Detection of COVID-19 in smartphone-based breathing recordings: A pre-screening deep learning tool

被引:35
作者
Alkhodari, Mohanad [1 ]
Khandoker, Ahsan H. [1 ]
机构
[1] Khalifa Univ, Healthcare Engn Innovat Ctr HEIC, Dept Biomed Engn, Abu Dhabi, U Arab Emirates
来源
PLOS ONE | 2022年 / 17卷 / 01期
关键词
NEURAL-NETWORKS; DATA AUGMENTATION; LUNG SOUNDS; CT IMAGES; CLASSIFICATION; FRAMEWORK; DIAGNOSIS; DISEASES; MODEL; TERM;
D O I
10.1371/journal.pone.0262448
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
This study was sought to investigate the feasibility of using smartphone-based breathing sounds within a deep learning framework to discriminate between COVID-19, including asymptomatic, and healthy subjects. A total of 480 breathing sounds (240 shallow and 240 deep) were obtained from a publicly available database named Coswara. These sounds were recorded by 120 COVID-19 and 120 healthy subjects via a smartphone microphone through a website application. A deep learning framework was proposed herein that relies on hand-crafted features extracted from the original recordings and from the mel-frequency cepstral coefficients (MFCC) as well as deep-activated features learned by a combination of convolutional neural network and bi-directional long short-term memory units (CNN-BiLSTM). The statistical analysis of patient profiles has shown a significant difference (p-value: 0.041) for ischemic heart disease between COVID-19 and healthy subjects. The Analysis of the normal distribution of the combined MFCC values showed that COVID-19 subjects tended to have a distribution that is skewed more towards the right side of the zero mean (shallow: 0.59 +/- 1.74, deep: 0.65 +/- 4.35, p-value: <0.001). In addition, the proposed deep learning approach had an overall discrimination accuracy of 94.58% and 92.08% using shallow and deep recordings, respectively. Furthermore, it detected COVID-19 subjects successfully with a maximum sensitivity of 94.21%, specificity of 94.96%, and area under the receiver operating characteristic (AUROC) curves of 0.90. Among the 120 COVID-19 participants, asymptomatic subjects (18 subjects) were successfully detected with 100.00% accuracy using shallow recordings and 88.89% using deep recordings. This study paves the way towards utilizing smartphone-based breathing sounds for the purpose of COVID-19 detection. The observations found in this study were promising to suggest deep learning and smartphone-based breathing sounds as an effective pre-screening tool for COVID-19 alongside the current reverse-transcription polymerase chain reaction (RT-PCR) assay. It can be considered as an early, rapid, easily distributed, time-efficient, and almost no-cost diagnosis technique complying with social distancing restrictions during COVID-19 pandemic.
引用
收藏
页数:25
相关论文
共 87 条
[1]  
Ai T., 2020, RADIOLOGY 200642, V296, P200642, DOI DOI 10.1148/RADIOL.2020200642
[2]   A novel framework for rapid diagnosis of COVID-19 on computed tomography scans [J].
Akram, Tallha ;
Attique, Muhammad ;
Gul, Salma ;
Shahzad, Aamir ;
Altaf, Muhammad ;
Naqvi, S. Syed Rameez ;
Damasevicius, Robertas ;
Maskeliunas, Rytis .
PATTERN ANALYSIS AND APPLICATIONS, 2021, 24 (03) :951-964
[3]   Estimating Left Ventricle Ejection Fraction Levels Using Circadian Heart Rate Variability Features and Support Vector Regression Models [J].
Alkhodari, Mohanad ;
Jelinek, Herbert F. ;
Werghi, Naoufel ;
Hadjileontiadis, Leontios J. ;
Khandoker, Ahsan H. .
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, 2021, 25 (03) :746-754
[4]   Convolutional and recurrent neural networks for the detection of valvular heart diseases in phonocardiogram recordings [J].
Alkhodari, Mohanad ;
Fraiwan, Luay .
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE, 2021, 200
[5]   Engaging nurses in health policy in the era of COVID-19 [J].
Anders, Robert L. .
NURSING FORUM, 2021, 56 (01) :89-94
[6]  
Andres E, 2018, J Med Life, V11, P89
[7]  
Bagad P., 2020, ARXIV PREPRINT ARXIV
[8]  
Barrett KE, 2016, Ganong's review of medical physiology
[9]   The use of the area under the roc curve in the evaluation of machine learning algorithms [J].
Bradley, AP .
PATTERN RECOGNITION, 1997, 30 (07) :1145-1159
[10]   Exploring Automatic Diagnosis of COVID-19 from Crowdsourced Respiratory Sound Data [J].
Brown, Chloe ;
Chauhan, Jagmohan ;
Grammenos, Andreas ;
Han, Jing ;
Hasthanasombat, Apinan ;
Spathis, Dimitris ;
Xia, Tong ;
Cicuta, Pietro ;
Mascolo, Cecilia .
KDD '20: PROCEEDINGS OF THE 26TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING, 2020, :3474-3484