Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction

Casper O. da Costa-Luis; Andrew J. Reader

doi:10.1109/NSSMIC.2017.8532624

Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction

Casper O. da Costa-Luis, Andrew J. Reader

Biomedical Engineering Department

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

16 Citations (Scopus)

2 Downloads (Pure)

Abstract

Resolution modelling in maximum likelihood expectation maximisation (MLEM) image reconstruction recovers resolution but at the cost of introducing ringing artefacts. Under-modelling, post-smoothing (PS) and regularisation methods which aim to suppress these artefacts nearly all result in a loss of resolution. This work proposes the use of deep convolutional neural networks (DCNNs) as a post-reconstruction image processing step to reduce reconstruction artefacts without compromising the resolution recovery.The DCNN results successfully suppress ringing arte-facts and furthermore result in an 80% lower normalised root mean squared error (NRMSE) versus MLEM, compared to a best decrease of only 0.2% when an optimal level of PS of MLEM is performed. The resultant images from the DCNN have lower noise, reduced ringing and partial volume effects, as well as sharper edges and improved resolution.

Original language	English
Title of host publication	2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	978-1-5386-2282-7
ISBN (Print)	978-1-5386-2283-4
DOIs	https://doi.org/10.1109/NSSMIC.2017.8532624
Publication status	Published - 15 Nov 2018
Event	2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC) - Atlanta, United States Duration: 21 Oct 2017 → 28 Oct 2017 http://www.nss-mic.org/2017/News.asp

Conference

Conference	2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)
Abbreviated title	IEEE NSS/MIC 2017
Country/Territory	United States
City	Atlanta
Period	21/10/2017 → 28/10/2017
Internet address	http://www.nss-mic.org/2017/News.asp

Keywords

Deep learning
PET
MR
CNN
Convolutional neural networks

Access to Document

10.1109/NSSMIC.2017.8532624

2017 IEEE NPSS Paul Phelps Continuing Education Grant
da Costa-Luis, C. (Recipient), 2017
Prize: Prize (including medals and awards)

Cite this

@inbook{a69bb28bac0d445ba630da5e24f5e4ef,

title = "Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction",

abstract = "Resolution modelling in maximum likelihood expectation maximisation (MLEM) image reconstruction recovers resolution but at the cost of introducing ringing artefacts. Under-modelling, post-smoothing (PS) and regularisation methods which aim to suppress these artefacts nearly all result in a loss of resolution. This work proposes the use of deep convolutional neural networks (DCNNs) as a post-reconstruction image processing step to reduce reconstruction artefacts without compromising the resolution recovery.The DCNN results successfully suppress ringing arte-facts and furthermore result in an 80% lower normalised root mean squared error (NRMSE) versus MLEM, compared to a best decrease of only 0.2% when an optimal level of PS of MLEM is performed. The resultant images from the DCNN have lower noise, reduced ringing and partial volume effects, as well as sharper edges and improved resolution.",

keywords = "Deep learning, PET, MR, CNN, Convolutional neural networks",

author = "{da Costa-Luis}, {Casper O.} and Reader, {Andrew J.}",

year = "2018",

month = nov,

day = "15",

doi = "10.1109/NSSMIC.2017.8532624",

language = "English",

isbn = "978-1-5386-2283-4",

booktitle = "2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

note = "2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), IEEE NSS/MIC 2017 ; Conference date: 21-10-2017 Through 28-10-2017",

url = "http://www.nss-mic.org/2017/News.asp",

}

da Costa-Luis, CO & Reader, AJ 2018, Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction. in 2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC). Institute of Electrical and Electronics Engineers Inc., 2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), Atlanta, United States, 21/10/2017. https://doi.org/10.1109/NSSMIC.2017.8532624

Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction. / da Costa-Luis, Casper O.; Reader, Andrew J.
2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC). Institute of Electrical and Electronics Engineers Inc., 2018.

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

TY - CHAP

T1 - Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction

AU - da Costa-Luis, Casper O.

AU - Reader, Andrew J.

PY - 2018/11/15

Y1 - 2018/11/15

N2 - Resolution modelling in maximum likelihood expectation maximisation (MLEM) image reconstruction recovers resolution but at the cost of introducing ringing artefacts. Under-modelling, post-smoothing (PS) and regularisation methods which aim to suppress these artefacts nearly all result in a loss of resolution. This work proposes the use of deep convolutional neural networks (DCNNs) as a post-reconstruction image processing step to reduce reconstruction artefacts without compromising the resolution recovery.The DCNN results successfully suppress ringing arte-facts and furthermore result in an 80% lower normalised root mean squared error (NRMSE) versus MLEM, compared to a best decrease of only 0.2% when an optimal level of PS of MLEM is performed. The resultant images from the DCNN have lower noise, reduced ringing and partial volume effects, as well as sharper edges and improved resolution.

AB - Resolution modelling in maximum likelihood expectation maximisation (MLEM) image reconstruction recovers resolution but at the cost of introducing ringing artefacts. Under-modelling, post-smoothing (PS) and regularisation methods which aim to suppress these artefacts nearly all result in a loss of resolution. This work proposes the use of deep convolutional neural networks (DCNNs) as a post-reconstruction image processing step to reduce reconstruction artefacts without compromising the resolution recovery.The DCNN results successfully suppress ringing arte-facts and furthermore result in an 80% lower normalised root mean squared error (NRMSE) versus MLEM, compared to a best decrease of only 0.2% when an optimal level of PS of MLEM is performed. The resultant images from the DCNN have lower noise, reduced ringing and partial volume effects, as well as sharper edges and improved resolution.

KW - Deep learning

KW - PET

KW - MR

KW - CNN

KW - Convolutional neural networks

U2 - 10.1109/NSSMIC.2017.8532624

DO - 10.1109/NSSMIC.2017.8532624

M3 - Conference paper

SN - 978-1-5386-2283-4

BT - 2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2017 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)

Y2 - 21 October 2017 through 28 October 2017

ER -

Deep Learning for Suppression of Resolution-Recovery Artefacts in MLEM PET Image Reconstruction

Abstract

Conference

Keywords

Access to Document

Fingerprint

Prizes

2017 IEEE NPSS Paul Phelps Continuing Education Grant

Cite this