Acoustic detection of Korotkoff sounds using non-linear analysis

Coronary artery disease (CAD), a leading cause of death in the United States, has few early diagnostic mechanisms. Acoustic detection of sound associated with turbulent blood flow in coronary arteries has been shown to discriminate normal and abnormal subjects; however, improved performance is required for clinical use. Nonlinear analysis methods sensitive to turbulence could improve acoustic detection of CAD. To assess the ability of these methods to detect turbulent flow, Korotkoff sounds generated by turbulent flow through the brachial artery were used. The nonlinear methods studied include largest Lyapunov and Hurst exponents and Shannon entropy. Data were taken on 3 subjects using hardware developed by SonoMedica (McLean, Virginia). All methods were sensitive to changes in pressure applied to the brachial artery. The Lyapunov exponents decreased with increasing pressure. Entropy and Hurst exponent methods showed the opposite trend. In particular, the entropy measurements showed a 50% decrease from high to low pressure. All Lyapunov exponents were positive, indicating the presence of chaotic components in the signal. Hurst exponents near 1.0 are indicative of long term persistence in measured high pressure signals. These results indicate that nonlinear methods may be suitable for detecting CAD using heart sound recordings from diseased subjects.