Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth

Weijie Gan; Cihat Eldeniz; Jiaming Liu; Sihao Chen; Hongyu An; Ulugbek S. Kamilov

doi:10.1109/IEEECONF51394.2020.9443403

Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth

Weijie Gan, Cihat Eldeniz, Jiaming Liu, Sihao Chen, Hongyu An, Ulugbek S. Kamilov

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

2 Scopus citations

Abstract

We propose a new plug-and-play priors (PnP) based MR image reconstruction method that systematically enforces data consistency while also exploiting deep-learning priors. Our prior is specified through a convolutional neural network (CNN) trained without any artifact-free ground truth to remove under-sampling artifacts from MR images. The results on reconstructing free-breathing MRI data into ten respiratory phases show that the method can form high-quality 4D images from severely undersampled measurements corresponding to acquisitions of about 1 and 2 minutes in length. The results also highlight the competitive performance of the method compared to several popular alternatives, including the TGV regularization and traditional UNet3D.

Original language	English
Title of host publication	Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020
Editors	Michael B. Matthews
Publisher	IEEE Computer Society
Pages	475-479
Number of pages	5
ISBN (Electronic)	9780738131269
DOIs	https://doi.org/10.1109/IEEECONF51394.2020.9443403
State	Published - Nov 1 2020
Event	54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020 - Pacific Grove, United States Duration: Nov 1 2020 → Nov 5 2020

Publication series

Name	Conference Record - Asilomar Conference on Signals, Systems and Computers
Volume	2020-November
ISSN (Print)	1058-6393

Conference

Conference	54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020
Country/Territory	United States
City	Pacific Grove
Period	11/1/20 → 11/5/20

Keywords

Image reconstruction
deep learning
magnetic resonance imaging
plug-and-play priors

Access to Document

10.1109/IEEECONF51394.2020.9443403

Cite this

Gan, W., Eldeniz, C., Liu, J., Chen, S., An, H., & Kamilov, U. S. (2020). Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth. In M. B. Matthews (Ed.), Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020 (pp. 475-479). [9443403] (Conference Record - Asilomar Conference on Signals, Systems and Computers; Vol. 2020-November). IEEE Computer Society. https://doi.org/10.1109/IEEECONF51394.2020.9443403

Gan, Weijie ; Eldeniz, Cihat ; Liu, Jiaming et al. / Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth. Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020. editor / Michael B. Matthews. IEEE Computer Society, 2020. pp. 475-479 (Conference Record - Asilomar Conference on Signals, Systems and Computers).

@inproceedings{adab00ed17df41a4a6c54d02d0a45852,

title = "Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth",

abstract = "We propose a new plug-and-play priors (PnP) based MR image reconstruction method that systematically enforces data consistency while also exploiting deep-learning priors. Our prior is specified through a convolutional neural network (CNN) trained without any artifact-free ground truth to remove under-sampling artifacts from MR images. The results on reconstructing free-breathing MRI data into ten respiratory phases show that the method can form high-quality 4D images from severely undersampled measurements corresponding to acquisitions of about 1 and 2 minutes in length. The results also highlight the competitive performance of the method compared to several popular alternatives, including the TGV regularization and traditional UNet3D.",

keywords = "Image reconstruction, deep learning, magnetic resonance imaging, plug-and-play priors",

author = "Weijie Gan and Cihat Eldeniz and Jiaming Liu and Sihao Chen and Hongyu An and Kamilov, {Ulugbek S.}",

note = "Funding Information: Research reported in this publication was supported by the Washington University Institute of Clinical and Translational Sciences grant UL1TR002345 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH). The content is solely the responsibility of the authors and does not necessarily represent the official view of the NIH. Publisher Copyright: {\textcopyright} 2020 IEEE.; 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020 ; Conference date: 01-11-2020 Through 05-11-2020",

year = "2020",

month = nov,

day = "1",

doi = "10.1109/IEEECONF51394.2020.9443403",

language = "English",

series = "Conference Record - Asilomar Conference on Signals, Systems and Computers",

publisher = "IEEE Computer Society",

pages = "475--479",

editor = "Matthews, {Michael B.}",

booktitle = "Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020",

}

Gan, W, Eldeniz, C, Liu, J, Chen, S, An, H & Kamilov, US 2020, Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth. in MB Matthews (ed.), Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020., 9443403, Conference Record - Asilomar Conference on Signals, Systems and Computers, vol. 2020-November, IEEE Computer Society, pp. 475-479, 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020, Pacific Grove, United States, 11/1/20. https://doi.org/10.1109/IEEECONF51394.2020.9443403

Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth. / Gan, Weijie; Eldeniz, Cihat; Liu, Jiaming et al.
Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020. ed. / Michael B. Matthews. IEEE Computer Society, 2020. p. 475-479 9443403 (Conference Record - Asilomar Conference on Signals, Systems and Computers; Vol. 2020-November).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth

AU - Gan, Weijie

AU - Eldeniz, Cihat

AU - Liu, Jiaming

AU - Chen, Sihao

AU - An, Hongyu

AU - Kamilov, Ulugbek S.

N1 - Funding Information: Research reported in this publication was supported by the Washington University Institute of Clinical and Translational Sciences grant UL1TR002345 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH). The content is solely the responsibility of the authors and does not necessarily represent the official view of the NIH. Publisher Copyright: © 2020 IEEE.

PY - 2020/11/1

Y1 - 2020/11/1

N2 - We propose a new plug-and-play priors (PnP) based MR image reconstruction method that systematically enforces data consistency while also exploiting deep-learning priors. Our prior is specified through a convolutional neural network (CNN) trained without any artifact-free ground truth to remove under-sampling artifacts from MR images. The results on reconstructing free-breathing MRI data into ten respiratory phases show that the method can form high-quality 4D images from severely undersampled measurements corresponding to acquisitions of about 1 and 2 minutes in length. The results also highlight the competitive performance of the method compared to several popular alternatives, including the TGV regularization and traditional UNet3D.

AB - We propose a new plug-and-play priors (PnP) based MR image reconstruction method that systematically enforces data consistency while also exploiting deep-learning priors. Our prior is specified through a convolutional neural network (CNN) trained without any artifact-free ground truth to remove under-sampling artifacts from MR images. The results on reconstructing free-breathing MRI data into ten respiratory phases show that the method can form high-quality 4D images from severely undersampled measurements corresponding to acquisitions of about 1 and 2 minutes in length. The results also highlight the competitive performance of the method compared to several popular alternatives, including the TGV regularization and traditional UNet3D.

KW - Image reconstruction

KW - deep learning

KW - magnetic resonance imaging

KW - plug-and-play priors

UR - http://www.scopus.com/inward/record.url?scp=85107773004&partnerID=8YFLogxK

U2 - 10.1109/IEEECONF51394.2020.9443403

DO - 10.1109/IEEECONF51394.2020.9443403

M3 - Conference contribution

AN - SCOPUS:85107773004

T3 - Conference Record - Asilomar Conference on Signals, Systems and Computers

SP - 475

EP - 479

BT - Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020

A2 - Matthews, Michael B.

PB - IEEE Computer Society

T2 - 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020

Y2 - 1 November 2020 through 5 November 2020

ER -

Gan W, Eldeniz C, Liu J, Chen S, An H, Kamilov US. Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth. In Matthews MB, editor, Conference Record of the 54th Asilomar Conference on Signals, Systems and Computers, ACSSC 2020. IEEE Computer Society. 2020. p. 475-479. 9443403. (Conference Record - Asilomar Conference on Signals, Systems and Computers). doi: 10.1109/IEEECONF51394.2020.9443403

Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this