L2G-ECG: Learning to Generate Missing Leads in ECG Signals using Adversarial Autoencoder

An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at presc...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE journal of biomedical and health informatics Ročník PP; s. 1 - 11
Hlavní autoři: Srivastava, Apoorva, Sheet, Debdoot, Patra, Amit
Médium: Journal Article
Jazyk:angličtina
Vydáno: United States IEEE 09.07.2025
Témata:
ISSN:2168-2194, 2168-2208, 2168-2208
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at prescribed locations. In this paper, we propose an adversarial learning-based method to generate multiple missing leads in ECG signals using a lesser number of available leads. A convolution neural network-based encoder-decoder-like network is trained to generate the missing lead signals, guided by a Visual Turing Test discriminator to enhance signal realism. We have experimentally demonstrated it on a publicly available Physionet Challenge 2021 dataset consisting of subjects from 3 different nationalities and genetic predispositions, including healthy individuals and those with multiple comorbidities across 29 different arrhythmia pathologies. We have exhaustively checked the ability to generate 12-lead ECG from 1,<inline-formula><tex-math notation="LaTeX">\dots</tex-math></inline-formula>, 7 available leads, with generalization assessed on a held-out test set. Quantitative analysis of the synthesized ECG signals is done by computing Mean Square Error (MSE), Structural Similarity Measure (SSIM), information-based metrics, and morphological feature comparison of real and generated Lead-II signals. Our study shows that as available leads increase from 1-7, average <inline-formula><tex-math notation="LaTeX">AUP_{S}</tex-math></inline-formula> values improved from 0.81 to 0.98 and <inline-formula><tex-math notation="LaTeX">AUP_{M}</tex-math></inline-formula> decreased from 30.96<inline-formula><tex-math notation="LaTeX">\times 10^{-5}</tex-math></inline-formula> to 1.18<inline-formula><tex-math notation="LaTeX">\times 10^{-5}</tex-math></inline-formula>. Notably, 12-lead ECG reconstruction is achievable with just 2 limb and 3 chest leads, supported by low MSE, high SSIM, minimal information loss, and strong Lead-II feature generation, highlighting the model's real-world applicability.
AbstractList An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at prescribed locations. In this paper, we propose an adversarial learning-based method to generate multiple missing leads in ECG signals using a lesser number of available leads. A convolution neural network-based encoder-decoder-like network is trained to generate the missing lead signals, guided by a Visual Turing Test discriminator to enhance signal realism. We have experimentally demonstrated it on a publicly available Physionet Challenge 2021 dataset consisting of subjects from 3 different nationalities and genetic predispositions, including healthy individuals and those with multiple comorbidities across 29 different arrhythmia pathologies. We have exhaustively checked the ability to generate 12-lead ECG from 1,<inline-formula><tex-math notation="LaTeX">\dots</tex-math></inline-formula>, 7 available leads, with generalization assessed on a held-out test set. Quantitative analysis of the synthesized ECG signals is done by computing Mean Square Error (MSE), Structural Similarity Measure (SSIM), information-based metrics, and morphological feature comparison of real and generated Lead-II signals. Our study shows that as available leads increase from 1-7, average <inline-formula><tex-math notation="LaTeX">AUP_{S}</tex-math></inline-formula> values improved from 0.81 to 0.98 and <inline-formula><tex-math notation="LaTeX">AUP_{M}</tex-math></inline-formula> decreased from 30.96<inline-formula><tex-math notation="LaTeX">\times 10^{-5}</tex-math></inline-formula> to 1.18<inline-formula><tex-math notation="LaTeX">\times 10^{-5}</tex-math></inline-formula>. Notably, 12-lead ECG reconstruction is achievable with just 2 limb and 3 chest leads, supported by low MSE, high SSIM, minimal information loss, and strong Lead-II feature generation, highlighting the model's real-world applicability.
An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at prescribed locations. In this paper, we propose an adversarial learning-based method to generate multiple missing leads in ECG signals using a lesser number of available leads. A convolution neural network-based encoder-decoder-like network is trained to generate the missing lead signals, guided by a Visual Turing Test discriminator to enhance signal realism. We have experimentally demonstrated it on a publicly available Physionet Challenge 2021 dataset consisting of subjects from 3 different nationalities and genetic predispositions, including healthy individuals and those with multiple comorbidities across 29 different arrhythmia pathologies. We have exhaustively checked the ability to generate 12-lead ECG from 1,$\dots$, 7 available leads, with generalization assessed on a held-out test set. Quantitative analysis of the synthesized ECG signals is done by computing Mean Square Error (MSE), Structural Similarity Measure (SSIM), information-based metrics, and morphological feature comparison of real and generated Lead-II signals. Our study shows that as available leads increase from 1-7, average $AUP_{S}$ values improved from 0.81 to 0.98 and $AUP_{M}$ decreased from 30.96$\times 10^{-5}$ to 1.18$\times 10^{-5}$. Notably, 12-lead ECG reconstruction is achievable with just 2 limb and 3 chest leads, supported by low MSE, high SSIM, minimal information loss, and strong Lead-II feature generation, highlighting the model's real-world applicability.
An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at prescribed locations. In this paper, we propose an adversarial learning-based method to generate multiple missing leads in ECG signals using a lesser number of available leads. A convolution neural network-based encoder-decoder-like network is trained to generate the missing lead signals, guided by a Visual Turing Test discriminator to enhance signal realism. We have experimentally demonstrated it on a publicly available Physionet Challenge 2021 dataset consisting of subjects from 3 different nationalities and genetic predispositions, including healthy individuals and those with multiple comorbidities across 29 different arrhythmia pathologies. We have exhaustively checked the ability to generate 12-lead ECG from 1,$\dots$, 7 available leads, with generalization assessed on a held-out test set. Quantitative analysis of the synthesized ECG signals is done by computing Mean Square Error (MSE), Structural Similarity Measure (SSIM), information-based metrics, and morphological feature comparison of real and generated Lead-II signals. Our study shows that as available leads increase from 1-7, average $AUP_{S}$ values improved from 0.81 to 0.98 and $AUP_{M}$ decreased from 30.96$\times 10^{-5}$ to 1.18$\times 10^{-5}$. Notably, 12-lead ECG reconstruction is achievable with just 2 limb and 3 chest leads, supported by low MSE, high SSIM, minimal information loss, and strong Lead-II feature generation, highlighting the model's real-world applicability.An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is to employ 12-lead ECG signals. Yet, in certain trauma patients or ones with physical injury, it is not possible to place electrodes at prescribed locations. In this paper, we propose an adversarial learning-based method to generate multiple missing leads in ECG signals using a lesser number of available leads. A convolution neural network-based encoder-decoder-like network is trained to generate the missing lead signals, guided by a Visual Turing Test discriminator to enhance signal realism. We have experimentally demonstrated it on a publicly available Physionet Challenge 2021 dataset consisting of subjects from 3 different nationalities and genetic predispositions, including healthy individuals and those with multiple comorbidities across 29 different arrhythmia pathologies. We have exhaustively checked the ability to generate 12-lead ECG from 1,$\dots$, 7 available leads, with generalization assessed on a held-out test set. Quantitative analysis of the synthesized ECG signals is done by computing Mean Square Error (MSE), Structural Similarity Measure (SSIM), information-based metrics, and morphological feature comparison of real and generated Lead-II signals. Our study shows that as available leads increase from 1-7, average $AUP_{S}$ values improved from 0.81 to 0.98 and $AUP_{M}$ decreased from 30.96$\times 10^{-5}$ to 1.18$\times 10^{-5}$. Notably, 12-lead ECG reconstruction is achievable with just 2 limb and 3 chest leads, supported by low MSE, high SSIM, minimal information loss, and strong Lead-II feature generation, highlighting the model's real-world applicability.
Author Patra, Amit
Sheet, Debdoot
Srivastava, Apoorva
Author_xml – sequence: 1
  givenname: Apoorva
  surname: Srivastava
  fullname: Srivastava, Apoorva
  email: apoorva.s.2311@gmail.com
  organization: Department of Electrical Engineering, Indian Institute of Technology, Kharagpur, India
– sequence: 2
  givenname: Debdoot
  orcidid: 0000-0001-9046-149X
  surname: Sheet
  fullname: Sheet, Debdoot
  email: debdoot@ee.iitkgp.ac.in
  organization: Department of Electrical Engineering, Indian Institute of Technology, Kharagpur, India
– sequence: 3
  givenname: Amit
  orcidid: 0000-0002-3996-5761
  surname: Patra
  fullname: Patra, Amit
  email: amit.patra@ieee.org
  organization: Department of Electrical Engineering, Indian Institute of Technology, Kharagpur, India
BackLink https://www.ncbi.nlm.nih.gov/pubmed/40633044$$D View this record in MEDLINE/PubMed
BookMark eNpFkM9PwjAUxxuDEUT-ABNjevQy7K9tnTckODAYD2ribenaNzIDHbabif-9RUBfD6957_N9h8856tnGAkKXlIwpJdnt4_18MWaExWMey1QweYIGjCYyYozI3vFPM9FHI-8_SCgZRllyhvqCJJwTIQbofcnyaDbN7_ASlLO1XeG2wTlYcKoF_FR7v5uFpfG4tjig-KVeWbX2uPtdTcwXOK9crdZ40rUNWN0YcBfotAoQjA59iN4eZq_TebR8zhfTyTLSjAsZaUoUzRhJhCFZnJZVSbkWQkktDC8VlyyjlAqloZSZqcrSKFPpRIs0pAkp-RDd7O9uXfPZgW-LTe01rNfKQtP5gjMmJeMxTwN6fUC7cgOm2Lp6o9x3cbQRALoHtGu8d1D9IZQUO-nFTnqxk14cpIfM1T5TA8A_T0kqwuM_H6h6nw
CODEN IJBHA9
ContentType Journal Article
DBID 97E
ESBDL
RIA
RIE
AAYXX
CITATION
NPM
7X8
DOI 10.1109/JBHI.2025.3587428
DatabaseName IEEE Xplore (IEEE)
IEEE Xplore Open Access Journals
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Electronic Library (IEL)
CrossRef
PubMed
MEDLINE - Academic
DatabaseTitle CrossRef
PubMed
MEDLINE - Academic
DatabaseTitleList
PubMed
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 Electronic Library (IEL)
  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 11
ExternalDocumentID 40633044
10_1109_JBHI_2025_3587428
11074747
Genre orig-research
Journal Article
GroupedDBID 0R~
4.4
6IF
6IH
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACIWK
ACPRK
AENEX
AFRAH
AGQYO
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
EBS
ESBDL
HZ~
IFIPE
IPLJI
JAVBF
M43
O9-
OCL
PQQKQ
RIA
RIE
RNS
AAYXX
AGSQL
CITATION
EJD
NPM
7X8
ID FETCH-LOGICAL-c2348-c10a192064d0957bfb13c44a8c4d3ba38291114aceb89dfbbdadfc6c4734800b3
IEDL.DBID RIE
ISSN 2168-2194
2168-2208
IngestDate Thu Oct 02 22:25:19 EDT 2025
Mon Jul 21 05:58:39 EDT 2025
Sat Nov 29 07:45:50 EST 2025
Wed Jul 23 05:50:26 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Language English
License https://creativecommons.org/licenses/by/4.0/legalcode
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c2348-c10a192064d0957bfb13c44a8c4d3ba38291114aceb89dfbbdadfc6c4734800b3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ORCID 0000-0001-9046-149X
0000-0002-3996-5761
OpenAccessLink https://ieeexplore.ieee.org/document/11074747
PMID 40633044
PQID 3228823537
PQPubID 23479
PageCount 11
ParticipantIDs proquest_miscellaneous_3228823537
pubmed_primary_40633044
ieee_primary_11074747
crossref_primary_10_1109_JBHI_2025_3587428
PublicationCentury 2000
PublicationDate 2025-Jul-09
PublicationDateYYYYMMDD 2025-07-09
PublicationDate_xml – month: 07
  year: 2025
  text: 2025-Jul-09
  day: 09
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
SSID ssj0000816896
Score 2.4413767
Snippet An electrocardiogram (ECG) is a clinically accepted record used for the diagnosis of cardiovascular diseases. The standard guideline to monitor heart health is...
SourceID proquest
pubmed
crossref
ieee
SourceType Aggregation Database
Index Database
Publisher
StartPage 1
SubjectTerms Adversarial learning
Autoencoders
Bioinformatics
Convolution
Deep learning
Electrocardiogram
Electrocardiography
Electrodes
Generation
Generative adversarial networks
Lead
Limbs
Missing leads
Myocardium
Training
Title L2G-ECG: Learning to Generate Missing Leads in ECG Signals using Adversarial Autoencoder
URI https://ieeexplore.ieee.org/document/11074747
https://www.ncbi.nlm.nih.gov/pubmed/40633044
https://www.proquest.com/docview/3228823537
Volume PP
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE Electronic Library (IEL)
  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/eLvHCXMwlV1LS8QwEA4qIl58P9YXETwJ2W2bdJt4W5f1hS6CCnsraZLKXlrZbf39zqRdxYMH6SWk6YN8k2Qmk_mGkAswSTKrw5xZmxgmnMiZigLBJGjXoh-BxpF7EtfHZDyWk4l6boPVfSyMc84fPnNdLHpfvi1NjVtlPbRVBFzLZDlJ-k2w1veGis8g4fNxRVBgMBJF68UMA9V7uL67B2swirs8lmAOYp4-WMvQmhe_liSfY-VvddMvOzeb__zhLbLR6pd00AjENllyxQ5Ze2o96Ltk8hjdstHw9oq2zKrvtCppQz5dOfoEOGAdZt6c02lBoSl9mb4jyzKt_S2fwnmuUXDpoK5KZMK0brZH3m5Gr8M71mZXYCbiQjITBhrUO1BJLKhZSZZnITdCaGmE5ZnmMsJ5UGjjMqlsngGkNjd9I5APJwgyvk9WirJwh4TqRCsOVZGySmiYA0KLAbm5FHEseSY75HLRwelHQ6KReuMjUCkCkyIwaQtMh-xhR_40bPuwQ84XmKQwBNCvoQtX1vMU5iSwE3jMoc1BA9b30wuMj_546zFZx4_7A7jqhKxUs9qdklXzWU3nszOQs4k883L2BY1dyT4
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lb9QwEB5BQcCFZ4HlaSROSG4Te9K1uZWq7RZ2V0gUaW-RYzvVXhK0m_D7mXGyRRx6QLlYjhNZ8_nxjcczA_CRVJIquLyWIUy9xIi1tCpDaYhd45EixlGnIK7z6XJpViv7fXRWT74wMcZ0-SwecDHZ8kPrez4qO2RdBem5DXcKRJUN7lrXRyoph0TKyKWoIGku4mjHzDN7-PXL7IL0QVUc6MKQQsiZ-mg3Y30e_9mUUpaVmwln2njOHv1nlx_Dw5FhiuNhSDyBW7F5CvcWow39Gazm6lyenpx_FmNs1SvRtWIIP91FsSAkuI5zb27FuhHUVPxYX3GcZdGnVymJ89bx0BXHfddyLMwQN_vw8-z08mQmx_wK0iuNRvo8c0TwiJQEIlrTqq5y7RGd8Rh05bRRvBKi87EyNtQVgRpqf-SRI-JkWaWfw17TNvElCDd1VlOVssGio1UgD-ySWxssCqMrM4FPOwGXv4YwGmVSPzJbMjAlA1OOwExgnwX5t-Eowwl82GFS0iRgy4ZrYttvS1qVSFPQhaY2Lwawrr_eYfzqhr--h_uzy8W8nF8sv72GB9yRdB3XvoG9btPHt3DX_-7W2827NNr-AHUOy50
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=L2G-ECG%3A+Learning+to+Generate+Missing+Leads+in+ECG+Signals+using+Adversarial+Autoencoder&rft.jtitle=IEEE+journal+of+biomedical+and+health+informatics&rft.au=Srivastava%2C+Apoorva&rft.au=Sheet%2C+Debdoot&rft.au=Patra%2C+Amit&rft.date=2025-07-09&rft.pub=IEEE&rft.issn=2168-2194&rft.spage=1&rft.epage=11&rft_id=info:doi/10.1109%2FJBHI.2025.3587428&rft_id=info%3Apmid%2F40633044&rft.externalDocID=11074747
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