Swin Deformable Attention U-Net Transformer (SDAUT) for Explainable Fast MRI
This work addresses computational bottlenecks for researchers and practitioners in medical imaging, offering an incremental improvement in fast MRI reconstruction.
The authors tackled the problem of high computational cost in Transformer-based models for fast MRI reconstruction by proposing a new architecture that combines Shifted Windows Transformer with U-Net and incorporates deformable attention for explainability. Their method achieves consistently superior performance with fewer network parameters compared to state-of-the-art Transformer models.
Fast MRI aims to reconstruct a high fidelity image from partially observed measurements. Exuberant development in fast MRI using deep learning has been witnessed recently. Meanwhile, novel deep learning paradigms, e.g., Transformer based models, are fast-growing in natural language processing and promptly developed for computer vision and medical image analysis due to their prominent performance. Nevertheless, due to the complexity of the Transformer, the application of fast MRI may not be straightforward. The main obstacle is the computational cost of the self-attention layer, which is the core part of the Transformer, can be expensive for high resolution MRI inputs. In this study, we propose a new Transformer architecture for solving fast MRI that coupled Shifted Windows Transformer with U-Net to reduce the network complexity. We incorporate deformable attention to construe the explainability of our reconstruction model. We empirically demonstrate that our method achieves consistently superior performance on the fast MRI task. Besides, compared to state-of-the-art Transformer models, our method has fewer network parameters while revealing explainability. The code is publicly available at https://github.com/ayanglab/SDAUT.