Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction
Abstract
Recent years have seen a resurgence of interest in inexpensive low magnetic field (< 0.3 T) MRI systems mainly due to advances in magnet, coil and gradient set designs. Most of these advances have focused on improving hardware and signal acquisition strategies, and far less on the use of advanced image reconstruction methods to improve attainable image quality at low field. We describe here the use of our end-to-end deep neural network approach (AUTOMAP) to improve the image quality of highly noise-corrupted low-field MRI data. We compare the performance of this approach to two additional state-of-the-art denoising pipelines. We find that AUTOMAP improves image reconstruction of data acquired on two very different low-field MRI systems: human brain data acquired at 6.5 mT, and plant root data acquired at 47 mT, demonstrating SNR gains above Fourier reconstruction by factors of 1.5- to 4.5-fold, and 3-fold, respectively. In these applications, AUTOMAP outperformed two different contemporary image-based denoising algorithms, and suppressed noise-like spike artifacts in the reconstructed images. The impact of domain-specific training corpora on the reconstruction performance is discussed. The AUTOMAP approach to image reconstruction will enable significant image quality improvements at low-field, especially in highly noise-corrupted environments.
- Authors:
-
- Massachusetts General Hospital, Boston, MA (United States). Dept. of Radiology. Athinoula A. Martinos Center for Biomedical Imaging; Harvard Medical School, Boston, MA (United States)
- Texas A & M Univ., College Station, TX (United States). Dept. of Biological and Agricultural Engineering
- Massachusetts General Hospital, Boston, MA (United States). Dept. of Radiology. Athinoula A. Martinos Center for Biomedical Imaging; Boston Univ., MA (United States). Dept. of Biomedical Engineering
- Massachusetts General Hospital, Boston, MA (United States). Dept. of Radiology. Athinoula A. Martinos Center for Biomedical Imaging; Harvard Medical School, Boston, MA (United States); Harvard Univ., Cambridge, MA (United States). Div. of Physics
- Publication Date:
- Research Org.:
- Texas A & M Univ., College Station, TX (United States)
- Sponsoring Org.:
- USDOE Advanced Research Projects Agency - Energy (ARPA-E)
- OSTI Identifier:
- 1816731
- Grant/Contract Number:
- AR0000823
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Scientific Reports
- Additional Journal Information:
- Journal Volume: 11; Journal Issue: 1; Journal ID: ISSN 2045-2322
- Publisher:
- Nature Publishing Group
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 47 OTHER INSTRUMENTATION; 42 ENGINEERING; magnetic resonance imaging; medical imaging
Citation Formats
Koonjoo, N., Zhu, B., Bagnall, G. Cody, Bhutto, D., and Rosen, M. S. Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction. United States: N. p., 2021.
Web. doi:10.1038/s41598-021-87482-7.
Koonjoo, N., Zhu, B., Bagnall, G. Cody, Bhutto, D., & Rosen, M. S. Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction. United States. https://doi.org/10.1038/s41598-021-87482-7
Koonjoo, N., Zhu, B., Bagnall, G. Cody, Bhutto, D., and Rosen, M. S. Thu .
"Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction". United States. https://doi.org/10.1038/s41598-021-87482-7. https://www.osti.gov/servlets/purl/1816731.
@article{osti_1816731,
title = {Boosting the signal-to-noise of low-field MRI with deep learning image reconstruction},
author = {Koonjoo, N. and Zhu, B. and Bagnall, G. Cody and Bhutto, D. and Rosen, M. S.},
abstractNote = {Recent years have seen a resurgence of interest in inexpensive low magnetic field (< 0.3 T) MRI systems mainly due to advances in magnet, coil and gradient set designs. Most of these advances have focused on improving hardware and signal acquisition strategies, and far less on the use of advanced image reconstruction methods to improve attainable image quality at low field. We describe here the use of our end-to-end deep neural network approach (AUTOMAP) to improve the image quality of highly noise-corrupted low-field MRI data. We compare the performance of this approach to two additional state-of-the-art denoising pipelines. We find that AUTOMAP improves image reconstruction of data acquired on two very different low-field MRI systems: human brain data acquired at 6.5 mT, and plant root data acquired at 47 mT, demonstrating SNR gains above Fourier reconstruction by factors of 1.5- to 4.5-fold, and 3-fold, respectively. In these applications, AUTOMAP outperformed two different contemporary image-based denoising algorithms, and suppressed noise-like spike artifacts in the reconstructed images. The impact of domain-specific training corpora on the reconstruction performance is discussed. The AUTOMAP approach to image reconstruction will enable significant image quality improvements at low-field, especially in highly noise-corrupted environments.},
doi = {10.1038/s41598-021-87482-7},
journal = {Scientific Reports},
number = 1,
volume = 11,
place = {United States},
year = {Thu Apr 15 00:00:00 EDT 2021},
month = {Thu Apr 15 00:00:00 EDT 2021}
}
Works referenced in this record:
Assessment of Brain Injury Using Portable, Low-Field Magnetic Resonance Imaging at the Bedside of Critically Ill Patients
journal, January 2021
- Sheth, Kevin N.; Mazurek, Mercy H.; Yuen, Matthew M.
- JAMA Neurology, Vol. 78, Issue 1
Learning a variational network for reconstruction of accelerated MRI data: Learning a Variational Network for Reconstruction of Accelerated MRI Data
journal, November 2017
- Hammernik, Kerstin; Klatzer, Teresa; Kobler, Erich
- Magnetic Resonance in Medicine, Vol. 79, Issue 6
Low-Cost High-Performance MRI
journal, October 2015
- Sarracanie, Mathieu; LaPierre, Cristen D.; Salameh, Najat
- Scientific Reports, Vol. 5, Issue 1
In Situ Nuclear Magnetic Resonance Imaging of Roots: Influence of Soil Type, Ferromagnetic Particle Content, and Soil Water 1
journal, November 1987
- Rogers, Hugo H.; Bottomley, Paul A.
- Agronomy Journal, Vol. 79, Issue 6
VascuSynth: Simulating vascular trees for generating volumetric image data with ground-truth segmentation and tree analysis
journal, December 2010
- Hamarneh, Ghassan; Jassi, Preet
- Computerized Medical Imaging and Graphics, Vol. 34, Issue 8
Using the robust principal component analysis algorithm to remove RF spike artifacts from MR images: Removal of k-Space Spikes Using RPCA
journal, July 2015
- Campbell-Washburn, Adrienne E.; Atkinson, David; Nagy, Zoltan
- Magnetic Resonance in Medicine, Vol. 75, Issue 6
High-sensitivity in vivo contrast for ultra-low field magnetic resonance imaging using superparamagnetic iron oxide nanoparticles
journal, July 2020
- Waddington, David E. J.; Boele, Thomas; Maschmeyer, Richard
- Science Advances, Vol. 6, Issue 29
MGH–USC Human Connectome Project datasets with ultra-high b-value diffusion MRI
journal, January 2016
- Fan, Qiuyun; Witzel, Thomas; Nummenmaa, Aapo
- NeuroImage, Vol. 124
Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering
journal, August 2007
- Dabov, Kostadin; Foi, Alessandro; Katkovnik, Vladimir
- IEEE Transactions on Image Processing, Vol. 16, Issue 8
Image reconstruction by domain-transform manifold learning
journal, March 2018
- Zhu, Bo; Liu, Jeremiah Z.; Cauley, Stephen F.
- Nature, Vol. 555, Issue 7697
Low‐field MRI: An MR physics perspective
journal, January 2019
- Marques, José P.; Simonis, Frank F. J.; Webb, Andrew G.
- Journal of Magnetic Resonance Imaging, Vol. 49, Issue 6
Variable-Density Single-Shot Fast Spin-Echo MRI with Deep Learning Reconstruction by Using Variational Networks
journal, November 2018
- Chen, Feiyu; Taviani, Valentina; Malkiel, Itzik
- Radiology, Vol. 289, Issue 2
Image Reconstruction: From Sparsity to Data-Adaptive Methods and Machine Learning
journal, January 2020
- Ravishankar, Saiprasad; Ye, Jong Chul; Fessler, Jeffrey A.
- Proceedings of the IEEE, Vol. 108, Issue 1
Availability and Quality of Computed Tomography and Magnetic Resonance Imaging Equipment in U.S. Emergency Departments
journal, August 2008
- Ginde, Adit A.; Foianini, Anthony; Renner, Daniel M.
- Academic Emergency Medicine, Vol. 15, Issue 8
Deep learning for undersampled MRI reconstruction
journal, June 2018
- Hyun, Chang Min; Kim, Hwa Pyung; Lee, Sung Min
- Physics in Medicine & Biology, Vol. 63, Issue 13
Convolutional Recurrent Neural Networks for Dynamic MR Image Reconstruction
journal, January 2019
- Qin, Chen; Schlemper, Jo; Caballero, Jose
- IEEE Transactions on Medical Imaging, Vol. 38, Issue 1
Non-invasive imaging of plant roots in different soils using magnetic resonance imaging (MRI)
journal, November 2017
- Pflugfelder, Daniel; Metzner, Ralf; van Dusschoten, Dagmar
- Plant Methods, Vol. 13, Issue 1
Benchmarking MRI Reconstruction Neural Networks on Large Public Datasets
journal, March 2020
- Ramzi, Zaccharie; Ciuciu, Philippe; Starck, Jean-Luc
- Applied Sciences, Vol. 10, Issue 5
Low-field magnetic resonance imaging of roots in intact clayey and silty soils
journal, July 2020
- Bagnall, G. Cody; Koonjoo, Neha; Altobelli, Stephen A.
- Geoderma, Vol. 370
Deep learning on image denoising: An overview
journal, November 2020
- Tian, Chunwei; Fei, Lunke; Zheng, Wenxian
- Neural Networks, Vol. 131
KIKI ‐net: cross‐domain convolutional neural networks for reconstructing undersampled magnetic resonance images
journal, April 2018
- Eo, Taejoon; Jun, Yohan; Kim, Taeseong
- Magnetic Resonance in Medicine, Vol. 80, Issue 5
MRI artifact correction using sparse + low-rank decomposition of annihilating filter-based hankel matrix: MR Artifact Correction Using Robust ALOHA
journal, July 2016
- Jin, Kyong Hwan; Um, Ji-Yong; Lee, Dongwook
- Magnetic Resonance in Medicine, Vol. 78, Issue 1
Deep Convolutional Neural Network for Inverse Problems in Imaging
journal, September 2017
- Jin, Kyong Hwan; McCann, Michael T.; Froustey, Emmanuel
- IEEE Transactions on Image Processing, Vol. 26, Issue 9
Two-dimensional imaging in a lightweight portable MRI scanner without gradient coils: Lightweight MRI Scanner without Gradient Coils
journal, March 2014
- Cooley, Clarissa Zimmerman; Stockmann, Jason P.; Armstrong, Brandon D.
- Magnetic Resonance in Medicine, Vol. 73, Issue 2
Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
journal, July 2017
- Zhang, Kai; Zuo, Wangmeng; Chen, Yunjin
- IEEE Transactions on Image Processing, Vol. 26, Issue 7
Weighted Nuclear Norm Minimization with Application to Image Denoising
conference, June 2014
- Gu, Shuhang; Zhang, Lei; Zuo, Wangmeng
- 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Trainable Nonlinear Reaction Diffusion: A Flexible Framework for Fast and Effective Image Restoration
journal, June 2017
- Chen, Yunjin; Pock, Thomas
- IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 39, Issue 6
On instabilities of deep learning in image reconstruction and the potential costs of AI
journal, May 2020
- Antun, Vegard; Renna, Francesco; Poon, Clarice
- Proceedings of the National Academy of Sciences, Vol. 117, Issue 48
Denoising of 3D magnetic resonance images with multi-channel residual learning of convolutional neural network
journal, July 2018
- Jiang, Dongsheng; Dou, Weiqiang; Vosters, Luc
- Japanese Journal of Radiology, Vol. 36, Issue 9
Brief review of image denoising techniques
journal, July 2019
- Fan, Linwei; Zhang, Fan; Fan, Hui
- Visual Computing for Industry, Biomedicine, and Art, Vol. 2, Issue 1
Design of a sustainable prepolarizing magnetic resonance imaging system for infant hydrocephalus
journal, April 2018
- Obungoloch, Johnes; Harper, Joshua R.; Consevage, Steven
- Magnetic Resonance Materials in Physics, Biology and Medicine, Vol. 31, Issue 5
Modelling root–soil interactions using three–dimensional models of root growth, architecture and function
journal, June 2013
- Dunbabin, Vanessa M.; Postma, Johannes A.; Schnepf, Andrea
- Plant and Soil, Vol. 372, Issue 1-2