Heart Sound Classification Based on Feature Analysis and Selection

The stethoscope is a commonly used medical device for diagnosing a patient’s condition by listening to their heartbeat. This diagnostic technique is known as auscultation, where each heart sound heard through the stethoscope has a distinct pattern that depends on the person’s heart condition. Researchers have developed computational methods that automatically analyze heart sounds to overcome the subjective nature of auscultation. In this study, we utilized the Hjorth method for feature extraction and the Long Short-Term RNN algorithm for classifying the normal and abnormal heart sounds. With 100 epochs, two layers of LSTM-RNN, and one layer of Dense, the classification accuracy for distinguishing normal and abnormal sounds reached 71.95%. This research aims to contribute to the development of an accurate system for detecting normal and abnormal heart sounds.

We purpose a novel method that combines modified frequency slice wavelet transform (MFSWT) and convolutional neural network (CNN) for classifying normal and abnormal heart sounds. A hidden Markov model is used to find the position of each cardiac cycle in the heart sound signal and determine the exact position of the four periods of S1, S2, systole, and diastole. Then the one-dimensional cardiac cycle signal was converted into a two-dimensional time-frequency picture using the MFSWT. Finally, two CNN models are trained using the aforementioned pictures. We combine two CNN models using sample entropy (SampEn) to determine which model is used to classify the heart sound signal. We evaluated our model on the heart sound public dataset provided by the PhysioNet Computing in Cardiology Challenge 2016. Experimental classification performance from a 10-fold cross-validation indicated that sensitivity (Se), specificity (Sp) and mean accuracy (MAcc) were 0.95, 0.93, and 0.94, respectively. The results showed the proposed method can classify normal and abnormal heart sounds with efficiency and high accuracy. Graphical abstract Block diagram of heart sound classification.

Auscultation of heart sounds is very important to identify cardiovascular diseases (CVD). In this paper new approach for classification of heart sounds is introduced. This approach depends on getting features from heart sounds using statistical calculations of coefficients of new mother wavelet transform. Then, these features are classified using support vector machine (SVM). Number of features is 40 features. Number of heart sounds used for training is 90 heart sounds. Number of heart sounds used for tested data is 64 heart sounds. The obtained accuracy percent is 92.29%. The obtained specificity is 95.38%. The obtained sensitivity is 90%. So, this new tool can help physicians to diagnose patients of CVD easily.

Heart hemodynamic status and detection of a cardiovascular disease can be evaluated by analyzing and visualizing the heart waveform through graphs called the Phonocardiogram (PCG). The normal sounds of the heart generate signals that are in the audible frequency range of the human ear. Due to the significance of cardiac auscultation for recognizing pathological cardiac status, there has been special interest in automating the classification of heart sounds in the past years. The objective of this research is to present an automatic classification algorithm for anomaly (normal vs. abnormal heart status) of PCG recordings. For this purpose, distinctive time and frequency features are extracted out of heart sound signals to build a learning model using random forest. The accuracy of the proposed algorithm is about 12% better than state of the art.

Analysis of heart sound signals for automatic segmentation and classification has revealed in recent decades that it has the potential to detect pathology accurately in clinical applications. Various audio signal processing techniques have been used to reduce the subjectivity of heart sound analysis. This study aims to classify normal and abnormal heart sound signals. The feature extraction process was optimized by EMD and calculated using five first-order statistical parameters: mean, variance, kurtosis, skewness, and entropy. The classification system is optimized with a mutual information algorithm to select traits that can significantly improve system performance. In addition, the selection of the optimal system configuration also includes the k-fold cross-validation and kNN methods with k values ​​and the proper distance type. Based on the test results, the highest accuracy of 98.2% was obtained when the value of k = 1 and the type of cosine distance on kNN with a five-fold cross-validation system evaluation model.

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

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

Many pathological conditions of the cardiovascular system cause murmurs and aberrations in heart sounds. Phonocardiography provides the clinician with a complementary tool to record the heart sounds heard during auscultation. The advancement of intracardiac phonocardiography combined with modern digital signal processing techniques has strongly renewed researchers' interest in studying heart sounds and murmurs. The aim of this work is to investigate the applicability of different spectral analysis methods to heart sound signals and explore their suitability for PDA-based implementation. Fourier transform (FT), short-time Fourier transform (STFT) and wavelet transform (WT) are used to perform spectral analysis on heart sounds. A segmentation algorithm based on Shannon energy is used to differentiate between first and second heart sounds. Then wavelet transform is deployed again to extract 64 features of heart sounds. The FT provides valuable frequency information but the timing information is lost during the transformation process. The STFT or spectrogram provides valuable time-frequency information but there is a trade-off between time and frequency resolution. Wavelet analysis, however, does not suffer from limitations of the STFT and provides adequate time and frequency resolution to accurately characterize the normal and pathological heart sounds. The results show that the wavelet-based segmentation algorithm is quite effective in localizing the important components of both normal and abnormal heart sounds. They also demonstrate that wavelet-based feature extraction provides suitable feature vectors which are clearly differentiable and useful for automatic classification of heart sounds.

Heart sounds are critical for early detection of cardiovascular diseases, yet existing studies mostly focus on traditional signal segmentation, feature extraction, and shallow classifiers, which often fail to sufficiently capture the dynamic and nonlinear characteristics of heart sounds, limit recognition of complex heart sound patterns, and are sensitive to data imbalance, resulting in poor classification performance. To address these limitations, this study proposes a novel heart sound classification method that integrates improved Mel-frequency cepstral coefficients (MFCC) for feature extraction with a convolutional neural network (CNN) and a deep Transformer model. In the preprocessing stage, a Butterworth filter is applied for denoising, and continuous heart sound signals are directly processed without segmenting the cardiac cycles, allowing the improved MFCC features to better capture dynamic characteristics. These features are then fed into a CNN for feature learning, followed by global average pooling (GAP) to reduce model complexity and mitigate overfitting. Lastly, a deep Transformer module is employed to further extract and fuse features, completing the heart sound classification. To handle data imbalance, the model uses focal loss as the objective function. Experiments on two public datasets demonstrate that the proposed method performs effectively in both binary and multi-class classification tasks. This approach enables efficient classification of continuous heart sound signals, provides a reference methodology for future heart sound research for disease classification, and supports the development of wearable devices and home monitoring systems.

The ability of clinicians, ie, 10 veterinary students, 10 general practitioners, and 10 board certified internists, to describe and interpret common normal and abnormal heart sounds was assessed. Recordings of heart sounds from 7 horses with a variety of normal and abnormal rhythms, heart sounds, and murmurs were analyzed by digital sonography. The perception of the presence or absence of the heart sounds S1, S2, and S4 was similar for clinicians irrespective of their level of training and was in agreement with the sonographic interpretation on 89, 82, and 78% of occasions, respectively. However, practitioners were less likely to correctly describe the presence of S3. The heart rhythm was correctly described as being regular or irregular on 89% of occasions, and this outcome was not affected by level of training. Differentiation of the type of irregularity was less reliable. The perception of the intensity of a heart murmur was accurate and correlated with the grade assigned in the living horses,R2= .68, and with sonographic measurements of the murmur's intensity,R2= .69. Clinicians overestimated the duration of cardiac murmurs, particularly that of the loud systolic murmur. Only diplomates could reliably differentiate systolic from diastolic murmurs. The ability to diagnose the underlying cardiac problem was significantly affected by training; diplomates, practitioners, and undergraduates made the correct diagnosis on 53, 33, and 29% of occasions, respectively. The poor diagnostic ability of practitioners and the lack of improvement in diagnostic skill after the 2nd year of veterinary school emphasizes the need for better teaching of these skills. Digital sonograms that combine sound files with synchronous visual interpretations may be useful in this regard.

A design policy was established for a specific data flow and learning method for the automatic heart-sound diagnosis system under development. The production of each part becomes possible, and auscultation and learning begin. It can be used over clothes as long as it is applied well to the skin surface of the chest. It would be nice to be able to set multiple auscultation positions, but there is a limit to what ordinary people can be asked to do, so this should be considered while having AI learn. We analyzed normal heart sounds to explore learning strategies. Sounds I and II are considered to be important anchor information sources for identifying other heart sounds. Abnormal heart sounds may not be heard at every beat and the rhythm may be abnormal. AI refers to the multiple beats of heart sounds during auscultation. Heartbeat analysis is a multidimensional information analysis related to time and space, and heart sounds are factored if normal and abnormal heart sounds can be organized based on the score. For pitch that tends to depend on individuals and devices, a relative discussion would be more appropriate.

CardioPHON: Quality assessment and self-supervised pretraining for screening of cardiac function based on phonocardiogram recordings

In this study, classification of Normal and Extra systolic heart sounds (HS) have been carried out using in PASCAL Heart Sounds (HS) data base. The extrasystole is the HS that is produced by performing an extra beat in each heart cycle, unlike the heartbeat normal cycle. It can be felt by people as palpitations. Occurrence of these sounds in certain age groups may be the indication of tachycardia. In this study, firstly HS have been normalized at first. Then an elliptic filter has used for noise reduction. HS features have been obtained using Mel-Frequency Cepstrum Coefficients. These features have classified using Artificial Neural Network. In this study, 45 extra systoles heart sounds have used. 30 of them have been used as training data for classification while remaining 15 ones have been used for the test. Certainty, sensitivity, accuracy values have been calculated using confusion matrix. Classification success has been calculated as 90%.

Heart sounds classification plays an important role in cardiovascular disease detection. Currently, deep learning methods for heart sound classification with heavy parameters consumption cannot be deployed in environments with limited memory and computational budgets. Besides, de-noising of heart sound signals (HSSs) can affect accuracy of heart sound classification, because erroneous removal of meaningful components may lead to heart sound distortion. In this paper, an automated heart sound classification method using adaptive wavelet threshold and 1D LDCNN (One-dimensional Lightweight Deep Convolutional Neural Network) is proposed. In this method, we exploit WT (Wavelet Transform) with an adaptive threshold to de-noise heart sound signals (HSSs). Furthermore, we utilize 1D LDCNN to realize automatic feature extraction and classification for de-noised heart sounds. Experiments on PhysioNet/CinC 2016 show that our proposed method achieves the superior classification results and excels in consumption of parameter comparing to state-of-the-art methods.

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.