Evaluation of Strategies for PET Motion Correction-Manifold Learning vs. Deep Learning

James Richard Clough; Daniel Robert Malcolm Balfour; Claudia Prieto Vasquez; Andrew Jonathan Reader; Paul Kenneth Marsden; Andrew Peter King

doi:10.1007/978-3-030-02628-8_7

Evaluation of Strategies for PET Motion Correction-Manifold Learning vs. Deep Learning

James Richard Clough, Daniel Robert Malcolm Balfour, Claudia Prieto Vasquez, Andrew Jonathan Reader, Paul Kenneth Marsden, Andrew Peter King

Research output: Contribution to journal › Conference paper › peer-review

4 Citations (Scopus)

318 Downloads (Pure)

Abstract

Image quality in abdominal PET is degraded by respiratory motion. In this paper we compare existing data-driven gating methods for motion correction which are based on manifold learning, with a proposed method in which a convolutional neural network learns estimated motion fields in an end-to-end manner, and then uses those estimated motion fields to motion correct the PET frames. We find that this proposed network approach is unable to outperform manifold learning methods in the literature, in terms of the image quality of the motion corrected volumes. We investigate possible explanations for this negative result and discuss the benefits of these unsupervised approaches which remain the state of the art.

Original language	English
Pages (from-to)	61-69
Journal	Lecture Notes in Computer Science
Volume	11038
Early online date	24 Oct 2018
DOIs	https://doi.org/10.1007/978-3-030-02628-8_7
Publication status	E-pub ahead of print - 24 Oct 2018

Access to Document

10.1007/978-3-030-02628-8_7

Evaluation of strategies for_CLOUGH_Accepted20August2018_GREEN AAMAccepted author manuscript, 1.29 MB

Cite this

@article{fbcfb9cac18b4b29a9509f6dab2e02bf,

title = "Evaluation of Strategies for PET Motion Correction-Manifold Learning vs. Deep Learning",

abstract = "Image quality in abdominal PET is degraded by respiratory motion. In this paper we compare existing data-driven gating methods for motion correction which are based on manifold learning, with a proposed method in which a convolutional neural network learns estimated motion fields in an end-to-end manner, and then uses those estimated motion fields to motion correct the PET frames. We find that this proposed network approach is unable to outperform manifold learning methods in the literature, in terms of the image quality of the motion corrected volumes. We investigate possible explanations for this negative result and discuss the benefits of these unsupervised approaches which remain the state of the art.",

author = "Clough, {James Richard} and Balfour, {Daniel Robert Malcolm} and {Prieto Vasquez}, Claudia and Reader, {Andrew Jonathan} and Marsden, {Paul Kenneth} and King, {Andrew Peter}",

year = "2018",

month = oct,

day = "24",

doi = "10.1007/978-3-030-02628-8_7",

language = "English",

volume = "11038",

pages = "61--69",

journal = "Lecture Notes in Computer Science",

issn = "0302-9743",

publisher = "Springer Verlag",

}

TY - JOUR

T1 - Evaluation of Strategies for PET Motion Correction-Manifold Learning vs. Deep Learning

AU - Clough, James Richard

AU - Balfour, Daniel Robert Malcolm

AU - Prieto Vasquez, Claudia

AU - Reader, Andrew Jonathan

AU - Marsden, Paul Kenneth

AU - King, Andrew Peter

PY - 2018/10/24

Y1 - 2018/10/24

N2 - Image quality in abdominal PET is degraded by respiratory motion. In this paper we compare existing data-driven gating methods for motion correction which are based on manifold learning, with a proposed method in which a convolutional neural network learns estimated motion fields in an end-to-end manner, and then uses those estimated motion fields to motion correct the PET frames. We find that this proposed network approach is unable to outperform manifold learning methods in the literature, in terms of the image quality of the motion corrected volumes. We investigate possible explanations for this negative result and discuss the benefits of these unsupervised approaches which remain the state of the art.

AB - Image quality in abdominal PET is degraded by respiratory motion. In this paper we compare existing data-driven gating methods for motion correction which are based on manifold learning, with a proposed method in which a convolutional neural network learns estimated motion fields in an end-to-end manner, and then uses those estimated motion fields to motion correct the PET frames. We find that this proposed network approach is unable to outperform manifold learning methods in the literature, in terms of the image quality of the motion corrected volumes. We investigate possible explanations for this negative result and discuss the benefits of these unsupervised approaches which remain the state of the art.

UR - http://www.scopus.com/inward/record.url?scp=85056466680&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-02628-8_7

DO - 10.1007/978-3-030-02628-8_7

M3 - Conference paper

SN - 0302-9743

VL - 11038

SP - 61

EP - 69

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

ER -

Evaluation of Strategies for PET Motion Correction-Manifold Learning vs. Deep Learning

Abstract

Access to Document

Other files and links

Fingerprint

Cite this