TY - GEN
T1 - Revisiting COVID-19 Diagnosis From Cough Sound
T2 - 10th International Conference on Biomedical and Bioinformatics Engineering, ICBBE 2023
AU - Guan, Yijia
AU - Chan, Rosa H.M.
PY - 2023/11/9
Y1 - 2023/11/9
N2 - The global outbreak of the COVID-19 pandemic has driven the development of effective and low-cost detection technologies. With an emphasis on methods’ economic viability and detection efficacy, researchers have been actively exploring novel technologies in response to it. To address the issue, we have revisited a deep learning-based framework to facilitate the diagnosis of COVID-19 solely through the analysis of cough sounds. We utilized the label of expert diagnoses and employed time stretching as an augmentation method on the combination of a convolutional neural network (CNN) and long short-term memory (LSTM). Trained and Tested on the largest publicly-available COUGHVID dataset, our proposed hybrid CNN-LSTM model showed its performance with a short training period, demonstrating proficiency in discerning between COVID-19-related cough sounds and those of a healthy nature. Our classification model achieved an accuracy of 99.19%, a precision of 94.92%, a recall of 88.61%, a F1 Score of 91.66%, and an AUC score of 96%. © 2023 Copyright held by the owner/author(s).
AB - The global outbreak of the COVID-19 pandemic has driven the development of effective and low-cost detection technologies. With an emphasis on methods’ economic viability and detection efficacy, researchers have been actively exploring novel technologies in response to it. To address the issue, we have revisited a deep learning-based framework to facilitate the diagnosis of COVID-19 solely through the analysis of cough sounds. We utilized the label of expert diagnoses and employed time stretching as an augmentation method on the combination of a convolutional neural network (CNN) and long short-term memory (LSTM). Trained and Tested on the largest publicly-available COUGHVID dataset, our proposed hybrid CNN-LSTM model showed its performance with a short training period, demonstrating proficiency in discerning between COVID-19-related cough sounds and those of a healthy nature. Our classification model achieved an accuracy of 99.19%, a precision of 94.92%, a recall of 88.61%, a F1 Score of 91.66%, and an AUC score of 96%. © 2023 Copyright held by the owner/author(s).
KW - Convolutional Neural Network
KW - Cough Sound
KW - COVID-19
KW - Deep Learning
KW - Long Short-Term Memory
UR - http://www.scopus.com/inward/record.url?scp=85187700250&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85187700250&origin=recordpage
U2 - 10.1145/3637732.3637746
DO - 10.1145/3637732.3637746
M3 - RGC 32 - Refereed conference paper (with host publication)
SN - 979-8-4007-0834-3
T3 - ACM International Conference Proceeding Series
SP - 230
EP - 237
BT - ICBBE '23
PB - Association for Computing Machinery
Y2 - 9 November 2023 through 12 November 2023
ER -
