Digital Subtraction Phonocardiography (DSP) applied to the detection and characterization of heart murmurs
© Akbari et al; licensee BioMed Central Ltd. 2011
Received: 25 October 2011
Accepted: 20 December 2011
Published: 20 December 2011
During the cardiac cycle, the heart normally produces repeatable physiological sounds. However, under pathologic conditions, such as with heart valve stenosis or a ventricular septal defect, blood flow turbulence leads to the production of additional sounds, called murmurs. Murmurs are random in nature, while the underlying heart sounds are not (being deterministic).
We show that a new analytical technique, which we call Digital Subtraction Phonocardiography (DSP), can be used to separate the random murmur component of the phonocardiogram from the underlying deterministic heart sounds.
We digitally recorded the phonocardiogram from the anterior chest wall in 60 infants and adults using a high-speed USB interface and the program Gold Wave http://www.goldwave.com. The recordings included individuals with cardiac structural disease as well as recordings from normal individuals and from individuals with innocent heart murmurs. Digital Subtraction Analysis of the signal was performed using a custom computer program called Murmurgram. In essence, this program subtracts the recorded sound from two adjacent cardiac cycles to produce a difference signal, herein called a "murmurgram". Other software used included Spectrogram (Version 16), GoldWave (Version 5.55) as well as custom MATLAB code.
Our preliminary data is presented as a series of eight cases. These cases show how advanced signal processing techniques can be used to separate heart sounds from murmurs. Note that these results are preliminary in that normal ranges for obtained test results have not yet been established.
Cardiac murmurs can be separated from underlying deterministic heart sounds using DSP. DSP has the potential to become a reliable and economical new diagnostic approach to screening for structural heart disease. However, DSP must be further evaluated in a large series of patients with well-characterized pathology to determine its clinical potential.
In the healthy cardiovascular system, blood flow is generally laminar in character. Under certain pathologic conditions, such as a narrowing of a heart valve or a small hole in the ventricular septum, blood flow becomes turbulent, and can be heard as a noise known as a murmur . One difficulty for clinicians is that this murmur is only part of the total sound signal emitted from the heart, which also contains underlying regular heart sounds. This fact necessarily complicates the listening process.
Listening to the emitted sounds from the heart using a stethoscope (auscultation) is a frequent first step in diagnosis. It is often followed by echocardiography when the auscultatory findings are abnormal. However, the lack of reliability of ordinary auscultation and the expense and awkwardness of echocardiography make it desirable to develop a more practical, inexpensive, reliable, non-invasive approach to auscultation, one that could also be adapted for continuous monitoring [2–6]. In order to better identify any pathology found and make as detailed as possible any diagnosis, it would also appear to be helpful if the murmurs could somehow be separated from the underlying deterministic heart sounds.
The present study seeks to apply a new (unpublished, unpatented) analytical technique, known as Digital Subtraction Phonocardiography (DSP), to develop a noninvasive means for the detection and characterization of heart murmurs, such as those resulting from heart valve lesions or other types of cardiac pathology. The proposed technique is fundamentally different from previous phonocardiographic signal processing efforts (Rangayyan and Lehner 1988 ; Khadra et al 1991 ; Bentley and McDonnell 1994 ; Durand et al 1993 ; Guo et al 1994 ; Durand and Pibarot 1995 ; Tranulis et al 2002 ).
The DSP technique is fundamentally different from those efforts that it starts by constructing a difference signal between two time-adjacent heart cycles, which we herein call a "murmurgram". Furthermore, based on a deterministic plus random component phonocardiogram model that seems quite plausible to us, we show through both mathematical reasoning and computer simulation how murmurgrams would be expected to behave. In addition, it is our empirical (although necessarily preliminary) observation that murmurgrams in patients with abnormalities like mitral regurgitation or aortic stenosis are different from normal controls.
Recorded heart sounds for some of the patients studied
Equipment list which was used in this project
Precordial Chest Piece
Latex Rubber Tubing
Mini Electret Microphone
Sound Level Calibrator, 94dB
Commercial Audio Amplifier with volume adjustment and meter
USB Audio Interface
Microphone calibration was performed using a Extech Sound Level Calibrator Model 407744 which produces a sinusoidal wave at 1 KHz with 94 dB SPL intensity. By comparing any recorded sounds to the calibration recordings, it is then possible to obtain absolute sound intensity measurements. In order to achieve high-quality recordings, the clinical recording environment was kept completely silent, with the patients lying in the supine position. Each recording was divided into five sections of 3 minutes each. Care was taken to ensure that borborygmi sounds (from stomach and intestines) and other artifacts were not present.
All sounds were recorded using the Goldwave software (Version 5.55), which includes tools for recording, playing, filtering, and analyzing sounds. Using this software, we also deleted any electrical utility frequency (50 Hz) from the ECG recordings. Although GoldWave presents a wide range of digital filters, further filtering was done in MATLAB including Cancellation DC drift and Low & High Pass Filtering. The detailed code can be seen in the additional file 1. As we generate the murmurgram by subtracting two consecutive cycles, the only issue in children in how to find the exact location of the R-peaks. Pan-Tompkins algorithm was a good choice and great sampling rate of the recording system was helpful in this regard.
Furthermore, The rate of heart beats has not any influence on the sound signals from our point of view and based on the experiments which were done by us. The detail method of recording the data can be found in reference .
Technical and Analytical Issues
Note that this process is complicated by the fact that, due to the physiological variability, not all cardiac cycles are of the same length. Consequently, prior to data analysis and the construction of the murmurgrams, any data collection must first be subjected to a preliminary analysis to determine the longest duration cardiac cycle present. A raw murmurgram for a legal cycle i is then formed as the difference between PCG (cycle i) and PCG (cycle i +1).
Following the murmurgram construction, we apply color spectrographic analysis to supplement the time-domain murmurgram. Signal intensity to color is mapped as follows: Red > Orange > Yellow > Green > Blue > Black. These color spectrograms use colors to denote signal intensity at a particular time and frequency. Our preliminary observations suggest that a normal murmurgram is fairly "flat" (uniform in character) and "low in intensity" in both the time and frequency domains, while this is expected not to be the case for patients with heart murmurs. (We expect that once a large corpus of cases have been collected that we will be able to replace vague concepts such as "flat" and "low in intensity" with something far more specific).
Finally, it is often important to be able to characterize murmurs in terms of their distribution over the cardiac cycle; e.g., systolic or diastolic. For instance, almost all murmurs that occur during diastole are abnormal.
Simulated Case 1
Simulated Case 2
Simulated Case 3
Clinical Cases 1 and 2
Clinical Case 3
Clinical Case 4
Clinical Case 5
Clinical Case 6
Clinical Case 7
Clinical Case 8
In this study ECG-synchronized digital subtraction and spectrographic analysis is used to study heart sounds and murmurs in an entirely new way. While older phonocardiograph designs remain useful to study cardiac disorders, this system offers two new dimensions to conventional graphical and auscultatory methods. These are: (1) the ability to separate the deterministic component of heart sounds from murmurs by digital subtraction; and (2) the ability to apply spectrographic analysis to the extracted murmur signals.
In this system we used digital signal processing techniques to construct a "murmurgram", defined as the resulting signal when one subtracts PCG cycle i+1 from PCG cycle i. A murmurgram is thus simply the difference between the acoustic emissions of two successive heart beats. In practice, the QRS complex of the ECG is used as a marker of the beginning of each cardiac cycle so that any two successive PCGs can be aligned and subtracted to produce a murmurgram. Also note that another series of murmurgrams could be obtained, for example, by subtracting PCG cycle i+2 from PCG cycle i and so on. For now we are concentrating on the use of "nearest neighbor" PCG cycles. Based on this model, a normal murmurgram should be more or less flat across the cardiac cycle (within the limits of system noise effects and biological variability) while increases in the murmurgram signal are expected to occur in regions of the cardiac cycle associated with intra cardiac turbulent blood flow resulting from cardiac structural pathology.
Our approach is an improvement in the state of the art; it produces a murmur signal free of the underlying deterministic heart sounds. By isolating the cardiac murmur from the phonocardiogram in this manner, the system may allow physicians to detect and characterize cardiac murmurs that may be rather difficult to reliably detect by traditional means using a stethoscope, such as when the heart is beating quickly or when the heart sounds are faint.
Cardiac murmurs can be separated from underlying deterministic heart sounds using Digital Subtraction Analysis. Digital Subtraction Phonocardiography has the potential to become a reliable and economical new diagnostic approach to screening for structural heart disease.
- Wierwille L: Pediatric heart murmurs: evaluation and management in primary care. Nurse Pract 2011, 36(3):22–8. 10.1097/01.NPR.0000393968.36683.f0View ArticleGoogle Scholar
- Rangayyan RM, Lehner RJ: Phonocardiogram Signal Analysis: A Review. CRC Critical Reviews in Biomedical Engineering 1988, 15: 3.Google Scholar
- Durand LG, Blanchard M, Cloutier G, Sabbah HN, Stein PD: Comparison of pattern recognition methods for computer-assisted classification of spectra of heart sounds in patients with a porcine bioprosthetic valve implanted in the mitral position. IEEE Trans Biomed Eng 1990, 37(12):1121–1129. 10.1109/10.64456View ArticleGoogle Scholar
- Durand LG, Blanchard M, Sabbah HN, Stein PD: Comparison of spectral techniques for computer assisted classification of spectra of heart sounds in patients with a porcine bioprosthetic valves. Med Biol Eng Comput 1991, 31: 229.View ArticleGoogle Scholar
- Durand LG, Pibarot P: Digital Signal Processing of the honocardiogram: Review of the Most Recent Advancements. Critical Reviews in Biomedical Engineering 1995, 23(3/4):163–219.View ArticleGoogle Scholar
- Sarbandi RR, Doyle DD, Navidbakhsh M, Hassani K, Torabiyan H: Color spectrographic phonocardiography (CSP) applied to the detection and characterization of heart murmurs: preliminary results. BioMedical Engineering On Line 2011, 10: 42.Google Scholar
- Rangayyan RM, Lehner RJ: Phonocardiogram signal processing: A review. CRC Critical Reviews in Biomedical Eng 1988, 15(3):211–236.Google Scholar
- Khadra L, Matalgah M, El-Asir B, Mavagdeh S: The wavelet transform and its application to phonocardiogram signal analysis. Med.Informat 1991, 16: 271–277. 10.3109/14639239109025301View ArticleGoogle Scholar
- Bentley PM, McDonnell JTE: 1994: Wavelet transforms: an introduction. Electron Commun Eng J 1994, 6(4):175–186. 10.1049/ecej:19940401View ArticleGoogle Scholar
- Durand LG, Guo Z, Sabbah HN, Stein PD: Comparison of spectral techniques for computer-assisted classification of spectra of heart sounds in patients with porcine bioprosthetic valves. Medical and Biological Engineering and Computing 1993, 31(3):229–236. 10.1007/BF02458041View ArticleGoogle Scholar
- Guo Z, Durand LG, Lee HC: Comparison of time-frequency distribution techniques for analysis of simulated Doppler ultrasound signals of the femoral artery. IEEE Trans On Biomedical Eng 1994, 41(4):332–342. 10.1109/10.284961View ArticleGoogle Scholar
- Durand LG, Pibarot P: Digital signal processing of the phonocardiogram: review of the most recent advancements. Crit Rev Biomed Eng 1995, 23: 163–219.View ArticleGoogle Scholar
- Tranulis C, Durand LG, Senhadji L, Pibarot P: Estimation of pulmonary arterial pressure by a neural network analysis using features based on time-frequency representations of the second heart sound. Med Biol Eng Comput 2002, 40(2):205–212. 10.1007/BF02348126View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.