key: cord-0478031-78nzh719 authors: Li, Guangyuan; Lv, Jun; Tong, Xiangrong; Wang, Chengyan; Yang, Guang title: High-Resolution Pelvic MRI Reconstruction Using a Generative Adversarial Network with Attention and Cyclic Loss date: 2021-07-21 journal: nan DOI: nan sha: 78f481335e5c7014cf7a2398a28cc4e4458f5406 doc_id: 478031 cord_uid: 78nzh719 Magnetic resonance imaging (MRI) is an important medical imaging modality, but its acquisition speed is quite slow due to the physiological limitations. Recently, super-resolution methods have shown excellent performance in accelerating MRI. In some circumstances, it is difficult to obtain high-resolution images even with prolonged scan time. Therefore, we proposed a novel super-resolution method that uses a generative adversarial network (GAN) with cyclic loss and attention mechanism to generate high-resolution MR images from low-resolution MR images by a factor of 2. We implemented our model on pelvic images from healthy subjects as training and validation data, while those data from patients were used for testing. The MR dataset was obtained using different imaging sequences, including T2, T2W SPAIR, and mDIXON-W. Four methods, i.e., BICUBIC, SRCNN, SRGAN, and EDSR were used for comparison. Structural similarity, peak signal to noise ratio, root mean square error, and variance inflation factor were used as calculation indicators to evaluate the performances of the proposed method. Various experimental results showed that our method can better restore the details of the high-resolution MR image as compared to the other methods. In addition, the reconstructed high-resolution MR image can provide better lesion textures in the tumor patients, which is promising to be used in clinical diagnosis. Magnetic resonance imaging (MRI) has the characteristics of non-invasive, non-radiation, and high contrast, etc. It is an important medical imaging modality. However, its imaging speed is relatively slow, which is mainly limited by the physiological factors including magnetic field strength, slew rate, nerve stimulation, etc. Compressed sensing mri Compressed sensing Kiki-net: crossdomain convolutional neural networks for reconstructing undersampled magnetic resonance images Deeplearning methods for parallel magnetic resonance imaging reconstruction: A survey of the current approaches, trends, and issues Reducing acquisition time in clinical mri by data undersampling and compressed sensing reconstruction Learning a deep convolutional network for image super-resolution Accelerating the super-resolution convolutional neural network Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network Accurate image super-resolution using very deep convolutional networks Deeply-recursive convolutional network for image super-resolution Image restoration using convolutional auto-encoders with symmetric skip connections Image super-resolution via deep recursive residual network Image super-resolution using dense skip connections Photo-realistic single image super-resolution using a generative adversarial network Generative adversarial nets Enhanced deep residual networks for single image superresolution Super-resolution reconstruction of knee magnetic resonance imaging based on deep learning Mri super-resolution with ensemble learning and complementary priors Rapid whole-heart cmr with single volume super-resolution U-net: Convolutional networks for biomedical image segmentation A hybrid convolutional neural network for super-resolution reconstruction of mr images Superresolution reconstruction of single anisotropic 3d mr images using residual convolutional neural network Backpropagation applied to handwritten zip code recognition High-resolution breast mri reconstruction using a deep convolutional generative adversarial network Unsupervised representation learning with deep convolutional generative adversarial networks Super-resolution musculoskeletal mri using deep learning Multi-contrast super-resolution mri through a progressive network Dagan: Deep de-aliasing generative adversarial networks for fast compressed sensing mri reconstruction Deep generative adversarial neural networks for compressive sensing mri Compressed sensing mri reconstruction using a generative adversarial network with a cyclic loss Parallel imaging with a combination of sensitivity encoding and generative adversarial networks Pic-gan: A parallel imaging coupled generative adversarial network for accelerated multi-channel mri reconstruction Unsupervised mri reconstruction via zero-shot learned adversarial transformers Prior-guided image reconstruction for accelerated multi-contrast mri via generative adversarial networks Image synthesis in multi-contrast mri with conditional generative adversarial networks Semi-supervised learning of mutually accelerated multicontrast mri synthesis without fully-sampled ground-truths Image synthesis with adversarial networks: A comprehensive survey and case studies Wasserstein gan Unpaired image-to-image translation using cycle-consistent adversarial networks Deep attentive wasserstein generative adversarial networks for mri reconstruction with recurrent contextawareness Mri reconstruction via cascaded channel-wise attention network Self-attention convolutional neural network for improved mr image reconstruction Sara-gan: Self-attention and relative average discriminator based generative adversarial networks for fast compressed sensing mri reconstruction Cagan: a cycle-consistent generative adversarial network with attention for low-dose ct imaging Deep residual learning for image recognition Deep admm-net for compressive sensing mri Admm-net: A deep learning approach for compressive sensing mri Sca-cnn: Spatial and channel-wise attention in convolutional networks for image captioning SUPPLEMENTARY MATERIALS Figure 1. When the upsampling factor is 2, the histogram of the HR image generated by each method under the T2 healthy subjects dataset in Figure 3 When the upsampling factor is 2, the histogram of the HR image generated by each method under the T2W SPAIR healthy subjects dataset in Figure 3 When the upsampling factor is 2, the histogram of the HR image generated by each method under the mDIXON-W healthy subjects dataset in Figure 3 When the upsampling factor is 2, the histogram of the HR image generated by each method under the T2W SPAIR patients Kurt: kurtosis; skew: skewness When the upsampling factor is 2, the histogram of the HR image generated by each method under the mDIXON-W patients Kurt: kurtosis; skew: skewness