Non-invasive diagnosis methods of coronary disease based on wavelet denoising and sound analyzing

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Heart sounds are crucial indicators of cardiovascular health.
• Accurate analysis of heart sounds is often hindered by noise.
• Advanced signal processing techniques are needed for reliable heart sound interpretation.

Purpose of the Study:

• To investigate the effectiveness of Wavelet Transform in denoising heart sound signals.
• To evaluate the performance of bispectrum estimation for classifying heart sounds.
• To explore the correlation between noise reduction and classification accuracy.

Main Methods:

• Heart sound signals were decomposed using Wavelet Transform to analyze different frequency ranges.
• Four wavelets (Haar, db6, sym8, coif6) were compared for denoising performance, with db6 showing optimal results.
• Bispectrum estimation was applied to denoised signals using an ARMA coefficients model for classification.

Main Results:

• The db6 wavelet demonstrated superior denoising capabilities for heart sound signals.
• Decomposition into five layers using the db6 wavelet yielded optimal noise reduction.
• The method successfully distinguished between normal and abnormal heart sound signals in clinical data.

Conclusions:

• Wavelet Transform, particularly the db6 wavelet with five layers, is effective for denoising heart sounds.
• Bispectrum estimation applied to denoised signals provides a reliable method for classifying cardiac health status.
• This approach holds promise for improving non-invasive diagnosis of cardiovascular conditions.