What are the four databases that have been placed exclusively in the training or test sets?

there are four collected databases that have been semi-randomly placed exclusively in either the training or test sets (to ensure there are ‘novel’ recording types and to reduce over-fitting on the recording methods).

What is the common step in the analysis of heart sounds?

Automated analysis of the heart sound in clinical applications usually consists of three steps shown in figure 3; pre-processing, segmentation and classification.

What are the weights for calculating Se?

The modified sensitivity (Se) and specificity (Sp) are defined as:where wa1 and wa2 are the percentages of good signal quality and poor signal quality recordings in all abnormal recordings respectively, and are used as weights for calculating Se, wn1 and wn2 are the percentages of good signal quality and poor signal quality recordings in all normal recordings respectively, and are used as weights for calculating Sp.

What makes the identification of the latter difficult?

The morphological similarity of the noise to normal and abnormal heart sounds makes identification of the latter also extremely difficult in the time domain.

What are the common locations used to listen to and transduce the heart sounds?

Four locations are most often used to listen to and transduce the heart sounds, which are named according to the positions in which the valves can be best heard (Springer 2016):• Aortic area—centred at the second right intercostal space.

Open AccessJournal Article10.1088/0967-3334/37/12/2181

An open access database for the evaluation of heart sound algorithms.

Chengyu Liu, +25 more

- 21 Nov 2016

- Physiological Measurement

- Vol. 37, Iss: 12, pp 2181-2213

723

TL;DR: A public heart sound database, assembled for an international competition, the PhysioNet/Computing in Cardiology (CinC) Challenge 2016, which comprises nine different heart sound databases sourced from multiple research groups around the world is described.

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

AI Agents for this Paper

Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps

Most frequently asked questions

1. How many recordings of normal and congenital heart sounds were tested?

A total of 823 cycles from 40 recordings of normal children and 80 recordings of children with congenital heart diseases were tested and an accuracy of 93.6% was achieved when splitting the recordings equally between training and test datasets.

2. How did Chen et al achieve a 92.1% accuracy?

Chen et al used a K-means clustering and a threshold method to identify the heart sounds, achieving 92.1% sensitivity and 88.4% positive predictivity tested on 27 recordings from healthy subjects (Chen et al 2009).

3. How many recordings were found in the Challenge test set?

The Challenge test set also included data from six databases (b through e, plus g and i) containing a total of 1277 heart sound recordings from 308 subjects/patients, lasting from 6 s to 104 s.

4. What did they use to differentiate between normal and pathological murmurs?

Using a separate test set of 53 patients they reported 83% sensitivity and 90% specificity when differentiating between innocent and pathological murmurs.

Figure 1. Phonocardiograms (above) from normal and abnormal heart sounds with pressure diagrams (below). Red indicates aortic pressure, green ventricular pressure and blue atrial pressure. Reproduced under the CC BY-SA 3.0 license and adapted from Madhero (2010).

Figure 2. General spectral regions for different heart sounds, and other physiologicalsounds during heart sound recordings. Adapted from Leatham (1975) and Springer (2016).

Figure 4. Example of an ECG-labelled PCG, with the ECG, PCG and four states of the heart cycle (S1, systole, S2 and diastole) shown. The R-peak and end-T-wave are labelled as references for defining the approximate positions of S1 and S2 respectively. Mid-systolic clicks, typical of mitral valve prolapse, can be seen in the systole states. Adapted from Springer (2016).

Figure 3. Typical three steps for automated analysis of heart sound in clinical applications.

Table 2. Summary of the major heart sound segmentation works.

Table 4. Summary of the training and test sets used in 2016 PhysioNet/CinC Challenge.

Citations

Journal Article•10.1166/JMIHI.2018.2442

An open access database for evaluating the algorithms of electrocardiogram rhythm and morphology abnormality detection

Eddie Y. K. Ng, +11 more

- 01 Sep 2018

- Journal of Medical Imaging and Health In...

563

•Proceedings Article•10.22489/CINC.2016.182-399

Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds

Cristhian Potes, +3 more

- 14 Sep 2016

TL;DR: The authors' classifier ensemble approach obtained the highest score of the competition with a sensitivity, specificity, and overall score of 0.9424, 0.7781, and 0.8602, respectively.

...read moreread less

352

Journal Article•10.1016/J.NEUNET.2020.06.015

Heart sound classification based on improved MFCC features and convolutional recurrent neural networks.

Muqing Deng, +5 more

- 01 Oct 2020

- Neural Networks

TL;DR: This paper proposes a new heart sound classification method based on improved Mel-frequency cepstrum coefficient (MFCC) features and convolutional recurrent neural networks and comprehensive studies on the performance of different network parameters and different network connection strategies are presented.

...read moreread less

284

•Proceedings Article•10.22489/CINC.2016.179-154

Classification of normal/abnormal heart sound recordings: The PhysioNet/Computing in Cardiology Challenge 2016

Gari D. Clifford, +6 more

- 14 Sep 2016

TL;DR: The PhysioNet/Computing in Cardiology (CinC) Challenge 2016 addresses the issue of a large and open database of heart sound recordings by assembling the largest public heart sound database, aggregated from eight sources obtained by seven independent research groups around the world.

...read moreread less

247

•Proceedings Article•10.22489/CINC.2016.180-213

Heart sound anomaly and quality detection using ensemble of neural networks without segmentation

Morteza Zabihi, +4 more

- 01 Mar 2016

TL;DR: The objective of this study is to develop an automatic classification method for anomaly and quality detection of PCG recordings without segmentation in order to evaluate the heart hemodynamic status and to detect a cardiovascular disease.

...read moreread less

202

...

Expand

References

•Journal Article•10.1088/0967-3334/31/4/004

Segmentation of heart sound recordings by a duration-dependent hidden Markov model

Samuel Emil Schmidt, +4 more

- 05 Mar 2010

- Physiological Measurement

TL;DR: In this article, a duration-dependent hidden Markov model (DHMM) was proposed for robust segmentation of heart sounds, based on duration of the events, amplitude of the signal envelope and a predefined model structure.

...read moreread less

256

•Proceedings Article•10.22489/CINC.2016.179-154

Classification of normal/abnormal heart sound recordings: The PhysioNet/Computing in Cardiology Challenge 2016

Gari D. Clifford, +6 more

- 14 Sep 2016

...read moreread less

247

Proceedings Article•10.1109/IEMBS.1997.757028

A heart sound segmentation algorithm using wavelet decomposition and reconstruction

Liang Huiying, +2 more

- 30 Oct 1997

TL;DR: A heart sound segmentation algorithm, which separates the heart sound signal into four parts (the first heart sound, the systolic period, the second heart sound and the diastolic period), has been developed and shown to perform correctly in over 93% of cases.

...read moreread less

181

•Journal Article•10.1016/J.BSPC.2012.11.008

A Robust Heart Sounds Segmentation Module Based on S-Transform

Ali Moukadem, +3 more

- 01 May 2013

- Biomedical Signal Processing and Control

TL;DR: A new module for heart sounds segmentation based on S-Transform is presented, which can be divided into three main blocks: localization of heart sounds, boundaries detection of the localized heart sounds and classification block to distinguish between S1 and S2.

...read moreread less

168

Journal Article•10.1016/J.ESWA.2010.05.088

Detection of cardiac abnormality from PCG signal using LMS based least square SVM classifier

Samit Ari, +2 more

- 01 Dec 2010

- Expert Systems With Applications

TL;DR: A technique to improve the performance of the Least Square Support Vector Machine (LSSVM) is proposed for classification of normal and abnormal heart sounds using wavelet based feature set using Lagrange multiplier and weight vector.

...read moreread less

162

...

Expand

An open access database for the evaluation of heart sound algorithms.

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. How many recordings of normal and congenital heart sounds were tested?

2. How did Chen et al achieve a 92.1% accuracy?

3. How many recordings were found in the Challenge test set?

4. What did they use to differentiate between normal and pathological murmurs?

5. What are the four databases that have been placed exclusively in the training or test sets?

6. What is the common step in the analysis of heart sounds?

7. What are the weights for calculating Se?

8. What makes the identification of the latter difficult?

9. What are the common locations used to listen to and transduce the heart sounds?

Figures

Citations

An open access database for evaluating the algorithms of electrocardiogram rhythm and morphology abnormality detection

Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds

Heart sound classification based on improved MFCC features and convolutional recurrent neural networks.

Classification of normal/abnormal heart sound recordings: The PhysioNet/Computing in Cardiology Challenge 2016

Heart sound anomaly and quality detection using ensemble of neural networks without segmentation

References

Segmentation of heart sound recordings by a duration-dependent hidden Markov model

Classification of normal/abnormal heart sound recordings: The PhysioNet/Computing in Cardiology Challenge 2016

A heart sound segmentation algorithm using wavelet decomposition and reconstruction

A Robust Heart Sounds Segmentation Module Based on S-Transform

Detection of cardiac abnormality from PCG signal using LMS based least square SVM classifier

Related Papers (5)

Logistic Regression-HSMM-Based Heart Sound Segmentation

Ensemble of feature-based and deep learning-based classifiers for detection of abnormal heart sounds

Heart sound anomaly and quality detection using ensemble of neural networks without segmentation

Classification of normal/abnormal heart sound recordings: The PhysioNet/Computing in Cardiology Challenge 2016

PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals.