Fast MRI Reconstruction: How Powerful Transformers Are?

TL;DR

This paper investigates the effectiveness of transformer-based neural networks, especially GAN-enhanced Swin transformers, for accelerating MRI reconstruction from undersampled data, showing improved image quality under various conditions.

Contribution

It introduces novel GAN-based Swin transformer architectures for MRI reconstruction, including edge and texture enhancements, and compares their performance with other models.

Findings

Transformers perform well for MRI reconstruction from undersampled data.

GAN structures enhance image quality when undersampling is 30% or higher.

Proposed models outperform some existing CNN-based methods in key metrics.

Abstract

Magnetic resonance imaging (MRI) is a widely used non-radiative and non-invasive method for clinical interrogation of organ structures and metabolism, with an inherently long scanning time. Methods by k-space undersampling and deep learning based reconstruction have been popularised to accelerate the scanning process. This work focuses on investigating how powerful transformers are for fast MRI by exploiting and comparing different novel network architectures. In particular, a generative adversarial network (GAN) based Swin transformer (ST-GAN) was introduced for the fast MRI reconstruction. To further preserve the edge and texture information, edge enhanced GAN based Swin transformer (EES-GAN) and texture enhanced GAN based Swin transformer (TES-GAN) were also developed, where a dual-discriminator GAN structure was applied. We compared our proposed GAN based transformers, standalone Swin transformer and other convolutional neural networks based GAN model in terms of the evaluation metrics PSNR, SSIM and FID. We showed that transformers work well for the MRI reconstruction from different undersampling conditions. The utilisation of GAN's adversarial structure improves the quality of images reconstructed when undersampled for 30% or higher. The code is publicly available at https://github.com/ayanglab/SwinGANMR.