Computer-Based Detection and Analysis of Heart Sound and Murmur

A novel heart sound activity detection framework for automated heart sound analysis

U.S. Army helicopters used to provide aeromedical transport are extremely noisy (105 dB). This prevents auscultation with current stethoscopes. Ten subjects used four different stethoscopes in noise at 70–100 dB in order to determine the detection threshold of a body sound. The stethoscopes differed in performance (p=0.001); the best being the acoustic, followed by the three electronic stethoscopes. The threshold of noise for the detection of heart and breath sounds was 80 dB and 70–75 dB, respectively. A comparison of a standard stethoscope and one modified with Communications Ear Plugs revealed a difference (p=0.05) at 70 and 80 dB that disappeared at higher noise levels, implying that it was not simple masking, but that noise was amplified after entering the sensing head. Despite the poorer performance of the electronic stethoscopes, the need for a gain of 30–35 dB indicated that this was the preferred route for development. A sensor with an acoustic impedance match close to body tissue, but poorly matched to air-coupled noise, was developed, and trials indicate that heart sounds can be heard at 105 dB. Development continues with the aim of reducing or eliminating the need for signal processing.

In this paper, we accurately detect the state-sequence first heart sound(S1)-systole-second heart sound(S2)-diastole, i.e., the positions of S1 and S2, in heart sound recordings. We propose an event detection approach without explicitly incorporating a priori information of the state duration. This renders it also applicable to recordings with cardiac arrhythmia and extendable to the detection of extra heart sounds (third and fourth heart sound), heart murmurs, as well as other acoustic events. We use data from the 2016 PhysioNet/CinC Challenge, containing heart sound recordings and annotations of the heart sound states. From the recordings, we extract spectral and envelope features and investigate the performance of different deep recurrent neural network(DRNN) architectures to detect the state sequence. We use virtual adversarial training, dropout, and data augmentation for regularization. We compare our results with the state-of-the-art method and achieve an average score for the four events of the state sequence of ${\bf F}_{1}\approx 96$% on an independent test set. Our approach shows state-of-the-art performance carefully evaluated on the 2016 PhysioNet/CinC Challenge dataset. In this work, we introduce a new methodology for the segmentation of heart sounds, suggesting an event detection approach with DRNNs using spectral or envelope features.

Phonocardiography (PCG) is a noninvasive method for obtaining recordings of cardiovascular sound, that is, acoustic phenomena that are perceivable by auscultation at the cardiac region on the chest wall. Essentially, vibrations of the chest surface are observed that are generated in cardiac structures and transmitted to the chest wall. In physiological terms, heart sounds and murmurs are mentioned: Heart sounds have a transient, musical character; murmurs, except for the musical types, have a random, noisy character. Normal heart sounds are initiated by valve closing (the first and the second sound) or by filling phenomena (the third and the fourth sound); besides the normal ones pathological sounds exist, for example, generated at opening of degenerated valves. Heart murmurs of the random type generally result from turbulence in blood flow; the latter can be caused by flow through a narrowed valve opening or a leaking valve. Besides a small number of normal types, most heart murmurs are pathological and many of these can be related to malfunctioning of cardiac valves. Phonocardiography and auscultation thus aim at the recognition of pathological sounds and murmurs and their relation to malfunctioning heart structures. Recording of heart sounds at the chest surface, as for any vibration recording, requires a vibration transducer (also called a heart vibration pickup, a heart sound microphone), a suitable amplifier with specific frequency characteristics and a storage, visualization, and processing device. Contact transducers (accelerometers) and air‐coupled transducers are used. The first type can be indicated as absolute, the second as relative as it measures a weighted difference in vibration between the air chamber surface and the surface at the edge of the rigid housing of the transducer. The sensing chest piece of the (electronic) stethoscope accords with the second type. Generally, quantitative measurement requires a transducer that does not affect the vibration to be measured. This is hard to achieve in vibration measurement on soft tissue as is the case for the chest wall. Affecting the vibrating surface is, for example, a technique applied in heart auscultation: with the use of a stiff membrane and by applying firm pressure on the stethoscope chest piece higher frequencies are better observed. In order to obtain equivalent results as from auscultation (due to the qualities of the human hearing sense) high pass frequency filter sets have to be applied on recorded vibrations in phonocardiography. A single signal representation does not provide the same information as heard with the stethoscope. On the other hand, heart sound recording can exceed the hearing abilities and inaudible sounds can be recorded. In conventional phonocardiography, signals are stored on paper with a suitable graphic recorder; Evidently, modern systems make use of information technology for storage, processing, and visualization. As a result of different transducers, an important interference with the chest wall vibration and different signal conditioning, heart sound recording is not standardized, as compared with electrocardiography, which has a clear millivolt scale on its ordinate axis: The method can be indicated semiquantitative as results depend on the measuring apparatus used. Nevertheless, knowledge of heart sounds and murmurs has been greatly increased with this technique. Signal analysis, more specifically time–frequency analysis, has proven to be very useful in the identification and classification of heart sound components and murmurs, and their relation to cardiac structures and hemodynamic variables. Whereas auscultation still holds its position as a diagnostic tool for general physicians and cardiologists, phonocardiography in its classical form has lost interest mostly as a result of the availability of techniques as Doppler echocardiography and cardiac imaging methods, which give more direct information on cardiac structures and hemodynamic variables. The historical value of conventional phonocardiography has to be stressed. On the other hand, recording and processing of heart sounds continue to be subjects of scientific research and remain beneficial for training and for supporting of diagnosis of physicians. Electronic stethoscopes coupled to a laptop and connected to the internet for automated or for remote diagnosis by a specialist may grow in importance in the next few years.

Heart sounds are a fundamental physiological variable that provide a unique insight into cardiac semiotics. However a deterministic and unambiguous association between noises in cardiac dynamics is far from being accomplished yet due to many and different overlapping events which contribute to the acoustic emission. The current computer-based capacities in terms of signal detection and processing allow one to move from the standard cardiac auscultation, even in its improved forms like electronic stethoscopes or hi-tech phonocardiography, to the extraction of information on the cardiac activity previously unexplored. In this report, we present a new equipment for the detection of heart sounds, based on a set of accelerometric sensors placed in contact with the chest skin on the precordial area, and are able to measure simultaneously the vibration induced on the chest surface by the heart's mechanical activity. By utilizing advanced algorithms for the data treatment, such as wavelet decomposition and principal component analysis, we are able to condense the spatially extended acoustic information and to provide a synthetical representation of the heart activity. We applied our approach to 30 adults, mixed per gender, age and healthiness, and correlated our results with standard echocardiographic examinations. We obtained a 93% concordance rate with echocardiography between healthy and unhealthy hearts, including minor abnormalities such as mitral valve prolapse.

Phonocardiogram (PCG), the digital recording of heart sounds is becoming increasingly popular as a primary detection system for diagnosing heart disorders and it is relatively inexpensive. Electrocardiogram (ECG) is used during the PCG in order to identify the systolic and diastolic parts manually. In this study a heart sound segmentation algorithm has been developed which separates the heart sound signal into these parts automatically. This study was carried out on 100 patients with normal and abnormal heart sounds. The algorithm uses discrete wavelet decomposition and reconstruction to produce PCG intensity envelopes and separates that into four parts: the first heart sound, the systolic period, the second heart sound and the diastolic period. The performance of the algorithm has been evaluated using 14,000 cardiac periods from 100 digital PCG recordings, including normal and abnormal heart sounds. In tests, the algorithm was over 93% correct in detecting the first and second heart sounds. The presented automatic segmentation algorithm using wavelet decomposition and reconstruction to select suitable frequency band for envelope calculations has been found to be effective to segment PCG signals into four parts without using an ECG.

An early detection of heart disease can save lives, caution individuals and also help to determine the type of treatment to be given to the patients. The first test of diagnosing a heart disease is through auscultation - listening to the heart sounds. The interpretation of heart sounds is subjective and requires a professional skill to identify the abnormalities in these sounds. A medical practitioner uses a stethoscope to perform an initial screening by listening for irregular sounds from the patient's chest. Later, echocardiography and electrocardiography tests are taken for further diagnosis. However, these tests are expensive and require specialized technicians to operate. A simple and economical way is vital for monitoring in homecare or rural hospitals and urban clinics. This dissertation is focused on developing a patient-centered device for initial screening of the heart sounds that is both low cost and can be used by the users on themselves, and later share the readings with the healthcare providers. An innovative mobile health service platform is created for analyzing and classifying heart sounds. Certain properties of heart sounds have to be evaluated to identify the irregularities such as the number of heart beats and gallops, intensity, frequency, and duration. Since heart sounds are generated in low frequencies, human ears tend to miss certain sounds as the high frequency sounds mask the lower ones. Therefore, this dissertation provides a solution to process the heart sounds using several signal processing techniques, identifies the features in the heart sounds and finally classifies them. This dissertation enables remote patient monitoring through the integration of advanced wireless communications and a customized low-cost stethoscope. It also permits remote management of patients' cardiac status while maximizing patient mobility. The smartphone application facilities recording, processing, visualizing, listening, and classifying heart sounds. The application also generates an electronic medical record, which is encrypted using the efficient elliptic curve cryptography and sent to the cloud, facilitating access to physicians for further analysis. Thus, this dissertation results in a patient-centered device that is essential for initial screening of the heart sounds, and could be shared for further diagnosis with the medical care practitioners.

Phonocardiograms (PCG) are recordings of the acoustic waves produced by the mechanical action of the heart. They generally consist of two kinds of acoustic vibrations: heart sounds and heart murmurs. Heart murmurs are often the first signs of pathological changes of the heart valves, and are usually found during auscultation in primary health care. Heart auscultation has been recognized for a long time as an important tool for the diagnosis of heart disease, although its accuracy is still insufficient to diagnose some heart diseases. It does not enable the analyst to obtain both qualitative and quantitative characteristics of the PCG signals. The efficiency of diagnosis can be improved considerably by using modern digital signal processing techniques. Therefore, these last can provide useful and valuable information on these signals. The aim of this study is to analyse PCG signals using wavelet transform. This analysis is based on an algorithm for the detection of heart sounds (the first and second sounds, S1 and S2) and heart murmurs using the PCG signal as the only source. The segmentation algorithm, which separates the components of the heart signal, is based on denoising by wavelet transform (DWT). This algorithm makes it possible to isolate individual sounds (S1 or S2) and murmurs. Thus, the analysis of various PCGs signals using wavelet transform can provide a wide range of statistical parameters related to the phonocardiogram signal.

Changes and murmurs of heart sounds often provide the earliest information of abnormal heart status, which makes heart sound monitor and diagnose an important mean for cardiovascular disease. Compared to other heart monitoring approaches, such as EGG, Color Doppler and Magnetic Resonance Imaging Diagnosis, heart sound detection is more portable, much safer and cheaper, and needs lower professional requirements, thus has a broader applicability. This paper aims to design a system for monitoring and analyzing heart sounds in noisy environments, which can achieve signal acquisition, automatic analysis and classification. For the signal acquisition part, a novel equipment prototype is designed by separately using the two channels of a digital sound, one to record mixed signals of heart sounds and some environmental noises, and the other to record only environmental noises as a reference, then the environmental noise can be filtered out based on Least Mean Square (LMS) adaptive noise cancellation to get purified heart sound. By disassembling and assembling a stethoscope and a recorder, this design is realized and verified. For the signal analysis and classification part, with the features extracted by wavelet decomposition and Normalized Average Shannon Energy, a novel fuzzy neural network method with structure learning is used for the heart sound classification. Experiments with real and simulated data demonstrated that our approach can correctly classify all the tested heart sounds with noises even for the ones with previous unseen heart diseases.

A heart sound segmentation algorithm has been developed which separates the heart sound signal into four parts: the first heart sound, the systolic period, the second heart sound, and the diastolic period. The algorithm uses discrete wavelet decomposition and reconstruction to produce phonocardiographic intensity envelopes. The performance of the algorithm has been evaluated using 14000 cardiac periods from 100 digital phonocardiographic recordings, including normal and abnormal heart sounds. In tests, the algorithm was over 93 percent correct in detecting the first and second heart sounds.Keywordsphonocardiographyauscultationfirst heart soundsecond heart soundsystolic perioddiastolic periodmurmurswavelet decompositionwavelet reconstructionsegmentationnormalized average Shannon energy

Developing explainable machine intelligence (XAI) models for heart sound abnormality detection is a crucial area of research aimed at improving the interpretability and transparency of machine learning algorithms in medical diagnostics. In this study, we propose a framework for building XAI models that can effectively detect abnormalities in heart sounds while providing interpretable explanations for their predictions. We leverage techniques such as SHapley additive exPlanations (SHAP) and local interpretable model-agnostic explanations (LIME) to generate explanations for model predictions, enabling clinicians to understand the rationale behind the algorithm’s decisions. Our approach involves preprocessing heart sound data, training machine learning models, and integrating XAI techniques to enhance the interpretability of the models. We evaluate the performance of our XAI models using standard metrics and demonstrate their effectiveness in accurately detecting heart sound abnormalities while providing insightful explanations for their predictions. This research contributes to the advancement of XAI in medical applications, particularly in the domain of cardiac diagnostics, where interpretability is crucial for clinical decision-making.

Analysis of PCG signals using quality assessment and homomorphic filters for localization and classification of heart sounds

The third (S3) and fourth (S4) heart sounds detected by phonocardiography are considered to represent the criterion standards of the gallop sounds, but their test characteristics have not been explored. To determine the diagnostic test characteristics of the S3 and S4 for prediction of left ventricular dysfunction using a computerized heart sound detection algorithm. Prospective study of 90 adult patients undergoing elective left-sided heart catheterization at a single US teaching hospital between August 2003 and June 2004. The mean age was 62 (SD, 13) years (range, 24-90 years) and 61 (68%) were male. Within a 4-hour period, participants underwent computerized heart sound phonocardiographic analysis, cardiac catheterization, transthoracic echocardiography, and blood sampling for assessment of an S3/S4, left ventricular end-diastolic pressure (LVEDP), left ventricular ejection fraction (LVEF), and B-type natriuretic peptide (BNP), respectively. Diagnostic test characteristics of the computerized phonocardiographic S3 and S4 using markers of left ventricular function as criterion standards. Mean (SD) LVEDP was significantly elevated (18.4 [6.9] mm Hg vs 12.1 [7.3] mm Hg; P<.001), mean (SD) LVEF was reduced (49.4% [20.2%] vs 63.6% [14.8%]; P<.001), and median (interquartile range) BNP was elevated (330 [98-1155] pg/mL vs 86 [41-192] pg/mL; P<.001) in those with an S3, S4, or both compared with patients without a diastolic heart sound. The sensitivities of these heart sounds to detect an elevated LVEDP, reduced LVEF, or elevated BNP were 41%, 52%, and 32% for an S3, and 46%, 43%, and 40% for an S4, respectively. For abnormal levels of the same markers of ventricular function, the specificities of the S3 were 92%, 87%, and 92%, while the specificities of the S4 were 80%, 72%, and 78%, respectively. Neither the phonocardiographic S3 nor the S4 is a sensitive marker of left ventricular dysfunction. The phonocardiographic S3 is specific for left ventricular dysfunction and appears to be superior to the moderate specificity of the phonocardiographic S4.

A novel and robust method for heart sound activity detection (HSAD) is presented. In this method, a new discriminative feature is developed based on the lag-1 autocorrelation coefficient and feature smoothing is introduced to identify the endpoints of low-level heart murmurs. Several experiments on a large scale phonocardiogram database show that the method significantly outperforms the other HSAD methods under varying levels of heart sounds and different types of noise. For a signal-to-noise ratio of 5 dB, the method achieves a sensitivity of 99.56% and a positive predictivity of 99.41% and has a smallest average time error value of 4.26 ms.

Cardiovascular diseases have become one of the most prevalent threats to human health throughout the world. As a noninvasive assistant diagnostic tool, the heart sound detection techniques play an important role in the prediction of cardiovascular diseases. In this paper, the latest development of the computer-aided heart sound detection techniques over the last five years has been reviewed. There are mainly the following aspects: the theories of heart sounds and the relationship between heart sounds and cardiovascular diseases; the key technologies used in the processing and analysis of heart sound signals, including denoising, segmentation, feature extraction and classification; with emphasis, the applications of deep learning algorithm in heart sound processing. In the end, some areas for future research in computer-aided heart sound detection techniques are explored, hoping to provide reference to the prediction of cardiovascular diseases.