Abstract
Magnetic resonance imaging (MRI) is a significant non-invasive clinical tool that can provide multi-contrast images of the same anatomical structures. Multi-contrast super-resolution (SR) utilizes the complementary information of MRI from different contrasts to achieve higher quality images. However, diffusion models used for image generation and reconstruction usually require large computational resources due to extensive iterative computations, which hinders adaptability in resource-constrained environments. To solve this problem, we propose a Multi-Contrast Enhanced Lightweight Diffusion Model (MCLDM) for MRI. The MCLDM model includes the developed Multi-Contrast Dual Attention (MCDA) module and the designed cross-channel attention mechanism modules. Specifically, the MCDA module integrates multi-contrast images into the diffusion model and applies a dual attention mechanism to these features, effectively compensating for the limitations of single-modality information and enhancing the capability of focusing on complementary information. Additionally, the designed cross-channel attention module improves feature extraction capability and computational efficiency. Experimental results show that MCLDM achieves high-quality reconstruction and balances performance and computational complexity, reducing image generation time compared to other diffusion methods.