TY - JOUR
T1 - MRI super-resolution via realistic downsampling with adversarial learning
AU - Huang, Bangyan
AU - Xiao, Haonan
AU - Liu, Weiwei
AU - Zhang, Yibao
AU - Wu, Hao
AU - Wang, Weihu
AU - Yang, Yunhuan
AU - Yang, Yidong
AU - Miller, G. Wilson
AU - Li, Tian
AU - Cai, Jing
N1 - Publisher Copyright:
© 2021 Institute of Physics and Engineering in Medicine.
PY - 2021/10/21
Y1 - 2021/10/21
N2 - Many deep learning (DL) frameworks have demonstrated state-of-the-art performance in the super-resolution (SR) task of magnetic resonance imaging, but most performances have been achieved with simulated low-resolution (LR) images rather than LR images from real acquisition. Due to the limited generalizability of the SR network, enhancement is not guaranteed for real LR images because of the unreality of the training LR images. In this study, we proposed a DL-based SR framework with an emphasis on data construction to achieve better performance on real LR MR images. The framework comprised two steps: (a) downsampling training using a generative adversarial network (GAN) to construct more realistic and perfectly matched LR/high-resolution (HR) pairs. The downsampling GAN input was real LR and HR images. The generator translated the HR images to LR images and the discriminator distinguished the patch-level difference between the synthetic and real LR images. (b) SR training was performed using an enhance4d deep super-resolution network (EDSR). In the controlled experiments, three EDSRs were trained using our proposed method, Gaussian blur, and k-space zero-filling. As for the data, liver MR images were obtained from 24 patients using breath-hold serial LR and HR scans (only HR images were used in the conventional methods). The k-space zero-filling group delivered almost zero enhancement on the real LR images and the Gaussian group produced a considerable number of artifacts. The proposed method exhibited significantly better resolution enhancement and fewer artifacts compared with the other two networks. Our method outperformed the Gaussian method by an improvement of 0.111 0.016 in the structural similarity index and 2.76 0.98 dB in the peak signal-to-noise ratio. The blind/reference-less image spatial quality evaluator metric of the conventional Gaussian method and proposed method were 46.6 4.2 and 34.1 2.4, respectively.
AB - Many deep learning (DL) frameworks have demonstrated state-of-the-art performance in the super-resolution (SR) task of magnetic resonance imaging, but most performances have been achieved with simulated low-resolution (LR) images rather than LR images from real acquisition. Due to the limited generalizability of the SR network, enhancement is not guaranteed for real LR images because of the unreality of the training LR images. In this study, we proposed a DL-based SR framework with an emphasis on data construction to achieve better performance on real LR MR images. The framework comprised two steps: (a) downsampling training using a generative adversarial network (GAN) to construct more realistic and perfectly matched LR/high-resolution (HR) pairs. The downsampling GAN input was real LR and HR images. The generator translated the HR images to LR images and the discriminator distinguished the patch-level difference between the synthetic and real LR images. (b) SR training was performed using an enhance4d deep super-resolution network (EDSR). In the controlled experiments, three EDSRs were trained using our proposed method, Gaussian blur, and k-space zero-filling. As for the data, liver MR images were obtained from 24 patients using breath-hold serial LR and HR scans (only HR images were used in the conventional methods). The k-space zero-filling group delivered almost zero enhancement on the real LR images and the Gaussian group produced a considerable number of artifacts. The proposed method exhibited significantly better resolution enhancement and fewer artifacts compared with the other two networks. Our method outperformed the Gaussian method by an improvement of 0.111 0.016 in the structural similarity index and 2.76 0.98 dB in the peak signal-to-noise ratio. The blind/reference-less image spatial quality evaluator metric of the conventional Gaussian method and proposed method were 46.6 4.2 and 34.1 2.4, respectively.
UR - http://www.scopus.com/inward/record.url?scp=85117222487&partnerID=8YFLogxK
U2 - 10.1088/1361-6560/ac232e
DO - 10.1088/1361-6560/ac232e
M3 - Journal article
AN - SCOPUS:85117222487
SN - 0031-9155
VL - 66
JO - Physics in Medicine and Biology
JF - Physics in Medicine and Biology
IS - 20
M1 - 205004
ER -