Detection of patients considering observation frequency of continuous and discontinuous adventitious sounds in lung sounds

Nakamura, Naoki; Yamashita, Masaru; Matsunaga, Shoichi

doi:10.1109/embc.2016.7591472

Cited by 12 publications

(9 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to the literature on lung sound analysis, the most common respiratory diseases that are studied include COPD , 53,54,65,75,79,83,86,87,94,101,105,144 asthma, 53,54,65,71,72,85,87,105,107,110,117 and pulmonary emphysema. 58,70,93,99,106,112 However, lung sounds recorded during pneumonia, 54,65,76,79,103 pulmonary fibrosis, 56,70,76 chronic bronchitis, 56 idiopathic pulmonary fibrosis, 63 congestive heart failure, 76,79 parenchymal pathology, 75,90 and interstitial lung disease 79,87,104 appear less frequently in the literature. Regarding heart sound analysis, it is observed that murmurs 26,121,<...>…”

Section: Discussionmentioning

confidence: 99%

Lung and Heart Sounds Analysis: State-of-the-Art and Future Trends

Padilla-Ortiz

Ibarra

2018

Crit Rev Biomed Eng

View full text Add to dashboard Cite

Lung sounds, which include all sounds that are produced during the mechanism of respiration, may be classified into normal breath sounds and adventitious sounds. Normal breath sounds occur when no respiratory problems exist, whereas adventitious lung sounds (wheeze, rhonchi, crackle, etc.) are usually associated with certain pulmonary pathologies. Heart and lung sounds that are heard using a stethoscope are the result of mechanical interactions that indicate operation of cardiac and respiratory systems, respectively. In this article, we review the research conducted during the last six years on lung and heart sounds, instrumentation and data sources (sensors and databases), technological advances, and perspectives in processing and data analysis. Our review suggests that chronic obstructive pulmonary disease (COPD) and asthma are the most common respiratory diseases reported on in the literature; related diseases that are less analyzed include chronic bronchitis, idiopathic pulmonary fibrosis, congestive heart failure, and parenchymal pathology. Some new findings regarding the methodologies associated with advances in the electronic stethoscope have been presented for the auscultatory heart sound signaling process, including analysis and clarification of resulting sounds to create a diagnosis based on a quantifiable medical assessment. The availability of automatic interpretation of high precision of heart and lung sounds opens interesting possibilities for cardiovascular diagnosis as well as potential for intelligent diagnosis of heart and lung diseases.

show abstract

Section: Discussionmentioning

confidence: 99%

Lung and Heart Sounds Analysis: State-of-the-Art and Future Trends

Padilla-Ortiz

Ibarra

2018

Crit Rev Biomed Eng

View full text Add to dashboard Cite

show abstract

“…In most systems, suitable features are extracted from the signal and are subsequently used to classify ARS (i.e., crackles and wheezes). The most common features and machine learning algorithms employed in the literature to detect or classify ARS have been reported [ 6 ], including spectral features [ 25 ], mel-frequency cepstral coefficients (MFCCs) [ 26 ], entropy [ 27 ], wavelet coefficients [ 28 ], rule-based models [ 29 ], logistic regression models [ 30 ], support vector machines (SVM) [ 31 ], and artificial neural networks [ 32 ]. More recently, deep learning strategies have also been introduced, where the feature extraction and classification steps are merged into the learning algorithm [ 33 , 34 , 35 ].…”

Section: Related Workmentioning

confidence: 99%

Automatic Classification of Adventitious Respiratory Sounds: A (Un)Solved Problem?

Rocha

Pessoa

Marques

et al. 2020

Sensors

View full text Add to dashboard Cite

(1) Background: Patients with respiratory conditions typically exhibit adventitious respiratory sounds (ARS), such as wheezes and crackles. ARS events have variable duration. In this work we studied the influence of event duration on automatic ARS classification, namely, how the creation of the Other class (negative class) affected the classifiers’ performance. (2) Methods: We conducted a set of experiments where we varied the durations of the other events on three tasks: crackle vs. wheeze vs. other (3 Class); crackle vs. other (2 Class Crackles); and wheeze vs. other (2 Class Wheezes). Four classifiers (linear discriminant analysis, support vector machines, boosted trees, and convolutional neural networks) were evaluated on those tasks using an open access respiratory sound database. (3) Results: While on the 3 Class task with fixed durations, the best classifier achieved an accuracy of 96.9%, the same classifier reached an accuracy of 81.8% on the more realistic 3 Class task with variable durations. (4) Conclusion: These results demonstrate the importance of experimental design on the assessment of the performance of automatic ARS classification algorithms. Furthermore, they also indicate, unlike what is stated in the literature, that the automatic classification of ARS is not a solved problem, as the algorithms’ performance decreases substantially under complex evaluation scenarios.

show abstract

“…A series of experiments were performed using this novel dataset and performance was compared with the BodyBeat [6] system. Several other shallow and deep learning based pulmonary activity detection works like wheeze detection [7][8][9][10][10][11][12][13][14][15] and cough detection [16][17][18] exist in literature. However, they often use limited training data which is not collected with a commodity smartphone.…”

Section: Introductionmentioning

confidence: 99%

DeepLung: Smartphone Convolutional Neural Network-Based Inference of Lung Anomalies for Pulmonary Patients

2019

View full text Add to dashboard Cite

DeepLung is an end-to-end deep learning based audio sensing and classification framework for lung anomaly (e.g. cough, wheeze) detection for pulmonary patients from streaming audio and inertial sensor data from a chest-held smartphone. We design and develop 1-D and 2-D convolutional neural networks for DeepLung, and train them using the Interspeech 2010 Paralinguistic Challenge features. Two different audio windowing schemes: i) real-time respiration cycle based natural windowing, and ii) static length windowing are compared and experimented with. Classifiers are developed considering 2 different system architectures: i) mobile-cloud hybrid architecture, and ii) mobile in-situ architecture. Patient privacy is preserved in the phone by filtering speech with a shallow classifier. To evaluate DeepLung, a novel and rigorous lung activity dataset is made by collecting audio and inertial sensor data from more than 131 real pulmonary patients and healthy subjects and annotated accurately by professional crowdsourcing. Experimental results show that the best combination of DeepLung convolutional neural network is 15-27% more accurate when compared to a stateof-the-art smartphone based body sound detection system, with a best F1 score of 98%.

show abstract

Detection of patients considering observation frequency of continuous and discontinuous adventitious sounds in lung sounds

Cited by 12 publications

References 6 publications

Lung and Heart Sounds Analysis: State-of-the-Art and Future Trends

Lung and Heart Sounds Analysis: State-of-the-Art and Future Trends

Automatic Classification of Adventitious Respiratory Sounds: A (Un)Solved Problem?

DeepLung: Smartphone Convolutional Neural Network-Based Inference of Lung Anomalies for Pulmonary Patients

Contact Info

Product

Resources

About