Learning the Sampling Pattern for MRI

Ferdia Sherry, Martin Benning, Juan Carlos De los Reyes, Martin J. Graves, Georg Maierhofer, Guy Williams, Carola-Bibiane Schönlieb, Matthias J. Ehrhardt

Research output: Working paper

9 Downloads (Pure)

Abstract

The discovery of the theory of compressed sensing brought the realisation that many inverse problems can be solved even when measurements are "incomplete". This is particularly interesting in magnetic resonance imaging (MRI), where long acquisition times can limit its use. In this work, we consider the problem of learning a sparse sampling pattern that can be used to optimally balance acquisition time versus quality of the reconstructed image. We use a supervised learning approach, making the assumption that our training data is representative enough of new data acquisitions. We demonstrate that this is indeed the case, even if the training data consists of just 5 training pairs of measurements and ground-truth images; with a training set of brain images of size 192 by 192, for instance, one of the learned patterns samples only 32% of k-space, however results in reconstructions with mean SSIM 0.956 on a test set of similar images. The proposed framework is general enough to learn arbitrary sampling patterns, including common patterns such as Cartesian, spiral and radial sampling.
Original languageEnglish
Publication statusPublished - 20 Jun 2019

Keywords

  • eess.IV
  • cs.CV
  • cs.NA
  • math.NA
  • math.OC

Cite this

Sherry, F., Benning, M., Reyes, J. C. D. L., Graves, M. J., Maierhofer, G., Williams, G., ... Ehrhardt, M. J. (2019). Learning the Sampling Pattern for MRI.

Learning the Sampling Pattern for MRI. / Sherry, Ferdia; Benning, Martin; Reyes, Juan Carlos De los; Graves, Martin J.; Maierhofer, Georg; Williams, Guy; Schönlieb, Carola-Bibiane; Ehrhardt, Matthias J.

2019.

Research output: Working paper

Sherry, F, Benning, M, Reyes, JCDL, Graves, MJ, Maierhofer, G, Williams, G, Schönlieb, C-B & Ehrhardt, MJ 2019 'Learning the Sampling Pattern for MRI'.
Sherry F, Benning M, Reyes JCDL, Graves MJ, Maierhofer G, Williams G et al. Learning the Sampling Pattern for MRI. 2019 Jun 20.
Sherry, Ferdia ; Benning, Martin ; Reyes, Juan Carlos De los ; Graves, Martin J. ; Maierhofer, Georg ; Williams, Guy ; Schönlieb, Carola-Bibiane ; Ehrhardt, Matthias J. / Learning the Sampling Pattern for MRI. 2019.
@techreport{0033a56c6bdc4748a5422d58a4ac2abd,
title = "Learning the Sampling Pattern for MRI",
abstract = "The discovery of the theory of compressed sensing brought the realisation that many inverse problems can be solved even when measurements are {"}incomplete{"}. This is particularly interesting in magnetic resonance imaging (MRI), where long acquisition times can limit its use. In this work, we consider the problem of learning a sparse sampling pattern that can be used to optimally balance acquisition time versus quality of the reconstructed image. We use a supervised learning approach, making the assumption that our training data is representative enough of new data acquisitions. We demonstrate that this is indeed the case, even if the training data consists of just 5 training pairs of measurements and ground-truth images; with a training set of brain images of size 192 by 192, for instance, one of the learned patterns samples only 32{\%} of k-space, however results in reconstructions with mean SSIM 0.956 on a test set of similar images. The proposed framework is general enough to learn arbitrary sampling patterns, including common patterns such as Cartesian, spiral and radial sampling.",
keywords = "eess.IV, cs.CV, cs.NA, math.NA, math.OC",
author = "Ferdia Sherry and Martin Benning and Reyes, {Juan Carlos De los} and Graves, {Martin J.} and Georg Maierhofer and Guy Williams and Carola-Bibiane Sch{\"o}nlieb and Ehrhardt, {Matthias J.}",
note = "The main document is 10 pages, the supporting document is 2 pages and attached at the end of the main document",
year = "2019",
month = "6",
day = "20",
language = "English",
type = "WorkingPaper",

}

TY - UNPB

T1 - Learning the Sampling Pattern for MRI

AU - Sherry, Ferdia

AU - Benning, Martin

AU - Reyes, Juan Carlos De los

AU - Graves, Martin J.

AU - Maierhofer, Georg

AU - Williams, Guy

AU - Schönlieb, Carola-Bibiane

AU - Ehrhardt, Matthias J.

N1 - The main document is 10 pages, the supporting document is 2 pages and attached at the end of the main document

PY - 2019/6/20

Y1 - 2019/6/20

N2 - The discovery of the theory of compressed sensing brought the realisation that many inverse problems can be solved even when measurements are "incomplete". This is particularly interesting in magnetic resonance imaging (MRI), where long acquisition times can limit its use. In this work, we consider the problem of learning a sparse sampling pattern that can be used to optimally balance acquisition time versus quality of the reconstructed image. We use a supervised learning approach, making the assumption that our training data is representative enough of new data acquisitions. We demonstrate that this is indeed the case, even if the training data consists of just 5 training pairs of measurements and ground-truth images; with a training set of brain images of size 192 by 192, for instance, one of the learned patterns samples only 32% of k-space, however results in reconstructions with mean SSIM 0.956 on a test set of similar images. The proposed framework is general enough to learn arbitrary sampling patterns, including common patterns such as Cartesian, spiral and radial sampling.

AB - The discovery of the theory of compressed sensing brought the realisation that many inverse problems can be solved even when measurements are "incomplete". This is particularly interesting in magnetic resonance imaging (MRI), where long acquisition times can limit its use. In this work, we consider the problem of learning a sparse sampling pattern that can be used to optimally balance acquisition time versus quality of the reconstructed image. We use a supervised learning approach, making the assumption that our training data is representative enough of new data acquisitions. We demonstrate that this is indeed the case, even if the training data consists of just 5 training pairs of measurements and ground-truth images; with a training set of brain images of size 192 by 192, for instance, one of the learned patterns samples only 32% of k-space, however results in reconstructions with mean SSIM 0.956 on a test set of similar images. The proposed framework is general enough to learn arbitrary sampling patterns, including common patterns such as Cartesian, spiral and radial sampling.

KW - eess.IV

KW - cs.CV

KW - cs.NA

KW - math.NA

KW - math.OC

M3 - Working paper

BT - Learning the Sampling Pattern for MRI

ER -