PT  - JOURNAL ARTICLE
AU  - Richard Fletcher
AU  - Christian Infante
AU  - Yogesh Thorat
AU  - Vandana Vincent
AU  - Sundeep Salvi
TI  - The Value of Voluntary Cough Sounds for Diagnosis of Pulmonary Disease
AID  - 10.1183/13993003.congress-2018.PA5035
DP  - 2018 Sep 15
TA  - European Respiratory Journal
PG  - PA5035
VI  - 52
IP  - suppl 62
4099  - http://erj.ersjournals.com/content/52/suppl_62/PA5035.short
4100  - http://erj.ersjournals.com/content/52/suppl_62/PA5035.full
SO  - Eur Respir J2018 Sep 15; 52
AB  - Introduction: Cough sound analysis is of interest for computer-aided diagnostics and telemedicine, but few rigorous studies exist. We present our results from a recent study which examines the microstructure of voluntary cough sounds and explore the potential value for diagnosis of specific respiratory diseases.Methods: Using a smartphone electronic stethoscope (Chamberlain et al. 2015), we recorded voluntary cough sounds from 54 patients exhibiting a combination of Asthma, COPD, and Allergic Rhinitis (AR), as well cough sounds from 33 healthy subjects. All patients were diagnosed by an expert pulmonologist and confirmed with a full PFT consisting of spirometry, body pleth, DLCO, and impulse oscillometry. Using the recorded waveform, approx 60 different feature parameters were extracted, based on the microstructure of the cough. These features were then used to train a logistic regression model and determine how well specific diseases could be predicted from the coughs alone.Results: Using cough sounds alone, the AUC (area under the curve) of the ROC curve was found to be: AUC = 74% for the Healthy vs Unhealthy classification; AUC = 80% for Obstructive vs non-Obstructive classification; and 81% for Asthma vs COPD discrimination. Compared to other diagnostic methods, cough sounds was found to perform better than lung sounds alone. Cough sound analysis was also found to provide some small improvement to disease prediction when used in combination with other low-cost tools.Conclusion: Within the disease categories which we studied (Asthma, COPD, and AR), , we conclude that voluntary cough sounds can have some value for computer-aided diagnostic systems, with primary use as a crude low-cost screening tool.FootnotesCite this article as: European Respiratory Journal 2018 52: Suppl. 62, PA5035.This is an ERS International Congress abstract. No full-text version is available. Further material to accompany this abstract may be available at www.ers-education.org (ERS member access only).