Quantile vectors based verification of normal lung sounds

Abstract

Anthropogenic activities associated to population growth impact overall health and contribute to elevated rates of cardiovascular and respiratory diseases. In this paper we propose the Lung Sound Automatic Verification (LSAV), and other modalities to represent acoustic lung signals obtained by auscultation using a digital stethoscope. The utilization of quantiles allowed a) an easier and objective assessment with smaller computational demand, b) building of less-complex Gaussian Mixed Models (GMM) than those reported previously, and c) to reach practically negligible error in healthy LS verification. These approaches relate the lung sound energy to its characteristic frequency components, which in addition to a reliable verification technique simplified the normal lung sound recognition. Once the LS are evaluated, it would be possible to simplify classification if an individual auscultatory evaluation falls into the category of normal or abnormal indicators thus providing a tool for broader medical evaluation which does not rely, as it is often the case, on a qualitative and subjective description of these sounds.