Screening of COVID-19 using Cough Audio Frequencies

Abstract

Clinicians routinely use biomedical and audio signals (e.g. sighs, breathing, pulse, digestion, sounds of vibration) as markers to diagnose diseases or to evaluate the progression of diseases. Until recently, these signals were normally obtained during scheduled visits by manual auscultation. With the advancement of technologies, digital methods are used to collect the body sounds for cardiovascular or respiratory testing (e.g. digital stethoscopes to predict the progression of diseases. A few early studies showed promising results for the detection of COVID-19 using voice and diagnostic signals. In the proposed model, an effective analysis is performed through the collection of large, multi-group, airborne acoustic sound data to perform the screening of COVID-19. The technique uses cough and breathing patterns to show the distinctive features of COVID-19 and it is reported that the cough patterns of COVID-19 are identifiable from asthma cough patterns. Using machine learning algorithms, an efficient classification model is developed for the screening of COVID-19.The area below the curve (AUC) of our proposed model exceeds 80%. The present study also explores the analysis of air patterns that can be recorded using the breathing styles of the infected persons to enhance the efficiency of the proposed screening techniques.