A Lightweight Multi-Stage Fusion Network for Multi-Contrast MRI Super-Resolution

Magnetic resonance imaging (MRI) offers multi-contrast imaging, allowing super-resolution (SR) methods to utilize complementary modality information for improved quality. However, current multi-contrast SR approaches remain computationally heavy with limited lightweight solutions, limiting their practical use. To address this, we present a lightweight Multi-Stage Fusion Network (MSFN) for multi-contrast MRI SR. The proposed model employs hierarchical feature learning and cross-modality feature refinement stage to efficiently integrate anatomical details from low-resolution and reference images. Experimental results show that our method achieves superior performance with significantly lower computational costs, making it well-suited for clinical applications requiring fast and efficient SR reconstruction.