification by heart sound signals using murmur likelihood and hidden Markov model state likelihood”, IET Journals,Vol. 6, No. 4, pp. 326-334, 2012. Kiran Kumari Patil, B.S. Nagbhushan and Vijaya Kumar B.P., “Psychoacoustic models For heart sounds”, in Proceedings of Advances in Networks and Communications, Communications in Computer & Information Science, Vol. 132, Part-2, pp. 555-556, 2011. Kiran Kumari Patil, B.S. Nagbhushan and Vijaya Kumar B.P., “An efficient modeling technique for heart sounds and murmurs”, published in International Journal of Scientific & Engineering Research, Vol. 2, No. 2, 2010. Lee, Y.M. andMoghavvemi, M, “Remote heart rate monitoring system based on phonocardiography”, Research and Development, SCOReD, pp. 823-832, 2002. Mostafa A. Salama, Omar S. Soliman, Ilias Maglogiannis, Aboul Ella Hassanien and Aly A. Fahmy, “Rough Set-Based Identification of Heart Valve Diseases Using Heart Sounds”, Vol. 43, pp. 475-491, 2013. Muthalif, F.A., Selvarathnam, M. andLiyanaarachchi, M.R, “Automated diagnosis of cardiac abnormalities using heart sounds”, IEEE Conference, Bangalore, pp. 252-255, 16-18 Jan. 2013. Naseri, H. and M.R. Homaeinezhad, “Detection and boundaryidentification of phonocardiogram sounds using an expert frequency-energy based metric”,Annals of biomedical engineering, Vol. 41, No. 2, pp. 279-92, Feb. 2013. Stein, P., “A physical and physiological basis for the interpretation of cardiac auscultation: evaluations based primarily on the second sound and ejection murmurs”, Futura Publishing Company Rabiner, L. and Juang, “Fundamentals of Speech Recognition, Prentice-Hall, 1981. Yildiz, O., “Cardiac arrhythmia analysis using Hidden Markov Model and murmur diagnosis”, IEEE Conference, Trabzon, pp. 2031-2034, 23-25 April 2014.