Motion-compensated 4DCT reconstruction from single-beat cardiac CT scans using convolutional networks

  • Zhenyao Yan
  • , Zhennong Chen
  • , Li Zhang
  • , Quanzheng Li
  • , Dufan Wu*
  • *Corresponding author for this work

Research output: Chapter in Book or Report/Conference proceedingConference Proceedingpeer-review

1 Citation (Scopus)

Abstract

We proposed a deep lea rning-ba sed method for single-hea rtbea t 4D ca rdia c CT reconstruction, where a single ca rdiac cycle wa s split into multiple pha ses for reconstruction. First, we pre-reconstruct ea ch pha se using the projection da ta from itself a nd the neighboring pha ses. The pre-reconstructions a re fed into a supervised registra tion network to genera te the deforma tion fields between different pha ses. The deformation fields a re tra ined so tha t it ca n match the ground truth images from the corresponding pha ses. The deforma tion fields a re then used in the FBP-a nd-wra p method for motion-compensated reconstruction, where a subsequent network is used to remove residua l a rtifa cts. The proposed method wa s va lida ted with simula tion da ta from 40 4D ca rdia c CT sca ns a nd demonstra ted improved RMSE a nd SSIM a nd less blurring compared to FBP a nd PICCS.

Original languageEnglish
Title of host publicationMedical Imaging 2024
Subtitle of host publicationPhysics of Medical Imaging
EditorsRebecca Fahrig, John M. Sabol, Ke Li
PublisherSPIE
ISBN (Electronic)9781510671546
DOIs
Publication statusPublished - 2024
Externally publishedYes
EventMedical Imaging 2024: Physics of Medical Imaging - San Diego, United States
Duration: 19 Feb 202422 Feb 2024

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume12925
ISSN (Print)1605-7422

Conference

ConferenceMedical Imaging 2024: Physics of Medical Imaging
Country/TerritoryUnited States
CitySan Diego
Period19/02/2422/02/24

Keywords

  • 4DCT
  • cardiac
  • CNN
  • deformable image registration
  • motion-compensated reconstruction

Fingerprint

Dive into the research topics of 'Motion-compensated 4DCT reconstruction from single-beat cardiac CT scans using convolutional networks'. Together they form a unique fingerprint.

Cite this