Infection detection in cystic fibrosis patients based on tunable Q-factor wavelet transform of respiratory sound signal and ensemble decision

Most adult Cystic Fibrosis (CF) patients frequently suffer from Pseudomonas aeruginosa (PA) infection, which is strongly associated with inflammation, lung destruction, and increased mortality. Diagnosis of PA infection in the primary stage is essential to initiate the treatment and reduce the risk of chronic infection. Sputum culture is the gold standard ‎for infection detection, but it is time-consuming. The objective of this study was to suggest and examine a method to decide about PA infection status in CF patients based only on their respiratory sound. Respiratory sounds were recorded from 36 CF patients. Some features which were generated from Tunable Q-factor wavelet transform(TQWT) components, were investigated. The features were fed into Support Vector Machine and also Ensemble classifier. The proposed method achieved an accuracy of 90.3% in identifying PA infection in CF patients. Furthermore, the probability of categorizing respiratory sounds as PA CF decreased significantly after the treatment of PA infection(P-value<0.003). Moreover, the method had a satisfactory performance in the presence of noises and artifacts. The developed method represents a novel approach to the diagnosis of PA infection in CF patients based only on respiratory sound signals, which is a necessary and innovative approach for early diagnosis of PA infection.