Self super-resolution for magnetic resonance images using deep networks

Can Zhao, Aaron Carass, Blake E. Dewey, Jerry Ladd Prince

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

High resolution magnetic resonance (MR) imaging (MRI) is desirable in many clinical applications; however, there is a trade-off between resolution, speed of acquisition, and noise. It is common for MR images to have worse through-plane resolution (slice thickness) than in-plane resolution. In these MRI images, high frequency information in the through-plane direction is not acquired, and cannot be resolved through interpolation. To address this issue, super-resolution methods have been developed to enhance spatial resolution. As an ill-posed problem, state-of-the-art super-resolution methods rely on the presence of external/training atlases to learn the transform from low resolution (LR) images to high resolution (HR) images. For several reasons, such HR atlas images are often not available for MRI sequences. This paper presents a self super-resolution (SSR) algorithm, which does not use any external atlas images, yet can still resolve HR images only reliant on the acquired LR image. We use a blurred version of the input image to create training data for a state-of-the-art super-resolution deep network. The trained network is applied to the original input image to estimate the HR image. Our SSR result shows a significant improvement on through-plane resolution compared to competing SSR methods.

Original languageEnglish (US)
Title of host publication2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018
PublisherIEEE Computer Society
Pages365-368
Number of pages4
Volume2018-April
ISBN (Electronic)9781538636367
DOIs
StatePublished - May 23 2018
Event15th IEEE International Symposium on Biomedical Imaging, ISBI 2018 - Washington, United States
Duration: Apr 4 2018Apr 7 2018

Other

Other15th IEEE International Symposium on Biomedical Imaging, ISBI 2018
CountryUnited States
CityWashington
Period4/4/184/7/18

Fingerprint

Atlases
Magnetic resonance
Image resolution
Magnetic Resonance Spectroscopy
Imaging techniques
Noise
Magnetic Resonance Imaging
Interpolation

Keywords

  • CNN
  • Deep network
  • MRI
  • Self super-resolution

ASJC Scopus subject areas

  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging

Cite this

Zhao, C., Carass, A., Dewey, B. E., & Prince, J. L. (2018). Self super-resolution for magnetic resonance images using deep networks. In 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018 (Vol. 2018-April, pp. 365-368). IEEE Computer Society. https://doi.org/10.1109/ISBI.2018.8363594

Self super-resolution for magnetic resonance images using deep networks. / Zhao, Can; Carass, Aaron; Dewey, Blake E.; Prince, Jerry Ladd.

2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018. Vol. 2018-April IEEE Computer Society, 2018. p. 365-368.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhao, C, Carass, A, Dewey, BE & Prince, JL 2018, Self super-resolution for magnetic resonance images using deep networks. in 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018. vol. 2018-April, IEEE Computer Society, pp. 365-368, 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018, Washington, United States, 4/4/18. https://doi.org/10.1109/ISBI.2018.8363594
Zhao C, Carass A, Dewey BE, Prince JL. Self super-resolution for magnetic resonance images using deep networks. In 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018. Vol. 2018-April. IEEE Computer Society. 2018. p. 365-368 https://doi.org/10.1109/ISBI.2018.8363594
Zhao, Can ; Carass, Aaron ; Dewey, Blake E. ; Prince, Jerry Ladd. / Self super-resolution for magnetic resonance images using deep networks. 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018. Vol. 2018-April IEEE Computer Society, 2018. pp. 365-368
@inproceedings{a66c2c0a61f24ea98e9409ca5df47be3,
title = "Self super-resolution for magnetic resonance images using deep networks",
abstract = "High resolution magnetic resonance (MR) imaging (MRI) is desirable in many clinical applications; however, there is a trade-off between resolution, speed of acquisition, and noise. It is common for MR images to have worse through-plane resolution (slice thickness) than in-plane resolution. In these MRI images, high frequency information in the through-plane direction is not acquired, and cannot be resolved through interpolation. To address this issue, super-resolution methods have been developed to enhance spatial resolution. As an ill-posed problem, state-of-the-art super-resolution methods rely on the presence of external/training atlases to learn the transform from low resolution (LR) images to high resolution (HR) images. For several reasons, such HR atlas images are often not available for MRI sequences. This paper presents a self super-resolution (SSR) algorithm, which does not use any external atlas images, yet can still resolve HR images only reliant on the acquired LR image. We use a blurred version of the input image to create training data for a state-of-the-art super-resolution deep network. The trained network is applied to the original input image to estimate the HR image. Our SSR result shows a significant improvement on through-plane resolution compared to competing SSR methods.",
keywords = "CNN, Deep network, MRI, Self super-resolution",
author = "Can Zhao and Aaron Carass and Dewey, {Blake E.} and Prince, {Jerry Ladd}",
year = "2018",
month = "5",
day = "23",
doi = "10.1109/ISBI.2018.8363594",
language = "English (US)",
volume = "2018-April",
pages = "365--368",
booktitle = "2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Self super-resolution for magnetic resonance images using deep networks

AU - Zhao, Can

AU - Carass, Aaron

AU - Dewey, Blake E.

AU - Prince, Jerry Ladd

PY - 2018/5/23

Y1 - 2018/5/23

N2 - High resolution magnetic resonance (MR) imaging (MRI) is desirable in many clinical applications; however, there is a trade-off between resolution, speed of acquisition, and noise. It is common for MR images to have worse through-plane resolution (slice thickness) than in-plane resolution. In these MRI images, high frequency information in the through-plane direction is not acquired, and cannot be resolved through interpolation. To address this issue, super-resolution methods have been developed to enhance spatial resolution. As an ill-posed problem, state-of-the-art super-resolution methods rely on the presence of external/training atlases to learn the transform from low resolution (LR) images to high resolution (HR) images. For several reasons, such HR atlas images are often not available for MRI sequences. This paper presents a self super-resolution (SSR) algorithm, which does not use any external atlas images, yet can still resolve HR images only reliant on the acquired LR image. We use a blurred version of the input image to create training data for a state-of-the-art super-resolution deep network. The trained network is applied to the original input image to estimate the HR image. Our SSR result shows a significant improvement on through-plane resolution compared to competing SSR methods.

AB - High resolution magnetic resonance (MR) imaging (MRI) is desirable in many clinical applications; however, there is a trade-off between resolution, speed of acquisition, and noise. It is common for MR images to have worse through-plane resolution (slice thickness) than in-plane resolution. In these MRI images, high frequency information in the through-plane direction is not acquired, and cannot be resolved through interpolation. To address this issue, super-resolution methods have been developed to enhance spatial resolution. As an ill-posed problem, state-of-the-art super-resolution methods rely on the presence of external/training atlases to learn the transform from low resolution (LR) images to high resolution (HR) images. For several reasons, such HR atlas images are often not available for MRI sequences. This paper presents a self super-resolution (SSR) algorithm, which does not use any external atlas images, yet can still resolve HR images only reliant on the acquired LR image. We use a blurred version of the input image to create training data for a state-of-the-art super-resolution deep network. The trained network is applied to the original input image to estimate the HR image. Our SSR result shows a significant improvement on through-plane resolution compared to competing SSR methods.

KW - CNN

KW - Deep network

KW - MRI

KW - Self super-resolution

UR - http://www.scopus.com/inward/record.url?scp=85048136024&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048136024&partnerID=8YFLogxK

U2 - 10.1109/ISBI.2018.8363594

DO - 10.1109/ISBI.2018.8363594

M3 - Conference contribution

AN - SCOPUS:85048136024

VL - 2018-April

SP - 365

EP - 368

BT - 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018

PB - IEEE Computer Society

ER -