Classification of lung sounds during bronchial provocation using waveform fractal dimensions

Area of Science:

• Pediatric Pulmonology
• Respiratory Acoustics
• Biomedical Signal Processing

Background:

• Lung sounds (LS) are expected to change in amplitude and pattern after induced bronchoconstriction.
• Assessing these changes is crucial for understanding pediatric airway dynamics.
• Previous methods may not fully capture the complexity of LS alterations.

Purpose of the Study:

• To investigate if lung sounds differ in children after bronchoconstriction compared to baseline.
• To evaluate the effectiveness of time-domain and fractal-based analyses for detecting bronchoconstriction.
• To determine the optimal combination of signal analysis features for classifying airway narrowing.

Main Methods:

• Acquired lung sounds from eight children (ages 9-15) pre- and post-methacholine challenge (MCh).
• Applied time-domain analysis (RMS-SNR) and two fractal dimension (FD) algorithms (variance-based and morphology-based).
• Utilized 1-nearest-neighbor classification with feature vectors including FD and RMS values.

Main Results:

• The combination of RMS-SNR and morphology-based FD achieved 90.3% true positive classification of bronchoconstriction.
• This approach outperformed RMS-SNR with variance-based FDs (63.5%) and RMS-SNR alone (58.3%).
• Both RMS-SNR and FD values provided significant insights into LS changes following bronchoconstriction.

Conclusions:

• Morphology-based fractal dimension combined with RMS-SNR is a highly effective method for classifying pediatric bronchoconstriction using lung sounds.
• This advanced signal analysis technique offers improved diagnostic potential for airway diseases in children.
• Further research can explore broader applications of these LS analysis methods.