Geometry of Deep Learning for Magnetic Resonance Fingerprinting

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

Research output: Contribution to journalConference article

1 Citation (Scopus)

Abstract

Current popular methods for Magnetic Resonance Fingerprint (MRF) recovery are bottlenecked by the heavy storage and computation requirements of a dictionary-matching (DM) step due to the growing size and complexity of the fingerprint dictionaries in multi-parametric quantitative MRI applications. In this paper we study a deep learning approach to address these shortcomings. Coupled with a dimensionality reduction first layer, the proposed MRF-Net is able to reconstruct quantitative maps by saving more than 60 times in memory and computations required for a DM baseline. Fine-grid manifold enumeration i.e. the MRF dictionary is only used for training the network and not during image reconstruction. We show that the MRF-Net provides a piece-wise affine approximation to the Bloch response manifold projection and that rather than memorizing the dictionary, the network efficiently clusters this manifold and learns a set of hierarchical matched-filters for affine regression of the NMR characteristics in each segment.
Original languageEnglish
Pages (from-to)7825-7829
Number of pages5
JournalIEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Volume2019
DOIs
Publication statusPublished - 1 May 2019

Keywords

  • Magnetic resonance fingerprinting
  • deep learning
  • dictionary
  • inverse problem
  • manifold compressed sensing

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Cite this

Geometry of Deep Learning for Magnetic Resonance Fingerprinting. / Golbabaee, Mohammad; Chen, Dongdong; Gómez, Pedro A. ; Menzel, Marion I.; Davies, Mike.

In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Vol. 2019, 01.05.2019, p. 7825-7829.

Research output: Contribution to journalConference article

Golbabaee, Mohammad ; Chen, Dongdong ; Gómez, Pedro A. ; Menzel, Marion I. ; Davies, Mike. / Geometry of Deep Learning for Magnetic Resonance Fingerprinting. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2019 ; Vol. 2019. pp. 7825-7829.
@article{c8b2d90d5dfc4a89ab4f87f1412ed6c2,
title = "Geometry of Deep Learning for Magnetic Resonance Fingerprinting",
abstract = "Current popular methods for Magnetic Resonance Fingerprint (MRF) recovery are bottlenecked by the heavy storage and computation requirements of a dictionary-matching (DM) step due to the growing size and complexity of the fingerprint dictionaries in multi-parametric quantitative MRI applications. In this paper we study a deep learning approach to address these shortcomings. Coupled with a dimensionality reduction first layer, the proposed MRF-Net is able to reconstruct quantitative maps by saving more than 60 times in memory and computations required for a DM baseline. Fine-grid manifold enumeration i.e. the MRF dictionary is only used for training the network and not during image reconstruction. We show that the MRF-Net provides a piece-wise affine approximation to the Bloch response manifold projection and that rather than memorizing the dictionary, the network efficiently clusters this manifold and learns a set of hierarchical matched-filters for affine regression of the NMR characteristics in each segment.",
keywords = "Magnetic resonance fingerprinting, deep learning, dictionary, inverse problem, manifold compressed sensing",
author = "Mohammad Golbabaee and Dongdong Chen and G{\'o}mez, {Pedro A.} and Menzel, {Marion I.} and Mike Davies",
year = "2019",
month = "5",
day = "1",
doi = "10.1109/ICASSP.2019.8683549",
language = "English",
volume = "2019",
pages = "7825--7829",
journal = "IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)",
issn = "1520-6149",
publisher = "IEEE",

}

TY - JOUR

T1 - Geometry of Deep Learning for Magnetic Resonance Fingerprinting

AU - Golbabaee, Mohammad

AU - Chen, Dongdong

AU - Gómez, Pedro A.

AU - Menzel, Marion I.

AU - Davies, Mike

PY - 2019/5/1

Y1 - 2019/5/1

N2 - Current popular methods for Magnetic Resonance Fingerprint (MRF) recovery are bottlenecked by the heavy storage and computation requirements of a dictionary-matching (DM) step due to the growing size and complexity of the fingerprint dictionaries in multi-parametric quantitative MRI applications. In this paper we study a deep learning approach to address these shortcomings. Coupled with a dimensionality reduction first layer, the proposed MRF-Net is able to reconstruct quantitative maps by saving more than 60 times in memory and computations required for a DM baseline. Fine-grid manifold enumeration i.e. the MRF dictionary is only used for training the network and not during image reconstruction. We show that the MRF-Net provides a piece-wise affine approximation to the Bloch response manifold projection and that rather than memorizing the dictionary, the network efficiently clusters this manifold and learns a set of hierarchical matched-filters for affine regression of the NMR characteristics in each segment.

AB - Current popular methods for Magnetic Resonance Fingerprint (MRF) recovery are bottlenecked by the heavy storage and computation requirements of a dictionary-matching (DM) step due to the growing size and complexity of the fingerprint dictionaries in multi-parametric quantitative MRI applications. In this paper we study a deep learning approach to address these shortcomings. Coupled with a dimensionality reduction first layer, the proposed MRF-Net is able to reconstruct quantitative maps by saving more than 60 times in memory and computations required for a DM baseline. Fine-grid manifold enumeration i.e. the MRF dictionary is only used for training the network and not during image reconstruction. We show that the MRF-Net provides a piece-wise affine approximation to the Bloch response manifold projection and that rather than memorizing the dictionary, the network efficiently clusters this manifold and learns a set of hierarchical matched-filters for affine regression of the NMR characteristics in each segment.

KW - Magnetic resonance fingerprinting

KW - deep learning

KW - dictionary

KW - inverse problem

KW - manifold compressed sensing

UR - http://www.scopus.com/inward/record.url?scp=85068996054&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2019.8683549

DO - 10.1109/ICASSP.2019.8683549

M3 - Conference article

VL - 2019

SP - 7825

EP - 7829

JO - IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

JF - IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

SN - 1520-6149

ER -