# Revisiting Global Token Mixing in Task-Dependent MRI Restoration: Insights from Minimal Gated CNN Baselines

arXiv: https://arxiv.org/abs/2603.01449 · HTML: https://mlsanity.com/paper/2603.01449/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 32 references that are in the map.
- First posted on arXiv 2 Mar 2026.

## Builds on (most cited first, 24 of 24)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [SwinIR: Image Restoration Using Swin Transformer](https://mlsanity.com/paper/2108.10257/) (2021)
- [Restormer: Efficient Transformer for High-Resolution Image Restoration](https://mlsanity.com/paper/2111.09881/) (2021)
- [Simple Baselines for Image Restoration](https://mlsanity.com/paper/2204.04676/) (2022)
- [fastMRI: An Open Dataset and Benchmarks for Accelerated MRI](https://mlsanity.com/paper/1811.08839/) (2018)
- [A Deep Cascade of Convolutional Neural Networks for Dynamic MR Image Reconstruction](https://mlsanity.com/paper/1704.02422/) (2017)
- [End-to-End Variational Networks for Accelerated MRI Reconstruction](https://mlsanity.com/paper/2004.06688/) (2020)
- [MambaOut: Do We Really Need Mamba for Vision?](https://mlsanity.com/paper/2405.07992/) (2024)
- [A Comparative Study of Image Restoration Networks for General Backbone Network Design](https://mlsanity.com/paper/2310.11881/) (2023)
- [Swin Transformer for Fast MRI](https://mlsanity.com/paper/2201.03230/) (2022)
- [ReconFormer: Accelerated MRI Reconstruction Using Recurrent Transformer](https://mlsanity.com/paper/2201.09376/) (2022)
- [KiU-Net: Towards Accurate Segmentation of Biomedical Images using Over-complete Representations](https://mlsanity.com/paper/2006.04878/) (2020)
- [Xformer: Hybrid X-Shaped Transformer for Image Denoising](https://mlsanity.com/paper/2303.06440/) (2023)
- [Restore-RWKV: Efficient and Effective Medical Image Restoration with RWKV](https://mlsanity.com/paper/2407.11087/) (2024)
- [HUMUS-Net: Hybrid unrolled multi-scale network architecture for accelerated MRI reconstruction](https://mlsanity.com/paper/2203.08213/) (2022)
- [Over-and-Under Complete Convolutional RNN for MRI Reconstruction](https://mlsanity.com/paper/2106.08886/) (2021)
- [MambaMIR: An Arbitrary-Masked Mamba for Joint Medical Image Reconstruction and Uncertainty Estimation](https://mlsanity.com/paper/2402.18451/) (2024)
- [Deform-Mamba Network for MRI Super-Resolution](https://mlsanity.com/paper/2407.05969/) (2024)
- [MMR-Mamba: Multi-Modal MRI Reconstruction with Mamba and Spatial-Frequency Information Fusion](https://mlsanity.com/paper/2406.18950/) (2024)
- [LSNet: See Large, Focus Small](https://mlsanity.com/paper/2503.23135/) (2025)
- [Progressive Divide-and-Conquer via Subsampling Decomposition for Accelerated MRI](https://mlsanity.com/paper/2403.10064/) (2024)
- [Exploring the Low-Pass Filtering Behavior in Image Super-Resolution](https://mlsanity.com/paper/2405.07919/) (2024)
- [Boosting ViT-based MRI Reconstruction from the Perspectives of Frequency Modulation, Spatial Purification, and Scale Diversification](https://mlsanity.com/paper/2412.10776/) (2024)
- [DH-Mamba: Exploring Dual-domain Hierarchical State Space Models for MRI Reconstruction](https://mlsanity.com/paper/2501.08163/) (2025)

## Abstract (arXiv)

Global token mixing, implemented via self-attention or state-space sequence models, has become a popular model design choice for MRI restoration. However, MRI restoration tasks differ substantially in how their degradations vary over image and k-space domains, and in the degree to which global coupling is already imposed by physics-driven data consistency terms. In this work, we ask the question whether global token mixing is actually beneficial in each individual task across three representative settings: accelerated MRI reconstruction with explicit data consistency, MRI super-resolution with k-space center cropping, and denoising of clinical carotid MRI data with spatially heteroscedastic noise. To reduce confounding factors, we establish a controlled testbed comparing a minimal local gated CNN and its large-field variant, benchmarking them directly against state-of-the-art global models under aligned training and evaluation protocols. For accelerated MRI reconstruction, the minimal unrolled gated-CNN baseline is already highly competitive compared to recent token-mixing approaches in public reconstruction benchmarks, suggesting limited additional benefits when the forward model and data-consistency steps provide strong global constraints. For super-resolution, where low-frequency k-space data are largely preserved by the controlled low-pass degradation, local gated models remain competitive, and a lightweight large-field variant yields only modest improvements. In contrast, for denoising with pronounced spatially heteroscedastic noise, token-mixing models achieve the strongest overall performance, consistent with the need to estimate spatially varying reliability. In conclusion, our results demonstrate that the utility of global token mixing in MRI restoration is task-dependent, and it should be tailored to the underlying imaging physics and degradation structure.