Detection of Adventitious Respiratory Sounds Using Empirical Mode Decomposition

Respiratory sound analysis using stethoscope continues to be the mostly used method for the diagnosis of respiratory diseases. This technique depends on detection of symptomatic adventitious sounds present with normal vesicular sounds. However, some factors such as dependence on the practitioner doctor’s experience, frequency distortion of the stethoscope and frequency response of the human ear limits the usefulness of this technique [1]. Some computerized systems have been developed to overcome these limitations. In this study, use of Empirical Mode Decomposition (EMD) have been explored for the detection of discontinuous (i.e. crackles) and continuous (i.e. wheezes) adventitious sounds that are present in the lung sounds when the patient has a respiratory disease. EMD is an adaptive technique for nonstationary signals and nonlinear signals [2]. It decomposes signals into oscillatory intrinsic modes called Intrinsic Mode Functions (IMFs) using a data driven algorithm. It is shown on the real lung sounds that these IMFs are helpful for easy detection of both the crackles and the wheezes.