Salient Representation for Lung Sound Analysis Based on the JAMF Transform

Abstract

Obstructive lung diseases are a family of diseases affecting the respiratory tract. By auscultation of the lungs, trained physicians can determine abnormal pathologies by listening to these lung sounds. When these lung sounds are recorded, the analysis can be performed by means of digital signal processing, often referred to as Computer-Aided Lung Sound Analysis (CALSA). The human hear is somewhat limited in it’s capabilities to detect and quantify small perceptual changes to the lung sound scene. The goal of CALSA is to be able to make objective conclusions on the state of the patient. In order to achieve this, the data recorded from the stethoscope must be transformed into a representation that allows efficient interpretation, either by a trained physician or a computer algorithm. In this paper, we propose the Joint Acoustic and Frequency Modulation representation as a useful tool for lung sound analysis, and illustrate the representation of various lung sounds in the transformed domain.