Logistic Regression-HSMM-Based Heart Sound Segmentation

Algorithm for predicting valvular heart disease from heart sounds in an unselected cohort

Objective

This study aims to assess the ability of state-of-the-art machine learning algorithms to detect valvular heart disease (VHD) from digital heart sound recordings in a general population that includes asymptomatic cases and intermediate stages of disease progression.

Methods

We trained a recurrent neural network to predict murmurs from heart sound audio using annotated recordings collected with digital stethoscopes from four auscultation positions in 2,124 participants from the Tromsø7 study. The predicted murmurs were used to predict VHD as determined by echocardiography.

Results

The presence of aortic stenosis (AS) was detected with a sensitivity of 90.9%, a specificity of 94.5%, and an area under the curve (AUC) of 0.979 (CI: 0.963-0.995). At least moderate AS was detected with an AUC of 0.993 (CI: 0.989-0.997). Moderate or greater aortic and mitral regurgitation (AR and MR) were predicted with AUC values of 0.634 (CI: 0.565-703) and 0.549 (CI: 0.506-0.593), respectively, which increased to 0.766 and 0.677 when clinical variables were added as predictors. The AUC for predicting symptomatic cases was higher for AR and MR, 0.756 and 0.711, respectively. Screening jointly for symptomatic regurgitation or presence of stenosis resulted in an AUC of 0.86, with 97.7% of AS cases (n = 44) and all 12 MS cases detected.

Conclusions

The algorithm demonstrated excellent performance in detecting AS in a general cohort, surpassing observations from similar studies on selected cohorts. The detection of AR and MR based on HS audio was poor, but accuracy was considerably higher for symptomatic cases, and the inclusion of clinical variables improved the performance of the model significantly.

Markov-Based Neural Networks for Heart Sound Segmentation: Using Domain Knowledge in a Principled Way

This work considers the problem of segmenting heart sounds into their fundamental components. We unify statistical and data-driven solutions by introducing Markov-based Neural Networks (MNNs), a hybrid end-to-end framework that exploits Markov models as statistical inductive biases for an Artificial Neural Network (ANN) discriminator. We show that an MNN leveraging a simple one-dimensional Convolutional ANN significantly outperforms two recent purely data-driven solutions for this task in two publicly available datasets: PhysioNet 2016 (Sensitivity: 0.947 ±0.02; Positive Predictive Value : 0.937 ±0.025) and the CirCor DigiScope 2022 (Sensitivity: 0.950 ±0.008; Positive Predictive Value: 0.943 ±0.012). We also propose a novel gradient-based unsupervised learning algorithm that effectively makes the MNN adaptive to unseen datum sampled from unknown distributions. We perform a cross dataset analysis and show that an MNN pre-trained in the CirCor DigiScope 2022 can benefit from an average improvement of 3.90% Positive Predictive Value on unseen observations from the PhysioNet 2016 dataset using this method.

Deep Learning Algorithms to Detect Murmurs Associated With Structural Heart Disease

Background The success of cardiac auscultation varies widely among medical professionals, which can lead to missed treatments for structural heart disease. Applying machine learning to cardiac auscultation could address this problem, but despite recent interest, few algorithms have been brought to clinical practice. We evaluated a novel suite of Food and Drug Administration-cleared algorithms trained via deep learning on >15 000 heart sound recordings. Methods and Results We validated the algorithms on a data set of 2375 recordings from 615 unique subjects. This data set was collected in real clinical environments using commercially available digital stethoscopes, annotated by board-certified cardiologists, and paired with echocardiograms as the gold standard. To model the algorithm in clinical practice, we compared its performance against 10 clinicians on a subset of the validation database. Our algorithm reliably detected structural murmurs with a sensitivity of 85.6% and specificity of 84.4%. When limiting the analysis to clearly audible murmurs in adults, performance improved to a sensitivity of 97.9% and specificity of 90.6%. The algorithm also reported timing within the cardiac cycle, differentiating between systolic and diastolic murmurs. Despite optimizing acoustics for the clinicians, the algorithm substantially outperformed the clinicians (average clinician accuracy, 77.9%; algorithm accuracy, 84.7%.) Conclusions The algorithms accurately identified murmurs associated with structural heart disease. Our results illustrate a marked contrast between the consistency of the algorithm and the substantial interobserver variability of clinicians. Our results suggest that adopting machine learning algorithms into clinical practice could improve the detection of structural heart disease to facilitate patient care.

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.

Parameter-Efficient Densely Connected Dual Attention Network for Phonocardiogram Classification

Cardiac auscultation, exhibited by phonocardiogram (PCG), is a non-invasive and low-cost diagnostic method for cardiovascular diseases (CVDs). However, deploying it in practice is quite challenging, due to the inherent murmurs and a limited number of supervised samples in heart sound data. To solve these problems, not only heart sound analysis based on handcrafted features, but also computer-aided heart sound analysis based on deep learning have been extensively studied in recent years. Though with elaborate design, most of these methods still use additional pre-processing to improve classification performance, which heavily relies on time-consuming experienced engineering. In this article, we propose a parameter-efficient densely connected dual attention network (DDA) for heart sound classification. It combines two advantages simultaneously of the purely end-to-end architecture and enriched contextual representations of the self-attention mechanism. Specifically, the densely connected structure can automatically extract the information flow of heart sound features hierarchically. Alongside, improving contextual modeling capabilities, the dual attention mechanism adaptively aggregates local features with global dependencies via a self-attention mechanism, which captures the semantic interdependencies across position and channel axes respectively. Extensive experiments across stratified 10-fold cross-validation strongly evidence that our proposed DDA model surpasses current 1D deep models on the challenging Cinc2016 benchmark with significant computational efficiency.

Recognition of Pediatric Congenital Heart Diseases by Using Phonocardiogram Signals and Transformer-Based Neural Networks

The phonocardiogram (PCG) or heart sound auscultation is a low-cost and non-invasive method to diagnose Congenital Heart Disease (CHD). However, recognizing CHD in the pediatric population based on heart sounds is difficult because it requires high medical training and skills. Also, the dependency of PCG signal quality on sensor location and developing heart in children are challenging. This study proposed a deep learning model that classifies unprocessed or raw PCG signals to diagnose CHD using a one-dimensional Convolution Neural Network (1D-CNN) with an attention transformer. The model was built on the raw PCG data of 484 patients. The results showed that the attention transformer model had a good balance of accuracy of 0.923, a sensitivity of 0.973, and a specificity of 0.833. The Receiver Operating Characteristic (ROC) plot generated an Area Under Curve (AUC) value of 0.964, and the F1-score was 0.939. The suggested model could provide quick and appropriate real-time remote diagnosis application in classifying PCG of CHD from non-CHD subjects.Clinical Relevance- The suggested methodology can be utilized to analyze PCG signals more quickly and affordably for rural doctors as a first screening tool before sending the cases to experts.

On the Impact of Synchronous Electrocardiogram Signals for Heart Sounds Segmentation

In this paper we study the heart sound segmentation problem using Deep Neural Networks. The impact of available electrocardiogram (ECG) signals in addition to phonocardiogram (PCG) signals is evaluated. To incorporate ECG, two different models considered, which are built upon a 1D U-net - an early fusion one that fuses ECG in an early processing stage, and a late fusion one that averages the probabilities obtained by two networks applied independently on PCG and ECG data. Results show that, in contrast with traditional uses of ECG for PCG gating, early fusion of PCG and ECG information can provide more robust heart sound segmentation. As a proof of concept, we use the publicly available PhysioNet dataset. Validation results provide, on average, a sensitivity of 97.2%, 94.5%, and 95.6% and a Positive Predictive Value of 97.5%, 96.2%, and 96.1% for Early-fusion, Late-fusion, and unimodal (PCG only) models, respectively, showing the advantages of combining both signals at early stages to segment heart sounds.Clinical relevance- Cardiac auscultation is the first line of screening for cardiovascular diseases. Its low cost and simplicity are especially suitable for screening large populations in underprivileged countries. The proposed analysis and algorithm show the potential of effectively including electrocardiogram information to improve heart sound segmentation performance, thus enhancing the capacity of extracting useful information from heart sound recordings.

Impact of Biological Sex on Radar-Measured Heart Sound Quality

Radar based contact-free technology has number of potential applications for monitoring the cardiopulmonary functions of patients. However, no study has evaluated the effect of gender on the quality of the recordings. This study makes an attempt to distinguish radar based recording of male and female subjects. The study analysed a publicly available dataset of radar-recorded heart sound signals from both male and female subjects. Here, we exploit the reference signal-to-noise ratio (RSNR) to quantify the signal's quality. The results indicate that there is a significant difference in the signal quality between males and females, with males having a higher RSNR value compared to females. This could be a limitation in the widespread use of the current radar based cardiopulmonary recording techniques and overcoming this should be considered for future research.Clinical relevance- This work has highlighted the gender based difference. By considering this, the radar based cardiopulmonary device has the potential for being used for patients requiring long-term monitoring.

A Hardware-Software System for Accurate Segmentation of Phonocardiogram Signal

Background

Phonocardiogram (PCG) signal provides valuable information for diagnosing heart diseases. However, its applications in quantitative analyses of heart function are limited because the interpretation of this signal is difficult. A key step in quantitative PCG is the identification of the first and second sounds (S1 and S2) in this signal.

Objective

This study aims to develop a hardware-software system for synchronized acquisition of two signals electrocardiogram (ECG) and PCG and to segment the recorded PCG signal via the information provided in the acquired ECG signal.

Material and Methods

In this analytical study, we developed a hardware-software system for real-time identification of the first and second heart sounds in the PCG signal. A portable device to capture synchronized ECG and PCG signals was developed. Wavelet de-noising technique was used to remove noise from the signal. Finally, by fusing the information provided by the ECG signal (R-peaks and T-end) into a hidden Markov model (HMM), the first and second heart sounds were identified in the PCG signal.

Results

ECG and PCG signals from 15 healthy adults were acquired and analyzed using the developed system. The average accuracy of the system in correctly detecting the heart sounds was 95.6% for S1 and 93.4% for S2.

Conclusion

The presented system is cost-effective, user-friendly, and accurate in identifying S1 and S2 in PCG signals. Therefore, it might be effective in quantitative PCG and diagnosing heart diseases.

A novel heart sound segmentation algorithm via multi-feature input and neural network with attention mechanism

Objective. Heart sound segmentation (HSS), which aims to identify the exact positions of the first heart sound(S1), second heart sound(S2), the duration of S1, systole, S2, and diastole within a cardiac cycle of phonocardiogram (PCG), is an indispensable step to find out heart health. Recently, some neural network-based methods for heart sound segmentation have shown good performance.Approach. In this paper, a novel method was proposed for HSS exactly using One-Dimensional Convolution and Bidirectional Long-Short Term Memory neural network with Attention mechanism (C-LSTM-A) by incorporating the 0.5-order smooth Shannon entropy envelope and its instantaneous phase waveform (IPW), and third intrinsic mode function (IMF-3) of PCG signal to reduce the difficulty of neural network learning features.Main results. An average F1-score of 96.85 was achieved in the clinical research dataset (Fuwai Yunnan Cardiovascular Hospital heart sound dataset) and an average F1-score of 95.68 was achieved in 2016 PhysioNet/CinC Challenge dataset using the novel method.Significance. The experimental results show that this method has advantages for normal PCG signals and common pathological PCG signals, and the segmented fundamental heart sound(S1, S2), systole, and diastole signal components are beneficial to the study of subsequent heart sound classification.

sEMG-Based Continuous Hand Action Prediction by Using Key State Transition and Model Pruning

Conventional classification of hand motions and continuous joint angle estimation based on sEMG have been widely studied in recent years. The classification task focuses on discrete motion recognition and shows poor real-time performance, while continuous joint angle estimation evaluates the real-time joint angles by the continuity of the limb. Few researchers have investigated continuous hand action prediction based on hand motion continuity. In our study, we propose the key state transition as a condition for continuous hand action prediction and simulate the prediction process using a sliding window with long-term memory. Firstly, the key state modeled by GMM-HMMs is set as the condition. Then, the sliding window is used to dynamically look for the key state transition. The prediction results are given while finding the key state transition. To extend continuous multigesture action prediction, we use model pruning to improve reusability. Eight subjects participated in the experiment, and the results show that the average accuracy of continuous two-hand actions is 97% with a 70 ms time delay, which is better than LSTM (94.15%, 308 ms) and GRU (93.83%, 300 ms). In supplementary experiments with continuous four-hand actions, over 85% prediction accuracy is achieved with an average time delay of 90 ms.

Identification of decompensation episodes in chronic heart failure patients based solely on heart sounds

Decompensation episodes in chronic heart failure patients frequently result in unplanned outpatient or emergency room visits or even hospitalizations. Early detection of these episodes in their pre-symptomatic phase would likely enable the clinicians to manage this patient cohort with the appropriate modification of medical therapy which would in turn prevent the development of more severe heart failure decompensation thus avoiding the need for heart failure-related hospitalizations. Currently, heart failure worsening is recognized by the clinicians through characteristic changes of heart failure-related symptoms and signs, including the changes in heart sounds. The latter has proven to be largely unreliable as its interpretation is highly subjective and dependent on the clinicians’ skills and preferences. Previous studies have indicated that the algorithms of artificial intelligence are promising in distinguishing the heart sounds of heart failure patients from those of healthy individuals. In this manuscript, we focus on the analysis of heart sounds of chronic heart failure patients in their decompensated and recompensated phase. The data was recorded on 37 patients using two types of electronic stethoscopes. Using a combination of machine learning approaches, we obtained up to 72% classification accuracy between the two phases, which is better than the accuracy of the interpretation by cardiologists, which reached 50%. Our results demonstrate that machine learning algorithms are promising in improving early detection of heart failure decompensation episodes.

A Study on the Association between Korotkoff Sound Signaling and Chronic Heart Failure (CHF) Based on Computer-Assisted Diagnoses

Background

Korotkoff sound (KS) is an important indicator of hypertension when monitoring blood pressure. However, its utility in noninvasive diagnosis of Chronic heart failure (CHF) has rarely been studied.

Purpose

In this study, we proposed a method for signal denoising, segmentation, and feature extraction for KS, and a Bayesian optimization-based support vector machine algorithm for KS classification.

Methods

The acquired KS signal was resampled and denoised to extract 19 energy features, 12 statistical features, 2 entropy features, and 13 Mel Frequency Cepstrum Coefficient (MFCCs) features. A controlled trial based on the VALSAVA maneuver was carried out to investigate the relationship between cardiac function and KS. To classify these feature sets, the K-Nearest Neighbors (KNN), decision tree (DT), Naive Bayes (NB), ensemble (EM) classifiers, and the proposed BO-SVM were employed and evaluated using the accuracy (Acc), sensitivity (Se), specificity (Sp), Precision (Ps), and F1 score (F1).

Results

The ALSAVA maneuver indicated that the KS signal could play an important role in the diagnosis of CHF. Through comparative experiments, it was shown that the best performance of the classifier was obtained by BO-SVM, with Acc (85.0%), Se (85.3%), and Sp (84.6%).

Conclusions

In this study, a method for noise reduction, segmentation, and classification of KS was established. In the measured data set, our method performed well in terms of classification accuracy, sensitivity, and specificity. In light of this, we believed that the methods described in this paper can be applied to the early, noninvasive detection of heart disease as well as a supplementary monitoring technique for the prognosis of patients with CHF.

Design of Abnormal Heart Sound Recognition System Based on HSMM and Deep Neural Network

Introduction

Heart sound signal is an important physiological signal of human body, and the identification and research of heart sound signal is of great significance.

Methods

For abnormal heart sound signal recognition, an abnormal heart sound recognition system, combining hidden semi-Markov models (HSMM) with deep neural networks, is proposed. Firstly, HSMM is used to build a heart sound segmentation model to accurately segment the heart sound signal, and then the segmented heart sound signal is subjected to feature extraction. Finally, the trained deep neural network model is used for recognition.

Results

Compared with other methods, this method has a relatively small amount of input feature data and high accuracy, fast recognition speed.

Discussion

HSMM combined with deep neural network is expected to be deployed on smart mobile devices for telemedicine detection.

The robustness of Random Forest and Support Vector Machine Algorithms to a Faulty Heart Sound Segmentation

Cardiac auscultation is the key exam to screen cardiac diseases both in developed and developing countries. A heart sound auscultation procedure can detect the presence of murmurs and point to a diagnosis, thus it is an important first-line assessment and also cost-effective tool. The design automatic recommendation systems based on heart sound auscultation can play an important role in boosting the accuracy and the pervasiveness of screening tools. One such as step, consists in detecting the fundamental heart sound states, a process known as segmentation. A faulty segmentation or a wrong estimation of the heart rate might result in an incapability of heart sound classifiers to detect abnormal waves, such as murmurs. In the process of understanding the impact of a faulty segmentation, several common heart sound segmentation errors are studied in detail, namely those where the heart rate is badly estimated and those where S1/S2 and Systolic/Diastolic states are swapped in comparison with the ground truth state sequence. From the tested algorithms, support vector machine (SVMs) and random forest (RFs) shown to be more sensitive to a wrong estimation of the heart rate (an expected drop of 6% and 8% on the overall performance, respectively) than to a swap in the state sequence of events (an expected drop of 1.9% and 4.6%, respectively).

CNN-Based Heart Sound Classification with an Imbalance-Compensating Weighted Loss Function

Heart sound auscultation is an effective method for early-stage diagnosis of heart disease. The application of deep neural networks is gaining increasing attention in automated heart sound classification. This paper proposes deep Convolutional Neural Networks (CNNs) to classify normal/abnormal heart sounds, which takes two-dimensional Mel-scale features as input, including Mel frequency cepstral coefficients (MFCCs) and the Log Mel spectrum. We employ two weighted loss functions during the training to mitigate the class imbalance issue. The model was developed on the public PhysioNet/Computing in Cardiology Challenge (CinC) 2016 heart sound database. On the considered test set, the proposed model with Log Mel spectrum as features achieves an Unweighted Average Recall (UAR) of 89.6%, with sensitivity and specificity being 89.5% and 89.7% respectively. This work proposes a CNN-based model to enable automated heart sound classification, which can provide auxiliary assistance for heart auscultation and has the potential to screen for heart pathologies in clinical applications at a relatively low cost.

Automated Quality Assessment for Accelerometer-Based Heart Sounds Recorded with a Novel Subcutaneous Medical Implant

In the context of monitoring patients with heart failure conditions, the automated assessment of heart sound quality is of major importance to insure the relevance of the medical analysis of the heart sound data. We propose in this study a technique of quality classification based on the selection of a small set of representative features. The first features are chosen to characterize whether the periodicity, complexity or statistical nature of the heart sound recordings. After segmentation process, the latter features are probing the detectability of the heart sounds in cardiac cycles. Our method is applied on a novel subcutaneous medical implant that combines ECG and accelerometric-based heart sound measurements. The actual prototype is in pre-clinical phase and has been implanted on 4 pigs, which anatomy and activity constitute a challenging environment for obtaining clean heart sounds. As reference quality labeling, we have performed a three-class manual annotation of each recording, qualified as "good", "unsure" and "bad". Our method allows to retrieve good quality heart sounds with a sensitivity and an accuracy of 82% ± 2% and 88% ± 6% respectively. Clinical Relevance- By accurately recovering high quality heart sound sequences, our method will enable to monitor reliable physiological indicators of heart failure complications such as decompensation.

The CirCor DigiScope Dataset: From Murmur Detection to Murmur Classification

Cardiac auscultation is one of the most cost-effective techniques used to detect and identify many heart conditions. Computer-assisted decision systems based on auscultation can support physicians in their decisions. Unfortunately, the application of such systems in clinical trials is still minimal since most of them only aim to detect the presence of extra or abnormal waves in the phonocardiogram signal, i.e., only a binary ground truth variable (normal vs abnormal) is provided. This is mainly due to the lack of large publicly available datasets, where a more detailed description of such abnormal waves (e.g., cardiac murmurs) exists. To pave the way to more effective research on healthcare recommendation systems based on auscultation, our team has prepared the currently largest pediatric heart sound dataset. A total of 5282 recordings have been collected from the four main auscultation locations of 1568 patients, in the process, 215780 heart sounds have been manually annotated. Furthermore, and for the first time, each cardiac murmur has been manually annotated by an expert annotator according to its timing, shape, pitch, grading, and quality. In addition, the auscultation locations where the murmur is present were identified as well as the auscultation location where the murmur is detected more intensively. Such detailed description for a relatively large number of heart sounds may pave the way for new machine learning algorithms with a real-world application for the detection and analysis of murmur waves for diagnostic purposes.

A multi-scale and multi-domain heart sound feature-based machine learning model for ACC/AHA heart failure stage classification

OBJECTIVE

Heart sounds can reflect detrimental changes in cardiac mechanical activity that are common pathological characteristics of chronic heart failure (CHF). The ACC/AHA heart failure (HF) stage classification is essential for clinical decision-making and the management of CHF. Herein, a machine learning model that makes use of multi-scale and multi-domain heart sound features was proposed to provide an objective aid for ACC/AHA HF stage classification.

APPROACH

A dataset containing phonocardiogram (PCG) signals from 275 subjects was obtained from two medical institutions and used in this study. Complementary ensemble empirical mode decomposition and tunable-Q wavelet transform were used to construct self-adaptive sub-sequences and multi-level sub-band signals for PCG signals. Time-domain, frequency-domain and nonlinear feature extraction were then applied to the original PCG signal, heart sound sub-sequences and sub-band signals to construct multi-scale and multi-domain heart sound features. The features selected via the least absolute shrinkage and selection operator were fed into a machine learning classifier for ACC/AHA HF stage classification. Finally, mainstream machine learning classifiers, including least-squares support vector machine (LS-SVM), deep belief network (DBN) and random forest (RF), were compared to determine the optimal model.

MAIN RESULTS

The results showed that the LS-SVM, which utilized a combination of multi-scale and multi-domain features, achieved better classification performance than the DBN and RF using multi-scale or multi-domain features alone or together, with average sensitivity, specificity, and accuracy of 0.821, 0.955 and 0.820 on the testing set, respectively.

SIGNIFICANCE

PCG signal analysis provides efficient measurement information regarding CHF severity and is a promising noninvasive method for ACC/AHA HF stage classification.

Automatic Classification of Normal–Abnormal Heart Sounds Using Convolution Neural Network and Long-Short Term Memory

The phonocardiogram (PCG) is an important analysis method for the diagnosis of cardiovascular disease, which is usually performed by experienced medical experts. Due to the high ratio of patients to doctors, there is a pressing need for a real-time automated phonocardiogram classification system for the diagnosis of cardiovascular disease. This paper proposes a deep neural-network structure based on a one-dimensional convolutional neural network (1D-CNN) and a long short-term memory network (LSTM), which can directly classify unsegmented PCG to identify abnormal signal. The PCG data were filtered and put into the model for analysis. A total of 3099 pieces of heart-sound recordings were used, while another 100 patients’ heart-sound data collected by our group and diagnosed by doctors were used to test and verify the model. Results show that the CNN-LSTM model provided a good overall balanced accuracy of 0.86 ± 0.01 with a sensitivity of 0.87 ± 0.02, and specificity of 0.89 ± 0.02. The F1-score was 0.91 ± 0.01, and the receiver-operating characteristic (ROC) plot produced an area under the curve (AUC) value of 0.92 ± 0.01. The sensitivity, specificity and accuracy of the 100 patients’ data were 0.83 ± 0.02, 0.80 ± 0.02 and 0.85 ± 0.03, respectively. The proposed model does not require feature engineering and heart-sound segmentation, which possesses reliable performance in classification of abnormal PCG; and is fast and suitable for real-time diagnosis application.

An automatic approach for heart failure typing based on heart sounds and convolutional recurrent neural networks

Heart failure (HF) is a complex clinical syndrome that poses a major hazard to human health. Patients with different types of HF have great differences in pathogenesis and treatment options. Therefore, HF typing is of great significance for timely treatment of patients. In this paper, we proposed an automatic approach for HF typing based on heart sounds (HS) and convolutional recurrent neural networks, which provides a new non-invasive and convenient way for HF typing. Firstly, the collected HS signals were preprocessed with adaptive wavelet denoising. Then, the logistic regression based hidden semi-Markov model was utilized to segment HS frames. For the distinction between normal subjects and the HF patients with preserved ejection fraction or reduced ejection fraction, a model based on convolutional neural network and recurrent neural network was built. The model can automatically learn the spatial and temporal characteristics of HS signals. The results show that the proposed model achieved a superior performance with an accuracy of 97.64%. This study suggests the proposed method could be a useful tool for HF recognition and as a supplement for HF typing.

Transfer Learning Models for Detecting Six Categories of Phonocardiogram Recordings

BACKGROUND AND AIMS

Auscultation is a cheap and fundamental technique for detecting cardiovascular disease effectively. Doctors' abilities in auscultation are varied. Sometimes, there may be cases of misdiagnosis, even when auscultation is performed by an experienced doctor. Hence, it is necessary to propose accurate computational tools to assist auscultation, especially in developing countries. Artificial intelligence technology can be an efficient diagnostic tool for detecting cardiovascular disease. This work proposed an automatic multiple classification method for cardiovascular disease detection by heart sound signals.

METHODS AND RESULTS

In this work, a 1D heart sound signal is translated into its corresponding 3D spectrogram using continuous wavelet transform (CWT). In total, six classes of heart sound data are used in this experiment. We combine an open database (including five classes of heart sound data: aortic stenosis, mitral regurgitation, mitral stenosis, mitral valve prolapse and normal) with one class (pulmonary hypertension) of heart sound data collected by ourselves to perform the experiment. To make the method robust in a noisy environment, the background deformation technique is used before training. Then, 10 transfer learning networks (GoogleNet, SqueezeNet, DarkNet19, MobileNetv2, Inception-ResNetv2, DenseNet201, Inceptionv3, ResNet101, NasNet-Large, and Xception) are used for comparison. Furthermore, other models (LSTM and CNN) are also compared with our proposed algorithm. The experimental results show that four transfer learning networks (ResNet101, DenseNet201, DarkNet19 and GoogleNet) outperformed their peer models with an accuracy of 0.98 to detect the multiple heart diseases. The performances have been validated both in the original heart sound and the augmented heart sound using 10-fold cross validation. The results of these 10 folds are reported in this research.

CONCLUSIONS

Our method obtained high classification accuracy even under a noisy background, which suggests that the proposed classification method could be used in auxiliary diagnosis for cardiovascular diseases.

Hybrid Convolutional Networks for End-to-End Event Detection in Concurrent PPG and PCG Signals Affected by Motion Artifacts

The accurate detection of physiologically-related events in photopletismographic (PPG) and phocardiographic (PCG) signals, recorded by wearable sensors, is mandatory to perform the estimation of relevant cardiovascular parameters like the heart rate and the blood pressure. However, the measurement performed in uncontrolled conditions without clinical supervision leaves the detection quality particularly susceptible to noise and motion artifacts. The performed work proposed a new fully-automatic computational framework, based on convolutional networks, to identify and localize fiducial points in time as the foot, maximum slope and peak in PPG signal and the S1 sound in the PCG signal, both acquired by a custom chest sensor, described recently in the literature by our group. The novelty entailing a custom neural architecture to process sequentially the PPG and PCG signals. Tests were performed analysing four different acquisition conditions (rest, cycling, rest recovery and walking). Cross-validation results for the three PPG fiducial points showed identification accuracy greater than 93 % and localization error (RMSE) less than 10 ms. As expected, cycling and walking conditions provided worse results than rest and recovery, however reaching an accuracy greater than 90 % and a localization error lower than 15 ms. Likewise, the identification and localization error for S1 sound were greater than 90 % and lower than 25 ms. Overall, this study showcased the ability of the proposed technique to detect events with high accuracy not only for steady acquisitions but also during subject movements. We also showed that the proposed network outperformed traditional Shannon-energy-envelope method in the detection of S1 sound. Therefore, we argue that coupling chest sensors and deep learning processing techniques may disclose wearable devices to unobtrusively acquire health information, being less affected by noise and motion artifacts.

Ensembled Transfer Learning and Multiple Kernel Learning for Phonocardiogram based Atherosclerotic Coronary Artery Disease Detection

Conventional machine learning has paved the way for a simple, affordable, non-invasive approach for Coronary artery disease (CAD) detection using phonocardiogram (PCG). It leaves a scope to explore improvement of performance metrics by fusion of learned representations from deep learning. In this study, we propose a novel, multiple kernel learning (MKL) for their fusion using deep embeddings transferred from pre-trained convolutional neural network (CNN). The proposed MKL, finds optimal kernel combination by maximizing the similarity with ideal kernel and minimizing the redundancy with other basis kernels. Experiments are performed on 960 PCG epochs collected from 40 CAD and 40 normal subjects. The transferred embeddings attain maximum subject-level accuracy of 89.25% with kappa of 0.7850. Later, their fusion with handcrafted features using the proposed MKL gives an accuracy of 91.19% and kappa 0.8238. The study shows the potential of development of high accuracy CAD detection system by using easy to acquire, non-invasive PCG signal.

Deep Learning-Based Heart Sound Analysis for Left Ventricular Diastolic Dysfunction Diagnosis

The aggravation of left ventricular diastolic dysfunction (LVDD) could lead to ventricular remodeling, wall stiffness, reduced compliance, and progression to heart failure with a preserved ejection fraction. A non-invasive method based on convolutional neural networks (CNN) and heart sounds (HS) is presented for the early diagnosis of LVDD in this paper. A deep convolutional generative adversarial networks (DCGAN) model-based data augmentation (DA) method was proposed to expand a HS database of LVDD for model training. Firstly, the preprocessing of HS signals was performed using the improved wavelet denoising method. Secondly, the logistic regression based hidden semi-Markov model was utilized to segment HS signals, which were subsequently converted into spectrograms for DA using the short-time Fourier transform (STFT). Finally, the proposed method was compared with VGG-16, VGG-19, ResNet-18, ResNet-50, DenseNet-121, and AlexNet in terms of performance for LVDD diagnosis. The result shows that the proposed method has a reasonable performance with an accuracy of 0.987, a sensitivity of 0.986, and a specificity of 0.988, which proves the effectiveness of HS analysis for the early diagnosis of LVDD and demonstrates that the DCGAN-based DA method could effectively augment HS data.

Robust optimization of convolutional neural networks with a uniform experiment design method: a case of phonocardiogram testing in patients with heart diseases

Background

Heart sound measurement is crucial for analyzing and diagnosing patients with heart diseases. This study employed phonocardiogram signals as the input signal for heart disease analysis due to the accessibility of the respective method. This study referenced preprocessing techniques proposed by other researchers for the conversion of phonocardiogram signals into characteristic images composed using frequency subband. Image recognition was then conducted through the use of convolutional neural networks (CNNs), in order to classify the predicted of phonocardiogram signals as normal or abnormal. However, CNN requires the tuning of multiple hyperparameters, which entails an optimization problem for the hyperparameters in the model. To maximize CNN robustness, the uniform experiment design method and a science-based methodical experiment design were used to optimize CNN hyperparameters in this study.

Results

An artificial intelligence prediction model was constructed using CNN, and the uniform experiment design method was proposed to acquire hyperparameters for optimal CNN robustness. The results indicate Filters (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{1}$$\end{document}X1), Stride (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{3}$$\end{document}X3), Activation functions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{6}$$\end{document}X6), and Dropout (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{7}$$\end{document}X7) to be significant factors considerably influencing the ability of CNN to distinguish among heart sound states. Finally, the confirmation experiment was conducted, and the hyperparameter combination for optimal model robustness was Filters (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{1}$$\end{document}X1) = 32, Kernel Size (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{2})$$\end{document}X2) = 3 × 3, Stride (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{3}$$\end{document}X3) = (1,1), Padding (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{4})$$\end{document}X4) as same, Optimizer (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{5})$$\end{document}X5) as the stochastic gradient descent, Activation functions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{6}$$\end{document}X6) as relu, and Dropout (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X}_{7}$$\end{document}X7) = 0.544. With this combination of parameters, the model had an average prediction accuracy rate of 0.787 and standard deviation of 0.

Conclusion

In this study, phonocardiogram signals were used for the early prediction of heart diseases. The science-based and methodical uniform experiment design was used for the optimization of CNN hyperparameters to construct a CNN with optimal robustness. The results revealed that the constructed model exhibited robustness and an acceptable accuracy rate. Other literature has failed to address hyperparameter optimization problems in CNN; a method is subsequently proposed for robust CNN optimization, thereby solving this problem.

Automatic segmentation for neonatal phonocardiogram

This work addresses the automatic segmentation of neonatal phonocardiogram (PCG) to be used in the artificial intelligence-assisted diagnosis of abnormal heart sounds. The proposed novel algorithm has a single free parameter - the maximum heart rate. The algorithm is compared with the baseline algorithm, which was developed for adult PCG segmentation. When evaluated on a large clinical dataset of neonatal PCG with a total duration of over 7h, an F1 score of 0.94 is achieved. The main features relevant for the segmentation of neonatal PCG are identified and discussed. The algorithm is able to increase the number of cardiac cycles by a factor of 5 compared to manual segmentation, potentially allowing to improve the performance of heart abnormality detection algorithms.

Can Heart Sound Denoising be Beneficial in Phonocardiogram Classification Tasksƒ

The purpose of computer-aided diagnosis (CAD) systems is to improve the detection of diseases in a shorter time and with reduced subjectivity. A robust system frequently requires a noise-free input signal. For CADs which use heart sounds, this problem is critical as heart sounds are often low amplitude and affected by some unavoidable sources of noise such as movement artifacts and physiological sounds. Removing noises by using denoising algorithms can be beneficial in improving the diagnostics accuracy of CADs. In this study, four denoising algorithms were investigated. Each algorithm has been carefully adapted to fit the requirements of the phonocardiograph signal. The effect of the denoising algorithms was objectively compared based on the improvement it introduces in the classification performance of the heart sound dataset. According to the findings, using denoising methods directly before classification decreased the algorithm's classification performance because a murmur was also treated as noise and suppressed by the denoising process. However, when denoising using Wiener estimation-based spectral subtraction was used as a preprocessing step to improve the segmentation algorithm, it increased the system's classification performance with a sensitivity of 96.0%, a specificity of 74.0%, and an overall score of 85.0%. As a result, to improve performance, denoising can be added as a preprocessing step into heart sound classifiers that are based on heart sound segmentation.

Toward Automated Analysis of Fetal Phonocardiograms: Comparing Heartbeat Detection from Fetal Doppler and Digital Stethoscope Signals

Longitudinal fetal health monitoring is essential for high-risk pregnancies. Heart rate and heart rate variability are prime indicators of fetal health. In this work, we implemented two neural network architectures for heartbeat detection on a set of fetal phonocardiogram signals captured using fetal Doppler and a digital stethoscope. We test the efficacy of these networks using the raw signals and the hand-crafted energy from the signal. The results show a Convolutional Neural Network is the most efficient at identifying the S1 waveforms in a heartbeat, and its performance is improved when using the energy of the Doppler signals. We further discuss issues, such as low Signal-to-Noise Ratios (SNR), present in the training of a model based on the stethoscope signals. Finally, we show that we can improve the SNR, and subsequently the performance of the stethoscope, by matching the energy from the stethoscope to that of the Doppler signal.

Unsegmented Heart Sound Classification Using Hybrid CNN-LSTM Neural Networks

Cardiac Auscultation, an integral part of the physical examination of a patient, is essential for early diagnosis of cardiovascular diseases (CVDs). The ability to accurately diagnose the heart sounds requires experience and expertise, which is lacking in doctors in the early years of clinical practice. Thus, there is a need for an automatic diagnostic tool that would aid medical practitioners with their diagnosis. We propose novel hybrid architectures for classification of unsegmented heart sounds to normal and abnormal classes. We propose two methods, with and without the conventional feature extraction step in the classification pipeline. We demonstrate that the F score using the approach with conventional feature extraction is 1.25 (absolute) more than using a baseline implementation on the Physionet dataset. We also introduce a mechanism to tag predictions as unsure and compare results with a varying threshold.

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.

Vascular Auscultation of Carotid Artery: Towards Biometric Identification and Verification of Individuals

BACKGROUND

Biometric sensing is a security method for protecting information and property. State-of-the-art biometric traits are behavioral and physiological in nature. However, they are vulnerable to tampering and forgery.

METHODS

The proposed approach uses blood flow sounds in the carotid artery as a source of biometric information. A handheld sensing device and an associated desktop application were built. Between 80 and 160 carotid recordings of 11 s in length were acquired from seven individuals each. Wavelet-based signal analysis was performed to assess the potential for biometric applications.

RESULTS

The acquired signals per individual proved to be consistent within one carotid sound recording and between multiple recordings spaced by several weeks. The averaged continuous wavelet transform spectra for all cardiac cycles of one recording showed specific spectral characteristics in the time-frequency domain, allowing for the discrimination of individuals, which could potentially serve as an individual fingerprint of the carotid sound. This is also supported by the quantitative analysis consisting of a small convolutional neural network, which was able to differentiate between different users with over 95% accuracy.

CONCLUSION

The proposed approach and processing pipeline appeared promising for the discrimination of individuals. The biometrical recognition could clinically be used to obtain and highlight differences from a previously established personalized audio profile and subsequently could provide information on the source of the deviation as well as on its effects on the individual's health. The limited number of individuals and recordings require a study in a larger population along with an investigation of the long-term spectral stability of carotid sounds to assess its potential as a biometric marker. Nevertheless, the approach opens the perspective for automatic feature extraction and classification.

Hemodynamics-driven mathematical model of first and second heart sound generation

We propose a novel, two-degree of freedom mathematical model of mechanical vibrations of the heart that generates heart sounds in CircAdapt, a complete real-time model of the cardiovascular system. Heart sounds during rest, exercise, biventricular (BiVHF), left ventricular (LVHF) and right ventricular heart failure (RVHF) were simulated to examine model functionality in various conditions. Simulated and experimental heart sound components showed both qualitative and quantitative agreements in terms of heart sound morphology, frequency, and timing. Rate of left ventricular pressure (LV dp/dt_max) and first heart sound (S1) amplitude were proportional with exercise level. The relation of the second heart sound (S2) amplitude with exercise level was less significant. BiVHF resulted in amplitude reduction of S1. LVHF resulted in reverse splitting of S2 and an amplitude reduction of only the left-sided heart sound components, whereas RVHF resulted in a prolonged splitting of S2 and only a mild amplitude reduction of the right-sided heart sound components. In conclusion, our hemodynamics-driven mathematical model provides fast and realistic simulations of heart sounds under various conditions and may be helpful to find new indicators for diagnosis and prognosis of cardiac diseases.

Among various vital signals used for diagnosis and prognosis of cardiac diseases, heart sounds are not employed precisely because physicians subjectively assess their auscultatory findings. On the other hand, recorded heart sounds are also difficult to quantitatively relate to different cardiac conditions given the complex nature of their generation. We therefore employed cardiovascular modeling and developed a novel hemodynamics-driven mathematical model for heart sound generation to unravel the relationships between heart sounds and other vital signals. Simulated and experimental heart sound components showed qualitative and quantitative agreements in terms of heart sound morphology, frequency, and timing, not only during normal conditions, but also during simulated exercise and heart failure. Our model can be used to understand generation of heart sounds in more details and may be helpful to find new diagnostic indicators and treatment methods of cardiac disorders.

Discrimination of Patients with Varying Degrees of Coronary Artery Stenosis by ECG and PCG Signals Based on Entropy

Coronary heart disease (CHD) is the leading cause of cardiovascular death. This study aimed to propose an effective method for mining cardiac mechano-electric coupling information and to evaluate its ability to distinguish patients with varying degrees of coronary artery stenosis (VDCAS). Five minutes of electrocardiogram and phonocardiogram signals was collected synchronously from 191 VDCAS patients to construct heartbeat interval (RRI)-systolic time interval (STI), RRI-diastolic time interval (DTI), HR-corrected QT interval (QTcI)-STI, QTcI-DTI, Tpeak-Tend interval (TpeI)-STI, TpeI-DTI, Tpe/QT interval (Tpe/QTI)-STI, and Tpe/QTI-DTI series. Then, the cross sample entropy (XSampEn), cross fuzzy entropy (XFuzzyEn), joint distribution entropy (JDistEn), magnitude-squared coherence function, cross power spectral density, and mutual information were applied to evaluate the coupling of the series. Subsequently, support vector machine recursive feature elimination and XGBoost were utilized for feature selection and classification, respectively. Results showed that the joint analysis of XSampEn, XFuzzyEn, and JDistEn had the best ability to distinguish patients with VDCAS. The classification accuracy of severe CHD-mild-to-moderate CHD group, severe CHD-chest pain and normal coronary angiography (CPNCA) group, and mild-to-moderate CHD-CPNCA group were 0.8043, 0.7659, and 0.7500, respectively. The study indicates that the joint analysis of XSampEn, XFuzzyEn, and JDistEn can effectively capture the cardiac mechano-electric coupling information of patients with VDCAS, which can provide valuable information for clinicians to diagnose CHD.

Deep Learning Methods for Heart Sounds Classification: A Systematic Review

The automated classification of heart sounds plays a significant role in the diagnosis of cardiovascular diseases (CVDs). With the recent introduction of medical big data and artificial intelligence technology, there has been an increased focus on the development of deep learning approaches for heart sound classification. However, despite significant achievements in this field, there are still limitations due to insufficient data, inefficient training, and the unavailability of effective models. With the aim of improving the accuracy of heart sounds classification, an in-depth systematic review and an analysis of existing deep learning methods were performed in the present study, with an emphasis on the convolutional neural network (CNN) and recurrent neural network (RNN) methods developed over the last five years. This paper also discusses the challenges and expected future trends in the application of deep learning to heart sounds classification with the objective of providing an essential reference for further study.

Detection of Coronary Artery Disease Using Multi-Domain Feature Fusion of Multi-Channel Heart Sound Signals

Heart sound signals reflect valuable information about heart condition. Previous studies have suggested that the information contained in single-channel heart sound signals can be used to detect coronary artery disease (CAD). But accuracy based on single-channel heart sound signal is not satisfactory. This paper proposed a method based on multi-domain feature fusion of multi-channel heart sound signals, in which entropy features and cross entropy features are also included. A total of 36 subjects enrolled in the data collection, including 21 CAD patients and 15 non-CAD subjects. For each subject, five-channel heart sound signals were recorded synchronously for 5 min. After data segmentation and quality evaluation, 553 samples were left in the CAD group and 438 samples in the non-CAD group. The time-domain, frequency-domain, entropy, and cross entropy features were extracted. After feature selection, the optimal feature set was fed into the support vector machine for classification. The results showed that from single-channel to multi-channel, the classification accuracy has increased from 78.75% to 86.70%. After adding entropy features and cross entropy features, the classification accuracy continued to increase to 90.92%. The study indicated that the method based on multi-domain feature fusion of multi-channel heart sound signals could provide more information for CAD detection, and entropy features and cross entropy features played an important role in it.

Cross-Domain Transfer Learning for PCG Diagnosis Algorithm

Cardiechema is a way to reflect cardiovascular disease where the doctor uses a stethoscope to help determine the heart condition with a sound map. In this paper, phonocardiogram (PCG) is used as a diagnostic signal, and a deep learning diagnostic framework is proposed. By improving the architecture and modules, a new transfer learning and boosting architecture is mainly employed. In addition, a segmentation method is designed to improve on the existing signal segmentation methods, such as R wave to R wave interval segmentation and fixed segmentation. For the evaluation, the final diagnostic architecture achieved a sustainable performance with a public PCG database.

End-to-end heart sound segmentation using deep convolutional recurrent network

Heart sound segmentation (HSS) aims to detect the four stages (first sound, systole, second heart sound and diastole) from a heart cycle in a phonocardiogram (PCG), which is an essential step in automatic auscultation analysis. Traditional HSS methods need to manually extract the features before dealing with HSS tasks. These artificial features highly rely on extraction algorithms, which often result in poor performance due to the different operating environments. In addition, the high-dimension and frequency characteristics of audio also challenge the traditional methods in effectively addressing HSS tasks. This paper presents a novel end-to-end method based on convolutional long short-term memory (CLSTM), which directly uses audio recording as input to address HSS tasks. Particularly, the convolutional layers are designed to extract the meaningful features and perform the downsampling, and the LSTM layers are developed to conduct the sequence recognition. Both components collectively improve the robustness and adaptability in processing the HSS tasks. Furthermore, the proposed CLSTM algorithm is easily extended to other complex heart sound annotation tasks, as it does not need to extract the characteristics of corresponding tasks in advance. In addition, the proposed algorithm can also be regarded as a powerful feature extraction tool, which can be integrated into the existing models for HSS. Experimental results on real-world PCG datasets, through comparisons to peer competitors, demonstrate the outstanding performance of the proposed algorithm.

Heart Sound Analysis in Individuals Supported With Left Ventricular Assist Devices

OBJECTIVE

LVADs are surgically implanted mechanical pumps that improve survival rates of individuals with advanced heart failure. LVAD therapy is associated with high morbidity, which can be partially attributed to challenges with detecting LVAD complications before adverse events occur. Current methods used to monitor for complications with LVAD support require frequent clinical assessments at specialized LVAD centers. Analysis of recorded precordial sounds may enable real-time, remote monitoring of device and cardiac function for early detection of LVAD complications. The dominance of LVAD sounds in the precordium limits the utility of routine cardiac auscultation of LVAD recipients. In this work, we develop a signal processing pipeline to mitigate sounds generated by the LVAD.

METHODS

We collected in vivo precordial sounds from 17 LVAD recipients, and contemporaneous echocardiograms from 12 of these individuals, to validate heart valve closure timings.

RESULTS

We characterized various acoustic signatures of heart sounds extracted from in vivo recordings, and report preliminary findings linking fundamental heart sound characteristics and level of LVAD support.

CONCLUSION

Mitigation of LVAD sounds from precordial sound recordings of LVAD recipients enables analysis of intrinsic heart sounds.

SIGNIFICANCE

These findings provide proof-of-concept evidence of the clinical utility of heart sound analysis for bedside and remote monitoring of LVAD recipients.

A Framework for AI-Assisted Detection of Patent Ductus Arteriosus from Neonatal Phonocardiogram

The current diagnosis of Congenital Heart Disease (CHD) in neonates relies on echocardiography. Its limited availability requires alternative screening procedures to prioritise newborns awaiting ultrasound. The routine screening for CHD is performed using a multidimensional clinical examination including (but not limited to) auscultation and pulse oximetry. While auscultation might be subjective with some heart abnormalities not always audible it increases the ability to detect heart defects. This work aims at developing an objective clinical decision support tool based on machine learning (ML) to facilitate differentiation of sounds with signatures of Patent Ductus Arteriosus (PDA)/CHDs, in clinical settings. The heart sounds are pre-processed and segmented, followed by feature extraction. The features are fed into a boosted decision tree classifier to estimate the probability of PDA or CHDs. Several mechanisms to combine information from different auscultation points, as well as consecutive sound cycles, are presented. The system is evaluated on a large clinical dataset of heart sounds from 265 term and late-preterm newborns recorded within the first six days of life. The developed system reaches an area under the curve (AUC) of 78% at detecting CHD and 77% at detecting PDA. The obtained results for PDA detection compare favourably with the level of accuracy achieved by an experienced neonatologist when assessed on the same cohort.

Contactless analysis of heart rate variability during cold pressor test using radar interferometry and bidirectional LSTM networks

Contactless measurement of heart rate variability (HRV), which reflects changes of the autonomic nervous system (ANS) and provides crucial information on the health status of a person, would provide great benefits for both patients and doctors during prevention and aftercare. However, gold standard devices to record the HRV, such as the electrocardiograph, have the common disadvantage that they need permanent skin contact with the patient. Being connected to a monitoring device by cable reduces the mobility, comfort, and compliance by patients. Here, we present a contactless approach using a 24 GHz Six-Port-based radar system and an LSTM network for radar heart sound segmentation. The best scores are obtained using a two-layer bidirectional LSTM architecture. To verify the performance of the proposed system not only in a static measurement scenario but also during a dynamic change of HRV parameters, a stimulation of the ANS through a cold pressor test is integrated in the study design. A total of 638 minutes of data is gathered from 25 test subjects and is analysed extensively. High F-scores of over 95% are achieved for heartbeat detection. HRV indices such as HF norm are extracted with relative errors around 5%. Our proposed approach is capable to perform contactless and convenient HRV monitoring and is therefore suitable for long-term recordings in clinical environments and home-care scenarios.

Research on Segmentation and Classification of Heart Sound Signals Based on Deep Learning

The heart sound signal is one of the signals that reflect the health of the heart. Research on the heart sound signal contributes to the early diagnosis and prevention of cardiovascular diseases. As a commonly used deep learning network, convolutional neural network (CNN) has been widely used in images. In this paper, the method of analyzing heart sound through using CNN has been studied. Firstly, the original data set was preprocessed, and then the heart sounds were segmented on U-net, based on the deep CNN. Finally, the classification of heart sounds was completed through CNN. The data from 2016 PhysioNet/CinC Challenge was utilized for algorithm validation, and the following results were obtained. When the heart sound segmented, the overall accuracy rate was 0.991, the accuracy of the first heart sound was 0.991, the accuracy of the systolic period was 0.996, the accuracy of the second heart sound was 0.996, and the accuracy of the diastolic period was 0.997, and the average accuracy rate was 0.995; While in classification, the accuracy was 0.964, the sensitivity was 0.781, and the specificity was 0.873. These results show that deep learning based on CNN shows good performance in the segmentation and classification of the heart sound signal.

Neonatal Heart and Lung Sound Quality Assessment for Robust Heart and Breathing Rate Estimation for Telehealth Applications

With advances in digital stethoscopes, internet of things, signal processing and machine learning, chest sounds can be easily collected and transmitted to the cloud for remote monitoring and diagnosis. However, low quality of recordings complicates remote monitoring and diagnosis, particularly for neonatal care. This paper proposes a new method to objectively and automatically assess the signal quality to improve the accuracy and reliability of heart rate (HR) and breathing rate (BR) estimation from noisy neonatal chest sounds. A total of 88 10-second long chest sounds were taken from 76 preterm and full-term babies. Six annotators independently assessed the signal quality, number of detectable beats, and breathing periods from these recordings. For quality classification, 187 and 182 features were extracted from heart and lung sounds, respectively. After feature selection, class balancing, and hyperparameter optimization, a dynamic binary classification model was trained. Then HR and BR were automatically estimated from the chest sound and several approaches were compared.The results of subject-wise leave-one-out cross-validation, showed that the model distinguished high and low quality recordings in the test set with 96% specificity, 81% sensitivity and 93% accuracy for heart sounds, and 86% specificity, 69% sensitivity and 82% accuracy for lung sounds. The HR and BR estimated from high quality sounds resulted in significantly less median absolute error (4 bpm and 12 bpm difference, respectively) compared to those from low quality sounds. The methods presented in this work, facilitates automated neonatal chest sound auscultation for future telehealth applications.

Deep Layer Kernel Sparse Representation Network for the Detection of Heart Valve Ailments from the Time-Frequency Representation of PCG Recordings

The heart valve ailments (HVAs) are due to the defects in the valves of the heart and if untreated may cause heart failure, clots, and even sudden cardiac death. Automated early detection of HVAs is necessary in the hospitals for proper diagnosis of pathological cases, to provide timely treatment, and to reduce the mortality rate. The heart valve abnormalities will alter the heart sound and murmurs which can be faithfully captured by phonocardiogram (PCG) recordings. In this paper, a time-frequency based deep layer kernel sparse representation network (DLKSRN) is proposed for the detection of various HVAs using PCG signals. Spline kernel-based Chirplet transform (SCT) is used to evaluate the time-frequency representation of PCG recording, and the features like L1-norm (LN), sample entropy (SEN), and permutation entropy (PEN) are extracted from the different frequency components of the time-frequency representation of PCG recording. The DLKSRN formulated using the hidden layers of extreme learning machine- (ELM-) autoencoders and kernel sparse representation (KSR) is used for the classification of PCG recordings as normal, and pathology cases such as mitral valve prolapse (MVP), mitral regurgitation (MR), aortic stenosis (AS), and mitral stenosis (MS). The proposed approach has been evaluated using PCG recordings from both public and private databases, and the results demonstrated that an average sensitivity of 100%, 97.51%, 99.00%, 98.72%, and 99.13% are obtained for normal, MVP, MR, AS, and MS cases using the hold-out cross-validation (CV) method. The proposed approach is applicable for the Internet of Things- (IoT-) driven smart healthcare system for the accurate detection of HVAs.

Detection of coronary artery disease using multi-modal feature fusion and hybrid feature selection

Objective: Coronary artery disease (CAD) is a common fatal disease. At present, an accurate method to screen CAD is urgently needed. This study aims to provide optimal detection models for suspected CAD detection according to the differences in medical conditions, so as to assist physicians to make accurate judgments on suspected CAD patients.Approach: Electrocardiogram (ECG) and phonocardiogram (PCG) signals of 32 CAD patients and 30 patients with chest pain and normal coronary angiograms (CPNCA) were simultaneously collected for this paper. For each subject, the ECG and PCG multi-domain features were extracted, and the results of Holter monitoring, echocardiography (ECHO), and biomarker levels (BIO) were obtained to construct a multi-modal feature set. Then, a hybrid feature selection (HFS) method was developed using mutual information, recursive feature elimination, random forest, and weight of support vector machine to obtain the optimal feature subset. A support vector machine with nested cross-validation was used for classification.Main results: Results showed that the Holter model achieved the best performance as a single-modal feature model with an accuracy of 82.67%. In terms of multi-modal feature models, PCG-Holter, PCG-Holter-ECHO, PCG-Holter-ECHO-BIO, and ECG-PCG-Holter-ECHO-BIO were the optimal bimodal, three-modal, four-modal, and five-modal models, with accuracies of 90.38%, 91.92%, 95.25%, and 96.67%, respectively. Among them, the ECG-PCG-Holter-ECHO-BIO model, which was constructed by combining ECG and PCG signals features with Holter, ECHO, and BIO examination results, achieved the best classification results with an average accuracy, sensitivity, specificity, and F1-measure of 96.67%, 96.67%, 96.67%, and 96.64%, respectively.Significance: The study indicated that multi-modal feature fusion and HFS can obtain more effective information for CAD detection and provide a reference for physicians to diagnose CAD patients.

Temporal Convolutional Network Connected with an Anti-Arrhythmia Hidden Semi-Markov Model for Heart Sound Segmentation

Heart sound segmentation (HSS) is a critical step in heart sound processing, where it improves the interpretability of heart sound disease classification algorithms. In this study, we aimed to develop a real-time algorithm for HSS by combining the temporal convolutional network (TCN) and the hidden semi-Markov model (HSMM), and improve the performance of HSMM for heart sounds with arrhythmias. We experimented with TCN and determined the best parameters based on spectral features, envelopes, and one-dimensional CNN. However, the TCN results could contradict the natural fixed order of S1-systolic-S2-diastolic of heart sound, and thereby the Viterbi algorithm based on HSMM was connected to correct the order errors. On this basis, we improved the performance of the Viterbi algorithm when detecting heart sounds with cardiac arrhythmias by changing the distribution and weights of the state duration probabilities. The public PhysioNet Computing in Cardiology Challenge 2016 data set was employed to evaluate the performance of the proposed algorithm. The proposed algorithm achieved an F1 score of 97.02%, and this result was comparable with the current state-of-the-art segmentation algorithms. In addition, the proposed enhanced Viterbi algorithm for HSMM corrected 30 out of 30 arrhythmia errors after checking one by one in the dataset.

Unsupervised hidden semi-Markov model for automatic beat onset detection in 1D Doppler ultrasound

OBJECTIVE

One dimensional (1D) Doppler ultrasound (DUS) is commonly used for fetal health assessment, during both regular prenatal visits and labor. It is used in preference to ECG and other modalities because of its simplicity and cost. To date, all analysis of such data has been confined to a smoothed, windowed heart rate estimation derived from the 1D DUS signal, reducing the potential of short-term variability information. A first step in improving the assessment of short-term variability of the fetal heart rate (FHR) is through implementing an accurate beat detector for 1D DUS signals.

APPROACH

This work presents an unsupervised probabilistic segmentation method enabled by a hidden semi-Markov model (HSMM). The proposed method employs envelope and spectral features for an online segmentation of fetal 1D DUS signal. The beat onsets and fetal cardiac beat-to-beat intervals are then estimated from the segmentations. For this work, two data sets were used, including 1D DUS recordings from five fetuses recorded in Germany, comprising 6521 beats and 45.06 minutes of data (dataset 1). Simultaneous fetal ECG (fECG) was used as the reference for beat timing. Dataset 2, comprising 4044 beats captured from 17 subjects in the UK was hand scored for beat location and was used as an independent held-out test set. Leave-one-out subject cross-validation was used for parameter tuning on dataset 1. No retraining was performed for dataset 2. To assess the performance of the beat onset detection, the root mean square error (RMSE), F1 score, sensitivity, positive predictivity (PPV) and the error in several standard common heart rate variability metrics were used. These metrics were evaluated on three fiducial points: (1) beat onset, (2) beat offset, and (3) middle of beat interval.

MAIN RESULTS

In dataset 1, the proposed method provided an RMSE of 20 ms, F1 score of 97.5 %, a Se of 97.6%, and a PPV of 97.3%. In dataset 2, the proposed method achieved an RMSE of 26 ms, an F1 score of 98.5 %, a Se of 98.0 % and a PPV of 98.9 %. It was also determined that the best beat-to-beat interval was derived from the onset of each beat. For the dataset 2, significant correlations were found in all short term heart rate variability metrics tested, both in the time and frequency domain. Only the proportion of successive normal-to-normal interval differences greater than 20 ms (pNN20) exhibited a significant absolute difference.

SIGNIFICANCE

This work presents the first-ever description of an algorithm to identify cardiac beats with 1D DUS, closely matching the fetal ECG-derived beats, to enable short-term heart rate variability analysis. The novel algorithm proposed requires no human labeling of data, and could have applicability beyond 1D DUS to other similar highly variable time series.

A dataset of clinically recorded radar vital signs with synchronised reference sensor signals

Using Radar it is possible to measure vital signs through clothing or a mattress from the distance. This allows for a very comfortable way of continuous monitoring in hospitals or home environments. The dataset presented in this article consists of 24 h of synchronised data from a radar and a reference device. The implemented continuous wave radar system is based on the Six-Port technology and operates at 24 GHz in the ISM band. The reference device simultaneously measures electrocardiogram, impedance cardiogram and non-invasive continuous blood pressure. 30 healthy subjects were measured by physicians according to a predefined protocol. The radar was focused on the chest while the subjects were lying on a tilt table wired to the reference monitoring device. In this manner five scenarios were conducted, the majority of them aimed to trigger hemodynamics and the autonomic nervous system of the subjects. Using the database, algorithms for respiratory or cardiovascular analysis can be developed and a better understanding of the characteristics of the radar-recorded vital signs can be gained.

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.