Acoustic COVID-19 Detection Using Multiple Instance Learning

In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA chall...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	IEEE journal of biomedical and health informatics Ročník 29; číslo 1; s. 620 - 630
Hlavní autori:	Reiter, Michael, Pernkopf, Franz
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	United States IEEE 01.01.2025
Predmet:	Algorithms Annotations Antigens audio-based infection prediction Bioinformatics Biological system modeling Costs coswara COVID-19 COVID-19 - diagnosis COVID-19 - physiopathology crowdsourced datasets DiCOVA Feature extraction Humans Labeling Machine Learning multiple instance learning Multiple-Instance Learning Algorithms Pandemics Predictive models SARS-CoV-2 Signal Processing, Computer-Assisted
ISSN:	2168-2194, 2168-2208, 2168-2208
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Abstract	In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA challenge was created. It is based on the Coswara dataset offering the recording categories cough, speech, breath and vowel phonation. Recording durations vary greatly, ranging from one second to over a minute. A base model is pre-trained on random, short time intervals. Subsequently, a Multiple Instance Learning (MIL) model based on self-attention is incorporated to make collective predictions for multiple time segments within each audio recording, taking advantage of longer durations. In order to compete in the fusion category of the DiCOVA challenge, we utilize a linear regression approach among other fusion methods to combine predictions from the most successful models associated with each sound modality. The application of the MIL approach significantly improves generalizability, leading to an AUC ROC score of 86.6% in the fusion category. By incorporating previously unused data, including the sound modality 'sustained vowel phonation' and patient metadata, we were able to significantly improve our previous results reaching a score of 92.2%.
AbstractList	In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA challenge was created. It is based on the Coswara dataset offering the recording categories cough, speech, breath and vowel phonation. Recording durations vary greatly, ranging from one second to over a minute. A base model is pre-trained on random, short time intervals. Subsequently, a Multiple Instance Learning (MIL) model based on self-attention is incorporated to make collective predictions for multiple time segments within each audio recording, taking advantage of longer durations. In order to compete in the fusion category of the DiCOVA challenge, we utilize a linear regression approach among other fusion methods to combine predictions from the most successful models associated with each sound modality. The application of the MIL approach significantly improves generalizability, leading to an AUC ROC score of 86.6% in the fusion category. By incorporating previously unused data, including the sound modality 'sustained vowel phonation' and patient metadata, we were able to significantly improve our previous results reaching a score of 92.2%. In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA challenge was created. It is based on the Coswara dataset offering the recording categories cough, speech, breath and vowel phonation. Recording durations vary greatly, ranging from one second to over a minute. A base model is pre-trained on random, short time intervals. Subsequently, a Multiple Instance Learning (MIL) model based on self-attention is incorporated to make collective predictions for multiple time segments within each audio recording, taking advantage of longer durations. In order to compete in the fusion category of the DiCOVA challenge, we utilize a linear regression approach among other fusion methods to combine predictions from the most successful models associated with each sound modality. The application of the MIL approach significantly improves generalizability, leading to an AUC ROC score of 86.6% in the fusion category. By incorporating previously unused data, including the sound modality 'sustained vowel phonation' and patient metadata, we were able to significantly improve our previous results reaching a score of 92.2%.In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool for audio recordings could enable widespread testing at low costs. In order to achieve comparability between such algorithms, the DiCOVA challenge was created. It is based on the Coswara dataset offering the recording categories cough, speech, breath and vowel phonation. Recording durations vary greatly, ranging from one second to over a minute. A base model is pre-trained on random, short time intervals. Subsequently, a Multiple Instance Learning (MIL) model based on self-attention is incorporated to make collective predictions for multiple time segments within each audio recording, taking advantage of longer durations. In order to compete in the fusion category of the DiCOVA challenge, we utilize a linear regression approach among other fusion methods to combine predictions from the most successful models associated with each sound modality. The application of the MIL approach significantly improves generalizability, leading to an AUC ROC score of 86.6% in the fusion category. By incorporating previously unused data, including the sound modality 'sustained vowel phonation' and patient metadata, we were able to significantly improve our previous results reaching a score of 92.2%.
Author	Pernkopf, Franz Reiter, Michael
Author_xml	– sequence: 1 givenname: Michael surname: Reiter fullname: Reiter, Michael organization: Christian Doppler Laboratory for Dependable Intelligent Systems in Harsh Environments, Graz University of Technology, Graz, Austria – sequence: 2 givenname: Franz orcidid: 0000-0002-6356-3367 surname: Pernkopf fullname: Pernkopf, Franz email: pernkopf@tugraz.at organization: Christian Doppler Laboratory for Dependable Intelligent Systems in Harsh Environments, Graz University of Technology, Graz, Austria
BackLink	https://www.ncbi.nlm.nih.gov/pubmed/39365725$$D View this record in MEDLINE/PubMed
BookMark	eNpNkM9PwjAUxxuDEUT-ABNjdvQy7I-taxMvCCozGC7itWm7NzMzOly3g_-9JYCxl9f0fd5Lv59LNHCNA4SuCZ4SguX96-Myn1JMkylLskRm6RkaUcJFTCkWg9OdyGSIJt5_4XBEeJL8Ag2ZZDzNaDpCDzPb9L6rbDRff-SLmMhoAR3YrmpctPGV-4ze-rqrdjVEufOddhaiFejWhdYVOi917WFyrGO0eX56ny_j1foln89WsaVZ0sXCSgbGAMEgLBXMFiCZzAwxkhSESS55WnItBbUhj9CmSC01usSpKYnhCRuju8PeXdt89-A7ta28hbrWDsLvFSOEEZrSjAb09oj2ZguF2rXVVrc_6pQ4AOQA2LbxvoXyDyFY7cWqvVi1F6uOYsPMzWGmAoB_fIZ5sMp-ARy2cTQ
CODEN	IJBHA9
Cites_doi	10.1038/s41597-021-00937-4 10.1038/s41746-021-00553-x 10.1016/j.compbiomed.2021.104572 10.1109/JBHI.2023.3339700 10.21437/Interspeech.2022-10389 10.5555/3045118.3045167 10.1109/OJEMB.2020.3026928 10.1371/journal.pone.0177926 10.1007/s00521-024-10150-0 10.1109/ICCV.2017.324 10.1016/j.ijid.2021.07.010 10.1109/ICASSP43922.2022.9747188 10.1016/j.aej.2021.08.070 10.1007/s11263-019-01198-w 10.1142/S021821302350046X 10.1109/CVPR.2016.90 10.21437/Interspeech.2020-2768 10.1109/ICASSP.2017.7952132 10.23919/eusipco58844.2023.10289774 10.1007/978-3-031-16474-3_13 10.1109/TBME.2022.3156293 10.1145/3394486.3412865
ContentType	Journal Article
DBID	97E RIA RIE AAYXX CITATION CGR CUY CVF ECM EIF NPM 7X8
DOI	10.1109/JBHI.2024.3474975
DatabaseName	IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE/IET Electronic Library (IEL) (UW System Shared) CrossRef Medline MEDLINE MEDLINE (Ovid) MEDLINE MEDLINE PubMed MEDLINE - Academic
DatabaseTitle	CrossRef MEDLINE Medline Complete MEDLINE with Full Text PubMed MEDLINE (Ovid) MEDLINE - Academic
DatabaseTitleList	MEDLINE MEDLINE - Academic
Database_xml	– sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: RIE name: IEEE/IET Electronic Library (IEL) (UW System Shared) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher – sequence: 3 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database
DeliveryMethod	fulltext_linktorsrc
Discipline	Medicine
EISSN	2168-2208
EndPage	630
ExternalDocumentID	39365725 10_1109_JBHI_2024_3474975 10706000
Genre	orig-research Research Support, Non-U.S. Gov't Journal Article
GrantInformation_xml	– fundername: Austrian Federal Ministry of Labour and Economy – fundername: Österreichische Nationalstiftung für Forschung, Technologie und Entwicklung; National Foundation for Research, Technology and Development funderid: 10.13039/100010132 – fundername: Christian Doppler Forschungsgesellschaft; Christian Doppler Research Association funderid: 10.13039/501100006012
GroupedDBID	0R~ 4.4 6IF 6IH 6IK 97E AAJGR AARMG AASAJ AAWTH ABAZT ABQJQ ABVLG ACIWK ACPRK AENEX AFRAH AGQYO AGSQL AHBIQ AKJIK AKQYR ALMA_UNASSIGNED_HOLDINGS BEFXN BFFAM BGNUA BKEBE BPEOZ EBS EJD HZ~ IFIPE IPLJI JAVBF M43 O9- OCL PQQKQ RIA RIE RNS AAYXX CITATION CGR CUY CVF ECM EIF NPM RIG 7X8
ID	FETCH-LOGICAL-c274t-8c93ebbe10e8c283cde9397b1b91d1396965f6a982c2028abd5c2baf05bf1b643
IEDL.DBID	RIE
ISICitedReferencesCount	0
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001392851400002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	2168-2194 2168-2208
IngestDate	Sun Sep 28 06:02:07 EDT 2025 Thu Jul 24 02:16:48 EDT 2025 Sat Nov 29 04:18:40 EST 2025 Wed Aug 27 01:58:00 EDT 2025
IsPeerReviewed	true
IsScholarly	true
Issue	1
Language	English
License	https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c274t-8c93ebbe10e8c283cde9397b1b91d1396965f6a982c2028abd5c2baf05bf1b643
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23
ORCID	0000-0002-6356-3367
PMID	39365725
PQID	3113125272
PQPubID	23479
PageCount	11
ParticipantIDs	crossref_primary_10_1109_JBHI_2024_3474975 pubmed_primary_39365725 proquest_miscellaneous_3113125272 ieee_primary_10706000
PublicationCentury	2000
PublicationDate	2025-Jan. 2025-1-00 2025-01-00 20250101
PublicationDateYYYYMMDD	2025-01-01
PublicationDate_xml	– month: 01 year: 2025 text: 2025-Jan.
PublicationDecade	2020
PublicationPlace	United States
PublicationPlace_xml	– name: United States
PublicationTitle	IEEE journal of biomedical and health informatics
PublicationTitleAbbrev	JBHI
PublicationTitleAlternate	IEEE J Biomed Health Inform
PublicationYear	2025
Publisher	IEEE
Publisher_xml	– name: IEEE
References	Subirana (ref7) 2020 ref13 ref35 ref12 ref34 ref15 Sharma (ref6) 2020 ref30 ref33 ref2 ref1 Chowdhury (ref22) 2022; 145 ref16 ref19 ref18 Ulyanov (ref28) 2016 Xia (ref10) 2021; 1 Kim (ref14) 2023; 13 ref24 ref23 ref26 ref20 ref21 Kim (ref29) 2022 Messner (ref4) 2020; 122 Hartquist (ref32) 2022 ref27 ref8 Ilse (ref31) 2018 ref9 Chen (ref17) 2024; 5 ref3 Chaudhari (ref11) 2011 Ahmad (ref5) 2023; 10 Tompson (ref25) 2014
References_xml	– ident: ref8 doi: 10.1038/s41597-021-00937-4 – volume: 5 volume-title: Healthcare Analytics year: 2024 ident: ref17 article-title: A vision transformer machine learning model for COVID-19 diagnosis using chest X-ray images – ident: ref35 doi: 10.1038/s41746-021-00553-x – ident: ref21 doi: 10.1016/j.compbiomed.2021.104572 – volume: 1 volume-title: Proc. Int. Conf. Neural Inf. Process. Syst. Datasets Benchmarks year: 2021 ident: ref10 article-title: COVID-19 sounds: A large-scale audio dataset for digital respiratory screening – ident: ref12 doi: 10.1109/JBHI.2023.3339700 – ident: ref20 doi: 10.21437/Interspeech.2022-10389 – ident: ref26 doi: 10.5555/3045118.3045167 – ident: ref34 doi: 10.1109/OJEMB.2020.3026928 – ident: ref2 doi: 10.1371/journal.pone.0177926 – ident: ref16 doi: 10.1007/s00521-024-10150-0 – ident: ref33 doi: 10.1109/ICCV.2017.324 – ident: ref1 doi: 10.1016/j.ijid.2021.07.010 – year: 2022 ident: ref32 article-title: Fine-tuning resnet-18 for audio classification – ident: ref18 doi: 10.1109/ICASSP43922.2022.9747188 – year: 2016 ident: ref28 article-title: Instance normalization: The missing ingredient for fast stylization – ident: ref19 doi: 10.1016/j.aej.2021.08.070 – start-page: 2127 volume-title: Pro. 35th Int. Conf. Mach. Learn. year: 2018 ident: ref31 article-title: Attention-based deep multiple instance learning – start-page: 648 volume-title: Proc. 2015 IEEE Conf. Comput. Vis. Pattern Recognit. year: 2014 ident: ref25 article-title: Efficient object localization using convolutional networks – ident: ref27 doi: 10.1007/s11263-019-01198-w – ident: ref15 doi: 10.1142/S021821302350046X – ident: ref24 doi: 10.1109/CVPR.2016.90 – volume-title: Proc. Interspeech year: 2020 ident: ref6 article-title: Coswara a database of breathing, cough, and voice sounds for COVID-19 diagnosis doi: 10.21437/Interspeech.2020-2768 – year: 2011 ident: ref11 article-title: Virufy: Global applicability of crowdsourced and clinical datasets for AI detection of COVID-19 from cough – ident: ref13 doi: 10.1109/ICASSP.2017.7952132 – volume: 13 issue: 4 volume-title: Appl. Sci. year: 2023 ident: ref14 article-title: COVID-19 detection model with acoustic features from cough sound and its application – volume: 10 issue: 11 volume-title: Bioeng. year: 2023 ident: ref5 article-title: COVID-19 detection via ultra-low-dose X-ray images enabled by deep learning – year: 2020 ident: ref7 article-title: Hi sigma, do i have the coronavirus?: Call for a new artificial intelligence approach to support health care professionals dealing with the COVID-19 pandemic – year: 2022 ident: ref29 article-title: QTI submission to DCASE 2021: Residual normalization for device-imbalanced acoustic scene classification with efficient design – ident: ref30 doi: 10.23919/eusipco58844.2023.10289774 – volume: 145 volume-title: Comput. Biol. Med. year: 2022 ident: ref22 article-title: Machine learning for detecting COVID-19 from cough sounds: An ensemble-based MCDM method – volume: 122 volume-title: Comput. Biol. Med. year: 2020 ident: ref4 article-title: Multi-channel lung sound classification with convolutional recurrent neural networks – ident: ref23 doi: 10.1007/978-3-031-16474-3_13 – ident: ref3 doi: 10.1109/TBME.2022.3156293 – ident: ref9 doi: 10.1145/3394486.3412865
SSID	ssj0000816896
Score	2.4256747
Snippet	In the COVID-19 pandemic, a rigorous testing scheme was crucial. However, tests can be time-consuming and expensive. A machine learning-based diagnostic tool...
SourceID	proquest pubmed crossref ieee
SourceType	Aggregation Database Index Database Publisher
StartPage	620
SubjectTerms	Algorithms Annotations Antigens audio-based infection prediction Bioinformatics Biological system modeling Costs coswara COVID-19 COVID-19 - diagnosis COVID-19 - physiopathology crowdsourced datasets DiCOVA Feature extraction Humans Labeling Machine Learning multiple instance learning Multiple-Instance Learning Algorithms Pandemics Predictive models SARS-CoV-2 Signal Processing, Computer-Assisted
Title	Acoustic COVID-19 Detection Using Multiple Instance Learning
URI	https://ieeexplore.ieee.org/document/10706000 https://www.ncbi.nlm.nih.gov/pubmed/39365725 https://www.proquest.com/docview/3113125272
Volume	29
WOSCitedRecordID	wos001392851400002&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText	1
inHoldings	1
isFullTextHit
isPrint
journalDatabaseRights	– providerCode: PRVIEE databaseName: IEEE/IET Electronic Library (IEL) (UW System Shared) customDbUrl: eissn: 2168-2208 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000816896 issn: 2168-2194 databaseCode: RIE dateStart: 20130101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV07T8MwELagQoiFZ4HyqILEhJRix3lZYiktVYtUYADULbIvF8SSIgj8fs5Oirp0YPOQWNbns-67h_0xdikgBGVA-5rcqZUwk76mQMvHSIa5lMCxqMUmkoeHdDZTT81ldXcXBhFd8xn27NDV8vM5fNtUGZ1w-9YLpwh9PUni-rLWX0LFKUg4Pa6ABj6dxLCpYgquru9vxxOKBoOwJ8OElmQVa6SScZRYkewll-Q0VlbTTed2Rjv_XPAu2274pdevDWKPrWG5zzanTQX9gN30Ye4EvLzB4-tk6AvlDbFy_Vil5_oHvGnTY-hNHHUE9JpHWN_a7GV09zwY-42Cgg8UbVZ-CkqiMSg4pkBEAnJURECMMErkxP1iFUdFrFUaAGGTapNHEBhd8MgUwhBZOWStcl7iMfMUcowKmQM3eQga0lQihc8aoiBMTMQ77GoBYvZRP5SRuQCDq8yCn1nwswb8DmtbsJY-rHHqsIsF7hmZua1d6BIJlkwKIYmLBUnQYUf1hvz9vdjHkxWznrKtwKr2usTJGWtVn994zjbgp3r_-uySLc3SrrOlX6a4wO8
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NT-MwEB2t2BVwYZePXcp-ECROSAE7jptY4sLCohbawqEgbpE9maz2kqLS8vt37KSolx64-ZBY1vNYfs8z9gM4lpiicWhjy9uptzBTsWWhFZNWaakUCqoas4lsNMqfnsx9e1k93IUholB8Rqe-GXL55QTn_qiMV7h_60WwQv-oUxY-zXWttyOV4CERHLkSbsS8FtM2jymFObv53euzHkzSU5VmPCjvWaOM6urM22QvbUrBZWU14Qwbz_Xndw75C2y1DDO6aEJiGz5QvQPrwzaHvgvnFzgJFl7R5d1j_yqWJrqiWajIqqNQQRAN2yrDqB_II1LUPsP6dw8erv-ML3tx66EQI-vNWZyjUeQcSUE5MpXAkgxTECedkSWzv67p6qprTZ4gY5NbV2pMnK2EdpV0TFe-wlo9qWkfIkOCdKVKFK5M0WKeK2IBbVEnaea06MDJAsTiuXkqowgSQ5jCg1948IsW_A7sebCWPmxw6sDRAveCA91nL2xNDEuhpFTMxpIs6cC3ZkLe_l7M48GKXg9hozceDopBf3T7HTYT7-EbjlF-wNpsOqef8AlfZ_9epr9CRP0HU4XDTg
openUrl	ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Acoustic+COVID-19+Detection+Using+Multiple+Instance+Learning&rft.jtitle=IEEE+journal+of+biomedical+and+health+informatics&rft.au=Reiter%2C+Michael&rft.au=Pernkopf%2C+Franz&rft.date=2025-01-01&rft.issn=2168-2208&rft.eissn=2168-2208&rft.volume=29&rft.issue=1&rft.spage=620&rft_id=info:doi/10.1109%2FJBHI.2024.3474975&rft.externalDBID=NO_FULL_TEXT
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2168-2194&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2168-2194&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2168-2194&client=summon