Heart sound cancellation based on multiscale products and linear prediction

Area of Science:

• Biomedical Engineering
• Signal Processing
• Respiratory Medicine

Background:

• Lung sound (LS) auscultation is crucial for diagnosing respiratory conditions.
• Heart sounds (HS) often contaminate LS recordings, hindering accurate diagnosis.
• Existing methods for HS cancellation may introduce artifacts or remove important LS components.

Purpose of the Study:

• To develop and validate a novel method for effective Heart Sound (HS) cancellation from Lung Sound (LS) records.
• To ensure the preservation of vital lung sound components during HS removal.
• To eliminate audible artifacts in the reconstructed lung sound signal.

Main Methods:

• Utilizing the multiscale product of wavelet coefficients to detect HS-included segments in LS recordings.
• Removing identified HS segments from wavelet coefficients across multiple levels.
• Estimating and reconstructing the signal in the removed segments using linear prediction filters.

Main Results:

• The proposed method successfully detected and removed HS from LS records.
• Reconstructed signals exhibited no audible artifacts, such as clicks.
• Spectral analysis confirmed the preservation of main lung sound components.
• Qualitative listening tests validated the effectiveness of HS removal.

Conclusions:

• The novel wavelet-based method offers a promising solution for accurate HS cancellation in LS recordings.
• This technique enhances the diagnostic quality of lung sound auscultation by removing confounding heart sounds.
• The method's ability to preserve LS components and avoid artifacts makes it suitable for clinical applications.