Auscultation is crucial for the diagnosis of respiratory system diseases.However,traditional stethoscopes have inherent limitations,such as inter-listener variability and subjectivity,and they cannot record respirator...Auscultation is crucial for the diagnosis of respiratory system diseases.However,traditional stethoscopes have inherent limitations,such as inter-listener variability and subjectivity,and they cannot record respiratory sounds for offline/retrospective diagnosis or remote prescriptions in telemedicine.The emergence of digital stethoscopes has overcome these limitations by allowing physicians to store and share respiratory sounds for consultation and education.On this basis,machine learning,particularly deep learning,enables the fully-automatic analysis of lung sounds that may pave the way for intelligent stethoscopes.This review thus aims to provide a comprehensive overview of deep learning algorithms used for lung sound analysis to emphasize the significance of artificial intelligence(AI)in this field.We focus on each component of deep learning-based lung sound analysis systems,including the task categories,public datasets,denoising methods,and,most importantly,existing deep learning methods,i.e.,the state-of-the-art approaches to convert lung sounds into two-dimensional(2D)spectrograms and use convolutional neural networks for the end-to-end recognition of respiratory diseases or abnormal lung sounds.Additionally,this review highlights current challenges in this field,including the variety of devices,noise sensitivity,and poor interpretability of deep models.To address the poor reproducibility and variety of deep learning in this field,this review also provides a scalable and flexible open-source framework that aims to standardize the algorithmic workflow and provide a solid basis for replication and future extension:https://github.com/contactless-healthcare/Deep-Learning-for-Lung-Sound-Analysis.展开更多
Noise artifacts are one of the key obstacles in applying continuous monitoring and computer-assisted analysis of lung sounds. Traditional adaptive noise cancellation (ANC) methodologies work reasonably well when signa...Noise artifacts are one of the key obstacles in applying continuous monitoring and computer-assisted analysis of lung sounds. Traditional adaptive noise cancellation (ANC) methodologies work reasonably well when signal and noise are stationary and independent. Clinical lung sound auscultation encounters an acoustic environment in which breath sounds are not stationary and often correlate with noise. Consequendy, capability of ANC becomes significantly compromised. This paper introduces a new methodology for extracting authentic lung sounds from noise-corrupted measurements. Unlike traditional noise cancellation methods that rely on either frequency band separation or signal/noise independence to achieve noise reduction, this methodology combines the traditional noise canceling methods with the unique feature of time-split stages in breathing sounds. By employing a multi-sensor system, the method first employs a high-pass filter to eliminate the off-band noise, and then performs time-shared blind identification and noise cancellation with recursion from breathing cycle to cycle. Since no frequency separation or signal/noise independence is required, this method potentially has a robust and reliable capability of noise reduction, complementing the traditional methods.展开更多
Background Although acute congestive heart failure (CHF) patients typically present with abnormal auscultatory findings on lung examination, lung sounds are not normally subjected to rigorous analysis. The goals of ...Background Although acute congestive heart failure (CHF) patients typically present with abnormal auscultatory findings on lung examination, lung sounds are not normally subjected to rigorous analysis. The goals of this study were to use a computerized analytic acoustic tool to evaluate lung sound patterns in CHF patients during acute exacerbation and after clinical improvement and to compare CHF profiles with those of normal individuals.Methods Lung sounds throughout the respiratory cycle was captured using a computerized acoustic-based imaging technique. Thirty-two consecutive CHF patients were imaged at the time of presentation to the emergency department and after clinical improvement. Digital images were created, geographical area of the images and lung sound patterns were quantitatively analyzed.Results The geographical areas of the vibration energy image of acute CHF patients without and with radiographically evident pulmonary edema were (67.9±4.7) and (60.3±3.5) kilo-pixels, respectively (P 〈0.05). In CHF patients without and with radiographically evident pulmonary edema (REPE), after clinical improvement the geographical area of vibration energy image of lung sound increased to (74.5±4.4) and (73.9±3.9) kilo-pixels (P 〈0.05), respectively. Vibration energy decreased in CHF patients with REPE following clinical improvement by an average of (85±19)% (P 〈0.01). Conclusions With clinical improvement of acute CHF exacerbations, there was more homogenous distribution of lung vibration energy, as demonstrated by the increased geographical area of the vibration energy image. Lung sound analysis may be useful to track in acute CHF exacerbations.展开更多
Crackles are an important kind of abnormal and discontinuous lung sounds,which have been found to be correlated to types of pulmonary diseases.The purpose of this work is to show a new perspective to solve the problem...Crackles are an important kind of abnormal and discontinuous lung sounds,which have been found to be correlated to types of pulmonary diseases.The purpose of this work is to show a new perspective to solve the problem of crackle detection,based on an emerging theory of fractional Hilbert transform.By applying fractional Hilbert transform to lung sound signals,a two-dimension texture image can be generated.The texture features corresponding to crackles are quite easy to be extracted.Experiments illustrate the effectiveness of our method.展开更多
基金This work is supported by the National Key Research and Development Program of China(2022YFC2407800)the General Program of National Natural Science Foundation of China(62271241)+1 种基金the Guangdong Basic and Applied Basic Research Foundation(2023A1515012983)the Shenzhen Fundamental Research Program(JCYJ20220530112601003).
文摘Auscultation is crucial for the diagnosis of respiratory system diseases.However,traditional stethoscopes have inherent limitations,such as inter-listener variability and subjectivity,and they cannot record respiratory sounds for offline/retrospective diagnosis or remote prescriptions in telemedicine.The emergence of digital stethoscopes has overcome these limitations by allowing physicians to store and share respiratory sounds for consultation and education.On this basis,machine learning,particularly deep learning,enables the fully-automatic analysis of lung sounds that may pave the way for intelligent stethoscopes.This review thus aims to provide a comprehensive overview of deep learning algorithms used for lung sound analysis to emphasize the significance of artificial intelligence(AI)in this field.We focus on each component of deep learning-based lung sound analysis systems,including the task categories,public datasets,denoising methods,and,most importantly,existing deep learning methods,i.e.,the state-of-the-art approaches to convert lung sounds into two-dimensional(2D)spectrograms and use convolutional neural networks for the end-to-end recognition of respiratory diseases or abnormal lung sounds.Additionally,this review highlights current challenges in this field,including the variety of devices,noise sensitivity,and poor interpretability of deep models.To address the poor reproducibility and variety of deep learning in this field,this review also provides a scalable and flexible open-source framework that aims to standardize the algorithmic workflow and provide a solid basis for replication and future extension:https://github.com/contactless-healthcare/Deep-Learning-for-Lung-Sound-Analysis.
基金Hong Wang's research was supported in part by the Anesthesiology Department at Wayne State University and in part by Wayne State University Research Enhancement ProgramLeyi Wang" s research was supported in part by the National Science Foundation ( No.
文摘Noise artifacts are one of the key obstacles in applying continuous monitoring and computer-assisted analysis of lung sounds. Traditional adaptive noise cancellation (ANC) methodologies work reasonably well when signal and noise are stationary and independent. Clinical lung sound auscultation encounters an acoustic environment in which breath sounds are not stationary and often correlate with noise. Consequendy, capability of ANC becomes significantly compromised. This paper introduces a new methodology for extracting authentic lung sounds from noise-corrupted measurements. Unlike traditional noise cancellation methods that rely on either frequency band separation or signal/noise independence to achieve noise reduction, this methodology combines the traditional noise canceling methods with the unique feature of time-split stages in breathing sounds. By employing a multi-sensor system, the method first employs a high-pass filter to eliminate the off-band noise, and then performs time-shared blind identification and noise cancellation with recursion from breathing cycle to cycle. Since no frequency separation or signal/noise independence is required, this method potentially has a robust and reliable capability of noise reduction, complementing the traditional methods.
文摘Background Although acute congestive heart failure (CHF) patients typically present with abnormal auscultatory findings on lung examination, lung sounds are not normally subjected to rigorous analysis. The goals of this study were to use a computerized analytic acoustic tool to evaluate lung sound patterns in CHF patients during acute exacerbation and after clinical improvement and to compare CHF profiles with those of normal individuals.Methods Lung sounds throughout the respiratory cycle was captured using a computerized acoustic-based imaging technique. Thirty-two consecutive CHF patients were imaged at the time of presentation to the emergency department and after clinical improvement. Digital images were created, geographical area of the images and lung sound patterns were quantitatively analyzed.Results The geographical areas of the vibration energy image of acute CHF patients without and with radiographically evident pulmonary edema were (67.9±4.7) and (60.3±3.5) kilo-pixels, respectively (P 〈0.05). In CHF patients without and with radiographically evident pulmonary edema (REPE), after clinical improvement the geographical area of vibration energy image of lung sound increased to (74.5±4.4) and (73.9±3.9) kilo-pixels (P 〈0.05), respectively. Vibration energy decreased in CHF patients with REPE following clinical improvement by an average of (85±19)% (P 〈0.01). Conclusions With clinical improvement of acute CHF exacerbations, there was more homogenous distribution of lung vibration energy, as demonstrated by the increased geographical area of the vibration energy image. Lung sound analysis may be useful to track in acute CHF exacerbations.
文摘Crackles are an important kind of abnormal and discontinuous lung sounds,which have been found to be correlated to types of pulmonary diseases.The purpose of this work is to show a new perspective to solve the problem of crackle detection,based on an emerging theory of fractional Hilbert transform.By applying fractional Hilbert transform to lung sound signals,a two-dimension texture image can be generated.The texture features corresponding to crackles are quite easy to be extracted.Experiments illustrate the effectiveness of our method.