Dual-Masked Autoencoders: Application to Multi-labeled Pediatric Thoracic Diseases

Pediatric thoracic diseases present significant health risks to children. While chest X-rays are commonly used for diagnosing thoracic diseases, interpreting pediatric images comes with unique challenges such as anatomical variations, developmental differences, and potential artifacts. Deep learning...

Full description

Saved in:
Bibliographic Details
Published in:IEEE access Vol. 12; p. 1
Main Authors: Yoon, Taeyoung, Kang, Daesung
Format: Journal Article
Language:English
Published: IEEE 01.01.2024
Subjects:
ISSN:2169-3536, 2169-3536
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Pediatric thoracic diseases present significant health risks to children. While chest X-rays are commonly used for diagnosing thoracic diseases, interpreting pediatric images comes with unique challenges such as anatomical variations, developmental differences, and potential artifacts. Deep learning offers promise in addressing these challenges, yet its effectiveness is hindered by the limited availability of pediatric chest X-ray data. To overcome this limitation, we introduce the dual-masked autoencoders (dual-MAE) algorithm, consisting of online and target networks with encoder and decoder modules. These networks are optimized by minimizing three losses: between the reconstructed image of the online network and the target network, between the input image and the reconstructed image of the online network, and between the input image and the reconstructed image of the target network. To learn efficiently from pediatric chest X-rays, we employ a two-step training strategy: pretraining the dual-MAE model on adult chest X-rays, then fine-tuning it on pediatric X-rays for diagnosing multi-labeled pediatric thoracic diseases. The proposed model exhibited superior performance with the highest mean AUC score (0.752), surpassing the ResNet-34 (0.669) and ViT-S (0.645) trained from scratch. Additionally, the dual-MAE model outperformed the ResNet-34 (0.697) and ViT-S (0.638), both pretrained on the ImageNet dataset and then fine-tuned on pediatric chest X-rays. Despite being pretrained on a significantly smaller number of X-rays compared to the ImageNet dataset, our model demonstrated better performance. Furthermore, it outperformed the ResNet-34 (0.712), ViT-S (0.673), and vanilla MAE method (0.735), all pretrained on adult chest X-rays and fine-tuned on pediatric chest X-rays. Even with only 50% of labeled pediatric chest X-ray images, dual-MAE demonstrated comparable performance to that of the vanilla MAE method and outperformed ResNet-34 and ViT-S fine-tuned with 100% labeled pediatric chest X-ray images.
AbstractList Pediatric thoracic diseases present significant health risks to children. While chest X-rays are commonly used for diagnosing thoracic diseases, interpreting pediatric images comes with unique challenges such as anatomical variations, developmental differences, and potential artifacts. Deep learning offers promise in addressing these challenges, yet its effectiveness is hindered by the limited availability of pediatric chest X-ray data. To overcome this limitation, we introduce the dual-masked autoencoders (dual-MAE) algorithm, consisting of online and target networks with encoder and decoder modules. These networks are optimized by minimizing three losses: between the reconstructed image of the online network and the target network, between the input image and the reconstructed image of the online network, and between the input image and the reconstructed image of the target network. To learn efficiently from pediatric chest X-rays, we employ a two-step training strategy: pretraining the dual-MAE model on adult chest X-rays, then fine-tuning it on pediatric X-rays for diagnosing multi-labeled pediatric thoracic diseases. The proposed model exhibited superior performance with the highest mean AUC score (0.752), surpassing the ResNet-34 (0.669) and ViT-S (0.645) trained from scratch. Additionally, the dual-MAE model outperformed the ResNet-34 (0.697) and ViT-S (0.638), both pretrained on the ImageNet dataset and then fine-tuned on pediatric chest X-rays. Despite being pretrained on a significantly smaller number of X-rays compared to the ImageNet dataset, our model demonstrated better performance. Furthermore, it outperformed the ResNet-34 (0.712), ViT-S (0.673), and vanilla MAE method (0.735), all pretrained on adult chest X-rays and fine-tuned on pediatric chest X-rays. Even with only 50% of labeled pediatric chest X-ray images, dual-MAE demonstrated comparable performance to that of the vanilla MAE method and outperformed ResNet-34 and ViT-S fine-tuned with 100% labeled pediatric chest X-ray images.
Author Yoon, Taeyoung
Kang, Daesung
Author_xml – sequence: 1
  givenname: Taeyoung
  surname: Yoon
  fullname: Yoon, Taeyoung
  organization: Department of Healthcare Information Technology, Inje University, Gimhae-si, Republic of Korea
– sequence: 2
  givenname: Daesung
  orcidid: 0000-0001-6136-3698
  surname: Kang
  fullname: Kang, Daesung
  organization: Department of Medical Information Technology, Inje University, Gimhae-si, Republic of Korea
BookMark eNp9kMtOAzEMRSMEEq9-ASzmB6Ykk0knYVeVp9QKBGUdORkHUoamStIFf8_QFgmxwBtbVz5X9j0m-8uwRELOGB0yRtXFeDK5fn4eVrSqh7xmUkmxR44qNlIlF3y0_2s-JIOUFrQv2UuiOSJPV2voyhmkd2yL8ToHXNrQYkyXxXi16ryF7MOyyKGYrbvsyw4Mdv3qI7YecvS2mL-FCLYfrnxCSJhOyYGDLuFg10_Iy831fHJXTh9u7yfjaWm5ZLlsqDQMTK2oBEsBgKtKtQqqxhmsjHNc1MoKRZ00bUMFqtY2wlkhTGVljfyE3G992wALvYr-A-KnDuD1RgjxVUPM3naoHUcpW26Q9uSICilrEI1TwDkzgNB7qa2XjSGliE5bnzev5wi-04zq76z1Nmv9nbXeZd2z_A_7c8v_1PmW8oj4ixANrXnNvwDVuY3f
CODEN IAECCG
CitedBy_id crossref_primary_10_1038_s41598_025_15704_3
crossref_primary_10_1016_j_engappai_2025_112055
Cites_doi 10.1038/s41598-023-30208-8
10.1109/ACCESS.2023.3346315
10.1109/CVPR42600.2020.00975
10.1007/s00247-022-05368-w
10.1109/WACV56688.2023.00358
10.1109/CVPR.2016.90
10.1038/s41598-020-73831-5
10.1109/TKDE.2013.39
10.1038/s41598-024-56819-3
10.1038/s41597-023-02102-5
10.5555/3524938.3525087
10.1109/CVPR.2017.369
10.5555/3495724.3497510
10.1007/978-3-030-11149-6_10
10.1007/s10554-023-03039-x
10.1109/CVPR52688.2022.01553
10.1609/aaai.v33i01.3301590
10.15172/pneu.2015.6/621
10.3390/bioengineering10080901
ContentType Journal Article
DBID 97E
ESBDL
RIA
RIE
AAYXX
CITATION
DOA
DOI 10.1109/ACCESS.2024.3418985
DatabaseName IEEE All-Society Periodicals Package (ASPP) 2005–Present
IEEE Xplore Open Access Journals
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Electronic Library (IEL)
CrossRef
DOAJ Directory of Open Access Journals
DatabaseTitle CrossRef
DatabaseTitleList

Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
– sequence: 2
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
EISSN 2169-3536
EndPage 1
ExternalDocumentID oai_doaj_org_article_f3e88d3be0c84605884a57f9a331baea
10_1109_ACCESS_2024_3418985
10570434
Genre orig-research
GrantInformation_xml – fundername: National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT)
  grantid: RS-2023-00249104
GroupedDBID 0R~
5VS
6IK
97E
AAJGR
ABAZT
ABVLG
ACGFS
ADBBV
ALMA_UNASSIGNED_HOLDINGS
BCNDV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
EBS
ESBDL
GROUPED_DOAJ
IPLJI
JAVBF
KQ8
M43
M~E
O9-
OCL
OK1
RIA
RIE
RNS
4.4
AAYXX
AGSQL
CITATION
EJD
ID FETCH-LOGICAL-c381t-708b1ab4908ac0aaa3929d9a27fbe2bff3549c590f8bd705e9dc75fc55b2c84e3
IEDL.DBID DOA
ISICitedReferencesCount 4
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001258805800001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 2169-3536
IngestDate Fri Oct 03 12:52:06 EDT 2025
Tue Nov 18 22:35:00 EST 2025
Sat Nov 29 06:25:50 EST 2025
Wed Aug 27 02:07:23 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Language English
License https://creativecommons.org/licenses/by-nc-nd/4.0
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c381t-708b1ab4908ac0aaa3929d9a27fbe2bff3549c590f8bd705e9dc75fc55b2c84e3
ORCID 0000-0001-6136-3698
0009-0008-9454-8590
OpenAccessLink https://doaj.org/article/f3e88d3be0c84605884a57f9a331baea
PageCount 1
ParticipantIDs ieee_primary_10570434
doaj_primary_oai_doaj_org_article_f3e88d3be0c84605884a57f9a331baea
crossref_citationtrail_10_1109_ACCESS_2024_3418985
crossref_primary_10_1109_ACCESS_2024_3418985
PublicationCentury 2000
PublicationDate 2024-01-01
PublicationDateYYYYMMDD 2024-01-01
PublicationDate_xml – month: 01
  year: 2024
  text: 2024-01-01
  day: 01
PublicationDecade 2020
PublicationTitle IEEE access
PublicationTitleAbbrev Access
PublicationYear 2024
Publisher IEEE
Publisher_xml – name: IEEE
References ref13
ref15
(ref23) 2024
ref11
(ref25) 2024
ref2
Wu (ref22) 2023
ref17
ref16
ref19
Hart (ref1) 2019
ref26
Rajpurkar (ref9) 2017
(ref24) 2024
ref20
ref21
Raghu (ref12); 32
Bhusal (ref10) 2022
ref28
Devlin (ref27); 1
Zhou (ref14) 2022
ref29
ref8
ref7
ref4
ref3
ref6
ref5
Dosovitskiy (ref18) 2020
References_xml – ident: ref5
  doi: 10.1038/s41598-023-30208-8
– ident: ref6
  doi: 10.1109/ACCESS.2023.3346315
– ident: ref29
  doi: 10.1109/CVPR42600.2020.00975
– volume: 1
  start-page: 2
  volume-title: Proc. NAACL-HLT
  ident: ref27
  article-title: BERT: Pre-training of deep bidirectional transformers for language understanding
– ident: ref4
  doi: 10.1007/s00247-022-05368-w
– year: 2023
  ident: ref22
  article-title: K-diag: Knowledge-enhanced disease diagnosis in radiographic imaging
  publication-title: arXiv:2302.11557
– ident: ref20
  doi: 10.1109/WACV56688.2023.00358
– ident: ref26
  doi: 10.1109/CVPR.2016.90
– volume: 32
  start-page: 1
  volume-title: Proc. Adv. Neural Inf. Process. Syst.
  ident: ref12
  article-title: Transfusion: Understanding transfer learning for medical imaging
– ident: ref3
  doi: 10.1038/s41598-020-73831-5
– volume-title: CheXpert Dataset Repository
  year: 2024
  ident: ref24
– ident: ref19
  doi: 10.1109/TKDE.2013.39
– year: 2020
  ident: ref18
  article-title: An image is worth 16×16 words: Transformers for image recognition at scale
  publication-title: arXiv:2010.11929
– ident: ref16
  doi: 10.1038/s41598-024-56819-3
– ident: ref21
  doi: 10.1038/s41597-023-02102-5
– volume-title: PediCXR Dataset Repository
  year: 2024
  ident: ref25
– ident: ref28
  doi: 10.5555/3524938.3525087
– ident: ref8
  doi: 10.1109/CVPR.2017.369
– volume-title: ChestX-Ray14 Dataset Repository
  year: 2024
  ident: ref23
– ident: ref17
  doi: 10.5555/3495724.3497510
– start-page: 107
  volume-title: Diseases of the Chest, Breast, Heart and Vessels 2019–2022: Diagnostic and Interventional Imaging
  year: 2019
  ident: ref1
  article-title: Pediatric chest disorders: Practical imaging approach to diagnosis
  doi: 10.1007/978-3-030-11149-6_10
– ident: ref11
  doi: 10.1007/s10554-023-03039-x
– ident: ref13
  doi: 10.1109/CVPR52688.2022.01553
– year: 2022
  ident: ref14
  article-title: Self pre-training with masked autoencoders for medical image classification and segmentation
  publication-title: arXiv:2203.05573
– ident: ref7
  doi: 10.1609/aaai.v33i01.3301590
– year: 2022
  ident: ref10
  article-title: Multi-label classification of thoracic diseases using dense convolutional network on chest radiographs
  publication-title: arXiv:2202.03583
– ident: ref2
  doi: 10.15172/pneu.2015.6/621
– ident: ref15
  doi: 10.3390/bioengineering10080901
– year: 2017
  ident: ref9
  article-title: CheXNet: Radiologist-level pneumonia detection on chest X-rays with deep learning
  publication-title: arXiv:1711.05225
SSID ssj0000816957
Score 2.3261254
Snippet Pediatric thoracic diseases present significant health risks to children. While chest X-rays are commonly used for diagnosing thoracic diseases, interpreting...
SourceID doaj
crossref
ieee
SourceType Open Website
Enrichment Source
Index Database
Publisher
StartPage 1
SubjectTerms Accuracy
Adult chest X-rays
Biomedical imaging
Deep learning
Diseases
dual-masked autoencoders
Image reconstruction
masked autoencoders
pediatric chest X-rays
pediatric thoracic diseases
Pneumonia
X-ray imaging
SummonAdditionalLinks – databaseName: IEEE Electronic Library (IEL)
  dbid: RIE
  link: http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LT8QgECZqPOjBt3F9pQePdoXSFvBWV40HNcasxlvDY4jGza5xW3-_QGtdD5p4I4SmwEc7DMN8H0JHOVHCEAJxLlIdp7lzUFSObawAMsuo5SRc-X-8Zre3_OlJ3LXJ6iEXBgDC5TPo-2KI5ZuJrv1R2YnXpMUpTefRPGOsSdbqDlS8goTIWMssRLA4KQYDNwjnAyZp3_2sufCCyTPWJ5D0_1BVCUblcvWf3VlDK-3uMSoauNfRHIw30PIMp-Amuj-v5Si-kdNXMFFRVxPPVOlvK59GxXewOqomUci9jd0qcJbHRJ1mRzR8dstCu8J5E7yZbqGHy4vh4CpuhRNi7QxwFTPMFZHKx_SkxlJKvwkyQibMKkiUtdR5hToT2HJlGM5AGM0yq7NMJZqnQLfRwngyhh0UJVxxCjkBC26nAs4_cg4e0aCkMlpz0UPJ14SWumUV9-IWozJ4F1iUDQqlR6FsUeih4-6ht4ZU4-_mZx6prqlnxA4VDo2y_cBKS4FzQxVgNwIf6-WpzJgVklKiJMge2vIIzryvAW_3l_o9tOT70Jy27KOF6r2GA7SoP6qX6fthWHqfqjLXqw
  priority: 102
  providerName: IEEE
Title Dual-Masked Autoencoders: Application to Multi-labeled Pediatric Thoracic Diseases
URI https://ieeexplore.ieee.org/document/10570434
https://doaj.org/article/f3e88d3be0c84605884a57f9a331baea
Volume 12
WOSCitedRecordID wos001258805800001&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: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  customDbUrl:
  eissn: 2169-3536
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0000816957
  issn: 2169-3536
  databaseCode: DOA
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://www.doaj.org/
  providerName: Directory of Open Access Journals
– providerCode: PRVHPJ
  databaseName: ROAD: Directory of Open Access Scholarly Resources
  customDbUrl:
  eissn: 2169-3536
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0000816957
  issn: 2169-3536
  databaseCode: M~E
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://road.issn.org
  providerName: ISSN International Centre
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV09T8MwELUQYoAB8VFE-agyMBKw4zi22UpbxNBWCBXEFtnOWSBQi2jKyG_HdkIJCywsURQ5ifN88t3L2e8QOsmIlgUhEGcyNXGaOYKiM2xjDcAsp1aQsOT_fsjHY_HwIG8apb78mrBKHrgC7txSEKKgGrARIYcnUsW4lYpSohWE0MhFPQ0yFeZgQTLJeC0zRLA87_Z67oscIUzSMzdzC-mrJzdcUVDs_1FiJXiYqy20WYeGUbfq0jZagekO2mgIBu6i2_5CvcQjNX-GIuouypmXofRLkS-i7ncmOipnUdhYGw-Vdm6liJYFOaLJoxtz4076VWZm3kJ3V4NJ7zquqyLExnnXMuZYaKK0T9gpg5VSPsIppEq41ZBoa6mjfIZJbIUuOGYgC8OZNYzpxEEIdA-tTmdT2EdRIrSgkBGw4LAFR34ceyMGtNKFMUK2UfIFUG5qyXBfueIlD9QBy7xCNfeo5jWqbXS6vOm1Usz4vfmlR37Z1MtdhwvOCPLaCPK_jKCNWn7cGu9jHKc0PfiPhx-idd_h6r_LEVot3xZwjNbMe_k0f-sEu3PH0cegE3YPfgI49Nyn
linkProvider Directory of Open Access Journals
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3PT9swFH7a2KTBYbCtiMLYcthxATu2E5tbV0BMKxWauolb5B_PAoFaRFP-fmwndN1hk7hZlqPY_pw8Pz-_7wP4UlKjHKWYl4rbnJfBQTEl8blBFL5iXtJ05f_3qBqP5eWluuiS1VMuDCKmy2d4EIsplu9mdhGPyg6jJi3hjL-EV4LzgrbpWssjlaghoUTVcQtRog4Hw2EYRvACC34QftdSRcnkFfuTaPr_0lVJZuV085kd2oK33f4xG7SAv4MXOH0PGyusgh_g5_FC3-bnen6DLhssmlnkqoz3lY-ywZ9wddbMspR9m4d1EGyPy5aqHdnkKiwMGwrHbfhm3oNfpyeT4VneSSfkNpjgJq-INFSbGNXTlmit4zbIKV1U3mBhvGfBL7RCES-Nq4hA5WwlvBXCFFZyZNuwNp1NcQeyQhrJsKToMexVMHhIwcWjFo02zlqp-lA8TWhtO17xKG9xWyf_gqi6RaGOKNQdCn34unzorqXV-H_zbxGpZdPIiZ0qAhp194nVnqGUjhkkYQQx2iu5FpVXmjFqNOo-9CKCK-9rwdv9R_1neHM2OR_Vo-_jH3uwHvvTnr18hLXmfoH78No-NNfz-09pGT4CCAPa8g
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=Dual-Masked+Autoencoders%3A+Application+to+Multi-Labeled+Pediatric+Thoracic+Diseases&rft.jtitle=IEEE+access&rft.au=Taeyoung+Yoon&rft.au=Daesung+Kang&rft.date=2024-01-01&rft.pub=IEEE&rft.eissn=2169-3536&rft.volume=12&rft.spage=87981&rft.epage=87990&rft_id=info:doi/10.1109%2FACCESS.2024.3418985&rft.externalDBID=DOA&rft.externalDocID=oai_doaj_org_article_f3e88d3be0c84605884a57f9a331baea
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2169-3536&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2169-3536&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2169-3536&client=summon