ERS Respiratory Channel

CDN Connection Issue

It appears that your current network does not allow connection to the CDN (Content Delivery Network) required for this service to function properly.

To resolve this issue, please try the following actions :

• Contact your network administrator : Inform them of this restriction so they can adjust the network settings accordingly.

• Use another network : Try connecting from a different network, such as another Wi-Fi network or a mobile connection.

We apologize for the inconvenience and thank you for your understanding.

Abstract

Background

Auscultation is important for diagnosis of lung diseases. However, it is subjective and requires a high degree of expertise. Artificial intelligence models are being developed to overcome these limitations. We compared the methods for extraction of respiratory sound features and developed an optimal machine learning model to detect wheezing in children.

Method

In this prospective study, pediatric pulmonologists recorded and verified 103 wheeze sounds and 184 other respiratory sounds in 76 children who visited an University hospital in South Korea. Various methods were used to extract the features of respiratory sounds. The dimension was reduced using the t-distributed Stochastic Neighbor Embedding, and the performance of the model for detection of wheezing was compared using the kernel Support Vector Machine (SVM).

Results

Mel-spectrogram, Mel-frequency cepstral coefficient (MFCC), and spectral contrast were most suitable for expression of respiratory sounds and showed good performance in cluster classification. The SVM model using spectral contrast showed the best performance at accuracy of 0.897, precision 0.800, recall 0.952, and F-1 score of 0.869.

Conclusion

Mel-spectrogram, MFCC, and spectral contrast are useful to characterize respiratory sounds. A machine learning model using spectral contrast showed high performance for detection of wheezing in children. This high-performance model will be useful for accurate diagnosis of respiratory diseases in children.

Abstract

#{title}