Comparison of instantaneous frequency analysis methods for acoustic detection of coronary artery disease

Coronary artery disease (CAD) is a chronic illness of the cardiovascular system that progressively diminishes the flow of blood and oxygen to the heart. The acoustic method of CAD diagnosis is safe, inexpensive and easy to administer and could be suitable as an early screening tool. However, few studies have shown clinically significant accuracy using this method. The Hilbert-Huang transform, an instantaneous frequency (IF) analysis method, has been shown to distinguish normal and diseased subjects with high accuracy. Recently, more accurate methods of IF analysis, such as direct quadrature and a normalized HH transform, have been proposed. In this study the HH transform was compared with direct quadrature and the normalized HH transform to determine if this new IF estimation method improved signal classification. Each method was used to analyze acoustic signals from 16 normal and 15 diseased subjects. None of the tested analysis methods appeared superior. The highest accuracy measurement was skewness of the IF, as derived through the HH transform. This measure retrospectively classified subjects with 74% accuracy and a sensitivity-specificity of 67%-81%.