Abstract
Deep learning-based super-resolution (SR) mitigates resolution bottlenecks in brain MRI, yet existing CNNs and Transformers struggle to balance global context, computational efficiency, and high-frequency detail fidelity. We propose GF-BrainSR, an efficient linear-complexity State Space Model (SSM) framework. GF-BrainSR introduces a DWT-driven Frequency Adaptive Gate that explicitly decouples and enhances structural boundaries, while a depthwise spatial gate robustly suppresses noise in homogeneous soft tissues. Additionally, an Implicit Coordinate Upsampling module fuses continuous spatial coordinates with local features, eliminating discrete interpolation artifacts to reconstruct morphologically continuous tissue interfaces. Experiments demonstrate that GF-BrainSR achieves competitive performance, delivering superior anatomical clarity and spatial resolution with linear computational complexity. The source code will be released upon acceptance.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/EMERGE_040.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=XgEnneVBqZ
BibTex
@InProceedings{TonBor_GFBrainSR_MICCAISAT2026,
author = { Tong, Boren AND Fan, Linxuan AND Yang, Panhui AND Jiang, Juntao},
title = { { GF-BrainSR: Gated Frequency-Aware Selective State Space Model-Based Brain MRI Image Super-resolution } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17266},
month = {pending},
page = {pending}
}
