Efficient extraction of fine/coarse crackles and squawks from lung sound recordings using wavelet packets and higher-order statistics
I. K. Kitsas, L. J. Hadjileontiadis, V. Gross, T. Penzel, S. M. Panas (Thessaloniki, Greece; Marburg, Germany)
Source: Annual Congress 2002 - Airway obstruction measurement (FOT - NEP) sleep and lung sound analysis
Disease area: Interstitial lung diseases
Abstract The separation of pathological discontinuous adventitious sounds (DAS) from vesicular sounds (VS) is of great importance to the analysis of lung sounds, since DAS are related to certain pulmonary pathologies. Efficient extaction of DAS from VS helps the physician to accurately identify the associated pathology, i.e., pulmonary and interstitial fibrosis (PF, IF), chronic bronchitis (CB), allergic alveolitis (AA). In this study we have created an automated way of isolating DAS from VS, based on their nonstationarity and non-Gaussianity. The proposed method combines wavelet packets (WP) with higher-order statistics (HOS) in order to compose a WP/HOS stationary-nonstationary (WP/HOSST-NST) filter. The latter discriminates the non-Gaussian transients (DAS) from the Gaussian, stationary background noise (VS), by efficiently detecting nonstationarity and non-Gaussianity occurrences in the recorded signals. Experimental results from the analysis of pre-classified lung sounds (fine crackles: 6 cases-PF, IF; coarse crackles: 5 cases-CB; squawks: 5 cases IF, AA) drawn from two lung sound databases (Kraman, S.S. Lung Sounds: An Introduction to the Interpretation of the Auscultatory Finding 1993; Northbrook, Illinois, USA: ACCP; Lehrer, S. Understanding Lung Sounds 1993; Philadelphia, PA, USA: W. B. Saunders, Co.), show that the WP/HOSST-NST filter performs very accurately (100% detectability, p <0.001) when compared to experts' (clinicians) scoring. Due to its low computational cost, simple implementation, and noise robustness it can easily be used in intesive care units for real-time DAS screening, thus enhancing the non-invasive diagnostic character of lung sounds.
Rating:
You must login to grade this presentation.
Share or cite this content
Citations should be made in the following way:
I. K. Kitsas, L. J. Hadjileontiadis, V. Gross, T. Penzel, S. M. Panas (Thessaloniki, Greece; Marburg, Germany). Efficient extraction of fine/coarse crackles and squawks from lung sound recordings using wavelet packets and higher-order statistics. Eur Respir J 2002; 20: Suppl. 38, 326
You must login to share this Presentation/Article on Twitter, Facebook, LinkedIn or by email.
Member's Comments
Related content which might interest you:
Related content which might interest you: