Deep Learning Algorithm to Detect Cardiac Sarcoidosis From Echocardiographic Movies

Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies. Among the patients who underwent echocardiography from January 2015 to December 2019, we chos...

Full description

Saved in:
Bibliographic Details
Published in:Circulation journal : official journal of the Japanese Circulation Society Vol. 86; no. 1; p. 87
Main Authors: Katsushika, Susumu, Kodera, Satoshi, Nakamoto, Mitsuhiko, Ninomiya, Kota, Kakuda, Nobutaka, Shinohara, Hiroki, Matsuoka, Ryo, Ieki, Hirotaka, Uehara, Masae, Higashikuni, Yasutomi, Nakanishi, Koki, Nakao, Tomoko, Takeda, Norifumi, Fujiu, Katsuhito, Daimon, Masao, Ando, Jiro, Akazawa, Hiroshi, Morita, Hiroyuki, Komuro, Issei
Format: Journal Article
Language:English
Published: Japan 24.12.2021
Subjects:
ISSN:1347-4820, 1347-4820
Online Access:Get more information
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies. Among the patients who underwent echocardiography from January 2015 to December 2019, we chose 151 echocardiographic movies from 50 CS patients and 151 from 149 healthy subjects. We trained two 3D convolutional neural networks (3D-CNN) to identify CS patients using a dataset of 212 echocardiographic movies with and without a transfer learning method (Pretrained algorithm and Non-pretrained algorithm). On an independent set of 41 echocardiographic movies, the area under the receiver-operating characteristic curve (AUC) of the Pretrained algorithm was greater than that of Non-pretrained algorithm (0.842, 95% confidence interval (CI): 0.722-0.962 vs. 0.724, 95% CI: 0.566-0.882, P=0.253). The AUC from the interpretation of the same set of 41 echocardiographic movies by 5 cardiologists was not significantly different from that of the Pretrained algorithm (0.855, 95% CI: 0.735-0.975 vs. 0.842, 95% CI: 0.722-0.962, P=0.885). A sensitivity map demonstrated that the Pretrained algorithm focused on the area of the mitral valve. A 3D-CNN with a transfer learning method may be a promising tool for detecting CS using an echocardiographic movie.
AbstractList Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies. Among the patients who underwent echocardiography from January 2015 to December 2019, we chose 151 echocardiographic movies from 50 CS patients and 151 from 149 healthy subjects. We trained two 3D convolutional neural networks (3D-CNN) to identify CS patients using a dataset of 212 echocardiographic movies with and without a transfer learning method (Pretrained algorithm and Non-pretrained algorithm). On an independent set of 41 echocardiographic movies, the area under the receiver-operating characteristic curve (AUC) of the Pretrained algorithm was greater than that of Non-pretrained algorithm (0.842, 95% confidence interval (CI): 0.722-0.962 vs. 0.724, 95% CI: 0.566-0.882, P=0.253). The AUC from the interpretation of the same set of 41 echocardiographic movies by 5 cardiologists was not significantly different from that of the Pretrained algorithm (0.855, 95% CI: 0.735-0.975 vs. 0.842, 95% CI: 0.722-0.962, P=0.885). A sensitivity map demonstrated that the Pretrained algorithm focused on the area of the mitral valve. A 3D-CNN with a transfer learning method may be a promising tool for detecting CS using an echocardiographic movie.
Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies.BACKGROUNDBecause the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies.Among the patients who underwent echocardiography from January 2015 to December 2019, we chose 151 echocardiographic movies from 50 CS patients and 151 from 149 healthy subjects. We trained two 3D convolutional neural networks (3D-CNN) to identify CS patients using a dataset of 212 echocardiographic movies with and without a transfer learning method (Pretrained algorithm and Non-pretrained algorithm). On an independent set of 41 echocardiographic movies, the area under the receiver-operating characteristic curve (AUC) of the Pretrained algorithm was greater than that of Non-pretrained algorithm (0.842, 95% confidence interval (CI): 0.722-0.962 vs. 0.724, 95% CI: 0.566-0.882, P=0.253). The AUC from the interpretation of the same set of 41 echocardiographic movies by 5 cardiologists was not significantly different from that of the Pretrained algorithm (0.855, 95% CI: 0.735-0.975 vs. 0.842, 95% CI: 0.722-0.962, P=0.885). A sensitivity map demonstrated that the Pretrained algorithm focused on the area of the mitral valve.METHODS AND RESULTSAmong the patients who underwent echocardiography from January 2015 to December 2019, we chose 151 echocardiographic movies from 50 CS patients and 151 from 149 healthy subjects. We trained two 3D convolutional neural networks (3D-CNN) to identify CS patients using a dataset of 212 echocardiographic movies with and without a transfer learning method (Pretrained algorithm and Non-pretrained algorithm). On an independent set of 41 echocardiographic movies, the area under the receiver-operating characteristic curve (AUC) of the Pretrained algorithm was greater than that of Non-pretrained algorithm (0.842, 95% confidence interval (CI): 0.722-0.962 vs. 0.724, 95% CI: 0.566-0.882, P=0.253). The AUC from the interpretation of the same set of 41 echocardiographic movies by 5 cardiologists was not significantly different from that of the Pretrained algorithm (0.855, 95% CI: 0.735-0.975 vs. 0.842, 95% CI: 0.722-0.962, P=0.885). A sensitivity map demonstrated that the Pretrained algorithm focused on the area of the mitral valve.A 3D-CNN with a transfer learning method may be a promising tool for detecting CS using an echocardiographic movie.CONCLUSIONSA 3D-CNN with a transfer learning method may be a promising tool for detecting CS using an echocardiographic movie.
Author Uehara, Masae
Higashikuni, Yasutomi
Akazawa, Hiroshi
Ieki, Hirotaka
Nakao, Tomoko
Ninomiya, Kota
Morita, Hiroyuki
Matsuoka, Ryo
Daimon, Masao
Kodera, Satoshi
Komuro, Issei
Katsushika, Susumu
Nakamoto, Mitsuhiko
Takeda, Norifumi
Kakuda, Nobutaka
Shinohara, Hiroki
Fujiu, Katsuhito
Ando, Jiro
Nakanishi, Koki
Author_xml – sequence: 1
  givenname: Susumu
  surname: Katsushika
  fullname: Katsushika, Susumu
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 2
  givenname: Satoshi
  surname: Kodera
  fullname: Kodera, Satoshi
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 3
  givenname: Mitsuhiko
  surname: Nakamoto
  fullname: Nakamoto, Mitsuhiko
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 4
  givenname: Kota
  surname: Ninomiya
  fullname: Ninomiya, Kota
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 5
  givenname: Nobutaka
  surname: Kakuda
  fullname: Kakuda, Nobutaka
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 6
  givenname: Hiroki
  surname: Shinohara
  fullname: Shinohara, Hiroki
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 7
  givenname: Ryo
  surname: Matsuoka
  fullname: Matsuoka, Ryo
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 8
  givenname: Hirotaka
  surname: Ieki
  fullname: Ieki, Hirotaka
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 9
  givenname: Masae
  surname: Uehara
  fullname: Uehara, Masae
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 10
  givenname: Yasutomi
  surname: Higashikuni
  fullname: Higashikuni, Yasutomi
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 11
  givenname: Koki
  surname: Nakanishi
  fullname: Nakanishi, Koki
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 12
  givenname: Tomoko
  surname: Nakao
  fullname: Nakao, Tomoko
  organization: Department of Clinical Laboratory, The University of Tokyo Hospital
– sequence: 13
  givenname: Norifumi
  surname: Takeda
  fullname: Takeda, Norifumi
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 14
  givenname: Katsuhito
  surname: Fujiu
  fullname: Fujiu, Katsuhito
  organization: Department of Advanced Cardiology, The University of Tokyo
– sequence: 15
  givenname: Masao
  surname: Daimon
  fullname: Daimon, Masao
  organization: Department of Clinical Laboratory, The University of Tokyo Hospital
– sequence: 16
  givenname: Jiro
  surname: Ando
  fullname: Ando, Jiro
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 17
  givenname: Hiroshi
  surname: Akazawa
  fullname: Akazawa, Hiroshi
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 18
  givenname: Hiroyuki
  surname: Morita
  fullname: Morita, Hiroyuki
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
– sequence: 19
  givenname: Issei
  surname: Komuro
  fullname: Komuro, Issei
  organization: Department of Cardiovascular Medicine, The University of Tokyo Hospital
BackLink https://www.ncbi.nlm.nih.gov/pubmed/34176867$$D View this record in MEDLINE/PubMed
BookMark eNpNUDtPwzAYtFARfcDOhDyypNiOndhjlbY8VMRQmKMvjtO6SuJgJ0j8e1pRJKY76U6nu5uiUetag9AtJXPKRPygrdeHefYSMRoRlogLNKExTyMuGRn942M0DeFACFNEqCs0jjlNE5mkE7RdGtPhjQHf2naHF_XOedvvG9w7vDS90T3OwJcWNN6C186WLtiA1941eKX3Tp9Et_PQ7a3Gr-7LmnCNLiuog7k54wx9rFfv2VO0eXt8zhabSAsl-0hVpWSSCkYI0QBE0jI2MUClSBUTI2XMeXVsCkYVpeIVS4AUjArOGS9AEDZD97-5nXefgwl93tigTV1Da9wQcia4UKfF6dF6d7YORWPKvPO2Af-d_x3BfgBOCGHH
CitedBy_id crossref_primary_10_1038_s44325_025_00064_8
crossref_primary_10_1007_s40336_023_00595_z
crossref_primary_10_1016_j_autrev_2025_103916
crossref_primary_10_1016_j_jjcc_2021_10_016
crossref_primary_10_1093_eurheartj_ehae356
crossref_primary_10_1097_MCP_0000000000000902
crossref_primary_10_3390_cells11010059
crossref_primary_10_3390_jimaging9020050
crossref_primary_10_3390_life13081653
crossref_primary_10_1007_s00408_023_00641_7
crossref_primary_10_1111_echo_15417
crossref_primary_10_1007_s12410_025_09607_0
crossref_primary_10_1007_s11886_024_02159_7
crossref_primary_10_1016_j_rdc_2022_07_004
crossref_primary_10_1007_s11886_025_02250_7
crossref_primary_10_1097_MCP_0000000000001193
crossref_primary_10_3390_diagnostics13142426
crossref_primary_10_1053_j_semnuclmed_2024_02_004
crossref_primary_10_21518_ms2025_040
crossref_primary_10_1007_s11886_024_02088_5
crossref_primary_10_1007_s42452_025_06504_5
ContentType Journal Article
DBID CGR
CUY
CVF
ECM
EIF
NPM
7X8
DOI 10.1253/circj.CJ-21-0265
DatabaseName Medline
MEDLINE
MEDLINE (Ovid)
MEDLINE
MEDLINE
PubMed
MEDLINE - Academic
DatabaseTitle 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: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod no_fulltext_linktorsrc
Discipline Medicine
EISSN 1347-4820
ExternalDocumentID 34176867
Genre Research Support, Non-U.S. Gov't
Journal Article
GroupedDBID ---
.55
29B
2WC
53G
5GY
5RE
6J9
ACGFO
ADBBV
AENEX
ALMA_UNASSIGNED_HOLDINGS
BAWUL
CGR
CS3
CUY
CVF
DIK
DU5
E3Z
EBS
ECM
EIF
EJD
F5P
GX1
JSF
JSH
KQ8
M~E
NPM
OK1
P2P
RJT
RNS
RZJ
TR2
W2D
X7M
XSB
ZXP
7X8
OVT
ID FETCH-LOGICAL-c598t-9fd828152000caa081d3e3aaf90f30e88344f341ae9bd94f26a0b2154424ba502
IEDL.DBID 7X8
ISICitedReferencesCount 24
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000735434500017&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1347-4820
IngestDate Thu Jul 10 23:55:16 EDT 2025
Thu Jan 02 22:37:52 EST 2025
IsDoiOpenAccess false
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 1
Keywords Deep learning
Echocardiography
Transfer learning
Artificial intelligence
Cardiac sarcoidosis
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c598t-9fd828152000caa081d3e3aaf90f30e88344f341ae9bd94f26a0b2154424ba502
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
OpenAccessLink https://www.jstage.jst.go.jp/article/circj/86/1/86_CJ-21-0265/_article/-char/en
PMID 34176867
PQID 2545990597
PQPubID 23479
ParticipantIDs proquest_miscellaneous_2545990597
pubmed_primary_34176867
PublicationCentury 2000
PublicationDate 2021-12-24
PublicationDateYYYYMMDD 2021-12-24
PublicationDate_xml – month: 12
  year: 2021
  text: 2021-12-24
  day: 24
PublicationDecade 2020
PublicationPlace Japan
PublicationPlace_xml – name: Japan
PublicationTitle Circulation journal : official journal of the Japanese Circulation Society
PublicationTitleAlternate Circ J
PublicationYear 2021
References 34471070 - Circ J. 2021 Dec 24;86(1):96-97. doi: 10.1253/circj.CJ-21-0663
References_xml – reference: 34471070 - Circ J. 2021 Dec 24;86(1):96-97. doi: 10.1253/circj.CJ-21-0663
SSID ssj0029059
Score 2.4656377
Snippet Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from...
SourceID proquest
pubmed
SourceType Aggregation Database
Index Database
StartPage 87
SubjectTerms Algorithms
Deep Learning
Echocardiography
Humans
Motion Pictures
Myocarditis
Sarcoidosis - diagnostic imaging
Title Deep Learning Algorithm to Detect Cardiac Sarcoidosis From Echocardiographic Movies
URI https://www.ncbi.nlm.nih.gov/pubmed/34176867
https://www.proquest.com/docview/2545990597
Volume 86
WOSCitedRecordID wos000735434500017&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1JS8NAFB7UinhxX-rGCF7HJrNkOUnpgogthSr0Viaz1IhNalP9_b5JUz0JgpecMsnw5i3fm_f4HkI3QieWQn5MqAEj54xxEgVakIBpwPdMx5EueWYfw34_Go3iQXXhVlRtlSufWDpqnSt3R96AREaA5wT8ezd7J25qlKuuViM01lGNAZRxWh2OvqsI1L1eJlw8JBxCXVWmpII1VDpXr7etB0Ihm6aB-B1gloGmu_vfLe6hnQpi4uZSJ_bRmskO0FavKqIfomHbmBmuiFUnuPk2gY8sXqZ4keO2cUUF3Cr1RuEhmEGe6rxIC9yd51PcAW-pyhbWkuk6VbiXw66KI_Tc7Ty17kk1W4EoEUcLElt3Fr4jXfKUlAAMNDNMSht7lnkmcuM3LEQ4aeJEx9zSQHoJddQ9lCdSePQYbWR5Zk4RpkkUGQrLpW851wrcZxJILULftwIUpI6uV-Iag-66goTMTP5RjH8EVkcnS5mPZ0uSjTH8GzKhIDz7w-pztE1dq4lPCeUXqGbBcs0l2lSfi7SYX5VKAc_-oPcFsn_A1Q
linkProvider ProQuest
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=Deep+Learning+Algorithm+to+Detect+Cardiac+Sarcoidosis+From+Echocardiographic+Movies&rft.jtitle=Circulation+journal+%3A+official+journal+of+the+Japanese+Circulation+Society&rft.au=Katsushika%2C+Susumu&rft.au=Kodera%2C+Satoshi&rft.au=Nakamoto%2C+Mitsuhiko&rft.au=Ninomiya%2C+Kota&rft.date=2021-12-24&rft.issn=1347-4820&rft.eissn=1347-4820&rft.volume=86&rft.issue=1&rft.spage=87&rft_id=info:doi/10.1253%2Fcircj.CJ-21-0265&rft.externalDBID=NO_FULL_TEXT
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1347-4820&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1347-4820&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1347-4820&client=summon