Spatio-temporal regularization for deep MR Fingerprinting

Mohammad Golbabaee; Dongdong Chen; Marion I. Menzel; Mike Davies; Pedro A.  Gómez

Spatio-temporal regularization for deep MR Fingerprinting

Mohammad Golbabaee, Dongdong Chen, Marion I. Menzel, Mike Davies, Pedro A. Gómez

University of Edinburgh

Research output: Contribution to conference › Paper › peer-review

Abstract

We study a deep learning approach to address the heavy storage and computation requirements of the baseline dictionary-matching (DM) for Magnetic Resonance Fingerprinting (MRF) reconstruction. The MRF-Net provides a piece-wise affine approximation to the (temporal) Bloch response manifold projection. Fed with non-iterated back-projected images, the network alone is unable to fully resolve spatially-correlated artefacts which appear in highly undersampling regimes. We propose an accelerated iterative reconstruction to minimize these artefacts before feeding into the network. This is done through a convex regularization that jointly promotes spatio-temporal regularities of the MRF time-series.

Original language	English
Publication status	Published - 10 Jul 2019
Event	Medical Imaging with Deep Learning - London, UK United Kingdom Duration: 8 Jul 2019 → 10 Jul 2019 https://openreview.net/group?id=MIDL.io/2019/Conference/Abstract

Conference

Conference	Medical Imaging with Deep Learning
Abbreviated title	MIDL 2019
Country/Territory	UK United Kingdom
City	London
Period	8/07/19 → 10/07/19
Internet address	https://openreview.net/group?id=MIDL.io/2019/Conference/Abstract

Access to Document

https://openreview.net/pdf?id=ryx64UL6YE

Cite this

@conference{9140af23757040de91531c42c480f4c9,

title = "Spatio-temporal regularization for deep MR Fingerprinting",

abstract = "We study a deep learning approach to address the heavy storage and computation requirements of the baseline dictionary-matching (DM) for Magnetic Resonance Fingerprinting (MRF) reconstruction. The MRF-Net provides a piece-wise affine approximation to the (temporal) Bloch response manifold projection. Fed with non-iterated back-projected images, the network alone is unable to fully resolve spatially-correlated artefacts which appear in highly undersampling regimes. We propose an accelerated iterative reconstruction to minimize these artefacts before feeding into the network. This is done through a convex regularization that jointly promotes spatio-temporal regularities of the MRF time-series.",

author = "Mohammad Golbabaee and Dongdong Chen and Menzel, \{Marion I.\} and Mike Davies and G{\'o}mez, \{Pedro A.\}",

year = "2019",

month = jul,

day = "10",

language = "English",

note = "Medical Imaging with Deep Learning, MIDL 2019 ; Conference date: 08-07-2019 Through 10-07-2019",

url = "https://openreview.net/group?id=MIDL.io/2019/Conference/Abstract",

}

TY - CONF

T1 - Spatio-temporal regularization for deep MR Fingerprinting

AU - Golbabaee, Mohammad

AU - Chen, Dongdong

AU - Menzel, Marion I.

AU - Davies, Mike

AU - Gómez, Pedro A.

PY - 2019/7/10

Y1 - 2019/7/10

N2 - We study a deep learning approach to address the heavy storage and computation requirements of the baseline dictionary-matching (DM) for Magnetic Resonance Fingerprinting (MRF) reconstruction. The MRF-Net provides a piece-wise affine approximation to the (temporal) Bloch response manifold projection. Fed with non-iterated back-projected images, the network alone is unable to fully resolve spatially-correlated artefacts which appear in highly undersampling regimes. We propose an accelerated iterative reconstruction to minimize these artefacts before feeding into the network. This is done through a convex regularization that jointly promotes spatio-temporal regularities of the MRF time-series.

AB - We study a deep learning approach to address the heavy storage and computation requirements of the baseline dictionary-matching (DM) for Magnetic Resonance Fingerprinting (MRF) reconstruction. The MRF-Net provides a piece-wise affine approximation to the (temporal) Bloch response manifold projection. Fed with non-iterated back-projected images, the network alone is unable to fully resolve spatially-correlated artefacts which appear in highly undersampling regimes. We propose an accelerated iterative reconstruction to minimize these artefacts before feeding into the network. This is done through a convex regularization that jointly promotes spatio-temporal regularities of the MRF time-series.

M3 - Paper

T2 - Medical Imaging with Deep Learning

Y2 - 8 July 2019 through 10 July 2019

ER -