Conditional generative adversarial network for 3D rigid‐body motion correction in MRI

Magnetic Resonance in Medicine - Tập 82 Số 3 - Trang 901-910 - 2019
Patricia M. Johnson1,2, Maria Drangova1,2
1Department of Medical Biophysics, Schulich School of Medicine & Dentistry, The University of Western Ontario, London, Ontario, Canada
2Imaging Research Laboratories, Robarts Research Institute, the University of Western Ontario, London, Ontario, Canada

Tóm tắt

PurposeSubject motion in MRI remains an unsolved problem; motion during image acquisition may cause blurring and artifacts that severely degrade image quality. In this work, we approach motion correction as an image‐to‐image translation problem, which refers to the approach of training a deep neural network to predict an image in 1 domain from an image in another domain. Specifically, the purpose of this work was to develop and train a conditional generative adversarial network to predict artifact‐free brain images from motion‐corrupted data.MethodsAn open source MRI data set comprising T2*‐weighted, FLASH magnitude, and phase brain images for 53 patients was used to generate complex image data for motion simulation. To simulate rigid motion, rotations and translations were applied to the image data based on randomly generated motion profiles. A conditional generative adversarial network, comprising a generator and discriminator networks, was trained using the motion‐corrupted and corresponding ground truth (original) images as training pairs.ResultsThe images predicted by the conditional generative adversarial network have improved image quality compared to the motion‐corrupted images. The mean absolute error between the motion‐corrupted and ground‐truth images of the test set was 16.4% of the image mean value, whereas the mean absolute error between the conditional generative adversarial network‐predicted and ground‐truth images was 10.8% The network output also demonstrated improved peak SNR and structural similarity index for all test‐set images.ConclusionThe images predicted by the conditional generative adversarial network have quantitatively and qualitatively improved image quality compared to the motion‐corrupted images.

Từ khóa


Tài liệu tham khảo

10.1016/j.jacr.2015.03.007

10.1371/journal.pone.0133921

10.1002/mrm.27033

10.1002/mrm.26838

10.1002/mrm.26364

10.1016/j.neuroimage.2015.11.054

10.1002/mrm.25670

10.1002/mrm.21038

10.1016/j.mri.2016.06.006

10.1002/mrm.27381

10.1002/mrm.27106

10.1002/mrm.26977

10.1038/nature25988

10.1109/TMI.2018.2791721

10.1007/s00330-018-5595-8

10.1109/TMI.2018.2837502

10.1016/j.compbiomed.2018.05.027

10.1109/JBHI.2018.2843819

10.1088/1361-6560/aab9e9

Nie D, 2017, Medical image synthesis with context‐aware generative adversarial networks, Med Image Comput Comput Assist Interv, 10435, 417

10.1109/ICASSP.2017.7952268

10.1007/s10334-017-0650-z

2018 K Pawar Z Chen NJ Shah GF Egan Motion correction in MRI using deep convolutional neural network

10.1002/mrm.27096

10.1109/CISP-BMEI.2017.8302197

MirzaM OsinderoS.Conditional generative adverserial nets.2014. arXiv:14111784 [cs.LG].

IsolaP ZhuJ‐Y ZhouT EfrosA.Image‐to‐image translation with conditional adversarial networks.2017. arXiv:161107004 [csCV].

10.1038/sdata.2014.50

10.1002/mrm.23228

Goodfellow IJ, 2014, Proceedings of the 27th International Conference on Neural Information Processing Systems, (NIPS) Volume 2, 2672

10.1007/978-3-319-24574-4_28

CholletF.Keras GitHub.2015. Available at:https://github.com/fchollet/keras. GitHub repository. Accessed January 2018.

AbadiM BarhamP ChenJ et al.TensorFlow: Large‐scale machine learning on heterogeneous systems.2015. Available at:https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/45166.pdf. Accessed January 2018.

CholletF.Keras advanced activations GitHub.2015. Available at:https://github.com/fchollet/keras. Accessed January 2018.

CholletF.Keras activations GitHub.2015. Available at:https://github.com/fchollet/keras. Accessed January 2018.

KingmaDP BaJ.Adam: a method for stochastic optimization.2015. arXiv:14126980 [cs.LG].

10.1561/0600000037

10.1049/iet-ipr.2012.0489

10.1109/TIP.2003.819861

10.1109/TMI.2018.2820120

10.1002/mp.12945

10.1109/TBME.2018.2821699

CohenJP LuckM HonariS.Distribution matching losses can hallucinate features in medical image translation.2018. arXiv:180508841.

10.1002/mrm.27355

HanY YeJC.k‐space deep learning for accelerated MRI.2018. arXiv:180503779 [csCV].

2018 PM Johnson M Drangova Motion correction in MRI using deep learning 4098

Thông tin
Thông tin xuất bản

Magnetic Resonance in Medicine

Tập 82 Số 3

901-910

Thông tin tác giả