Inverse GANs for accelerated MRI reconstruction

Dominik Narnhofer, Kerstin Hammernik, Florian Knoll, Thomas Pock

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

State-of-the-art algorithms for accelerated magnetic resonance image (MRI) reconstruction are nowadays dominated by deep learning-based techniques. However, the majority of these methods require the respective sampling patterns in training, which limits their application to a specific problem class. We propose an iterative reconstruction approach that incorporates the implicit prior provided by a generative adversarial network (GAN), which learns the probability distribution of uncorrupted MRI data in an off-line step. Since the unsupervised training of the GAN is completely independent of the measurement process, our method is in principle able to address multiple sampling modalities using a single pre-trained model. However, it turns out that the desired target images potentially lie outside the range space of the learned GAN, leading to reconstructions that resemble the target images only at a coarse level of …
Original languageEnglish
Title of host publicationWavelets and Sparsity XVIII
Publication statusPublished - 2019

Fingerprint

Magnetic resonance
Image reconstruction
Sampling
Probability distributions
Deep learning

Cite this

Narnhofer, D., Hammernik, K., Knoll, F., & Pock, T. (2019). Inverse GANs for accelerated MRI reconstruction. In Wavelets and Sparsity XVIII

Inverse GANs for accelerated MRI reconstruction. / Narnhofer, Dominik; Hammernik, Kerstin; Knoll, Florian; Pock, Thomas.

Wavelets and Sparsity XVIII. 2019.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Narnhofer, D, Hammernik, K, Knoll, F & Pock, T 2019, Inverse GANs for accelerated MRI reconstruction. in Wavelets and Sparsity XVIII.
Narnhofer D, Hammernik K, Knoll F, Pock T. Inverse GANs for accelerated MRI reconstruction. In Wavelets and Sparsity XVIII. 2019
Narnhofer, Dominik ; Hammernik, Kerstin ; Knoll, Florian ; Pock, Thomas. / Inverse GANs for accelerated MRI reconstruction. Wavelets and Sparsity XVIII. 2019.
@inproceedings{1131e27f86d14376836cd5c301b4a04f,
title = "Inverse GANs for accelerated MRI reconstruction",
abstract = "State-of-the-art algorithms for accelerated magnetic resonance image (MRI) reconstruction are nowadays dominated by deep learning-based techniques. However, the majority of these methods require the respective sampling patterns in training, which limits their application to a specific problem class. We propose an iterative reconstruction approach that incorporates the implicit prior provided by a generative adversarial network (GAN), which learns the probability distribution of uncorrupted MRI data in an off-line step. Since the unsupervised training of the GAN is completely independent of the measurement process, our method is in principle able to address multiple sampling modalities using a single pre-trained model. However, it turns out that the desired target images potentially lie outside the range space of the learned GAN, leading to reconstructions that resemble the target images only at a coarse level of …",
author = "Dominik Narnhofer and Kerstin Hammernik and Florian Knoll and Thomas Pock",
year = "2019",
language = "English",
booktitle = "Wavelets and Sparsity XVIII",

}

TY - GEN

T1 - Inverse GANs for accelerated MRI reconstruction

AU - Narnhofer, Dominik

AU - Hammernik, Kerstin

AU - Knoll, Florian

AU - Pock, Thomas

PY - 2019

Y1 - 2019

N2 - State-of-the-art algorithms for accelerated magnetic resonance image (MRI) reconstruction are nowadays dominated by deep learning-based techniques. However, the majority of these methods require the respective sampling patterns in training, which limits their application to a specific problem class. We propose an iterative reconstruction approach that incorporates the implicit prior provided by a generative adversarial network (GAN), which learns the probability distribution of uncorrupted MRI data in an off-line step. Since the unsupervised training of the GAN is completely independent of the measurement process, our method is in principle able to address multiple sampling modalities using a single pre-trained model. However, it turns out that the desired target images potentially lie outside the range space of the learned GAN, leading to reconstructions that resemble the target images only at a coarse level of …

AB - State-of-the-art algorithms for accelerated magnetic resonance image (MRI) reconstruction are nowadays dominated by deep learning-based techniques. However, the majority of these methods require the respective sampling patterns in training, which limits their application to a specific problem class. We propose an iterative reconstruction approach that incorporates the implicit prior provided by a generative adversarial network (GAN), which learns the probability distribution of uncorrupted MRI data in an off-line step. Since the unsupervised training of the GAN is completely independent of the measurement process, our method is in principle able to address multiple sampling modalities using a single pre-trained model. However, it turns out that the desired target images potentially lie outside the range space of the learned GAN, leading to reconstructions that resemble the target images only at a coarse level of …

M3 - Conference contribution

BT - Wavelets and Sparsity XVIII

ER -