Localizing heart sounds in respiratory signals using singular spectrum analysis

Area of Science:

• Biomedical Engineering
• Signal Processing
• Respiratory Medicine

Background:

• Respiratory sounds are often obscured by heart sound interference.
• Accurate localization of heart sound components is crucial for effective heart sound cancellation.
• Existing methods face challenges due to frequency overlap between heart and lung sounds.

Purpose of the Study:

• To introduce Singular Spectrum Analysis (SSA) for localizing primary heart sound components in respiratory signals.
• To evaluate the performance of SSA against established signal processing techniques.
• To demonstrate the efficacy of SSA in improving heart sound cancellation preprocessing.

Main Methods:

• Singular Spectrum Analysis (SSA), a time series analysis technique, was applied to respiratory signals.
• The method identifies distinct trends in eigenvalue spectra to isolate heart sound information.
• Performance was evaluated using artificially mixed and real respiratory signals.

Main Results:

• SSA successfully identified a subspace rich in heart sound information despite frequency overlap.
• The method demonstrated good decomposition quality and low computational cost with optimal window length selection.
• Compared to wavelet transform, SSA showed lower false detection rates and higher correlation with heart sounds.
• SSA performance was slightly superior to entropy-based methods with significantly reduced execution time.

Conclusions:

• SSA is a robust and efficient technique for localizing heart sound components in respiratory signals.
• The proposed SSA method offers significant advantages over wavelet and entropy-based approaches in terms of accuracy and speed.
• This technique enhances the preprocessing stage for heart sound cancellation, paving the way for improved diagnostic tools.