Automatic heart sound segmentation and murmur detection in pediatric phonocardiograms

A Noise-Robust Heart Sound Segmentation Algorithm Based on Shannon Energy

Heart sound segmentation has been shown to improve the performance of artificial intelligence (AI)-based auscultation decision support systems increasingly viewed as a solution to compensate for eroding auscultatory skills and the associated subjectivity. Various segmentation approaches with demonstrated performance can be utilized for this task, but their robustness can suffer in the presence of noise. A noise-robust heart sound segmentation algorithm was developed and its accuracy was tested using two datasets: the CirCor DigiScope Phonocardiogram dataset and an in-house dataset - a heart murmur library collected at the Children's National Hospital (CNH). On the CirCor dataset, our segmentation algorithm marked the boundaries of the primary heart sounds S1 and S2 with an accuracy of 0.28 ms and 0.29 ms, respectively, and correctly identified the actual positive segments with a sensitivity of 97.44%. The algorithm also executed four times faster than a logistic regression hidden semi-Markov model. On the CNH dataset, the algorithm succeeded in 87.4% cases, achieving a 6% increase in segmentation success rate demonstrated by our original Shannon energy-based algorithm. Accurate heart sound segmentation is critical to supporting and accelerating AI research in cardiovascular diseases. The proposed algorithm increases the robustness of heart sound segmentation to noise and viability for clinical use.

Novel phonocardiography system for heartbeat detection from various locations

The paper presents evaluation of the proposed phonocardiography (PCG) measurement system designed primarily for heartbeat detection to estimate heart rate (HR). Typically, HR estimation is performed using electrocardiography (ECG) or pulse wave as one of the fundamental diagnostic methodologies for assessing cardiac function. The system includes novel both sensory part and data processing procedure, which is based on signal preprocessing using Wavelet Transform (WT) and Shannon energy computation and heart sounds classification using K-means. Due to the lack of standardization in the placement of PCG sensors, the study focuses on evaluating the signal quality obtained from 7 different sensor locations on the subject's chest and investigates which locations are most suitable for recording heart sounds. The suitability of sensor localization was examined in 27 subjects by detecting the first two heart sounds (S1, S2). The HR detection sensitivity related to reference ECG from all sensor positions reached values over 88.9 and 77.4% in detection of S1 and S2, respectively. The placement in the middle of sternum showed the higher signal quality with median of the proper S1 and S2 detection sensitivity of 98.5 and 97.5%, respectively.

Beyond Heart Murmur Detection: Automatic Murmur Grading From Phonocardiogram

OBJECTIVE

Murmurs are abnormal heart sounds, identified by experts through cardiac auscultation. The murmur grade, a quantitative measure of the murmur intensity, is strongly correlated with the patient's clinical condition. This work aims to estimate each patient's murmur grade (i.e., absent, soft, loud) from multiple auscultation location phonocardiograms (PCGs) of a large population of pediatric patients from a low-resource rural area.

METHODS

The Mel spectrogram representation of each PCG recording is given to an ensemble of 15 convolutional residual neural networks with channel-wise attention mechanisms to classify each PCG recording. The final murmur grade for each patient is derived based on the proposed decision rule and considering all estimated labels for available recordings. The proposed method is cross-validated on a dataset consisting of 3456 PCG recordings from 1007 patients using a stratified ten-fold cross-validation. Additionally, the method was tested on a hidden test set comprised of 1538 PCG recordings from 442 patients.

RESULTS

The overall cross-validation performances for patient-level murmur gradings are 86.3% and 81.6% in terms of the unweighted average of sensitivities and F1-scores, respectively. The sensitivities (and F1-scores) for absent, soft, and loud murmurs are 90.7% (93.6%), 75.8% (66.8%), and 92.3% (84.2%), respectively. On the test set, the algorithm achieves an unweighted average of sensitivities of 80.4% and an F1-score of 75.8%.

CONCLUSIONS

This study provides a potential approach for algorithmic pre-screening in low-resource settings with relatively high expert screening costs.

SIGNIFICANCE

The proposed method represents a significant step beyond detection of murmurs, providing characterization of intensity, which may provide an enhanced classification of clinical outcomes.

A New Method for Heart Disease Detection: Long Short-Term Feature Extraction from Heart Sound Data

Heart sounds have been extensively studied for heart disease diagnosis for several decades. Traditional machine learning algorithms applied in the literature have typically partitioned heart sounds into small windows and employed feature extraction methods to classify samples. However, as there is no optimal window length that can effectively represent the entire signal, windows may not provide a sufficient representation of the underlying data. To address this issue, this study proposes a novel approach that integrates window-based features with features extracted from the entire signal, thereby improving the overall accuracy of traditional machine learning algorithms. Specifically, feature extraction is carried out using two different time scales. Short-term features are computed from five-second fragments of heart sound instances, whereas long-term features are extracted from the entire signal. The long-term features are combined with the short-term features to create a feature pool known as long short-term features, which is then employed for classification. To evaluate the performance of the proposed method, various traditional machine learning algorithms with various models are applied to the PhysioNet/CinC Challenge 2016 dataset, which is a collection of diverse heart sound data. The experimental results demonstrate that the proposed feature extraction approach increases the accuracy of heart disease diagnosis by nearly 10%.

A Heart Segmentation Algorithm Based on Dynamic Ultrasound

The heart is one of the most important organs of the human body. The role of the heart is to promote blood flow and provide sufficient blood flow to organs and tissues. The research on the heart has important theoretical and clinical significance. Because of the noninvasive and intuitive display of ultrasound image, it can dynamically obtain the heart state and has become the main means to detect the heart dynamics. We analyze the characteristics of cardiac ultrasound image from the medical point of view and signal processing. The heart movement is periodic and rhythmic. The image signal can be decomposed. Firstly, the image is decomposed into high- and low-frequency signals to highlight different dimensional information. Then, the attention model was introduced, focusing on the heart region. Finally, the multidimensional network carrying model was established to achieve cardiac segmentation. The experimental results show that the AOM of the algorithm proposed in this paper reaches 92%, which has a certain degree of advancement and can assist doctors to make accurate diagnosis.

Unsupervised Heart Sound Decomposition and State Estimation with Generative Oscillation Models

This paper proposes a new generative probabilistic model for phonocardiograms (PCGs) that can simultaneously capture oscillatory factors and state transitions in cardiac cycles. Conventionally, PCGs have been modeled in two main aspects. One is a state space model that represents recurrent and frequently appearing state transitions. Another is a factor model that expresses the PCG as a non-stationary signal consisting of multiple oscillations. To model these perspectives in a unified framework, we combine an oscillation decomposition with a state space model. The proposed model can decompose the PCG into cardiac state dependent oscillations by reflecting the mechanism of cardiac sounds generation in an unsupervised manner. In the experiments, our model achieved better accuracy in the state estimation task compared to the empirical mode decomposition method. In addition, our model detected S2 onsets more accurately than the supervised segmentation method when distributions among PCG signals were different.

Cardiovascular Disease Recognition Based on Heartbeat Segmentation and Selection Process

Assessment of heart sounds which are generated by the beating heart and the resultant blood flow through it provides a valuable tool for cardiovascular disease (CVD) diagnostics. The cardiac auscultation using the classical stethoscope phonological cardiogram is known as the most famous exam method to detect heart anomalies. This exam requires a qualified cardiologist, who relies on the cardiac cycle vibration sound (heart muscle contractions and valves closure) to detect abnormalities in the heart during the pumping action. Phonocardiogram (PCG) signal represents the recording of sounds and murmurs resulting from the heart auscultation, typically with a stethoscope, as a part of medical diagnosis. For the sake of helping physicians in a clinical environment, a range of artificial intelligence methods was proposed to automatically analyze PCG signal to help in the preliminary diagnosis of different heart diseases. The aim of this research paper is providing an accurate CVD recognition model based on unsupervised and supervised machine learning methods relayed on convolutional neural network (CNN). The proposed approach is evaluated on heart sound signals from the well-known, publicly available PASCAL and PhysioNet datasets. Experimental results show that the heart cycle segmentation and segment selection processes have a direct impact on the validation accuracy, sensitivity (TPR), precision (PPV), and specificity (TNR). Based on PASCAL dataset, we obtained encouraging classification results with overall accuracy 0.87, overall precision 0.81, and overall sensitivity 0.83. Concerning Micro classification results, we obtained Micro accuracy 0.91, Micro sensitivity 0.83, Micro precision 0.84, and Micro specificity 0.92. Using PhysioNet dataset, we achieved very good results: 0.97 accuracy, 0.946 sensitivity, 0.944 precision, and 0.946 specificity.

Temporal-Framing Adaptive Network for Heart Sound Segmentation Without Prior Knowledge of State Duration

OBJECTIVE

This paper presents a novel heart sound segmentation algorithm based on Temporal-Framing Adaptive Network (TFAN), including state transition loss and dynamic inference.

METHODS

In contrast to previous state-of-the-art approaches, TFAN does not require any prior knowledge of the state duration of heart sounds and is therefore likely to generalize to non sinus rhythm. TFAN was trained on 50 recordings randomly chosen from Training set A of the 2016 PhysioNet/Computer in Cardiology Challenge and tested on the other 12 independent databases (2,099 recordings and 52,180 beats). And further testing of performance was conducted on databases with three levels of increasing difficulty (LEVEL-I, -II and -III).

RESULTS

TFAN achieved a superior F₁ score for all 12 databases except for 'Test-B,' with an average of 96.72%, compared to 94.56% for logistic regression hidden semi-Markov model (LR-HSMM) and 94.18% for bidirectional gated recurrent neural network (BiGRNN). Moreover, TFAN achieved an overall F₁ score of 99.21%, 94.17%, 91.31% on LEVEL-I, -II and -III databases respectively, compared to 98.37%, 87.56%, 78.46% for LR-HSMM and 99.01%, 92.63%, 88.45% for BiGRNN.

CONCLUSION

TFAN therefore provides a substantial improvement on heart sound segmentation while using less parameters compared to BiGRNN.

SIGNIFICANCE

The proposed method is highly flexible and likely to apply to other non-stationary time series. Further work is required to understand to what extent this approach will provide improved diagnostic performance, although it is logical to assume superior segmentation will lead to improved diagnostics.

Automatic heart sound classification from segmented/unsegmented phonocardiogram signals using time and frequency features

OBJECTIVE

Heart abnormality detection using heart sound signals (phonocardiogram (PCG)) has been an active research area for the last few decades. In this paper, automatic heart sound classification using segmented and unsegmented PCG signals is presented.

APPROACH

In this paper: (i) we perform an in-depth analysis of various time and frequency domain features, followed by experimental determination of effective feature subsets for improved classification performance; (ii) both segmented and unsegmented PCG signals are studied and important results concerning the respective feature subsets and their classification performances are reported; and (iii) different classification algorithms, including the support vector machine, kth nearest neighbor, decision tree, ensemble classifier, artificial neural network and long short-term memory network (LSTMs), are employed to evaluate the performance of the proposed feature subsets and their comparison with other established features and methods is presented.

MAIN RESULTS

It is observed that LSTM performs better on mel-frequency cepstral coefficient (MFCC) features extracted from unsegmented PCG data, with an area under curve (AUC) score of 91.39%, however, the MFCC features do not show a consistent performance with other classifiers (the second highest AUC score is 62.08% with the decision tree classifier). In contrast, in the case of time-frequency features from segmented data, the performance of all the classifiers is appreciable with AUC scores over 70%. In particular, the conventional machine learning techniques shows consistency in achieving over 80% in AUC scores. Significanc e: The results of this study highlight the importance of time and frequency domain features. Thus it is necessary to employ both the time and frequency features of segmented PCG signals to achieve improved classification.

Improved Segmentation with Dynamic Threshold Adjustment for Phonocardiography Recordings

In this study with the intent to segment phonocardiography (PCG) recordings, an algorithm that processes a feature vector obtained by jointly using wavelet transform and mel scaled energy spectrum of the PCG signal is proposed. The feature vector is processed by a peak detection algorithm which results in a set of peaks that meet some certain criteria and will be exploited in the succeeding stages of the algorithm. Heart sounds are labeled by convolving circularly a template and a fragment of the feature vector which is picked up with the guidance of the peaks. At the final stage, the algorithm tries to detect and correct erroneous labels. The performance of the algorithm was tested on both normal heart sounds and abnormal heart sounds; 80 records in total. As a result of these tests, for normal heart sounds S1 and S2 sounds were detected with 99,51% recall and 97,28% precision, while with 97,59% recall and 92,53% precision for abnormal heart sounds.

A Markov-Switching Model Approach to Heart Sound Segmentation and Classification

OBJECTIVE

We consider challenges in accurate segmentation of heart sound signals recorded under noisy clinical environments for subsequent classification of pathological events. Existing state-of-the-art solutions to heart sound segmentation use probabilistic models such as hidden Markov models (HMMs), which, however, are limited by its observation independence assumption and rely on pre-extraction of noise-robust features.

METHODS

We propose a Markov-switching autoregressive (MSAR) process to model the raw heart sound signals directly, which allows efficient segmentation of the cyclical heart sound states according to the distinct dependence structure in each state. To enhance robustness, we extend the MSAR model to a switching linear dynamic system (SLDS) that jointly model both the switching AR dynamics of underlying heart sound signals and the noise effects. We introduce a novel algorithm via fusion of switching Kalman filter and the duration-dependent Viterbi algorithm, which incorporates the duration of heart sound states to improve state decoding.

RESULTS

Evaluated on Physionet/CinC Challenge 2016 dataset, the proposed MSAR-SLDS approach significantly outperforms the hidden semi-Markov model (HSMM) in heart sound segmentation based on raw signals and comparable to a feature-based HSMM. The segmented labels were then used to train Gaussian-mixture HMM classifier for identification of abnormal beats, achieving high average precision of 86.1% on the same dataset including very noisy recordings.

CONCLUSION

The proposed approach shows noticeable performance in heart sound segmentation and classification on a large noisy dataset.

SIGNIFICANCE

It is potentially useful in developing automated heart monitoring systems for pre-screening of heart pathologies.

An automatic segmentation method for heart sounds

BACKGROUND

There are two major challenges in automated heart sound analysis: segmentation and classification. An efficient segmentation is capable of providing valuable diagnostic information of patients. In addition, it is crucial for some feature-extraction based classification methods. Therefore, the segmentation of heart sound is of significant value.

METHODS

This paper presents an automatic heart sound segmentation method that combines the time-domain analysis, frequency-domain analysis and time-frequency-domain analysis. Employing this method, the boundaries of heart sound components are first located, and the components are then recognized. Finally, the heart sounds are divided into several segments on the basis of the results of boundary localization and component identification.

RESULTS

In order to evaluate the performance of the proposed method, quantitative experiments are performed on an authoritative heart sound database. The experimental results show that the boundary localization has a sensitivity (Se) of 100%, a positive predictive value (PPV) of 99.3% and an accuracy (Acc) of 99.93%. Moreover, the Se, PPV and Acc of component identification reach 98.63, 99.86 and 98.49%, respectively.

CONCLUSION

The proposed method shows reliable performance on the segmentation of heart sounds. Compared with previous works, this method can be applied to not only normal heart sounds, but also the sounds with S3, S4 and murmurs, thus greatly increasing the applied range.

Heart Sound Segmentation-An Event Detection Approach Using Deep Recurrent Neural Networks

OBJECTIVE

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.

METHODS

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.

RESULTS

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.

CONCLUSION

Our approach shows state-of-the-art performance carefully evaluated on the 2016 PhysioNet/CinC Challenge dataset.

SIGNIFICANCE

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.

Residual Error Based Anomaly Detection Using Auto-Encoder in SMD Machine Sound

Detecting an anomaly or an abnormal situation from given noise is highly useful in an environment where constantly verifying and monitoring a machine is required. As deep learning algorithms are further developed, current studies have focused on this problem. However, there are too many variables to define anomalies, and the human annotation for a large collection of abnormal data labeled at the class-level is very labor-intensive. In this paper, we propose to detect abnormal operation sounds or outliers in a very complex machine along with reducing the data-driven annotation cost. The architecture of the proposed model is based on an auto-encoder, and it uses the residual error, which stands for its reconstruction quality, to identify the anomaly. We assess our model using Surface-Mounted Device (SMD) machine sound, which is very complex, as experimental data, and state-of-the-art performance is successfully achieved for anomaly detection.

The effect of data exchange protocols on decision support systems for heart sounds

Heart auscultation is one of the basic exams performed during a patient physical examination, but it is also one that has a high skill ceiling. Decision support systems can provide physicians with a tool that can help to reduce the demanding skill requirements of this exam. Nevertheless, this second opinion needs to be delivered in a timely interval in order to be truly useful for a physician. To do this we need not only optimized algorithms, but also a well designed system. In this paper, we have studied how two different data exchange protocols, that define how data should be transferred from an acquisition to a process module, can impact the celerity of delivering second opinion to a physician. With data collected from real exams, acquired in a field hospital initiative in Brazil, we recreated two use cases that allowed us to measure performance in the form of time and resources spent, as well as power consumption. Results have shown that different data exchange protocols can have a significant impact on a decision support system response time.

An open access database for the evaluation of heart sound algorithms

In the past few decades, analysis of heart sound signals (i.e. the phonocardiogram or PCG), especially for automated heart sound segmentation and classification, has been widely studied and has been reported to have the potential value to detect pathology accurately in clinical applications. However, comparative analyses of algorithms in the literature have been hindered by the lack of high-quality, rigorously validated, and standardized open databases of heart sound recordings. This paper describes a public heart sound database, assembled for an international competition, the PhysioNet/Computing in Cardiology (CinC) Challenge 2016. The archive comprises nine different heart sound databases sourced from multiple research groups around the world. It includes 2435 heart sound recordings in total collected from 1297 healthy subjects and patients with a variety of conditions, including heart valve disease and coronary artery disease. The recordings were collected from a variety of clinical or nonclinical (such as in-home visits) environments and equipment. The length of recording varied from several seconds to several minutes. This article reports detailed information about the subjects/patients including demographics (number, age, gender), recordings (number, location, state and time length), associated synchronously recorded signals, sampling frequency and sensor type used. We also provide a brief summary of the commonly used heart sound segmentation and classification methods, including open source code provided concurrently for the Challenge. A description of the PhysioNet/CinC Challenge 2016, including the main aims, the training and test sets, the hand corrected annotations for different heart sound states, the scoring mechanism, and associated open source code are provided. In addition, several potential benefits from the public heart sound database are discussed.

A multi-spot exploration of the topological structures of the reconstructed phase-space for the detection of cardiac murmurs

Acoustic heart signals are generated by a turbulence effect created when the heart valves snap shut, and therefore carrying significant information of the underlying functionality of the cardiovascular system. In this paper, we present a method for heart murmur classification divided into three major steps: a) features are extracted from the heart sound; b) features are selected using a Backward Feature Selection algorithm; c) signals are classified using a K-nearest neighbor's classifier. A new set of fractal features are proposed, which are based on the distinct signatures of complexity and self-similarity registered on the normal and pathogenic cases. The experimental results show that fractal features are the most capable of describing the non-linear structure and the underlying dynamics of heart sounds among the all feature families tested. The classification results achieved for the mitral auscultation spot (88% of accuracy) are in agreement with the current state of the art methods for heart murmur classification.

Comparison between users of a new methodology for heart sound auscultation

Auscultation is a routine exam and the first line of screening in heart pathologies. The objective of this study was to assess if using a new data collection system, the DigiScope Collector, with a guided and automatic annotation of heart auscultation, different levels of expertise/experience users could collect similar digital auscultations. Data were collected within the Heart Caravan Initiative (Paraíba, Brasil). Patients were divided into two study groups: Group 1 evaluated by a third year medical student (User 1), and an experienced nurse (User 2); Group 2 evaluated by User 2 and an Information Technology professional (User 3). Patients were auscultated sequentially by the two users, according to the randomization. Features extracted from each data set included the length (HR) of the audio files, the number of repetitions per auscultation area, heart rate, first (S1) and second (S2) heart sound amplitudes, S2/S1, and aortic (A2) and pulmonary (P2) components of the second heart sound and relative amplitudes (P2/A2). Features extracted were compared between users using paired-sample test Wilcoxon test, and Spearman correlations (P<;0.05 considered significant). Twenty-seven patients were included in the study (13 Group 1, and 14 Group 2). No statistical significant differences were found between groups, except in the time of auscultation (User 2 consistently presented longer auscultation time). Correlation analysis showed significant correlations between extracted features from both groups: S2/S1 in Group 1, and S1, S2, A2, P2, P2/A2 amplitudes, and HR in Group 2. Using the DigiScope Collector, we were able to collect similar digital auscultations, according to the features evaluated. This may indicate that in sites with limited access to specialized clinical care, auscultation files may be acquired and used in telemedicine for an expert evaluation.