Edge-enhanced dual discriminator generative adversarial network for fast MRI with parallel imaging using multi-view information

Abstract

In clinical medicine, magnetic resonance imaging (MRI) is one of the most important tools for diagnosis, triage, prognosis, and treatment planning. However, MRI suffers from an inherent slow data acquisition process because data is collected sequentially in k-space. In recent years, most MRI reconstruction methods proposed in the literature focus on holistic image reconstruction rather than enhancing the edge information. This work steps aside this general trend by elaborating on the enhancement of edge information. Specifically, we introduce a novel parallel imaging coupled dual discriminator generative adversarial network (PIDD-GAN) for fast multi-channel MRI reconstruction by incorporating multi-view information. The dual discriminator design aims to improve the edge information in MRI reconstruction. One discriminator is used for holistic image reconstruction, whereas the other one is responsible for enhancing edge information. An improved U-Net with local and global residual learning is proposed for the generator. Frequency channel attention blocks (FCA Blocks) are embedded in the generator for incorporating attention mechanisms. Content loss is introduced to train the generator for better reconstruction quality. We performed comprehensive experiments on Calgary-Campinas public brain MR dataset and compared our method with state-of-the-art MRI reconstruction methods. Ablation studies of residual learning were conducted on the MICCAI13 dataset to validate the proposed modules. Results show that our PIDD-GAN provides high-quality reconstructed MR images, with well-preserved edge information. The time of single-image reconstruction is below 5ms, which meets the demand of faster processing.

Promoting fast MR imaging pipeline by full-stack AI

“… 32 A pioneer work is to perform CS-MRI reconstruction with convolutional neural network (CNN). 2 Similarly, in, 33 the authors adopt Generative Adversarial Networks (GANs) to further refine the reconstructed images and improve the perceptual quality and diagnostic viability. Different from conventional network architectures, Transformer models 34 were introduced to capture long-range dependencies and hierarchical structures in image data and showed improved reconstruction of intricate anatomical structures.…”

Swin Deformable Attention U-Net Transformer (SDAUT) for Explainable Fast MRI

“…Our proposed SDAUT was compared with other fast MRI methods, e.g., DAGAN [26], nPIDD-GAN [9], Swin-UNet [1] and SwinMR [11] using Gaussian 1D 30% and radial 10% mask.…”

SIDM-WIR: A Sampling Improved DiffusionModel for Worksite Image Restoration

“…CNN-based IR methods [17,18] achieve impressive performance on IR. After that, GANs [19][20][21] became more popular compared with traditional IR methods. However, it is not worth that the training process of GANs is challenging as it requires achieving Nash equilibrium between the generator and discriminator, which may not always converge in practice and can lead to fluctuations or oscillations in model parameters.…”