Literature DB >> 32614100

Self-supervised learning of physics-guided reconstruction neural networks without fully sampled reference data.

Burhaneddin Yaman1,2, Seyed Amir Hossein Hosseini1,2, Steen Moeller2, Jutta Ellermann2, Kâmil Uğurbil2, Mehmet Akçakaya1,2.   

Abstract

PURPOSE: To develop a strategy for training a physics-guided MRI reconstruction neural network without a database of fully sampled data sets.
METHODS: Self-supervised learning via data undersampling (SSDU) for physics-guided deep learning reconstruction partitions available measurements into two disjoint sets, one of which is used in the data consistency (DC) units in the unrolled network and the other is used to define the loss for training. The proposed training without fully sampled data is compared with fully supervised training with ground-truth data, as well as conventional compressed-sensing and parallel imaging methods using the publicly available fastMRI knee database. The same physics-guided neural network is used for both proposed SSDU and supervised training. The SSDU training is also applied to prospectively two-fold accelerated high-resolution brain data sets at different acceleration rates, and compared with parallel imaging.
RESULTS: Results on five different knee sequences at an acceleration rate of 4 shows that the proposed self-supervised approach performs closely with supervised learning, while significantly outperforming conventional compressed-sensing and parallel imaging, as characterized by quantitative metrics and a clinical reader study. The results on prospectively subsampled brain data sets, in which supervised learning cannot be used due to lack of ground-truth reference, show that the proposed self-supervised approach successfully performs reconstruction at high acceleration rates (4, 6, and 8). Image readings indicate improved visual reconstruction quality with the proposed approach compared with parallel imaging at acquisition acceleration.
CONCLUSION: The proposed SSDU approach allows training of physics-guided deep learning MRI reconstruction without fully sampled data, while achieving comparable results with supervised deep learning MRI trained on fully sampled data.
© 2020 International Society for Magnetic Resonance in Medicine.

Keywords:  accelerated imaging; convolutional neural networks; deep learning; image reconstruction; parallel imaging; self-supervised learning

Mesh:

Year:  2020        PMID: 32614100      PMCID: PMC7811359          DOI: 10.1002/mrm.28378

Source DB:  PubMed          Journal:  Magn Reson Med        ISSN: 0740-3194            Impact factor:   4.668


  52 in total

1.  SENSE: sensitivity encoding for fast MRI.

Authors:  K P Pruessmann; M Weiger; M B Scheidegger; P Boesiger
Journal:  Magn Reson Med       Date:  1999-11       Impact factor: 4.668

2.  Regularization parameter selection for nonlinear iterative image restoration and MRI reconstruction using GCV and SURE-based methods.

Authors:  Sathish Ramani; Zhihao Liu; Jeffrey Rosen; Jon-Fredrik Nielsen; Jeffrey A Fessler
Journal:  IEEE Trans Image Process       Date:  2012-04-17       Impact factor: 10.856

3.  Undersampled radial MRI with multiple coils. Iterative image reconstruction using a total variation constraint.

Authors:  Kai Tobias Block; Martin Uecker; Jens Frahm
Journal:  Magn Reson Med       Date:  2007-06       Impact factor: 4.668

4.  Projection reconstruction MR imaging using FOCUSS.

Authors:  Jong Chul Ye; Sungho Tak; Yeji Han; Hyun Wook Park
Journal:  Magn Reson Med       Date:  2007-04       Impact factor: 4.668

5.  Compressed sensing in dynamic MRI.

Authors:  Urs Gamper; Peter Boesiger; Sebastian Kozerke
Journal:  Magn Reson Med       Date:  2008-02       Impact factor: 4.668

6.  k-t FOCUSS: a general compressed sensing framework for high resolution dynamic MRI.

Authors:  Hong Jung; Kyunghyun Sung; Krishna S Nayak; Eung Yeop Kim; Jong Chul Ye
Journal:  Magn Reson Med       Date:  2009-01       Impact factor: 4.668

7.  Highly undersampled magnetic resonance image reconstruction via homotopic l(0) -minimization.

Authors:  Joshua Trzasko; Armando Manduca
Journal:  IEEE Trans Med Imaging       Date:  2009-01       Impact factor: 10.048

8.  Image reconstruction by domain-transform manifold learning.

Authors:  Bo Zhu; Jeremiah Z Liu; Stephen F Cauley; Bruce R Rosen; Matthew S Rosen
Journal:  Nature       Date:  2018-03-21       Impact factor: 49.962

9.  Learning a variational network for reconstruction of accelerated MRI data.

Authors:  Kerstin Hammernik; Teresa Klatzer; Erich Kobler; Michael P Recht; Daniel K Sodickson; Thomas Pock; Florian Knoll
Journal:  Magn Reson Med       Date:  2017-11-08       Impact factor: 4.668

10.  Deep Generative Adversarial Neural Networks for Compressive Sensing MRI.

Authors:  Morteza Mardani; Enhao Gong; Joseph Y Cheng; Shreyas S Vasanawala; Greg Zaharchuk; Lei Xing; John M Pauly
Journal:  IEEE Trans Med Imaging       Date:  2018-07-23       Impact factor: 10.048

View more
  18 in total

1.  Dense Recurrent Neural Networks for Accelerated MRI: History-Cognizant Unrolling of Optimization Algorithms.

Authors:  Seyed Amir Hossein Hosseini; Burhaneddin Yaman; Steen Moeller; Mingyi Hong; Mehmet Akçakaya
Journal:  IEEE J Sel Top Signal Process       Date:  2020-06-17       Impact factor: 6.856

2.  Compressed Sensing MRI with ℓ1-Wavelet Reconstruction Revisited Using Modern Data Science Tools.

Authors:  Hongyi Gu; Burhaneddin Yaman; Kamil Ugurbil; Steen Moeller; Mehmet Akcakaya
Journal:  Annu Int Conf IEEE Eng Med Biol Soc       Date:  2021-11

3.  20-fold Accelerated 7T fMRI Using Referenceless Self-Supervised Deep Learning Reconstruction.

Authors:  Omer Burak Demirel; Burhaneddin Yaman; Logan Dowdle; Steen Moeller; Luca Vizioli; Essa Yacoub; John Strupp; Cheryl A Olman; Kamil Ugurbil; Mehmet Akcakaya
Journal:  Annu Int Conf IEEE Eng Med Biol Soc       Date:  2021-11

4.  Unsupervised Deep Learning Methods for Biological Image Reconstruction and Enhancement: An overview from a signal processing perspective.

Authors:  Mehmet Akçakaya; Burhaneddin Yaman; Hyungjin Chung; Jong Chul Ye
Journal:  IEEE Signal Process Mag       Date:  2022-02-24       Impact factor: 15.204

5.  MRI RECOVERY WITH A SELF-CALIBRATED DENOISER.

Authors:  Sizhuo Liu; Philip Schniter; Rizwan Ahmad
Journal:  Proc IEEE Int Conf Acoust Speech Signal Process       Date:  2022-04-27

6.  Deep denoising for multi-dimensional synchrotron X-ray tomography without high-quality reference data.

Authors:  Allard A Hendriksen; Minna Bührer; Laura Leone; Marco Merlini; Nicola Vigano; Daniël M Pelt; Federica Marone; Marco di Michiel; K Joost Batenburg
Journal:  Sci Rep       Date:  2021-06-04       Impact factor: 4.379

7.  Blind Primed Supervised (BLIPS) Learning for MR Image Reconstruction.

Authors:  Anish Lahiri; Guanhua Wang; Saiprasad Ravishankar; Jeffrey A Fessler
Journal:  IEEE Trans Med Imaging       Date:  2021-10-27       Impact factor: 10.048

8.  ENSURE: ENSEMBLE STEIN'S UNBIASED RISK ESTIMATOR FOR UNSUPERVISED LEARNING.

Authors:  Hemant Kumar Aggarwal; Aniket Pramanik; Mathews Jacob
Journal:  Proc IEEE Int Conf Acoust Speech Signal Process       Date:  2021-06

9.  ULTRAHIGH FIELD and ULTRAHIGH RESOLUTION fMRI.

Authors:  Kamil Uğurbil
Journal:  Curr Opin Biomed Eng       Date:  2021-04-14

10.  Accelerate gas diffusion-weighted MRI for lung morphometry with deep learning.

Authors:  Caohui Duan; He Deng; Sa Xiao; Junshuai Xie; Haidong Li; Xiuchao Zhao; Dongshan Han; Xianping Sun; Xin Lou; Chaohui Ye; Xin Zhou
Journal:  Eur Radiol       Date:  2021-07-13       Impact factor: 7.034

View more

北京卡尤迪生物科技股份有限公司 © 2022-2023.