Deep learning for undersampled MRI reconstruction

Chang Min Hyun, Hwa Pyung Kim, Sung Min Lee, Sungchul Lee, Jin Keun Seo

Research output: Contribution to journalArticle

30 Citations (Scopus)

Abstract

This paper presents a deep learning method for faster magnetic resonance imaging (MRI) by reducing k-space data with sub-Nyquist sampling strategies and provides a rationale for why the proposed approach works well. Uniform subsampling is used in the time-consuming phase-encoding direction to capture high-resolution image information, while permitting the image-folding problem dictated by the Poisson summation formula. To deal with the localization uncertainty due to image folding, a small number of low-frequency k-space data are added. Training the deep learning net involves input and output images that are pairs of the Fourier transforms of the subsampled and fully sampled k-space data. Our experiments show the remarkable performance of the proposed method; only 29 of the k-space data can generate images of high quality as effectively as standard MRI reconstruction with the fully sampled data.

Original languageEnglish
Article number135007
JournalPhysics in medicine and biology
Volume63
Issue number13
DOIs
Publication statusPublished - 2018 Jun 25

Fingerprint

Magnetic Resonance Imaging
Learning
Fourier Analysis
Uncertainty
Direction compound

All Science Journal Classification (ASJC) codes

  • Radiological and Ultrasound Technology
  • Radiology Nuclear Medicine and imaging

Cite this

Hyun, Chang Min ; Kim, Hwa Pyung ; Lee, Sung Min ; Lee, Sungchul ; Seo, Jin Keun. / Deep learning for undersampled MRI reconstruction. In: Physics in medicine and biology. 2018 ; Vol. 63, No. 13.
@article{aeb8599cbbab479e9fc537295391f414,
title = "Deep learning for undersampled MRI reconstruction",
abstract = "This paper presents a deep learning method for faster magnetic resonance imaging (MRI) by reducing k-space data with sub-Nyquist sampling strategies and provides a rationale for why the proposed approach works well. Uniform subsampling is used in the time-consuming phase-encoding direction to capture high-resolution image information, while permitting the image-folding problem dictated by the Poisson summation formula. To deal with the localization uncertainty due to image folding, a small number of low-frequency k-space data are added. Training the deep learning net involves input and output images that are pairs of the Fourier transforms of the subsampled and fully sampled k-space data. Our experiments show the remarkable performance of the proposed method; only 29 of the k-space data can generate images of high quality as effectively as standard MRI reconstruction with the fully sampled data.",
author = "Hyun, {Chang Min} and Kim, {Hwa Pyung} and Lee, {Sung Min} and Sungchul Lee and Seo, {Jin Keun}",
year = "2018",
month = "6",
day = "25",
doi = "10.1088/1361-6560/aac71a",
language = "English",
volume = "63",
journal = "Physics in Medicine and Biology",
issn = "0031-9155",
publisher = "IOP Publishing Ltd.",
number = "13",

}

Deep learning for undersampled MRI reconstruction. / Hyun, Chang Min; Kim, Hwa Pyung; Lee, Sung Min; Lee, Sungchul; Seo, Jin Keun.

In: Physics in medicine and biology, Vol. 63, No. 13, 135007, 25.06.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Deep learning for undersampled MRI reconstruction

AU - Hyun, Chang Min

AU - Kim, Hwa Pyung

AU - Lee, Sung Min

AU - Lee, Sungchul

AU - Seo, Jin Keun

PY - 2018/6/25

Y1 - 2018/6/25

N2 - This paper presents a deep learning method for faster magnetic resonance imaging (MRI) by reducing k-space data with sub-Nyquist sampling strategies and provides a rationale for why the proposed approach works well. Uniform subsampling is used in the time-consuming phase-encoding direction to capture high-resolution image information, while permitting the image-folding problem dictated by the Poisson summation formula. To deal with the localization uncertainty due to image folding, a small number of low-frequency k-space data are added. Training the deep learning net involves input and output images that are pairs of the Fourier transforms of the subsampled and fully sampled k-space data. Our experiments show the remarkable performance of the proposed method; only 29 of the k-space data can generate images of high quality as effectively as standard MRI reconstruction with the fully sampled data.

AB - This paper presents a deep learning method for faster magnetic resonance imaging (MRI) by reducing k-space data with sub-Nyquist sampling strategies and provides a rationale for why the proposed approach works well. Uniform subsampling is used in the time-consuming phase-encoding direction to capture high-resolution image information, while permitting the image-folding problem dictated by the Poisson summation formula. To deal with the localization uncertainty due to image folding, a small number of low-frequency k-space data are added. Training the deep learning net involves input and output images that are pairs of the Fourier transforms of the subsampled and fully sampled k-space data. Our experiments show the remarkable performance of the proposed method; only 29 of the k-space data can generate images of high quality as effectively as standard MRI reconstruction with the fully sampled data.

UR - http://www.scopus.com/inward/record.url?scp=85049973605&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85049973605&partnerID=8YFLogxK

U2 - 10.1088/1361-6560/aac71a

DO - 10.1088/1361-6560/aac71a

M3 - Article

C2 - 29787383

AN - SCOPUS:85049973605

VL - 63

JO - Physics in Medicine and Biology

JF - Physics in Medicine and Biology

SN - 0031-9155

IS - 13

M1 - 135007

ER -