A Real-Time Neural Representation via Algorithm-Hardware Synergy for Sparse-View CT Reconstruction

Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE transaction on neural networks and learning systems Ročník PP; s. 1 - 13
Hlavní autori: Li, Xin, Wan, Haochuan, Long, Kangjie, Wu, Qing, Du, Chenhe, Su, Ying, Luo, Zhe, Lou, Xin, Zhang, Yuyao
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: United States IEEE 04.11.2025
Predmet:
ISSN:2162-237X, 2162-2388, 2162-2388
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Abstract Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs) as a self-supervised paradigm for solving underdetermined inverse problems have demonstrated excellent performance in SVCT reconstruction. However, since INR-based approaches rely on subject-specific training, they require a significant investment of time to optimize from scratch. Consequently, previous INR methods have not been able to meet the requisite timeliness of reconstruction. In our work, we propose RTSyner, an algorithm-hardware collaboration framework that facilitates the real-time efficiency of CT reconstruction. On the algorithmic side, we introduce an efficient coordinate-based feature module that exploits the local latent features as a positional external condition, leveraging the limited structural information of corrupted images derived from the sensory domain. By fusing latent features and coordinate information, the model learns a neural representation of the final tomographic image. On the hardware side, we design a dedicated hardware architecture with a customized algorithm flow to improve reconstruction speed and reduce power consumption. Furthermore, we improve the efficiency of model inference through model quantization, which also facilitates the subsequent deployment of hardware. Our extensive experimental results demonstrate that the RTSyner based on neural representation has achieved real-time SVCT reconstruction through the synergistic acceleration of the algorithm and hardware. We further explore its application potential via volume reconstructions under more complex acquisition geometries.
AbstractList Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs) as a self-supervised paradigm for solving underdetermined inverse problems have demonstrated excellent performance in SVCT reconstruction. However, since INR-based approaches rely on subject-specific training, they require a significant investment of time to optimize from scratch. Consequently, previous INR methods have not been able to meet the requisite timeliness of reconstruction. In our work, we propose RTSyner, an algorithm-hardware collaboration framework that facilitates the real-time efficiency of CT reconstruction. On the algorithmic side, we introduce an efficient coordinate-based feature module that exploits the local latent features as a positional external condition, leveraging the limited structural information of corrupted images derived from the sensory domain. By fusing latent features and coordinate information, the model learns a neural representation of the final tomographic image. On the hardware side, we design a dedicated hardware architecture with a customized algorithm flow to improve reconstruction speed and reduce power consumption. Furthermore, we improve the efficiency of model inference through model quantization, which also facilitates the subsequent deployment of hardware. Our extensive experimental results demonstrate that the RTSyner based on neural representation has achieved real-time SVCT reconstruction through the synergistic acceleration of the algorithm and hardware. We further explore its application potential via volume reconstructions under more complex acquisition geometries.
Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs) as a self-supervised paradigm for solving underdetermined inverse problems have demonstrated excellent performance in SVCT reconstruction. However, since INR-based approaches rely on subject-specific training, they require a significant investment of time to optimize from scratch. Consequently, previous INR methods have not been able to meet the requisite timeliness of reconstruction. In our work, we propose RTSyner, an algorithm-hardware collaboration framework that facilitates the real-time efficiency of CT reconstruction. On the algorithmic side, we introduce an efficient coordinate-based feature module that exploits the local latent features as a positional external condition, leveraging the limited structural information of corrupted images derived from the sensory domain. By fusing latent features and coordinate information, the model learns a neural representation of the final tomographic image. On the hardware side, we design a dedicated hardware architecture with a customized algorithm flow to improve reconstruction speed and reduce power consumption. Furthermore, we improve the efficiency of model inference through model quantization, which also facilitates the subsequent deployment of hardware. Our extensive experimental results demonstrate that the RTSyner based on neural representation has achieved real-time SVCT reconstruction through the synergistic acceleration of the algorithm and hardware. We further explore its application potential via volume reconstructions under more complex acquisition geometries.Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs) as a self-supervised paradigm for solving underdetermined inverse problems have demonstrated excellent performance in SVCT reconstruction. However, since INR-based approaches rely on subject-specific training, they require a significant investment of time to optimize from scratch. Consequently, previous INR methods have not been able to meet the requisite timeliness of reconstruction. In our work, we propose RTSyner, an algorithm-hardware collaboration framework that facilitates the real-time efficiency of CT reconstruction. On the algorithmic side, we introduce an efficient coordinate-based feature module that exploits the local latent features as a positional external condition, leveraging the limited structural information of corrupted images derived from the sensory domain. By fusing latent features and coordinate information, the model learns a neural representation of the final tomographic image. On the hardware side, we design a dedicated hardware architecture with a customized algorithm flow to improve reconstruction speed and reduce power consumption. Furthermore, we improve the efficiency of model inference through model quantization, which also facilitates the subsequent deployment of hardware. Our extensive experimental results demonstrate that the RTSyner based on neural representation has achieved real-time SVCT reconstruction through the synergistic acceleration of the algorithm and hardware. We further explore its application potential via volume reconstructions under more complex acquisition geometries.
Author Zhang, Yuyao
Du, Chenhe
Su, Ying
Luo, Zhe
Wan, Haochuan
Lou, Xin
Long, Kangjie
Li, Xin
Wu, Qing
Author_xml – sequence: 1
  givenname: Xin
  orcidid: 0009-0007-2324-8120
  surname: Li
  fullname: Li, Xin
  email: lixin22023@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 2
  givenname: Haochuan
  surname: Wan
  fullname: Wan, Haochuan
  email: wanhch@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 3
  givenname: Kangjie
  orcidid: 0000-0001-9560-2132
  surname: Long
  fullname: Long, Kangjie
  email: longkj@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 4
  givenname: Qing
  orcidid: 0000-0001-9320-0332
  surname: Wu
  fullname: Wu, Qing
  email: wuqing@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 5
  givenname: Chenhe
  orcidid: 0000-0002-1834-4309
  surname: Du
  fullname: Du, Chenhe
  email: duchenhe@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 6
  givenname: Ying
  surname: Su
  fullname: Su, Ying
  email: su.ying@zs-hospital.sh.cn
  organization: Zhongshan Hospital, Cardiac Intensive Care Center, Fudan University, Shanghai, China
– sequence: 7
  givenname: Zhe
  surname: Luo
  fullname: Luo, Zhe
  email: luo.zhe@zs-hospital.sh.cn
  organization: Zhongshan Hospital, Cardiac Intensive Care Center, Fudan University, Shanghai, China
– sequence: 8
  givenname: Xin
  orcidid: 0000-0001-7196-9861
  surname: Lou
  fullname: Lou, Xin
  email: louxin@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
– sequence: 9
  givenname: Yuyao
  orcidid: 0000-0001-6706-4867
  surname: Zhang
  fullname: Zhang, Yuyao
  email: zhangyy8@shanghaitech.edu.cn
  organization: School of Information Science and Technology, ShanghaiTech University, Shanghai, China
BackLink https://www.ncbi.nlm.nih.gov/pubmed/41187032$$D View this record in MEDLINE/PubMed
BookMark eNpFkE1Lw0AQhhep2Kr9AyKSo5fU_Up2cyxFrVAq2Cjelt1kUiP5cjex9N-b2lrnMsPwvO_hOUeDqq4AoSuCJ4Tg6C5eLherCcU0mLCQhpjJEzSiJKQ-ZVIOjrd4H6Kxc5-4nxAHIY_O0JATIgVmdITM1HsBXfhxXoK3hM7qon80FhxUrW7zuvK-c-1Ni3Vt8_aj9OfaphttwVttK7DrrZfV1ls12jrw33LYeLO4L0jqyrW2S3YFl-g004WD8WFfoNeH-3g29xfPj0-z6cJPqBCtD2kAkRBZxhjOBI-YMQkXxuhA4igkAZZcSk0SbrSRqYYIM50BjygWUZJCwC7Q7b63sfVXB65VZe4SKApdQd05xWgoJeec0B69OaCdKSFVjc1Lbbfqz0sP0D2Q2No5C9kRIVjt_Ktf_2rnXx3896HrfSgHgP8AoZQLwdgPrB2BHQ
CODEN ITNNAL
ContentType Journal Article
DBID 97E
RIA
RIE
AAYXX
CITATION
NPM
7X8
DOI 10.1109/TNNLS.2025.3626038
DatabaseName IEEE All-Society Periodicals Package (ASPP) 2005–Present
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/IET Electronic Library
  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 Computer Science
EISSN 2162-2388
EndPage 13
ExternalDocumentID 41187032
10_1109_TNNLS_2025_3626038
11224773
Genre orig-research
Journal Article
GrantInformation_xml – fundername: National Natural Science Foundation of China
  grantid: 62571328
  funderid: 10.13039/501100001809
GroupedDBID 0R~
4.4
5VS
6IK
97E
AAJGR
AASAJ
AAWTH
ABQJQ
ABVLG
ACIWK
ACPRK
AENEX
AFRAH
AGQYO
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
EBS
IFIPE
IPLJI
JAVBF
M43
MS~
O9-
OCL
PQQKQ
RIA
RIE
RNS
AAYXX
AGSQL
CITATION
EJD
ABAZT
NPM
7X8
ID FETCH-LOGICAL-c277t-ed5e977ff330f7493bbc47bba580961508488a1c4bab8dae903afe492079cde53
IEDL.DBID RIE
ISICitedReferencesCount 0
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001610945700001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 2162-237X
2162-2388
IngestDate Wed Nov 05 17:16:31 EST 2025
Thu Nov 06 11:56:54 EST 2025
Sat Nov 29 06:53:32 EST 2025
Wed Nov 26 07:21:08 EST 2025
IsPeerReviewed false
IsScholarly true
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-c277t-ed5e977ff330f7493bbc47bba580961508488a1c4bab8dae903afe492079cde53
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ORCID 0000-0001-9320-0332
0000-0001-7196-9861
0000-0001-9560-2132
0000-0002-1834-4309
0009-0007-2324-8120
0000-0001-6706-4867
PMID 41187032
PQID 3268844412
PQPubID 23479
PageCount 13
ParticipantIDs pubmed_primary_41187032
ieee_primary_11224773
crossref_primary_10_1109_TNNLS_2025_3626038
proquest_miscellaneous_3268844412
PublicationCentury 2000
PublicationDate 2025-11-04
PublicationDateYYYYMMDD 2025-11-04
PublicationDate_xml – month: 11
  year: 2025
  text: 2025-11-04
  day: 04
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
PublicationTitle IEEE transaction on neural networks and learning systems
PublicationTitleAbbrev TNNLS
PublicationTitleAlternate IEEE Trans Neural Netw Learn Syst
PublicationYear 2025
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0000605649
Score 2.4847379
Snippet Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging....
SourceID proquest
pubmed
crossref
ieee
SourceType Aggregation Database
Index Database
Publisher
StartPage 1
SubjectTerms Computed tomography
Computed tomography (CT) reconstruction
Encoding
Feature extraction
Image reconstruction
implicit neural representation (INR)
Inference algorithms
Magnetic resonance imaging
Neural radiance field
real-time reconstruction
Real-time systems
Rendering (computer graphics)
software-hardware synergy
sparse-view CT
Training
Title A Real-Time Neural Representation via Algorithm-Hardware Synergy for Sparse-View CT Reconstruction
URI https://ieeexplore.ieee.org/document/11224773
https://www.ncbi.nlm.nih.gov/pubmed/41187032
https://www.proquest.com/docview/3268844412
Volume PP
WOSCitedRecordID wos001610945700001&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
  customDbUrl:
  eissn: 2162-2388
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0000605649
  issn: 2162-237X
  databaseCode: RIE
  dateStart: 20120101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1NT9wwEB0V1AMXPlqgWwpypd6QwbGdtX1coSIO1apit9XeItuZtCu1WbQsoP77erwJiAMHbjnETjTjjzcev3kAX4xRTeGj587WlmuUhoe60Wm6-0aamFbDILLYhBmP7Wzmvndk9cyFQcR8-QzP6DHn8utFvKOjsvOC0kCp9w3YMMasyVqPByoiAfNhhruyGEoulZn1JBnhzqfj8bdJCgdleUYFWIQioT5NWttCyWd7UhZZeRlv5n3ncueVf7wL2x3AZKP1iNiDN9i-g51evIF1c_k9hBG7TiCREweEUYmO1Og6X4vt2Egtu597Nvrza7Gcr37_5ZTjf_BLZJN_mS_IEtxlk5sUGCP_OccHdjFlFMw-laTdhx-XX6cXV7wTXOBRGrPiWJeY8GDTKCUao50KIWoTgi8tKcOUVHvf-iLq4IOtPTqhfIPaSWFcrLFUB7DZLlr8AMwPjVSliJGWhQRivFfETHRYqNI6EwZw2pu8ulnX1ahyPCJclX1Vka-qzlcD2CfbPr3ZmXUAn3s3VWlaUK7Dt7i4u60SKrVWJ6wnB3C49t9j697tH1_o9Qi26OOZcag_wWayGh7D23i_mt8uT9LYm9mTPPb-A7qo0xY
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3PTxQxFH5RNNGLgKIugpTEmyl02s62PW6IBMI6Iexq9jZpO290E50lywLxv6ftzEA8cOA2h2nTvNcf3-vr9z6AL0qJOrPeUqMrTSVyRV1Vy7Dcbc2VD7uhY0lsQhWFns3MeUdWT1wYREyPz_AgfqZcfrXw1_Gq7DCLaaDQ-3N4kUvJs5audX-lwgI0HybAy7Mhp1yoWU-TYeZwWhTjSQgIeX4QS7AwEaX6ZFTbZoL_dyolmZXHEWc6eY7XnzjmDXjTQUwyaufEJjzD5i2s9_INpFvN78CNyEWAiTSyQEgs0hEaXaSHsR0fqSE3c0tGf34tlvPV7780Zvlv7RLJ5F9iDJIAeMnkMoTGSH_O8ZYcTUkMZx-K0m7Bj-Nv06MT2kkuUM-VWlGscgyIsK6FYLWSRjjnpXLO5jpqw-Sx-r62mZfOOl1ZNEzYGqXhTBlfYS7ew1qzaPAjEDtUXOTM-7gxBBhjrYjcRIOZyLVRbgBfe5OXl21ljTJFJMyUyVdl9FXZ-WoAW9G2D392Zh3Afu-mMiyMmO2wDS6ur8qAS7WWAe3xAXxo_Xffunf79iO97sGrk-n3cTk-Lc4-wev2HnlImdyBtWBB3IWX_mY1v1p-TjPwDuX61XQ
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=A+Real-Time+Neural+Representation+via+Algorithm-Hardware+Synergy+for+Sparse-View+CT+Reconstruction&rft.jtitle=IEEE+transaction+on+neural+networks+and+learning+systems&rft.au=Li%2C+Xin&rft.au=Wan%2C+Haochuan&rft.au=Long%2C+Kangjie&rft.au=Wu%2C+Qing&rft.date=2025-11-04&rft.eissn=2162-2388&rft.volume=PP&rft_id=info:doi/10.1109%2FTNNLS.2025.3626038&rft_id=info%3Apmid%2F41187032&rft.externalDocID=41187032
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2162-237X&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2162-237X&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2162-237X&client=summon